The Law of Crime Concentration at Places: Editors’ Introduction
- 2.3k Downloads
The empirical observation that a small number of micro places generate the bulk of urban crime problems has become a criminological axiom. These micro places, commonly referred to as “hot spots” of crime, have been defined as clusters of street addresses, groupings of street blocks, and particular street intersections and street segments (Weisburd et al. 2009). High-activity crime places have been found in neighborhoods characterized by both low and high levels of social disadvantage (Weisburd and Green 1995; Curman et al. 2015). More recently, researchers have also demonstrated that the concentration of crime at particular places is stable over time (Weisburd et al. 2004; Braga et al. 2010; Andresen and Malleson 2011). The stability of crime at hot spots reinforces the logic that underlies place-based prevention strategies. Without intervention, crime hot spots are not likely to disappear in short time periods. Concentrating crime prevention resources on such places makes good sense both for short term and long term gains (Braga and Weisburd 2010; Weisburd et al. 2004).
Reflecting on the larger body of empirical evidence and his own analyses of crime in five larger cities and three smaller cities, Weisburd (2015, p. 133) suggests a “law of crime concentration” at places which posits, “that for a defined measure of crime at a specific microgeographic unit, the concentration of crime will fall within a narrow bandwidth of percentages for a defined cumulative proportion of crime.” He further specifies the term “bandwidth of percentages” as a specific cumulative proportion of crime, such as 25 or 50% of crime in a city, which would be very narrow for a standard unit of crime and geography. In his analysis of crime variability across smaller and larger cities, Weisburd (2015, p. 143) concludes that there was a tight bandwidth of crime concentration at places supporting the proposed law: for 50% concentration, that bandwidth is about 4% (from 2 to 6%), and for 25% concentration, that bandwidth is less than 1.5% (from 4 to 1.6%).
In light of the existing empirical evidence to support the law of crime concentration at places, Weisburd (2015) put forth a research agenda for those working in the crime and place literature. First, although there is support for the law of crime concentration at places in many cities of different sizes and on different continents, Weisburd (2015) refers to the relatively limited sample (compared to individual-level research) as a convenience sample. Consequently, Weisburd (2015) asks is the law of crime concentration at places is truly generalizable? Moreover, are there any situations in which the law does not apply? In a slightly different vein of thought, is the law of crime concentration at places a property of crime events, or simply an artifact of human activity patterns that are concentrated at places? Moving beyond the question of whether the law of crime concentration at places exists independently of human activity patterns, Weisburd (2015) also asks why crime concentrates in such a manner? Our traditional theories that explain the geography of crime (routine activity theory, rational choice theory, and the geometry of crime) all consider criminal events at the micro-spatial unit of analysis, but research has shown that neighborhood level characteristics help explain micro-spatial crime patterns (Weisburd et al. 2012). And finally, are these crime concentrations truly stable over time as indicated by much of the current research? If so, why is this the case? Are the social characteristics of places constant such that crime concentrations follow suit?
Despite a growing theoretical and scientific base, gaps remain in our knowledge of crime and place. Moreover, crime prevention theory and practice often continue to ignore the importance of place. The law of crime concentration at places suggests an analytical framework that can enhance our theoretical understanding of the crime problem and improve our capacity to prevent crime by focusing attention on persistent problem places. This special issue of the Journal of Quantitative Criminology is devoted to advancing the empirical study of the law of crime concentration at places. The ten papers published in this peer-reviewed volume use state-of-the-art statistical models to examine timely and important empirical questions related to key elements of the law put forth by Weisburd (2015).
In the opening article, Andresen, Curman, and Linning conduct longitudinal analyses of 16 years of crime incident data at street segments and intersections in Vancouver, BC, between 1991 and 2006. Importantly, the authors disaggregate their analyses of total crime data into seven specific types of crime: assault, burglary, robbery, theft, theft of vehicle, theft from vehicle, and other crimes. Their analyses test the law of crime concentration of crime at places and represent an extension of the group-based trajectory models used in the seminal Weisburd et al. (2004) paper examining the concentration and stability of crime at micro places over time. While they found strong overall support for the law in Vancouver, Andresen and colleagues also noted that place trajectories varied for disaggregated crime types, suggesting the importance of specificity when developing theory to understand and policy to address particular crime places.
Bernasco and Steenbeek suggest that a standard methodology should exist for measuring and reporting crime concentration. Indeed, much of the existing empirical evidence on the concentration of crime at places uses varying units of analysis and varying methodological approaches. Bernasco and Steenbeek argue that researchers should use the Lorenz curve as a method for providing a detailed description of crime concentrations and the Gini coefficient as a summary measure. Using crime incident data from The Hague, Netherlands, Bernasco and Steenbeek show that the generalized Gini coefficient and Lorenz curves provide better descriptions of crime concentration, especially in situations when the crime data are sparse. They conclude that these modified approaches not only provide a method for comparing distributions across different locales, but may also serve analysts as a way to better understand the underlying nature of crime concentrations.
Schnell, Braga, and Piza analyze violent crime incidents in Chicago, IL, 2001–2014, at multiple scales of analysis: street segments, neighborhood clusters, and community areas. Schnell and colleagues use these data to investigate crime concentrations with Lorenz curves, Gini coefficients, and linear mixed models to test aspects of the law of crime concentration at places. Overall, these authors found that roughly 5–7% of street segments accounted for 50% of violent crime incidents, depending on the year—approximately 20% of the neighborhood clusters and community areas were necessary to account for 50% of violent crime incidents. And the Lorenz curves that show the percentage of crime relative to the percentage of spatial units showed a high degree of concentration. Moreover, their use of Gini coefficients reinforced this finding by showing a high degree of spatial inequality/concentration for violent crime incidents. This use of multiple statistics is important because it shows the law of crime concentration at places is robust to various types of measurement. The core of their study, however, is to investigate the amount of variation in violent crime incidents that can be attributed to the 3 different spatial units of analysis, replicating and methodologically extending the work of Steenbeek and Weisburd (2015). Overall, Schnell and colleagues find that one-half to two-thirds of the variation in violent crime incidents in Chicago can be attributed to street segments. In other words, if one is to only consider larger areal units such as neighborhood clusters and community areas, the vast majority of spatial variation would be missed in any analyses. This shows the importance of the micro-spatial unit of analysis in any attempt to understand spatial patterns of crime.
Hibdon, Telep, and Groff investigate the law of crime concentration at places in Seattle, WA using calls-for-service police data and emergency medical services data for drug activity, 2009–2014. Using a variety of statistical techniques, Hibdon and colleagues test the degree of crime concentration and its temporal stability. Both data sets exhibit crime concentrations as with previous research, 50% of calls (police and medical) are accounted for with less than 2% of street segments. More interesting, however, is that they only find moderate levels of local-level stability over time with regard to the spatial patterns of the data from year to year. Specifically, they find that most of the stability present in the data emerges from zero and low-volume street segments. Moreover, Hibdon and colleagues find that calls locations for emergency services and not the same as those for police calls-for-service. This is important because the use of only one data set will miss important sites of drug activity, in this particular context. Of course, this should not come as a complete surprise because not all drug activity will necessitate the participation of both of these public safety services.
Gill, Wooditch, and Weisburd conduct a group-based trajectory analysis of crime incidents on street segments between 2000 and 2014 in Brooklyn Park, a suburb of Minneapolis, MN. They find very strong support for the law of crime concentration at places in this suburban setting: only 2% of street segments generated 50% of crime and 0.4% of street segments generated 25% of crime. These observed crime concentrations were highly stable over time. However, the authors also note two important differences relative to similar analyses in major urban areas. The concentration of crime in Brooklyn Park is much higher and there is less street-to-street variability in the spatial distribution of crime. Nonetheless, their findings suggest that place-based crime prevention initiatives can be used to good effect in non-urban settings.
Haberman, Sorg, and Ratcliffe use data on robberies at street blocks and intersections in Philadelphia, PA to test whether the bandwidth percentages suggested by the law of crime concentration hold using different theoretically-relevant temporal scales—hours of the day, days of the week, and seasons of the year. They find a very close match between Weisburd’s (2015) benchmarks and the cumulative percentages of blocks and intersections experiencing 25 and 50% of the street robberies across the three temporal scales. Their analyses also suggest that the spatial concentration of crime at micro places and the temporal concentration of crime within-day periods act separately from one another with some areas experiencing consistent patterns of crime, while other locations experiencing crime only during specific within-day periods. These findings suggest that more attention should be paid to the interaction of space and time in understanding the concentration of crime events.
Rosser, Davies, Bowers, Johnson, and Cheng extend the law of crime concentration to compare varying ways to predict future high-activity crime places. While predicting patterns of criminal behavior is not a new idea, prior research has often emphasized larger aggregate areas such as police beats or user-defined grid areas. Rosser and colleagues use a network-based analysis of street segments for the prospective identification of high crime areas and compare these findings to the more traditional grid-based analysis. Using property crime data from a large metropolitan police force in the United Kingdom, their comparative analysis suggests that the calibrated network-based model substantially outperforms a grid-based alternative in terms of predictive accuracy. Their street network models generated approximately 20% more crime identified at a coverage level of 5%. This suggests that network-based methods of crime forecasting should be used to guide place-based crime prevention initiatives such as hot spots policing.
Two articles in this special issue raise methodological and substantive challenges to the law of crime concentration at places. Hipp and Kim use two measures employing different temporal assumptions to analyze crime data at street segments between 2005 and 2012 for 42 cities in southern California with at least 40,000 residents. They find considerable variability in the top 5% of street segments across the cities and raise questions about the temporal scales applied in empirical investigations of crime concentrations at places over time. Levin, Rosenfeld, and Deckard analyze annual counts of violent and property crimes at street segments in St. Louis, MO, between 2000 and 2014 and specifically examine temporal stability in the spatial distribution of crime at places. Like Hipp and Kim, they find notable volatility in the places that comprise the top 5% of street segments with the highest crime frequencies each year. Levin et al. also find that the spatial concentration of crime is reduced when crime-free street segments were excluded from the crime distributions. While these studies produce results that support the fundamental premise of the law, the empirical analyses also suggest considerable mobility into and out of the category of places with the highest crime frequencies over the study time period.
In the final article, O’Brien and Winship use citizen calls-for-service and non-emergency request data in Boston, MA, to assess the value gained by the use of street addresses in understanding crime concentrations beyond the analysis of neighborhoods and street segments. The findings of a series of multi-level Poisson regression models supported the law of crime concentration at places, while also suggesting a more nuanced approach to understanding these crime clusters. Their analyses find that less than 1% of addresses generated 25% of citizens’ reports of crime and disorder and, though there was significant clustering at the street segment and neighborhood levels, some 95–99% of the variance was at the address level. While street segments should remain an important focus of crime prevention interventions, O’Brien and Winship encourage law enforcement agencies to identify and deal with problem properties on high-crime street segments that could account for a bulk of the crime at specific places.
Overall, the articles in this special issue offer strong support for the key components of Weisburd’s (2015) law of crime concentration at places. In general, the empirical evidence presented here reveals that the concentration of crime at micro places falls within narrow bandwidths of percentages for a defined cumulative proportion of crime across time. Depending on the temporal scales applied and the specific research setting, there may be some volatility in the places that comprise the highest crime concentration category over extended time periods. Nonetheless, the existing research shows that the law holds for a variety of jurisdictions, in urban and suburban environments, and across developed nations.
The implication for theoretical development is clear: micro spatial units of analysis are critical for understanding the patterns of crime. As outlined by Sherman et al. (1989, p. 43), dangerous neighborhoods are generally safe. Because of this fact, neighborhood level analyses stop short of understanding the spatial distribution of crime. This is supported, specifically, in articles within this special issue by showing the majority of spatial variation in crime is at the micro spatial level of analysis. Such findings are completely consistent with our existing spatial theories of crime: motivated offenders and suitable targets converge at specific (micro) places, activities spaces generally do not include entire neighborhoods, and rational choices are situationally (spatially and temporally) specific. We must, therefore, develop these theories at the micro place in order to move forward. To return to a variation of one of Weisburd’s (2015) questions, why does crime occur at this micro place and not that micro place?
It follows that crime prevention policy and practice should be focused on these generally-stable high crime places that generate the bulk of crime and disorder problems in jurisdictions. Rigorous evaluations of place-based crime prevention strategies, such as hot spots policing (Braga et al. 2014), reveal the significant crime reduction benefits of such an approach. Indeed, because the convergences, activity spaces, and choices that fundamentally lead to crime at relatively few micro places are generally, albeit not perfectly, stable over time, crime- and place-specific crime prevention activities have theoretical and empirical support. By focusing on these generally-stable high crime places that generate the bulk of crime and disorder, the benefit-to-cost ratios of crime prevention efforts can be expected to be high.
There are indeed many research questions that still need to be answered, including those posed by Weisburd (2015). However, the existing body of theoretical and empirical evidence assembled thus far suggests that “place-based criminology” has great promise for advancing our capacity to understand and address crime problems.