The Action Is Everywhere, But Greater at More Localized Spatial Scales: Comparing Concentrations of Crime across Addresses, Streets, and Neighborhoods

• Author: Daniel T. O’Brien
• DOI: 10.1177/0022427818806040
• Date: 01 May 2019
• Published date: 01 May 2019
• Subject Matter: Articles

Article

The Action Is

Everywhere, But

Greater at More

Localized Spatial

Scales: Comparing

Concentrations of Crime

across Addresses, Streets,

and Neighborhoods

Daniel T. O’Brien

1,2

Abstract

Objectives: Recent work has debated which geographic scale is most relevant

to understanding the clustering of crime and disorder across a city. This

study introduces nested Gini coefficients that help answer this question by

disentangling concentrations of crime at multiple scales in a single city while

also controlling for artifacts of arithmetic and urban form. Methods: The

study examines six indices of crime and disorder drawn from requests for

government services received by the City of Boston in 2011 for addresses

(N¼98,355) nested in street segments (N¼13,048) nested in census

1

Boston Area Research Initiative, Northeastern University, Boston, MA, USA

2

School of Public Policy and Urban Affairs and School of Criminology and Criminal Justice,

Northeastern University, Boston, MA, USA

Corresponding Author:

Daniel T. O’Brien, Boston Area Research Initiative, School of Public Policy and Urban Affairs

and School of Criminology and Criminal Justice, Northeastern University, 1135 Tremont St.,

360Y Renaissance Park, Boston, MA 02120, USA.

Email: d.obrien@neu.edu

Journal of Research in Crime and

Delinquency

2019, Vol. 56(3) 339-377

ªThe Author(s) 2018

Article reuse guidelines:

sagepub.com/journals-permissions

DOI: 10.1177/0022427818806040

journals.sagepub.com/home/jrc

tracts (N¼178). Nested Gini coefficients assessed the average concentra-

tion at each level independent of the higher geographic unit (e.g., the streets

of a single tract). Results: Concentrations were greatest at addresses, then at

streets,and then at tracts. Compared to whole-citycalculations, they showed

equal or greater levels of concentration of crime and disorderfor addresses,

but lower concentrations for streets.Controlling for the numberof locations

on a street or in a tract also markedly diminished concentrations. Conclusions:

The findings indicate a continued need to explain concentrations of crime,

especially at localized geographic scales.

Keywords

law of concentration of crime, computational social science, problem

properties, hotspot streets

One of the basic motivations for urban criminology is the observation that

crime and disorder are unevenly distributed across theurban landscape. Over

the years, though, multiple perspectives have emphasized the importance of

one geographic scale of analysis or another for observing and understanding

this variation.Classical thought emphasized “high-risk neighborhoods”(e.g.,

Park, Burgess, and McKenzie [1925] 1984; Sampson 2012; Shaw and

McKay [1942] 1969), but recent trends have redirected attention to more

localized contexts, including “hotspot streets” (e.g., Andresen and Malleson

2011; Braga, Papachristos, and Hureau 2010; Weisburd 2015; Weisburd,

Groff, and Yang 2012) and “problem properties” (or “hotdots”; Farrell and

Pease 2001; Johnson, Bowers, and Hirschfield 1997; O’Brien and Winship

2017; Sherman, Gartin, and Buerger 1989). Although considerable workhas

provided evidencefor the relevance of each of thesegeographic scales, only a

few studies have analyzed them simultaneously in order to compare their

relativelevels of concentration. Thisis a particularlytricky task because these

geographic scales are inherently interdependent—addresses sit on streets

which lie within neighborhoods—creating the need for analytic approaches

that can disentangle concentrations at one level from the others.

The current study introduces a new methodological approach of nested

Gini coefficients to disentangle concentrations of crime and disorder across

multiple geographic scales. It applies this technique to a database of

requests for government services (i.e., 911 and 311 cal ls) from Boston,

MA, to examine the relative concentration of six types of crime and disorder

at addresses, street segments, and census tracts. Gini coefficients quantify

340 Journal of Research in Crime and Delinquency 56(3)

inequality in a distribution, but nesting them within geographic levels

makes it possible to assess concentrations at each geographic scale inde-

pendently of concentrations that exist at higher levels of aggregation (e.g.,

separating concentrations at addresses from concentrations across the

streets that contain them). The study will also leverage addition al tech-

niques to address two other challenges to the interpretation of concentra-

tions of crime and disorder: (1) the arithmetic consequences of the rarity of

crime, which can inflate Gini estimates, and (2) aspects of urban form that

can influence the distribution of crime and disorder, including land use and

the density of locations where such events might occur. Altogether, the

approach permits a comprehensive assessment of concentration at all three

levels, controlling for multiple arithmetic and compositional artifacts that

have occluded previous work. Before the presentation of data and analyses,

the proceeding sections summarize in greater depth how the history of the

study of crime concentrations and the statistical and conceptual challenges

facing their examination shape the current study.

Concentrations of Crime: A Methodological Challenge

Scholars have sought to understand the uneven distribution of outcomes

across a city’s neighborhoods for at least 150 years (e.g., Booth 1903; May-

hew 1862), making it one of the oldest themes in urban science. In the early

twentiethcentury, the Chicago School of Sociologytook particular aim at this

subject, probing inequalities in crime, health, and education across commu-

nities (Park and Burgess 1925; Shaw and McKay [1942] 1969). This work

has since provided t he conceptual basis for decades of r esearch on the social,

demographic,and physical features of neighborhoods that can influencelocal

levels of crime (Browning, Soller, and Jackson 2015; Cohen et al. 2000;

Gibson et al. 2010; Kawachi and Berkman 2003; Leventhal and Brooks-

Gunn 2000; Raudenbush and Sampson 1999; Sampson, Raudenbush, and

Earls 1997). A more recentline of work, however, has highlighted the extent

to which microplaces, or streets and address es, contribute to t he distribution

of outcomes, most notably crime and disorder, across the city. This

“criminology of place” has debunked the assumption that neighborhoods are

homogenous regions, demonstrating that many streets in high-crime neigh-

borhoodsactually experience littlecrime, and that, conversely,there are high-

crime streets in low-crime neighborhoods.

Criminology of place began with two studies in the late 1980s in two

different cities, each demonstrating that *3 percent of addresses accounted

for 50 percent of crime events (Pierce, Spaar, and Briggs 1988; Sherman

O’Brien 341