ROLE OF THE STREET NETWORK IN BURGLARS' SPATIAL DECISION‐MAKING*

• Date: 01 May 2017
• Published date: 01 May 2017
• DOI: http://doi.org/10.1111/1745-9125.12133

ROLE OF THE STREET NETWORK IN BURGLARS’

SPATIAL DECISION-MAKING∗

MICHAEL J. FRITH,1SHANE D. JOHNSON,1

and HANNAH M. FRY2

1Department of Security and Crime Science, University College London

2Bartlett Centre for Advanced Spatial Analysis, University College London

KEYWORDS: burglary, road network, graph theory, discrete choice, mixed logit

Explaining why crime is spatially concentrated has been a central theme of much

criminological research. Although various theories focus on neighborhood social pro-

cesses, environmental criminology asserts that the physical environment plays a central

role by shaping people’s activity patterns and the opportunities for crime. Here, we test

theoretical expectations regarding the role of the road network in shaping the spatial

distribution of crime and, in contrast to prior research, disentangle how it might influ-

ence offender awareness of criminal opportunities and the supply of ambient guardian-

ship. With a mixed logit (discrete choice) model, we use data regarding (N=459)

residential burglaries (for the first time) to model offender spatial decision-making at

the street segment level. Novel graph theory metrics are developed to estimate offender

awareness of street segments and to estimate levels of ambient guardianship, distin-

guishing between local and nonlocal guardianship. As predicted by crime pattern the-

ory, novel metrics concerning offender familiarity and effort were significant predictors

of residential burglary location choices. And, in line with Newman’s (1972) concept of

defensible space, nonlocal (local) pedestrian traffic was found to be associated with an

increase (decrease) in burglary risk. Our findings also demonstrate that “taste” prefer-

ences vary across offenders, which presents a challenge for future research to explain.

That crime is spatially concentrated now seems incontestable (e.g., Eck and Weisburd,

1995). Explaining why this is so, however, is still a central theme of criminological research

and a matter of some debate. Several prevailing theories assert that the environment

plays a central role in shaping the distribution of crime by facilitating the convergence

in space and time of offenders and suitable targets, in the absence of capable guardians

(Cohen and Felson, 1979). One fundamental determinant of this is the road network

because it defines how people move through the urban environment. In so doing, it serves

∗Additional supporting information can be found in the listing for this article in the Wiley Online

Library at http://onlinelibrary.wiley.com/doi/10.1111/crim.2017.55.issue-2/issuetoc.

This project was funded by EPSRC grant no. EP/G037264/1 as part of the UCL Department of

Security and Crime Science’s Doctoral Training Centre.

Direct correspondence to Shane D. Johnson, Department of Security and Crime Sci-

ence, University College London, 35 Tavistock Square, London WC1H 9EZ, U.K. (e-mail:

shane.johnson@ucl.ac.uk).

This is an open access article under the terms of the Creative Commons Attribution License, which per-

mits use, distribution and reproduction in any medium, provided the original work is properly cited.

C2017 The Authors. Criminology published by Wiley Periodicals, Inc. on behalf of American Society of

Criminology doi: 10.1111/1745-9125.12133

CRIMINOLOGY Volume 55 Number 2 344–376 2017 344

OFFENDER SPATIAL DECISION-MAKING 345

a “dual function,” determining the opportunities for crime offenders encounter and be-

come aware of (e.g., Beavon, Brantingham, and Brantingham, 1994), and influencing the

locations through which ordinary citizens move and provide ambient guardianship. The

aim of prior research has been to examine the role of the road network on crime, but it

has failed to isolate the influence of these two mechanisms. This is largely because analy-

ses have been conducted to examine only the location of crime events, without reference

to the offenders involved. In addition, most researchers have tended to employ crude

metrics to describe the network, and their findings have often lacked statistical rigor.

To address these shortcomings, in this article, we make several novel contributions.

We use a discrete choice approach (McFadden, 1974), or more accurately an offense lo-

cation choice approach (Bernasco and Nieuwbeerta, 2005), to estimate empirically how

the opportunities that are targeted by burglars differ from those that are not. Our re-

search differs from previous studies of offender spatial decision-making in two important

ways. First, in most previous studies, scholars have examined offender location choice

at the area level (for the exceptions, see Bernasco, 2010b, and Vandeviver et al., 2015).

Here, consistent with contemporary theory (Weisburd, Groff, and Yang, 2012) and the

research questions at hand, we do so at the street segment level. Second, we build on

previous work (Davies and Johnson, 2015) and introduce to the discrete choice literature

a methodological approach that aims to estimate independently how the road network

influences offender awareness of crime opportunities, on the one hand, and guardian po-

tential at particular locations, on the other. By following Townsley, Birks, Ruiter, et al.

(2015), we also recognize that offenders may vary in the extent to which their spatial

decision-making is affected by different factors, and so we employ mixed logit statistical

models to estimate parameters and how they vary across offenders. In combination, our

approach allows us to examine the dual role that the road network might play in shaping

burglar spatial decision-making and the extent to which this varies across offenders.

The remainder of this article is organized as follows. In the next section, theoretical per-

spectives and research are reviewed to introduce the theoretical model. The second and

third sections describe the data and analytic strategy, respectively. The latter includes a

discussion of the application of graph theory to quantify the character of the road net-

work, as well as the statistical model employed. The fourth and fifth sections present and

discuss the results.

BACKGROUND

OFFENDERS AND THE ROAD NETWORK

The rational choice perspective (Cornish and Clarke, 1986) describes offenders, such

as burglars, as nonarbitrary decision-makers who consider (however briefly) the costs

and benefits of action alternatives, including the decision of where to offend (Clarke and

Felson, 1993). It is suggested that such decision-making will be bounded by imperfect and

incomplete information, and that future choices will be informed by the outcome of previ-

ous ones (Cornish and Clarke, 1986; see also Bennett and Wright, 1984; Cromwell, Olson,

and Avary, 1991; Wright and Decker, 1994). Nevertheless, although choices made may

not appear rational to an observer, it is assumed that selections are made that aim to max-

imize the perceived utility of a decision and minimize the expected costs. In particular, the

distance an offender must travel to offend has consistently been shown by researchers to

influence offender location choice in both quantitative and ethnographic studies.

346 FRITH, JOHNSON, & FRY

For example, offenders have consistently reported (e.g., Brown and Altman, 1981;

Rengert and Wasilchick, 1985; Repetto, 1974), and the results of analyses of the jour-

ney to crime (Snook, 2004; Wiles and Costello, 2000) have shown, that the likelihood of

an offender selecting a location at which to offend is inversely proportional to the dis-

tance he or she must travel to reach it. The aim of most studies of the journey to crime

has been to examine the Euclidean distance between locations (e.g., Clare, Fernandez,

and Morgan, 2009), and even though this provides a good estimate of the likely cost of

travel, it is clearly imperfect. The cost of travel is intrinsically linked to the configuration

and properties (such as the vehicle speed limit) of the road network as this determines

how quick and how easy it is to travel between any two locations. Consequently, in the

current study, we expect to find the following:

Hypothesis 1: Street segments that are quicker to travel to (in terms of estimated

travel time) from the burglar’s home will be more likely to be selected for burglary.

In addition to the fact that the selection of targets near to an offender’s home location

minimizes the cost of travel, this preference is expected to emerge for other reasons en-

capsulated by crime pattern theory (e.g., Brantingham and Brantingham, 1993). Accord-

ing to the theory, like everyone else, offenders are assumed to frequent activity nodes

routinely, such as their home and workplace. As a consequence of doing so, they develop

an awareness of these places and the routes between them. This is not necessary for of-

fending, but it shapes offenders’ familiarity and awareness of the criminal opportunities

within these spaces. According to crime pattern theory, offenders are believed to prefer

such opportunities to alternative targets for two reasons. First, offenders cannot select tar-

gets of which they are not aware (Rengert and Wasilchick, 1985). And, second, targeting

locations about which something is known reduces uncertainty about the likely outcome

of that choice (Beavon, Brantingham, and Brantingham, 1994). As such, we predict the

following:

Hypothesis 2: Street segments that are more likely to be familiar to an offender are

more likely to be selected for burglary.

In previous studies, scholars have assumed that the distance between a location and an

offender’s routine activity nodes (e.g., the home) provides a reasonable estimate of that

location’s familiarity. This assumption is not unreasonable as the findings from qualita-

tive research with burglars have revealed that they are typically most familiar with areas

that are closest to their home locations (e.g., Rengert and Wasilchick, 1985). Neverthe-

less, at the street segment level, an individual’s familiarity with locations will be more

nuanced than this, and it is likely to be a function of how frequently he or she travels to

or through them (Brantingham and Brantingham, 1993; Rengert and Wasilchick, 1985).

Although distance will influence this awareness, so too will the configuration of the road

network because this affects the likelihood that an individual will travel along a particu-

lar street to reach that or other locations (e.g., Beavon, Brantingham, and Brantingham,

1994). Some streets (such as major roads) may feature in many journeys, whereas oth-

ers less so (cul de sacs and dead ends, for instance). On the whole, streets rarely trav-

eled will be less familiar than those commonly used. Even though it is not possible to

determine which streets offenders travel during their routine activities without collect-

ing extensive primary data, it is possible to employ techniques from graph theory to