Emergent Regularities of Interpersonal Victimization

• Author: Michael Townsley,Daniel Birks,Anna Stewart
• DOI: 10.1177/0022427813487353
• Published date: 01 February 2014
• Date: 01 February 2014
• Subject Matter: Articles

Article

Emergent

Regularities

of Interpersonal

Victimization:

An Agent-Based

Investigation

Daniel Birks

1

, Michael Townsley

1

,

and Anna Stewart

2

Abstract

Objectives: Apply computational agent-based modeling to explore the gen-

erative sufficiency of several mechanisms derived from the field of

environmental criminology in explaining commonly observed patterns of

interpersonal victimization. Method: Controlled simulation experiments

compared patterns of simulated interpersonal victimization to three empiri-

cally derived regularities of crime using established statistical techniques:

(1) spatial clustering (nearest neighbor index), (2) repeat victimization (Gini

coefficient), and (3) journeys to crime (Pearson’s coefficient of skewness).

Results: Large, statistically significant increases in spatial clustering, repeat

victimization, and journey to crime skewness are observed when virtual

offenders operate according to mechanisms proposed by the routine

1

ARC Centre of Excellence in Policing & Security, Griffith University, Brisbane, Australia

2

School of Criminology and Criminal Justice, Griffith University, Brisbane, Australia

Corresponding Author:

Daniel Birks, Griffith University, Mount Gravatt Campus, Building M07 Room 2.23, 176

Messines Ridge Road, Mount Gravatt, Brisbane 4122, Australia.

Email: d.birks@griffith.edu.au

Journal of Research in Crime and

Delinquency

2014, Vol 51(1) 119-140

ªThe Author(s) 2013

Reprints and permission:

sagepub.com/journalsPermissions.nav

DOI: 10.1177/0022427813487353

jrcd.sagepub.com

activity approach, rational choice perspective, and geometry/pattern

theories of crime. Conclusion: This research provides support for several

propositions of environmental criminology in explaining why interpersonal

victimization tends to be spatially concentrated, experienced by a small

number of repeat victims, and why aggregate journey to crime curves tend

to follow a distance decay relationship. By extending previous work in

agent-based modeling of property victimization, it also demonstrates that

the same core mechanisms are sufficient to generate plausible patterns of

crime when examining fundamentally different types of offending.

Keywords

computational criminology, interpersonal victimization, routine activity

theory, criminological theory, rational choice theory, geometry/pattern

theories of crime

Macro-Regularities and Micro-Specifications of the Crime Event

Empirical research has identified a number of well-established regularities

of crime, consistent across a range of offence types, localities, contexts, and

measurement instruments (Birks, Townsley, and Stewart 2012). Such regu-

larities include the observations that offending is not uniformly distributed

in space nor time, that a small number of people, places and targets expe-

rience the majority of victimization, and that the distances offenders travel

to offend are typically short and exhibit the characteristics of a distance

decay function. One can think of these regularities as the predictable, emer-

gent, macro-level outcomes of mechanisms operating at the micro level—

that is, the perception, cognition, and actions of all actors significant to the

crime event. It is these micro-level actions, and their interactions, that gen-

erate societal level patterns of victimization, and through this lens that such

crime regularities can be described as the output of a complex dynamical

system.

The criminological literature contains an abundance of theoretical depic-

tions of these interactions and, in turn, explanations for the observed pat-

terns of victimization we propose they give rise to. Unfortunately, our

ability to test the veracity of these hypotheses is severely constrained by

three key limitations in our ability to:

1. Observe: Widespread longitudinal observations of the interactions

between actors involved in crime events are simply unobtainable.

Moreover, making causal inferences about these interactions using

120 Journal of Research in Crime and Delinquency 51(1)