STREET NETWORK STRUCTURE AND CRIME RISK: AN AGENT‐BASED INVESTIGATION OF THE ENCOUNTER AND ENCLOSURE HYPOTHESES

• Published date: 01 November 2017
• DOI: http://doi.org/10.1111/1745-9125.12163
• Author: TOBY DAVIES,DANIEL BIRKS
• Date: 01 November 2017

STREET NETWORK STRUCTURE AND CRIME RISK:

AN AGENT-BASED INVESTIGATION OF THE

ENCOUNTER AND ENCLOSURE HYPOTHESES∗

DANIEL BIRKS1and TOBY DAVIES2

1Griffith Criminology Institute, Griffith University

2Department of Security and Crime Science, University College London

KEYWORDS: street networks, agent-based modeling, encounter/enclosure, residential

burglary, guardianship

Street networks shape day-to-day activities in complex ways, dictating where, when,

and in what contexts potential victims, offenders, and crime preventers interact with one

another. Identifying generalizable principles of such influence offers considerable util-

ity to theorists, policy makers, and practitioners. Unfortunately, key difficulties associ-

ated with the observation of these interactions, and control of the settings within which

they take place, limit traditional empirical approaches that aim to uncover mechanisms

linking street network structure with crime risk. By drawing on parallel advances in the

formal analyses of street networks and the computational modeling of crime events in-

teractions, we present a theoretically informed and empirically validated agent-based

model of residential burglary that permits investigation of the relationship between

street network structure and crime commission and prevention through guardianship.

Through the use of this model, we explore the validity of competing theoretical ac-

counts of street network permeability and crime risk—the encounter (eyes on the street)

and enclosure (defensible space) hypotheses. The results of our analyses provide sup-

port for both hypotheses, but in doing so, they reveal that the relationship between

street network permeability and crime is likely nonlinear. We discuss the ramifications

of these findings for both criminological theory and crime prevention practice.

One of the central principles of environmental criminology is the idea that crime can be

understood in terms of the interaction between the key actors involved in criminal events:

offenders, victims, and preventers (Brantingham and Brantingham, 1981). From a spatial

perspective, one of the most immediate corollaries of this is that the distribution of crime

should be influenced by urban morphology: The structure of the built environment deter-

mines the places people visit in the course of their day-to-day activities, the routes they

take in moving between them, and the interactions that they experience as they do so.

Motivated by this, a substantial volume of research has been aimed at examining spatial

∗We are grateful to Danielle Reynald, Mike Townsley, and Henk Elffers for their insightful feed-

back on earlier drafts of this article. In addition, we would like to thank Wayne Osgood and the

four anonymous reviewers for their perceptive and valuable comments.

Direct correspondence to Daniel Birks, Griffith Criminology Institute, Griffith University, 176

Messines Ridge Road, Room 3.01N, Technology (M10), Mt Gravatt Queensland 4122, Australia

(e-mail: d.birks@griffith.edu.au).

C2017 American Society of Criminology doi: 10.1111/1745-9125.12163

CRIMINOLOGY Volume 55 Number 4 900–937 2017 900

STREET NETWORK STRUCTURE AND CRIME RISK 901

theories of crime by exploring the existence and form of such relationships, with partic-

ular emphasis on the role the street network plays in shaping activity. The results of this

research indicate that not only does the concentration of crime display pronounced regu-

larities at the network level (Weisburd, 2015), but also that its variation can be reconciled

with the structure of the network (e.g., Davies and Johnson, 2015).

Despite the apparent relationship between network configuration and crime, however,

the mechanism by which activity patterns give rise to crime is not well understood. Al-

though environmental theories—most notably Brantingham and Brantingham’s geomet-

ric (Patricia Brantingham and Paul Brantingham, 1981) and pattern (Paul Brantingham

and Patricia Brantingham, 1993) theories of crime—provide a compelling rationale from

the perspective of offender target awareness, most extant empirical evidence is corre-

lational in nature. Furthermore, theoretical perspectives concerned with the influence

of other elements in the crime event are much less clear cut. Most notably, the role of

guardianship—the third element of the “crime triangle”—is the subject of several com-

peting discourses (Jacobs, 1961; Newman, 1972), with differing implications for the rela-

tionship between movement patterns and crime. This tension is typified by the disparity

between the “encounter” and “enclosure” principles of urban design.

The encounter and enclosure hypotheses are both concerned with the relationship be-

tween movement patterns and crime, but they differ in their assessment of the relative

contributions of competing mechanisms. Put simply, in the encounter hypothesis, it is as-

serted that the movement of people through places confers a guardianship effect, and

that places that experience greater use will therefore be safer. In contrast, in the enclo-

sure hypothesis, it is suggested that such a guardianship effect is overstated, and in fact, it

is outweighed by the increased exposure to offenders that frequent use implies; it there-

fore predicts that less readily used—that is, more “enclosed”—places will be safer. As

the extent to which places are used is determined to a large extent (although not com-

pletely) by their position within the wider urban configuration, the street network is of

clear relevance in assessing the relative merits of these arguments.

Evidently, therefore, understanding the way in which the street network shapes the in-

teractions that lead to (or, indeed, prevent) crime is of substantial potential significance

for theory. Investigating this issue empirically, however, is highly challenging: Although

correlational studies can reveal associations, the identification of causal mechanisms is

hindered by several significant obstacles. Some of these are logistical: It is not feasible,

for example, to manipulate the structure of real-world street networks systematically to

the extent that would be necessary to enable causal inferences in an experimental setting.

There are, however, more fundamental barriers to traditional approaches. The inherent

difficulty of observing behaviors of interest (e.g., individual movements) means that it

typically cannot be established with certainty that a hypothesized mechanism is responsi-

ble for an observed spatial pattern (O’Sullivan, 2004). This can be seen most starkly when

considering the role of guardianship, which is a phenomenon defined by the absence,

rather than by the presence, of an event: Establishing a counterfactual in such cases is

highly problematic.

In this article, we aim to gain new insights into the encounter and enclosure hypothe-

ses by combining two recent methodological advances from the field of environmental

criminology: 1) a formal approach to network analysis that has recently been applied

in empirical studies of crime and 2) a computational model of offender behavior that has

been shown to reproduce many features of real-world offending. Through the use of these

902 BIRKS & DAVIES

techniques, we carry out a series of simulated experiments exploring the relationship be-

tween street network structure, individual activity patterns, and several theoretical propo-

sitions describing crime commission and guardianship, with the aim of addressing funda-

mental questions that are mostly inaccessible to traditional empirical techniques. In doing

so, our primary aim is to examine the theoretical causal sufficiency of the encounter and

enclosure hypotheses by investigating the extent to, and means by, which guardianship ef-

fects, as shaped by the street network, influence the volume and distribution of property

crime, in this case—mirroring much of the extant empirical research—focusing on resi-

dential burglary. In building toward this, we also explore the relationship between net-

work structure and offending more generally, seeking insight into the behavioral mecha-

nisms that offer a causally explicit account of crime concentration at the street segment

level.

The remainder of the article is structured as follows. We begin by describing extant the-

oretical and empirical research that is concerned with the relationship between street net-

work structure and crime risk. We then outline the computational agent-based modeling

approach we leverage to increase this understanding. Subsequently, we describe our sci-

entific instrument: an agent-based model of residential burglary and its explicit theoretical

underpinnings. We integrate these sections as a means of specifying both the model itself

and the decisions made in its construction. After describing our instrument, we present

several tests of its validity before proceeding to set out a series of simulation experiments

in which it is used to address our primary research question. Subsequently, we present

our findings and discuss their ramifications for both the theories that underlie our model

and crime prevention interventions that draw on them. We conclude by discussing several

potential weaknesses of our approach and by setting out how, in continuing our research

in this area, we aim to address them.

THEORETICAL BACKGROUND

Much of the theory concerned with the relationship between environment and crime is

grounded in the basic framework provided by routine activity theory (Cohen and Felson,

1979). This sets out the fundamental observation that, for a direct-contact crime to occur,

three elements must coincide in time and space: a suitable target, a motivated offender,

and the absence of a capable guardian. From this assertion, it follows that the overall

spatiotemporal characteristics of crime can be understood in terms of the movements and

behaviors of these three elements.

As the primary substrate for routine human activities, the street network plays a cru-

cial role in determining where the convergences of these elements occur. In many cases,

potential targets are located, and thus encountered by offenders, at some position on the

network: This is particularly apparent for crimes against fixed targets, such as burglary,

but also it applies to interpersonal crimes that take place in the urban environment. In

addition, offenders will typically use the street network when traveling to and from of-

fenses. Importantly, these principles apply not only to the movements and presence of

offenders but also to those of citizens in general, thereby influencing the supply of poten-

tial guardians at particular places. In all these cases, the locations and movements of the

key actors involved in the crime event are constrained to a large extent by the structure

of the street network.