Vous êtes sur la page 1sur 20

European Journal of

Criminology
http://euc.sagepub.com/

Usual suspects, ideal victims and vice versa: The relationship between
youth offending and victimization and the mediating influence of risky
lifestyles
Diederik Cops and Stefaan Pleysier
European Journal of Criminology 2014 11: 361 originally published online 20 September
2013
DOI: 10.1177/1477370813500886
The online version of this article can be found at:
http://euc.sagepub.com/content/11/3/361

Published by:
http://www.sagepublications.com

On behalf of:

European Society of Criminology

Additional services and information for European Journal of Criminology can be found at:
Email Alerts: http://euc.sagepub.com/cgi/alerts
Subscriptions: http://euc.sagepub.com/subscriptions
Reprints: http://www.sagepub.com/journalsReprints.nav
Permissions: http://www.sagepub.com/journalsPermissions.nav
Citations: http://euc.sagepub.com/content/11/3/361.refs.html
Downloaded from euc.sagepub.com at University of Bucharest on October 22, 2014

>> Version of Record - Apr 23, 2014


OnlineFirst Version of Record - Sep 20, 2013
What is This?

Downloaded from euc.sagepub.com at University of Bucharest on October 22, 2014

500886
2013

EUC11310.1177/1477370813500886European Journal of CriminologyCops and Pleysier

Article

Usual suspects, ideal


victims and vice versa: The
relationship between youth
offending and victimization and
the mediating influence of risky
lifestyles

European Journal of Criminology


2014, Vol. 11(3) 361378
The Author(s) 2013
Reprints and permissions:
sagepub.co.uk/journalsPermissions.nav
DOI: 10.1177/1477370813500886
euc.sagepub.com

Diederik Cops and Stefaan Pleysier

Leuven Institute of Criminology, University of Leuven, Belgium

Abstract
The fact that young people report disproportionately high levels of offending has resulted in
substantial criminological attention to the topic of young people as offenders. However, we tend
to forget that victims of those young delinquents are largely young people as well. Moreover,
not only are young people overrepresented in both offender and victim populations, there are
equally remarkable similarities between the two populations. Despite the generally accepted
homogeneity in victim and offender characteristics and their respective risk factors, little theorydriven empirical research has been conducted in an attempt to further explain this overlap. This
study tries to fill this gap by focusing on the relationship between offending and victimization in
a large sample of young people in the Brussels Capital Region (N = 2070). More specifically, the
influence of risky lifestyles on the relationship between offending and victimization is studied in
order to test the assumption that the offendingvictimization overlap is in reality the result of the
convergence in time and space of (groups of potential) offenders and victims.

Keywords
Adolescents, risky lifestyle, victimoffender overlap

Introduction
Numerous criminological theories and studies have tried to explain the delinquent
behaviour of young people. This focus largely results from the fact that this group,
compared with other age groups, is responsible for a disproportionately high number
Corresponding author:
Diederik Cops, Leuven Institute of Criminology, University of Leuven, 3000 Leuven, Belgium.
Email: diederik.cops@law.kuleuven.be

Downloaded from euc.sagepub.com at University of Bucharest on October 22, 2014

362

European Journal of Criminology 11(3)

of crimes. The general focus on the offender and the causes of offending that traditionally has dominated criminological research has been challenged by researchers pointing to the neglect of the victim (Maguire and Pointing, 1988; Young, 1986), leading to
the (re)discovery of victimology as a sub-discipline of criminology. At the policy level,
this has advanced the development of crime victim surveys, the first of which was
implemented in the United States at the end of the 1960s. European countries, such as
the United Kingdom and the Netherlands, followed the US example from the end of
the 1970s onwards by introducing similar nationwide large-scale victim surveys.
However, despite the continuous rise in attention to the victim, the risk of victimization
and differential victimization levels in society, relatively little research has been
directed towards the prevalence, causes and experiences of victimization among young
people. Equally, little attention has been given to the overlap between victimization
and offending. Research on offending on the one hand and victimization on the other
has developed rather independently, with only few interactions between the two. As a
consequence, these separate traditions are to a certain extent responsible for the dominant assumptions seeing young people as the usual suspects of delinquency and older
age groups as ideal victims, resulting in a one-sided focus on the role of young people as offenders and not as victims of crime.
By the end of the 1970s, the narrow focus on the offender on the one hand, or on the
victim on the other, gradually became challenged by a new perspective that focused on
the criminal event rather than on the offending or victimized individual. This results in a
new understanding of the aetiology of delinquency, in which the convergence in time and
space of potential offenders and potential victims is a key mechanism. Although the
routine activities theory (Cohen and Felson, 1979) adopts this perspective to explain
crime rates at a macro (societal) level, the lifestyle exposure theory (Hindelang et al.,
1978) has applied it to explain individual differences in victimization (risk). These theories suggest that the overlap between offenders and victims at a group and individual
level is in reality due to the convergence in time and place of separate groups of individuals. More recently, however, the victimoffender overlap is not merely seen as a
product of that convergence in time and space of potential offenders and victims, but
instead a true causal relationship between the two is assumed, suggesting that victimization or offending itself can cause subsequent offending or victimization (e.g. Wittebrood,
2007).
However, empirical research on the victimoffending overlap is relatively scarce
(Smith and Ecob, 2007), and most research remains limited to the identification of shared
socio-demographic characteristics of both groups (Fattah, 1991; Jennings et al., 2010;
Smith and Ecob, 2007). Victim and offender populations are similar in several ways:
young, male, unmarried and unemployed. An important practical reason for this relative
lack of empirical knowledge is the narrow focus of most criminological research on
either offending or victimization; studies combining items to measure both aspects are
relatively rare (Pauwels and Svensson, 2011). Although some studies investigate this
offendingvictimization overlap at an individual level (e.g. Lauritsen et al., 1991;
Pauwels and Svensson, 2011; Sampson and Lauritsen, 1990; Wittebrood and Nieuwbeerta;
1997), empirical knowledge about this relationship generally speaking remains confined

Downloaded from euc.sagepub.com at University of Bucharest on October 22, 2014

363

Cops and Pleysier

to the identification of its existence, and analyses of the validity of the theoretical
assumptions for this relationship are largely absent.
This study is therefore directed towards the relationship between offending and victimization among adolescents. The choice of this particular age group can be substantiated by the fact that both offending and victimization are more prominent during
adolescence compared with any other period in the course of life. More specifically, the
studys aim is to determine the validity of the assumption that the offendervictim relationship is in fact a spurious one and is in reality mediated by risky lifestyle, as is
assumed by the lifestyle exposure theory. In order to test these assumptions, multivariate
regression analyses and structural equation modelling are used, which allows us to
explore the mediating role of risky lifestyle in the victimoffender relationship. The
paper starts with a more in-depth description of the two competing perspectives with
regard to the relationship between offending and victimization

The offendingvictimization relationship


Research on the correlates of victimization has for a long time focused on demographic
factors, such as age, sex, ethnic origin and marital status (Lauritsen et al., 1991). One
of its key findings are the unequal distribution of victimization risks across the population. Some social groups are confronted more frequently with victimization than others. It was found that young peoples chances of being victimized are much higher than
the victim rates in older age groups. However, although juveniles and young adults
are clearly more likely to be victimized, surprisingly little research has directly examined their victimization experiences (Lauritsen et al., 1991: 266; see also Fattah, 1991;
Loader, 1996).
An important incentive for the growing attention to the relationship between victimization and offending are the routine activity and lifestyle theories, which have
caused a shift from a strong focus on either offenders or victims towards the study of
the exposure to criminogenic settings, situations and places (Cohen and Felson, 1979;
Hindelang et al., 1978). The lifestyle exposure theory was one of the first theories trying to explain individual differences in victimization risks (Hindelang et al., 1978).
This shift towards the influence of lifestyle differences equally stimulated more attention to the victimizationoffending overlap, suggesting that both offending and victimization are produced by similar processes (Gottfredson, 1981). This perspective is
described by Hindelang et al. (1978) as the principle of homogamy, which suggests
that the link between victimization and offending is the product of shared routine
activities and lifestyle characteristics of both victims and offenders. In other words,
this principle states that persons are more likely to be victimized when they come into
contact with members of demographic groups that contain a disproportionate share of
offenders (Lauritsen et al., 1991: 2678). The higher prevalence of victimization
among young males is, according to this principle, due to the higher likelihood of sharing routine activities and lifestyles with other young men who are engaged in delinquent behaviour. This principle therefore sees the link between offending and
victimization as the product of the mutual presence, at the same time and in the same

Downloaded from euc.sagepub.com at University of Bucharest on October 22, 2014

364

European Journal of Criminology 11(3)

(public) space, of separate groups of potential victims and potential offenders. An analogous interpretation is put forward by the routine activities approach (Cohen and
Felson, 1979), which states that routine activities patterns can lead to crime by affecting the convergence in time and space of motivated offenders, suitable targets and
the absence of capable guardians. Although this latter approach primarily focuses on
general crime trends, both perspectives have a similar view on the offendervictim
overlap, suggesting that this largely results from the fact that lifestyle and routine
activities patterns produce the likely pools of victims and the likely pools of offenders
and the circumstances that they are likely to come into contact with one another
(Gottfredson, 1981: 726).
Empirical research consistently found a remarkably strong relationship between
offending and victimization; individuals who report having committed a crime equally
have higher chances of being the victim of crime, compared with individuals who do
not commit crimes. However, empirical studies remain mostly limited to bivariate
analyses in which the victimization rates among offenders and non-offenders are compared (Gottfredson, 1984; Hindelang et al., 1978; Junger and Wittebrood, 1997). This
overlap is also being identified over time, using longitudinal designs, by, for example,
Jennings et al. (2010), who found similarities in group trajectories of offending and
victimization.

Causal relationships between offending and


victimization
An alternative approach with regard to the assumption of homogamy interprets offending as
a causal mechanism for the explanation of victimization, due to the motives, vulnerability or
culpability of the offenders themselves. Offenders engage in risky, criminogenic settings
and become attractive targets or ideal victims, because they may be victimized with relative impunity since offender victims might be more reluctant to report their victimization
to the police, and may also be victimized by their previous victims themselves (Sampson
and Lauritsen, 1990; Sparks, 1982; Wittebrood, 2007; Wittebrood and van Wilsem, 2008).
Victimization, in turn, can equally lead to offending, because it may result in feelings of
anger, revenge or injustice, causing victims to offend themselves. This suggests that the
relationship between offending and victimization is more than the overlap between separate
groups of (potential) offenders and (potential) victims in a particular geographical and criminogenic area, as is assumed by lifestyle theories.
However, this relationship is hardly substantiated by empirical research. As already
mentioned, most studies on the victimoffender overlap focus on the possible effect of
offending on victimization risk. It is only recently that both relationships from offending to victimization and vice versa have been studied in the same analyses. Pauwels
and Svensson (2011), for example, found that risky lifestyles and victimization are strong
predictors of offending, whereas having a more risky lifestyle and offending can to only
a minor degree explain variance in individual differences in victimization. Studies such
as this, using a cross-sectional design, are however hindered in the possibility of disentangling true causal mechanisms in the relationship between victimization and offending. Smith and Ecob (2007) identified the existence of a causal relationship between

Downloaded from euc.sagepub.com at University of Bucharest on October 22, 2014

365

Cops and Pleysier

offending and victimization using longitudinal data. They found significant relationships
between the level of offending at one point in time and a change in the level of victimization at a later time, and vice versa. Equally, they identified a more pronounced causal
relationship between offending and later victimization than between victimization and
subsequent offending (Smith and Ecob, 2007: 651).
In general, it can be concluded that research linking offending and victimization is
still relatively limited. Most studies in this niche solely focus on the identification of the
victimoffender overlap, while theory-driven empirical analyses aimed at understanding
and explaining this well-established overlap remain scarce. Moreover, most studies with
empirical research are limited to controlling for demographic characteristics as proxy
indicators of lifestyle, such as gender, age, professional and marital status, educational
level and urbanization (e.g. Jensen and Brownfield, 1986; Wittebrood and Nieuwbeerta,
1997), or use a combination of demographic factors and more direct indicators of lifestyle such as number of nights out (Sampson and Lauritsen, 1990) or drinking
(Lauritsen et al., 1991). In contrast, theoretically more relevant concepts, measured in a
valid and reliable way, are only to a very limited extent integrated in empirical studies.
Analyses that offer insight into the possibly spurious character of the victimoffender
relationship are virtually non-existent. Other important limitations are that most studies
use a sample of the general population (Jennings et al., 2010; Lauritsen et al., 1991;
Pauwels and Svensson, 2011; Smith and Ecob, 2007, are important exceptions). As a
consequence, very little is known about these processes in the group of adolescents. This
is relatively surprising since both young offenders and victims are disproportionately
overrepresented: young people have the highest risk of committing crimes, but also of
being the victim of a crime (Fattah, 1991).

The study
Based on the above-mentioned limitations of previous studies on this topic, our main
question in this study is whether differences in risky lifestyles can account for the victimoffender overlap among adolescents. To answer this query, data from the Youth
Monitor Brussels are used. This study was conducted by the Youth Research Platform,
an interdisciplinary and inter-university cooperation between the Social Welfare research
group (University of Ghent), the Department of Sociology of the University of Brussels
and the Leuven Institute of Criminology (University of Leuven), which is funded by the
Flemish government.
The Youth Monitor Brussels was administered in the Brussels Capital Region.
Brussels, the capital of Belgium, is the only city in the country with more than 1 million
inhabitants. The study was administered in 2010 in secondary schools that are funded by
the Flemish community and in which Dutch is therefore the official language.1 All
schools of the Dutch-speaking community were asked to participate; more than 70 percent (n = 32) of them agreed to do so. Classes were randomly selected based on grade
and type of education; of the pupils in these classes, 88 percent filled in the questionnaire. The non-response rate at the individual level was mainly the result of illness of the
respondents or absence of classes owing to a school trip. In sum, N = 2070 adolescents
between 13 and 19 years filled in the questionnaire.

Downloaded from euc.sagepub.com at University of Bucharest on October 22, 2014

366

European Journal of Criminology 11(3)

Operationalization of variables
Offending
In the Youth Monitor Brussels, the prevalence and incidence of eight offences is measured. Respondents were asked if and how many times during the previous 12 months
they committed the following crimes: vandalism, theft of something worth less than 5,
theft of something worth more than 5, physical violence, harassment, carrying a weapon
in public, dealing drugs, and burglary. Respondents could indicate whether they did not
commit that crime (0), committed it once (1), twice (2), three times (3), or more
than three times (4). All answers on these eight crimes were summed in order to create
a single general offending scale ranging from 0 to 32.

Victimization
The respondents were given a list of five crimes and were asked to report if and how many
times they became the victim of that crime in the previous 12 months. This list consists of
the following crimes: having something vandalized, having something stolen, being physically attacked, being threatened with a weapon, suffering extortion on the street.
Respondents could choose from the following options: not been the victim (0), victimized
once (1), twice (2), or more than twice (3). The frequencies of becoming the victim of
these crimes were summed to create a single victimization scale ranging from 0 to 15.

Lifestyle risk
It can be defined as the lifestyle risks individuals are confronted with in public space, and
are known to influence the risk both of victimization (Hindelang et al., 1978) and of
offending (Wikstrm, 2004). Although traditionally measured using demographic proxy
variables, recent studies have considered this concept to be multidimensional, consisting
of three dimensions: (1) where, or the places where young people spend their leisure
time; (2) who, or with whom they spend this leisure time; and (3) what, or the kind of
behaviour young people undertake during their leisure time (Pauwels and Svensson,
2011). These three dimensions are also operationalized in the Youth Monitor Brussels,
with frequency of participation in public space and night time activities (where), number of delinquent peers (who) and frequency of alcohol and use of soft drugs (what).
These three dimensions are used to create one scale, measuring lifestyle risk.

Social bonds
The extent to which young people establish social bonds is measured in two domains,
more specifically the family and the school. With regard to the bonds with the family
context, the perceived quality of the relationship with the mother and the perceived
degree of parental control are measured. To measure the bond with the school context, an
indicator of school well-being is used. All three scales are positively oriented, meaning
that a higher score refers to a stronger emotional bond with the mother, higher perceived
parental control and a higher level of school well-being.

Downloaded from euc.sagepub.com at University of Bucharest on October 22, 2014

367

Cops and Pleysier


Table 1. Descriptives of the variables.
Lowest

Highest

Mean

SD

Victimization
Offending
Risky lifestyle
Relationship with mother
Parental control
School well-being
Age

0
0
3.0
0
0
0
13

15.0
32.0
24.0
100.0
100.0
100.0
20

1.8
2.6
7.0
72.0
62.0
36.0
16

2.3
5.0
4.49
23.0
22.0
19.0
2

Degree of wealth

6.72

0.74

0.0043

0.99

Demographic variables
A set of demographic variables is integrated as control variables. This allows us to determine whether the relationships between offending, victimization and risky lifestyle are
valid irrespective of age, sex, ethnic origin, educational level and socioeconomic status.
Age is integrated as a metric variable; sex is a dichotomous variable, with boys being
coded as 0 and girls as 1; 47.2 percent of the respondents are male, 52.8 percent female.
The ethnic origin of the respondents is also operationalized as a dichotomous variable:
respondents whose parents are both of Belgian origin (28.5 percent of our respondents)
are coded as 0, youngsters with at least one non-Belgian parent (71.5 percent) are coded
as 1. The degree of wealth of the family is measured by six items in which the respondents
are asked, for example, whether or not their family is able to afford to replace broken
furniture, to go on a holiday once a year, or to keep the house warm during the winter.
Because of the dichotomous answer categories (yes/no), a categorical principal components analysis (CATPCA) is used to identify the latent concept. This analysis results in a
one-dimensional structure, with an eigenvalue of 2.48 and a Cronbachs of .72. A higher
score on the scale indicates a higher degree of family wealth. A last variable is type of
education, which is measured as a dummy variable. The reference category is general
secondary education, while technical and vocational secondary education are coded as
1.2 In our sample, 54.4 percent of the pupils were in general secondary education at the
time of the study and 45.6 percent on the technical or vocational educational track.
Table 1 describes the range, means and standard deviations of all the variables that are
used in this study.

Analytical strategy
Both theoretical approaches for the offendingvictimization overlap allow us to distil
contrasting expectations of the relationship between victimization, offending and risky
lifestyles. The lifestyle exposure theory (Hindelang et al., 1978) suggests that the bivariate relationship between offending and victimization disappears once lifestyle is controlled for. The relationship is therefore seen as spurious, caused by the overlap in routine
activities and lifestyle between separate groups of offenders and victims (Fagan et al.,

Downloaded from euc.sagepub.com at University of Bucharest on October 22, 2014

368

European Journal of Criminology 11(3)

1987). Moreover, this approach suggests that the victimoffender overlap is a result of
the common characteristics of victims and offenders, both socio-demographically as
well as with regard to the causes of offending and victimization, which are also seen as
similar (Fagan et al., 1987; Gottfredson, 1981, 1984). It is therefore expected that the
relationship between offending and victimization will disappear when the variable risky
lifestyle is integrated, because this variable is seen as a common cause for both
variables.
In the causal relationship approach, in contrast, it is expected that the victim
offender connection is not (completely) influenced by the risky lifestyle variable. We
must note, however, that the cross-sectional design of our data does not allow strict causality to be discerned. We therefore need to be careful when interpreting our results; we
will not be able to identify clear empirical evidence for the validity of the causality
hypothesis because we lack longitudinal data. What this perspective suggests, and what
we are able to test empirically in a cross-sectional design, is that the direct statistical
relationship between offending and victimization holds, even after controlling for possible common cause variables such as risky lifestyle, socio-demographic factors and
other relevant variables in a single model. In sum, the data allow us to determine whether
the principle of homogamy is a valid assumption by controlling for risky lifestyle; strict
empirical evidence on the expected causal relationship between offending and victimization will be not found, although a statistical relationship between both aspects after controlling for a set of possible common causes can be interpreted as an important indication
of the existence of the causal relationship.
The analyses conducted to test these hypotheses are divided into three steps. In a first
step, bivariate analyses are presented to identify the overlap between victimization and
offending in the data used. Then multivariate analyses are conducted to test these hypotheses. First, regression analyses are conducted with, respectively, offending and victimization as the dependent variables. Secondly, the relationship between offending,
victimization and risky lifestyle is investigated in more depth using structural equation
modelling.

Bivariate analyses
In the bivariate analyses, our main focus is to examine the strength and characteristics of
the link between offending and victimization. The potential overlap between offenders
and victims is examined in two ways. First, the frequencies of offending and victimization for five crimes are studied; of these five crimes, the prevalence both of offending
and of victimization is examined in the study. For each of these crimes, we identify the
number of respondents who have committed that crime and were also a victim of it. The
value is calculated in order to determine the difference between the observed overlap
and the expected overlap based on the assumption that there is no relationship between
victims and offenders. In a second step we analyse the victimoffender overlap at a more
general level and aim to identify the risk of victimization among offenders and nonoffenders, irrespective of the type of offence.
As the analyses in Table 2 indicate, there is a substantial overlap between offenders
and victims for each of the five types of crime. Offenders have a substantially higher risk

Downloaded from euc.sagepub.com at University of Bucharest on October 22, 2014

369

Cops and Pleysier


Table 2. Offendingvictimization overlap for five crimes.

Vandalism
Theft
Physical violence
Carrying a weapon /
threatened by weapon
Harassment / being
harassed

Observed
frequency of
victim-offenders

Expected
frequency of
victim-offenders

value

Odds ratio

186
341
88
62

137
291
26
14

35.11***
24.25***
197.13***
207.77***

2.00
1.62
8.66
10.86

60

34

25.24***

2.28

***p < .001

Table 3. Prevalence and incidence of victimization among offenders and non-offenders.


Prevalence of
victimization (%)

Incidence (mean number


of victimizations)

Offenders
Non-offenders

71.1
52.8

2.3
1.2

Significance

= 63.19***

t = 9.99***

***p < .001

of becoming a victim of that same crime than non-offenders. This relationship seems to
be stronger for crimes against the person than for property crimes (vandalism and theft).
The risk of becoming the victim of physical violence is 8.66 times higher among perpetrators of physical violence compared with non-perpetrators of this crime, as is indicated
by the odds ratios in the final column.
This victimoffender overlap can also be identified at a more general level, as is suggested in Table 3. This table indicates that both the prevalence and the incidence of
victimization are significantly higher among offenders compared with non-offenders.
With regard to victimization prevalence, 71.1 percent of the individuals in the offender
group had been victimized at least once in the previous year (out of the five types of
crime), whereas in the non-offender group only 52.8 percent had been victimized (
= 63.19, p < .001). A similar relationship is identified if the mean incidence of victimization is considered. Offenders had been the victim of on average 2.3 crimes, whereas
non-offenders had been victimized on average 1.2 times in the previous year (t = 9.99,
p < .001).
These results indicate the existence of an important victimoffender overlap, suggesting that offenders have a higher chance of being victimized compared with nonoffenders, and vice versa. However, two points are important. First, there is no one-onone relationship between offending and victimization: it is not only offenders who are
being victimized, but non-offenders too show relatively high prevalence and incidence

Downloaded from euc.sagepub.com at University of Bucharest on October 22, 2014

370

European Journal of Criminology 11(3)

Table 4. Frequencies of pure victims, pure offenders, victim-offenders and non-victim-nonoffenders in absolute numbers and percentages.
Pure offender

Pure victim

Both victim and


offender

Neither victim
nor offender

Vandalism
Theft
Physical violence
Weapon

190 (10.3%)
298 (16.0%)
209 (11.3%)
120 (6.5%)

484 (26.3%)
498 (27.0%)
72 (3.9%)
76 (4.1%)

186 (10.1%)
341 (18.5%)
88 (4.8%)
62 (3.3%)

983 (53.3%)
706 (38.5%)
1481 (80.1%)
1597 (86.1%)

Harassment

166 (9.0%)

222 (12.0%)

60 (3.3%)

1397 (75.7%)

rates of victimization. Second, the number of adolescents who have been both a victim
and an offender is for most crimes the smallest group (see Table 4). Overall, most adolescents have not been either an offender or a victim.

Multivariate analyses
The multivariate analyses consist of two steps. In the first step, we use multiple regression
analyses with, respectively, offending and victimization as dependent variables. In a second step, structural equation modelling (SEM) is used to estimate both the direct and
indirect effects of the different variables. The regression analyses try to take into account
as much of the variation in the dependent variables as possible, which is in contrast to
most studies on this topic in which mainly dichotomous operationalizations of victimization and offending are used. Because both dependent variables are strongly skewed and
therefore not normally distributed, a standard linear regression model cannot be executed.
Instead, Poisson regression is used, which fits closer to the distribution of the dependent
variable (Osgood, 2000). Because a standard Poisson regression is based on specific
assumptions about the mean and standard deviation of the dependent variable, with the
mean being 0 and with a standard deviation of 1, a specific type of Poisson regression in
which these assumptions are not implied the negative binomial regression is used.
In our analyses, we use a stepwise integration of the independent variables. The first
model includes the socio-demographic and the social bond variables. In the two subsequent models, risky lifestyle (Model 2) and offending or victimization (Model 3) are
added separately. This allows us to determine the direct influence of both variables on the
dependent variable. In a fourth and final model, the impact of the complete set of variables is tested. This model allows a first impression of the relationships between offending, victimization and risky lifestyles. Two separate regression analyses are run. In a first
analysis (Table 5), victimization is used as the dependent variable, and in the second
analysis (Table 6) offending operates as the dependent variable.
The results of Model 1 in Table 5, in which the socio-demographic and social bonds
variables are integrated, suggest that only sex, age and school well-being predict significant differences in victimization. Boys, younger respondents and respondents with a
lower level of school well-being are significantly more victimized. In contrast, the degree
of family wealth, type of education, ethnic origin and the two parental variables are not

Downloaded from euc.sagepub.com at University of Bucharest on October 22, 2014

371

Cops and Pleysier

Table 5. Negative binomial regression analysis with victimization as the dependent variable
(unstandardized coefficients).
Model 1
b

Model 2
b

Model 3
b

Model 4
b

Sex (ref. male)


Age (metric)
Degree of family wealth (scale)
Ethnic origin (ref. both parents
of Belgian origin)
Type of education (ref. general
secondary education)
Relationship with mother (scale)
Parental control (scale)
School well-being (scale)
Risky lifestyle (scale)

.248**
.045*
.069
.020

.100
.043*
.010
.043

.218**
.062**
.061
.038

.101
.042
.010
.047

.098

.010

.093

.010

.001
.002
.009***

.001
.001
.006**

.001
.002
.009***
.026**

.001
.001
.006**
.002

Offending (scale)

.200***

.201***

*p < .05; **p < .01; ***p < .001

Table 6. Negative binomial regression analysis with offending as the dependent variable
(unstandardized coefficients).
Model 1
b

Model 2
b

Model 3
b

Model 4
b

Sex (ref. male)


Age
Degree of family wealth
Ethnic origin (ref. both parents
of Belgian origin)
Type of education (ref. general
secondary education)
Relationship with mother
Parental control
School well-being
Risky lifestyle

.846***
.054*
.100**
.189*

.695***
.061*
.054
.532***

.793***
.061*
.034
.181

.643***
.058*
.005
.529***

.436***

.485***

.386***

.464***

.008***
.007***
.018***

.006**
.006**
.016***
.128***

.009***
.007**
.016***

.007***
.004
.014***
.129***

Victimization

.158***

.156***

*p < .05; **p < .01; ***p < .001

significantly related to victimization. The separate integration of risky lifestyle (Model 2)


and offending (Model 3) suggest that both variables are directly and significantly linked
to victimization. However, in a full model (Model 4), the initial direct effect of risky lifestyle disappears, whereas the effect of offending remains stable, which suggests that the
effect of risky lifestyle on victimization is indirect and mediated by offending. With regard
to the other variables in the model, only the effect of school well-being remains significant; the difference between boys and girls found in Model 1 becomes insignificant when

Downloaded from euc.sagepub.com at University of Bucharest on October 22, 2014

372

European Journal of Criminology 11(3)

offending is added, which suggests that differences in victimization rates are due to the
higher rate of offending among boys rather than to being a boy as such.
The first model in Table 6, with offending as the dependent variable, shows that all
socio-demographic variables are significantly linked to offending. Boys, older respondents, those living in a more economically deprived family, respondents with at least one
non-Belgian parent and adolescents in technical or vocational education commit significantly more crimes. With regard to the social bonds variables, the results suggest that a
less positive relationship with mother, stronger parental control and a lower level of school
well-being are significantly related to higher levels of offending. Furthermore, Models 2
and 3 indicate that risky lifestyle and victimization each have an additional effect on the
frequency of offending. Having a more risky lifestyle and being victimized more frequently are related to higher levels of offending. When all variables are integrated in the
final model (Model 4), the effect of risky lifestyle and victimization on offending are
independent of each other; the b-values for both variables remain stable and significant.
In other words, even when risky lifestyle is controlled for, there still is a direct and significant influence of victimization on the frequency of offending among adolescents.
A final step to explore the relationship between offending, victimization and risky
lifestyles is the use of structural equation models. SEM has several advantages compared
with the traditional regression analysis technique. First, it allows the relationship between
the independent and dependent variables to be estimated while at the same time taking
into account the relationships between the independent variables. Although the regression
analyses have offered an indication of the relationship between victimization, offending
and risky lifestyles, SEM offers the opportunity to measure and visualize all relationships
at once. Second, it allows nested models to be built and compared and to determine
whether adding or omitting specific relationships between the variables results in a better
fit of the model. Similar to the regression analyses, the models presented here are built
with, respectively, victimization and offending as dependent variables. In Figure 1,
offending is the dependent variable, and in Figure 2 victimization is used as the dependent
variable. Both figures consist of two nested models. In the first model, only two relationships are estimated: between risky lifestyle and victimization, and between victimization
and offending. In a second model, the relationship between risky lifestyle and offending
is added, which allows the mediating effects of the independent variables to be
determined.
Several methods exist to estimate the parameters in SEM. In these analyses, we will
use the weighted least squares method because it is more robust towards violations of the
assumptions of normality in the distribution of the observed variables (Diamantopoulos
and Siguaw, 2000). For the purpose of these analyses, the statistical software package
LISREL 8.7 is used (Jreskog and Srbom, 1993). In order to indicate the quality of fit
of the models, two measures are calculated. A first (traditional) measure is the value.
A significant value implies the rejection of the null hypothesis of a perfect fit of the
model. However, this test is strongly influenced by the sample size, the assumption of
normality and the assumption of perfect fit, which in reality is very difficult to achieve.
Therefore, instead of determining the perfect fit of a model, it is more relevant to determine the lack of fit, for which the Root Mean Square Error of Approximation (RMSEA)
is often used. RMSEA values under .050 are considered to indicate a good fit of the

Downloaded from euc.sagepub.com at University of Bucharest on October 22, 2014

373

Cops and Pleysier

Vicmizaon

.95

.96
Offending

Risky lifestyle

= 1785; df = 117; p < .001; RMSEA = .075

.66

Vicmizaon

.13

.85
Risky lifestyle

Offending

= 1550; df = 116; p < .001; RMSEA = .070

Figure 1. Structural equation model with offending as the dependent variable (unstandardized
coefficients).

.94

Offending

.70

Risky lifestyle

Vicmizaon

= 1551; df = 117; p < .001; RMSEA = .070

.94

Offending

.59

.11
Risky lifestyle

Vicmizaon

= 1550; df = 116; p < .001; RMSEA = .070

Figure 2. Structural equation model with victimization as the dependent variable


(unstandardized coefficients).

model, and values between .050 and .080 are interpreted as satisfactory (Diamantopolous
and Siguaw, 2000).
Different findings can be derived from these structural equation models (Figures 1
and 2). First, they suggest that a strong direct effect of risky lifestyle on offending and
victimization exists. In both models, a more frequent display of a risky lifestyle is
related to higher levels of both victimization and offending. There are equally strong
direct effects of victimization on offending (Figure 1) and of offending on victimization

Downloaded from euc.sagepub.com at University of Bucharest on October 22, 2014

374

European Journal of Criminology 11(3)

(Figure 2), although this relationship seems more pronounced in the former case than in
the latter.
Second and more important in light of the aim of SEM, contradictory findings are
found with regard to the spuriousness of the offendervictimization relationship. In the
first path model (Figure 1), in which offending is the dependent variable, the direct
effect of victimization on offending disappears almost completely, whereas the direct
effect of risky lifestyle on offending is very strong. This suggests that the effect of
victimization on offending is to a large extent due to a similarity in lifestyles between
victims and offenders. The added value of the additional effect of risky lifestyle in the
second sub-model of Figure 1 is also indicated by the goodness of fit of the model.
Both the value ( = 235, df = 1, p < .001) and the RMSEA measure (from .075
to .070) decrease when the relationship between risky lifestyle and offending is added
to the model.
In Figure 2, however, in which victimization is the dependent variable, the result of
the additional path from risky lifestyle to victimization differs from the initial model.
More specifically, the direct effect of risky lifestyle on victimization is rather small,
while the effect of offending on victimization remains relatively strong and is not, as in
Figure 1, mediated by this new direct effect. In other words, the direct link from offending to victimization remains valid after risky lifestyle is accounted for. The stability of
the statistical indicators of the model fit between both sub-models in Figure 2 is also
indicative of this conclusion.

Discussion and conclusion


This study focused on the offendingvictimization relationship among adolescents. Up
to now, criminological research has paid little attention to the overlap between offenders
and victims, although several studies have found that both groups share the same sociodemographic characteristics (e.g. Fattah, 1991). Using a large sample of young people in
the Brussels Capital Region (N = 2070) aged 13 to 19, the existence of the victim
offender overlap and the extent to which direct and indirect relationships between both
victimization and offending could be identified was studied.
First, the results indicate the validity of the offendingvictimization overlap; for each
of the offences, respondents reporting offending at least once in the previous year have a
higher risk of becoming the victim of that crime in the same period. This relationship was
more prominent for crimes against the person, such as physical violence, and drugrelated crimes, than it is for property-related offences. In general this suggests that young
people who offend have a higher risk of being victimized, and vice versa. However, some
important findings add necessary nuances to this conclusion. For each of the crimes in
this study, the largest proportion of young people have neither offended nor been victimized. In addition, the group of victim-offenders is in each case the smallest group (with
the notable exception of physical violence). A more fundamental point or limitation of
this study is that these results are undoubtedly influenced by the particular types of crime
that were used. The crimes in this study can all be described as forms of petty crime,
committed in public space. It may well be that the victimoffender overlap as identified
in this study would look different when others types of crime are examined.

Downloaded from euc.sagepub.com at University of Bucharest on October 22, 2014

375

Cops and Pleysier

A second aim of this paper was to determine whether a risky lifestyle can explain the
overlap between victimization and offending. The lifestyle exposure theory (Hindelang
et al., 1978) suggests that the relationship between victimization and offending is in reality spurious, resulting from the fact that offenders and victims share the same lifestyle.
Once lifestyle is controlled for, this significant effect would therefore disappear. The
results of both the regression analyses and the structural equation models indicate that
this assumption is valid as far as the relationship from victimization to offending is concerned. In other words, the original direct effect of victimization on offending is mediated by risky lifestyle. These results are in line with previous studies (Pauwels and
Svensson, 2011), which equally found that the victimizationoffending relationship was
less pronounced than the offendingvictimization relationship.
More generally, risky lifestyles and other explanatory variables in the regression analyses seem to be more important in explaining differences in offending than in explaining
levels of victimization. As a consequence, Gottfredsons conclusion that, by and large,
combinations of characteristics predictive of offending, are also predictive of victimization (1981: 7256) needs some important nuances. The analyses in this study suggest
that the socio-demographic and social bond variables are more predictive of differences
in offending than in victimization. This suggests that victimization is not as such a result
of having a risky lifestyle, but that this effect is mediated by offending behaviour: having
a more risky lifestyle is related to a higher prevalence of offending, which in turn leads
to higher levels of victimization. In contrast, the mediating effect of risky lifestyle on the
relationship from offending to victimization was largely absent. Although our crosssectional data do not allow us to identify a direct causal relationship, these findings do
seem to substantiate the suggestions that offending behaviour in itself may contain characteristics that render offenders attractive victims, irrespective of their lifestyle (Sampson
and Lauritsen, 1990; Wittebrood, 2007). Our results are moreover in line with the findings from previous longitudinal research (see Smith and Ecob, 2007), which found that
the effect of offending on later victimization was more pronounced than the opposite
relationship.
In general, this study, despite its limitations, demonstrates that the relationships
between offending, victimization and risky lifestyles are more complex than traditionally
has been explored. It therefore adds important insights to the debate on the victim
offender overlap and contains relevant information to develop a more nuanced view.
However, we also have to mention other limitations that are important in the interpretation and generalizability of the results. Besides the already mentioned cross-sectional
nature of the data (inhibiting the identification of strict causality) and the impact of the
specific types of crime used, we were also not able to build a multilevel model and to
integrate ecological variables, such as proximity to crime (the neighbourhood crime
level), which in previous studies was found to be an important influence on the offender
victim overlap. A third important point is that the specificity of the location in which the
data were collected a highly dense urban area with more than 1 million inhabitants and
a high proportion of young people may limit the generalizability of the results to a
broader population. The urban context of Brussels may lead to an overestimation of this
overlap between offending and victimization, at least in relation to more rural areas and
smaller cities. It may therefore be useful to test the relationships between offending,

Downloaded from euc.sagepub.com at University of Bucharest on October 22, 2014

376

European Journal of Criminology 11(3)

victimization and risky lifestyles in broader samples, incorporating respondents living in


more rural areas.
On a final note, it is important to warn against a blaming the victim conclusion.
Using these results to argue that victims are responsible for their victimization because
they were prior to their victimization offenders, is not a conclusion that can be justified.
Several arguments question this blaming the victim perspective. First, only a small
fraction of victims are involved in offending, and vice versa. Most respondents who
report being victimized have not committed any crime in the same period. Second,
offending was only able to explain a small part of the variance in the prevalence of victimization. Third, the theoretical model in general succeeded in explaining only a relatively small part of the variance in victimization. This implies that other variables, not
integrated in these analyses, are responsible for the unexplained variance in victimization. It therefore can be suggested that further research could focus more in depth on the
variables that account for differences in victimization among young people in particular.
Although it is accepted that young people have a higher risk of being victimized compared with other age groups, only limited knowledge exists on the explanatory factors of
victimization in this group. This does not imply that results from this study are not reliable, but instead offers additional incentives for future research in which these elements
are integrated in an attempt to further illuminate the findings presented here. More specifically, the use of longitudinal data in which not only changes in victimization and
offending (such as the study of Smith and Ecob, 2007) are linked, but also changes in
risky lifestyles are included, may be of great importance to further deconstruct the complex and reciprocal relationship between victimization and offending.
Funding
The Youth Monitor is administered by the Youth Research Platform, a cooperation between three
research groups of three Flemish universities, which is funded by the Youth Department of the
Flemish government.

Notes
1. The Brussels Capital Region is an officially bilingual area, in which both the Flemish community and the French community are competent to organize the educational system. This
results in two separate systems, with one system organized by the French community in
which French is the official language, and a system organized by the Flemish community in
which Dutch is the official language.
2. The Flemish educational system consists of these three types of education. Pupils in general
secondary education receive a theoretically oriented education and most of them will go on
to higher studies after leaving compulsory education (at the age of 18). Pupils in the technical
and vocational system on the other hand receive a practically oriented education, preparing
them for entrance into the labour market, which most of them do after leaving secondary
education.

References
Block R (1981) Victim-offender dynamics in violent crime. Journal of Criminal Law and
Criminology 72(2): 743761.

Downloaded from euc.sagepub.com at University of Bucharest on October 22, 2014

377

Cops and Pleysier

Cohen L and Felson M (1979) Social change and crime rate trends: A routine activity approach.
American Sociological Review 44(4): 588608.
Cops D and Op de Beeck H (2011) Criminaliteit in de grootstad. Dader- en slachtofferschap bij
Brusselse jongeren. In: Vettenburg N, Elchardus M and Put J (eds) Jong in Brussel. Bevindingen
uit de JOP-monitor Brussel. Leuven: Acco, 297328.
Diamantopoulos A and Siguaw J (2000) Introducing LISREL. London: Sage Publications.
Fagan J, Piper E and Cheng YT (1987) Contributions of victimization to delinquency in inner cities. Journal of Criminal Law and Criminology 78(3): 586613.
Fattah E (1991) Understanding Criminal Victimization. Ontario: Prentice-Hall Canada Inc.
Gottfredson M (1981) On the etiology of criminal victimization. Journal of Criminal Law and
Criminology 72: 714726.
Gottfredson M (1984) Victims of Crime: The Dimensions of Risk. London: Home Office.
Hindelang H, Gottfredson M and Garofalo S (1978) Victims of Personal Crime: An Empirical
Foundation for a Theory of Personal Victimisation. Cambridge: Ballinger.
Jennings W, Higgins G, Tewksbury R, Gover A and Piquero A (2010) A longitudinal assessment
of the victim-offender overlap. Journal of Interpersonal Violence 25(12): 21472174.
Jensen G and Brownfield D (1986) Gender, lifestyles, and victimization: Beyond routine activity.
Violence and Victims 1(2): 8599.
Jreskog K and Srbom D (1993) LISREL 8: Structural Equation Modeling with the SIMPLIS
Command Language. Chicago, IL: Scientific Software International Inc.
Junger M and Wittebrood K (1997) Daderschap, slachtofferschap en ongevallen. In: Wittebrood
K, Michon J and ter Voert M (eds) Nederlanders over criminaliteit en rechtshandhaving.
Deventer: Gouda Quint, 3342.
Lauritsen J, Sampson R and Laub J (1991) The link between offending and victimization among
adolescents. Criminology 29(2): 265292.
Loader I (1996) Youth, Policing and Democracy. London: Palgrave Macmillan.
Maguire M and Pointing J (eds) (1988) Victims of Crime. A New Deal? Milton Keynes: Open
University Press.
Osgood D (2000) Poisson-based regression analysis of aggregate crime rates. Journal of
Quantitative Criminology 16(1): 2143.
Pauwels L and Svensson R (2011) Exploring the relationship between offending and victimization: What is the role of risky lifestyles and low self-control? A test in two urban samples.
European Journal on Criminal Policy and Research 17(3): 163177.
Sampson R and Lauritsen J (1990) Deviant lifestyles, proximity to crime and the offendervictim link in personal violence. Journal of Research in Crime and Delinquency 27(2):
110139.
Smith D and Ecob R (2007) An investigation into causal links between victimization and offending in adolescents. British Journal of Sociology 58(4): 633659.
Sparks R (1982) Research on Victims of Crime. Washington, DC: U.S. Government Printing
Office.
Verdonck E, Cops D, Pleysier S and Put J (2011) Jongeren en geweld. Dader-en slachtofferschap
gemeten en beleefd. Leuven: Acco.
Wikstrm PO (2004) Crime as alternative. Towards a cross-level situational action theory of crime
causation. In: McCord J (ed.) Beyond Empiricism: Institutions and Intentions in the Study of
Crime. Advances in Criminological Theory, Vol. 13. New Brunswick, NJ: Transaction.
Wittebrood K (2007) Slachtoffers van criminaliteit. Een inleiding in de victimologie. The Hague:
Boom Juridische Uitgevers.
Wittebrood K and van Wilsem J (2008) Jongeren en geweld. De relatie tussen slachtofferschap,
daderschap en leefstijl. In: Bruinsma W, Huisman R and van Swaaningen R (eds) Basisteksten

Downloaded from euc.sagepub.com at University of Bucharest on October 22, 2014

378

European Journal of Criminology 11(3)

in de criminologie I: Aard, omvang en verklaringen. The Hague: Boom Juridische uitgevers,


205214.
Wittebrood K and Nieuwbeerta P (1997) Wie misdaden pleegt, kan klappen verwachten. De invloed van daderschap en leefpatronen op de kans slachtoffer van geweld te worden. Tijdschrift
voor Criminologie 39(4): 341356.
Young J (1986) The failure of criminology. In: Matthews R and Young J (eds) Confronting Crime.
London: Sage, 430.

Vous aimerez peut-être aussi