The Temporal (In)Stability of the Unemployment and Crime Relationship
| Author | Olivia K. Ha,Martin A. Andresen,Garth Davies |
| DOI | 10.1177/0306624X19896454 |
| Published date | 01 June 2020 |
| Date | 01 June 2020 |
| Subject Matter | Articles |
research-article2020
Article
International Journal of
Offender Therapy and
The Temporal (In)Stability
Comparative Criminology
2020, Vol. 64(8) 840 –859
of the Unemployment and
© The Author(s) 2020
Article reuse guidelines:
sagepub.com/journals-permissions
Crime Relationship
https://doi.org/10.1177/0306624X19896454
DOI: 10.1177/0306624X19896454
journals.sagepub.com/home/ijo
Olivia K. Ha1, Martin A. Andresen2 ,
and Garth Davies1
Abstract
The complex relationship between crime and economic change has had a long
pedigree in criminological research. This article considers the temporal stability of
the Cantor and Land model of unemployment and crime using a decomposition
model of Canadian provinces, 1981 to 2009. We include multiple economic measures
for a more comprehensive representation of economic performance, allowing for
the estimates of long- and short-run unemployment effects to vary over time. We
undertake this analysis considering 12 crime types, finding strong support for the
Cantor and Land model in both property and violent crimes. However, in a number
of cases, we find that there is significant variation of these relationships over time.
This result implies that support for this model depends on the time period analyzed
and that any policy derived from this model of unemployment and crime is time-
period dependent.
Keywords
unemployment, crime, panel data, temporal stability, decomposition model
Introduction
Understanding crime trends, including factors that drive fluctuations, is fundamental
to the study of crime. Because crime trends are dynamic, gaining insight into mecha-
nisms that contribute to variance in crime patterns is an important area of development
for crime scholars. Accordingly, an array of factors has been invoked to explain
1Simon Fraser University, Burnaby, British Columbia, Canada
2Griffith University, Southport, Queensland, Australia
Corresponding Author:
Martin A. Andresen, School of Criminology and Criminal Justice, Gold Coast Campus, Griffith University,
Parklands Dr, Southport, Queensland 4215, Australia.
Email: m.andresen@griffith.edu.au
Ha et al.
841
changes in crime trends, including economic-based determinants. Investigations into
the crime–economy relationship resulted in the emergence of a huge literature span-
ning a range of academic disciplines and drawing upon a variety of theories, most of
which relate to the motivational, opportunity, or rational choice perspectives (Cook &
Watson, 2014).
Early in the theoretical literature, scholars generally agreed that the unemploy-
ment–crime relationship was positive: An increase in unemployment leads to increases
in crime (Cook & Watson, 2014). However, a review of the empirical research por-
trays an inconsistency in previous research (Chiricos, 1987; Cook & Watson, 2014).
Limitations in earlier studies may have been attributed to the fact that researchers did
not consider theoretically similar mechanisms within a common framework, further
contributing to mixed findings. Cantor and Land (1985) recognized the constraints of
having disjointed theoretical components and sought to remedy this limitation. In
doing so, they formulated a model of unemployment and crime that effectively inte-
grated previously fragmented theoretical approaches. The result was a model that syn-
thesized two distinct and counterbalancing structural effects of unemployment on
crime: An increase in unemployment has a lagged positive effect on crime through
increased motivation and a contemporaneous negative effect on crime because of
increased guardianship and reduced opportunity. Currently, the bulk of crime–econ-
omy literature is either critical of, or premised on, the Cantor and Land model of
unemployment and crime.
Some of this literature has focussed on the use of longitudinal data and their cor-
responding statistical techniques. Longitudinal data enable the researcher to disen-
tangle the temporal ordering of variables, shedding light on the causal relationships to
address questions of variation or change over time (J. A. Phillips & Greenberg, 2008).
However, the statistical models within much of the current literature do not allow for
the simultaneous identification of the long- and short-run effects stated in the Cantor
and Land (1985) model. In addition, the previous research testing this model has
assumed time-stable relationships for the unemployment rate.
To address this issue, the current analysis considers the methodological and empiri-
cal issues previously mentioned using data from the 10 Canadian provinces between
1981 and 2009. Moving beyond previous inquiries, the Cantor and Land (1985) model
of unemployment and crime is extended through the use of the decomposition model
with time-varying effects for the unemployment rate on multiple crime types. The aim
in adopting a time-varying approach to specifying the Cantor and Land model is to
investigate whether the relationship between unemployment and crime is stable over
time, and if that (in)stability depends on the crime type under analysis.
Previous Research
Early research on the crime–economy relationship generally occupied one of two the-
oretical frameworks: motivation or opportunity, both of which were premised on the
conceptual framework of routine activity theory (Britt, 1997; Cantor & Land, 1985;
Chiricos, 1987; Cohen & Felson, 1979). All this changed when Cantor and Land
842
International Journal of Offender Therapy and Comparative Criminology 64(8)
(1985) formulated a model that would effectively integrate these frameworks. In their
model of unemployment and crime, Cantor and Land posited that by altering the con-
ditions of social strain and social control, economic change (unemployment rate)
would positively affect criminal motivation. Also, economic changes influence the
availability of vulnerable targets and, hence, increase the number of available criminal
targets, opportunities (J. Phillips & Land, 2012). Criminal motivation was theorized as
having a lagged effect, taking time to develop as individuals do not immediately turn
to illegitimate activity in the face of economic hardship. Conversely, the opportunity
effect occurs immediately because unemployment instantly alters the duration and
frequency that individuals are away from the home: Being unemployed leads to a shift
in routine activities toward the home that allows people to guard person and property,
making them less susceptible to victimization (Cantor & Land, 1985). Finally, Cantor
and Land found that both motivation and opportunity matter empirically, particularly
for property crime, but operated at different time frames. Motivation matters in the
long run, due to the lagged effect, whereas opportunity matters in the short run, as the
effect is immediate.
Since its inception, researchers have argued over the empirical validation of the
Cantor and Land (1985) model, more specifically whether opportunity actually domi-
nates motivation. However, the theory behind the model is not often questioned
(Andresen, 2013b). Still, two major issues have been raised concerning to the current
state of empirical inquiry on unemployment and crime: The first pertains to disagree-
ments on the empirical methods, primarily the statistical models used to test the Cantor
and Land model; and second is a set of issues relating to the appropriateness of using
unemployment as an isolated measure to test economic performance (Andresen,
2013b; Arvanites & Defina, 2006). Proper specification of the Cantor and Land model
will assist in bridging inconsistencies in the results due to discrepancies in method
selection. Likewise, because the empirical testing of a model is sensitive to both the
variables chosen and the statistical method employed, to make conclusive statements
on the model being tested and the theoretical relationships it represents, proper speci-
fication is crucial (Andresen, 2015).
The issue of whether the opportunity effect dominates the motivation effect has
been one of the most contentious within the literature. Hale and Sabbagh (1991)
argue that Cantor and Land’s (1985) results were likely invalid as the methodologi-
cal approach adopted by these researchers had fundamental flaws (see Cantor &
Land, 1991, for a response). Accordingly, Hale and Sabbagh focused their analysis
on determining opportunity and motivation effects using time series data from
England and Wales. The results indicated the presence of a motivational effect that
was positively associated with crime but did not find evidence for an opportunity
effect. Field (1990) demonstrated the importance of incorporating other economic
factors when evaluating the unemployment–crime relationship, as concentrating on
unemployment alone may lead to model misspecification. For instance, Field found
that for property crime, personal consumption better captures the economic–crime
relationships than unemployment; Rosenfeld and Fornango (2007) found similar
results in the context of consumer sentiment. Field concluded that in times when
Ha et al.
843
growth in personal consumption is on the decline, property offenses will decrease,
consistent with the Cantor and Land model.
Smith et al. (1992) conducted a time series analysis to determine relationships
among race- and gender-specific rates of unemployment and corresponding rates of
arrest. Specifically,...
To continue reading
Request your trial