Professional Documents
Culture Documents
Int egrat ing Criminal Careers and Ecological Research T he Import ance of Geographic Locat ion for Tar…
Troy Allard
European Journal of Criminology
http://euc.sagepub.com
Published by:
http://www.sagepublications.com
On behalf of:
Additional services and information for European Journal of Criminology can be found at:
Subscriptions: http://euc.sagepub.com/subscriptions
Reprints: http://www.sagepub.com/journalsReprints.nav
Permissions: http://www.sagepub.co.uk/journalsPermissions.nav
Citations http://euc.sagepub.com/cgi/content/refs/6/5/419
ABSTRACT
KEY WORDS
Introduction
The risk factor approach is one of the dominant paradigms for explaining
serious juvenile delinquency (e.g. Farrington 2005). This approach empha-
sizes long-term risk factors such as impulsivity, lack of social skills,
inadequate parental supervision and poor school performance. Serious and
persistent delinquency is primarily explained by the presence of an accu-
mulation of long-term risk factors and a lack of protective factors (see e.g.
Farrington 2005; Loeber et al. 2006). Individuals who display frequent
and/or serious delinquent behaviour – we will refer to these individuals as
serious delinquents – differ from other types of delinquents, because they
are usually influenced by an accumulation of long-term risk factors.
Conversely, the situational approach emphasizes situations in which
potential offenders find themselves when they commit offences (see e.g.
Clarke and Felson 1993; Osgood et al. 1996; Felson 1998). Several studies
show that some situations are more conducive to offending than others, in
terms of suitable targets or the absence of tangible guardians. Offending
may also be facilitated by the company of potential co-offenders or by using
alcohol or drugs. According to the situational approach, everybody is a
potential offender. Whether or not an individual commits an offence depends
primarily on the situation (e.g. Cohen and Felson 1979; Clarke and Felson
1993; Felson 1998). Farrington (2005) refers to factors that determine a
person’s behaviour in actual situations as short-term risk factors.
For many years, these different theoretical and empirical research
traditions seemed to be mutually exclusive. Hardly any empirical research
integrates both long-term and short-term risk factors. Recently, however,
theoretical attempts have been made to integrate both approaches.
Farrington in his Integrated Cognitive Antisocial Potential (ICAP) theory
(2005), for example, combines both long-term and short-term risk factors
in one theoretical model. The aim of this article is to fill the empirical gap in
the literature by examining both long-term and short-term risk factors for
serious delinquency, testing hypotheses from the ICAP theory and using data
from the WODC Youth Delinquency Survey. The following sections of this
article describe in more detail the empirical and theoretical background of
this study, the data and method, and empirical results. In the last section, we
discuss the relevance of these empirical findings for theory and research.
the family and the school (see e.g. Farrington 2005; Loeber et al. 2001,
2006). Examples of individual risk factors are impulsiveness and a lack of
social skills. Risk factors in the family domain include weak attachment
or inadequate parental supervision. Poor school performance and weak
attachment to school are risk factors in the school domain.
If we take several risk factors into account, the effect of one single
factor on serious delinquency is generally weak. A robust finding is that the
probability of serious problem behaviour increases if risk factors accumulate
across several domains (Bronfenbrenner 1979; Stouthamer-Loeber et al.
1993, 2002; Pollard et al. 1999; Loeber et al. 2001). According to the risk
factor approach, serious offenders differ from other offenders (and from
non-offenders) in that they are exposed to an accumulation of risk factors
and a lack of protective factors (see e.g. Farrington 2005; Rutter et al. 1998).
These include factors in the individual, family, neighbourhood and school
domains, which, regardless of the features of the actual situation, may influ-
ence an individual’s behaviour over a relatively long period: so-called long-
term risk factors (Farrington 2005; Wikström and Sampson 2003).
Critics, however, state that the empirical risk factor approach does
not explain why some individuals do in fact offend (or persistently offend),
since other individuals who are subject to similar long-term risks do not con-
tinue their criminal career or might even not commit any offences (Fattah
1993; Osgood et al. 1996). The focus on long-term risk factors means that
no account is taken of the situational characteristics in which offences are
committed. Some situations are more provocative or tempting, or are less
likely to deter individuals from offending than others, regardless of indi-
vidual characteristics (Wikström and Sampson 2003).
Situational theories assume that whether or not an individual will
commit an offence is primarily determined by specific features of the situ-
ation, also referred to as short-term risk factors (Wikström and Sampson
2003; Farrington 2005). The central hypothesis is that the motivation
to commit an offence lies in the situation. This means that, in principle,
anyone can be motivated to commit an offence (Briar and Piliavin 1965;
Cohen and Felson 1979; Clarke and Felson 1993; Fattah 1993; Osgood
et al. 1996; Felson 1998). Routine activities theory, for example, argues that
the presence of a motivated offender and a suitable target and the absence
of tangible guardians are sufficient conditions to commit an offence (Felson
1998). In other situational theories, the presence of co-offenders and being
drunk are regarded as situational factors facilitating the commission of
offences (Felson 1998; Warr 2002; Farrington 2005). A consistent finding
is that adolescents who commit offences often do so in groups (Warr 2002).
Within groups, friends encourage each other to commit offences: friends
may be keen to avoid losing face, want to gain respect or wish to be loyal
to the group (Warr 2002). Another precipitating factor is being under the
influence of alcohol or drugs, which may cause a loss of inhibitions and thus
increase the probability of offending (Warr 2002).
In addition to long-term risk factors, short-term risk factors can there-
fore also play a significant role in the committal of offences. It is striking
that it was not until recently that a number of theoretical attempts have been
made to integrate both approaches in order to gain a better understanding
of serious delinquency during adolescence (Wikström and Sampson 2003;
Farrington 2005; Wikström 2005). In his ICAP theory, Farrington (2003,
2005) combines the risk factor approach with the situational approach.
According to this theory, long-term risk factors increase an individual’s
predisposition towards antisocial behaviour in the long term. Long-term
risk factors, however, do not explain why individuals offend in specific situ-
ations. Short-term risk factors, such as lack of tangible guardianship, the
presence of co-offenders or being under the influence of alcohol or drugs,
may reduce constraints and increase the likelihood of offending. These im-
mediate situational factors decrease the subjective costs of offending, but
according to ICAP theory only those individuals with a predisposition
towards antisocial behaviour will be seduced by these factors to commit an
offence.
ICAP theory was developed very recently and empirical research
in this area is limited.1 In this article we examine both short-term and
long-term risk factors for serious delinquency. We derive hypotheses from
ICAP theory. More specifically, we study the extent to which, in addition
to long-term risk factors, short-term factors contribute toward explaining
serious delinquency, as compared with moderate delinquency, during adoles-
cence. We use the term ‘serious juvenile delinquents’ to refer to adolescents
who offend frequently and who commit serious offences.
The central question of our study is: ‘What is the relationship between
short-term risk factors and serious versus moderate delinquency during
adolescence, taking into account (an accumulation of) long-term risk factors?’
Based on the risk factor approach, we hypothesize that, as more long-term
risk factors accumulate, the probability of serious delinquency will increase,
compared with moderate delinquency. From this perspective, we also ex-
pect that situational factors will not provide any additional explanation
for serious delinquency. Following a situational approach, however, we
expect that, regardless of long-term risk factors, the circumstances in which
1 Some research has been done on the effects on delinquency of situational features related
to social control concepts or lifestyle (see e.g. Bernburg and Thorlindsson 2001; Osgood and
Anderson 2004). However, we are not aware of any empirical study that combines the effects
of long-term and short-term risk factors on delinquency.
serious delinquents commit offences will differ from those that apply to
moderate delinquents. According to Farrington’s ICAP theory, we expect
that both an accumulation of long-term risk factors and the existence of
short-term risk factors will be directly associated with a higher probability
of serious delinquency. In other words, we also hypothesize that, in addition
to differences in long-term risk factors, the circumstances in which serious
offenders commit offences will differ from those in which offences are
committed by moderate offenders. Finally, ICAP theory would also expect
that a combination of long- and short-term risk factors will be associated
with an even greater likelihood of serious delinquency.
Method
2 These offences were ‘forced sex’, ‘inflicting injury’, ‘intimidation’, ‘use of violence in order to
commit theft’, ‘the use of intimidation in order to commit theft’, ‘pick pocketing’, ‘burglary’,
‘theft of property from a vehicle’, ‘theft of a bicycle or scooter’, ‘shoplifting of goods to the
value of less than €10’, ‘shoplifting of goods to the value of more than €10’, ‘vehicle vandalism’,
‘house vandalism’, ‘vandalism in a bus/tram/metro’ and ‘other vandalism’.
[VMBO]) and 47.9 percent lived in urban areas.3 Below we will describe the
operationalization of self-reported delinquency and the long- and short-term
risk factors. A more detailed description of the psychometric characteristics
of the scales is provided in Appendix A.
Delinquency
Delinquency was measured using 31 self-reported offences, varying from
frequently occurring and minor offences to rare and serious offences. The
adolescents were asked how many times they had committed each offence
during the previous 12 months. On the basis of the total sample, a delinqu-
ency scale was constructed in order to distinguish between moderate juvenile
delinquents and serious juvenile delinquents, taking account of the serious-
ness, frequency and variety of offences. First, the offences were classified
according to seriousness – based on the applicable legal penalty – as a
minor offence (maximum custodial sentence of 48 months, e.g. vandalism;
score 1) or a serious offence (minimum custodial sentence of 48 months,
e.g. breaking and entering; score 2). Second, the offences were categorized
according to frequency. Minor offences were assigned code 0 (committed
0 times during the past 12 months), 1 (1–4 times) or 2 (5 times or more);
serious offences were coded 0 (0 times), 1 (once), 2 (2–4 times), 3 (5–10
times) or 4 (11 times or more). Third, for each offence the severity was multi-
plied by the frequency.4 Finally, the delinquency scale was the sum of these
multiplicative scores (31 items; range 0–55; M = 4.4; SD = 5.6; α = .74).
We found a clear distinction between two groups of adolescents
on the delinquency scale: a group of relatively moderate and a group of
relatively serious (45.3 percent) juvenile delinquents (the ‘cut-off point’ for
serious delinquents was ≥ 5).5 The majority (84 percent) of the delinquents
who had committed only minor offences and no serious offences were
classified as moderate delinquents. The majority (87 percent) of the
3 For five background characteristics (sex, age, origin, educational level and proximity to
urban centres), we examined to what extent the selection of 292 delinquents differed from the
remaining delinquents (who were all adolescents who reported only minor offences) within
the total sample. Significant differences were found for age (χ2 = 4,504, p = .034) and level of
education (χ2 = 7.202, p = .027).
4 Our choice for doing so is partly theoretically driven and partly based on empirical research
done in the Netherlands by Nijboer (1997), who used Mokken analysis to find out which
aspects of the offence should be weighted in which way. For example, if we had coded only
by seriousness (e.g. non-serious = 1; serious = 2) and not by frequency, youngsters who had
committed four non-serious offences such as shoplifting something worth less than €5 would
have had the same score on the delinquency scale as youngsters who had committed two
serious offences such as robbery.
5 The percentage of serious delinquents is high owing to the selection of the sample, which is
average performance over the previous year on the basis of one item (five-
point Likert scale). The second scale relates to his or her ‘attachment to
school’ (5 items; α = .76). The scale was taken from the NSCR school
project (Weerman et al. 2003).
All scales were recoded to dichotomous variables with values 0 (‘no
risk’) and 1 (‘risk’). In this context, a ‘cut-off point’ of below the 25th or
above the 75th percentile was used, depending on the meaning of the pole of
the scale (see Farrington and Loeber 2000). Although dichotomization may
lead to loss of variation, it has some important advantages. By dichotom-
izing the variables the interpretation is made easier because the scales of
all variables are equal. Next, dichotomization accommodates skewness
of variables (such as the delinquency scale) without the necessary trans-
formations that are complex to interpret. Furthermore, research has shown
that dichotomization does not necessarily lead to different interpretation of
the results (Farrington and Loeber 2000).
she committed an offence (≥ P50). This means that the individual variation
regarding situational features was incorporated into one score.
Control variables
Research has shown that young males are more likely to commit an offence
than young females. Furthermore, older adolescents are more likely to
offend than younger adolescents. Gender and age were therefore taken into
account in the analyses. Finally, research has shown that adolescents of
foreign heritage are more likely to be exposed to long-term risk factors than
adolescents of native Dutch heritage (Dekoviç et al. 2006). The analyses
therefore also controlled for heritage (0 for ‘persons of native Dutch
heritage’ and 1 for ‘persons of foreign heritage’).
In order to investigate the relationship between long- and short-term
risk factors and the degree of delinquency, we first produced bivariate logit
analyses using STATA (version 7.0). The association was examined on the
basis of the overall χ2 (significance level p < .05). Marginal effects were
calculated in order to establish the risk effect of a variable on delinquency
(Borooah 2002; Liao 1994). This is a relatively new method within the
field of criminology. A major advantage of marginal effects is that they are
easier to interpret than, for instance, odds ratios or regression coefficients,
because they are represented in terms of percentages. In our study, the
moderate delinquents are the reference category. For example, a marginal
effect of 0.34 means that the presence of a specific risk variable such as
having used substances is linked to a 34 percent increased probability of
belonging to the group of serious juvenile delinquents. Next, we carried
out stepwise multivariate logistical regression analyses in order to examine
the effects of short-term risk factors, taking into account control variables
and long-term risk factors.6 Only those variables that were significant at the
univariate level were included in the models.7 In order to determine whether
the extension of a model leads to a significant improvement, the difference
in (–2) log likelihood (LL) of the various models was examined.8
6 Analyses using continuous (dependent and independent) variables lead to the same conclusions
the likelihood that an individual will commit an offence, taking into account long-term
risk factors (measured at the level of the individual offender), a multilevel model might be
adequate. However, we try to answer questions regarding differences between individuals
and not between offences. Because the question is to what extent short-term risk factors help
to explain serious delinquency of individuals, we think a logistical regression model is more
appropriate.
Results
First we examined the bivariate correlations between the variables and the
type of delinquent. Table 1 shows that all long-term risk factors have a
significant correlation with serious delinquency, with the exception of high
hyperactivity, emotional problems and openness towards parents. The table
also shows that serious delinquency displays a significant correlation with
short-term risk factors. Only the perceived risk of being caught is unrelated
to serious delinquency. Variables that do not display a significant (p < .05)
bivariate correlation with serious delinquency were not included in the
follow-up analysis.
Table 1 shows the bivariate marginal effects of the variables on the
group of serious delinquents. In the individual domain, only low prosocial
behaviour appears to be a risk factor in relation to serious delinquency.
Low prosocial behaviour increases the probability of serious delinquency
Table 1 Bivariate relations between long-term and short-term risk factors and type of
delinquent (N = 292)
by 16 percent (p < .05), on top of the basic level of 45.3 percent. In the
family domain, risk effects are identified for low emotional support from
parents, low parental solicitation and inadequate parental supervision. In
the school domain, risk effects are found for poor performance at school
and weak attachment to school. Finally, the short-term risk factors of no
tangible guardians, the presence of co-offenders and having used alcohol
and/or drugs prior to committing the offence also increase the probability
that an individual will belong to the group ‘serious delinquents’.
Multivariate relationships
Stepwise multivariate logistic regression analyses were carried out in order
to examine what effect individual long-term risk factors have on serious
delinquency and whether or not short-term risk factors also play a role in
this regard. Table 2 presents the results. Model 1 includes only the control
variables. Gender and age show a significant relationship with serious
delinquency (young males are 15 percent more likely than young females
to belong to the group ‘serious delinquents’, and adolescents aged 14–17
are 22 percent more likely to belong to this group than those aged 10–13).
Entering Model 2, which includes long-term risk factors in addition to the
control variables, leads to a significant improvement (∆ –2LL(6) = 40.48,
p < .001). Low prosocial behaviour and inadequate parental supervision
both increase the likelihood of serious delinquency by 16 percent, and poor
performance at school increases the probability by 21 percent and weak
school attachment increases the probability by 24 percent.
The last step (Model 3) adds short-term risk factors to the model. This
also improves the model significantly (∆ –2LL(3) = 15.27, p < .01). The
absence of tangible guardians and having used substances both increase the
likelihood of serious delinquency, regardless of the long-term risk factors.
Table 2 Multivariate relations between long-term and short-term risk factors and type of
delinquent (N = 292)
Model 4 Model 5
Discussion
The aim of this study was to examine both short-term and long-term
risk factors for serious juvenile delinquency. The findings indicate that,
controlling for long-term risk factors, short-term risk factors do play a
role in explaining serious juvenile delinquency. The circumstances in which
serious delinquents commit offences differ substantially from those in which
relatively moderate delinquents offend.
Short-term risk factors have been largely ignored in the field of
developmental criminology. Serious delinquency is mainly explained by the
presence of long-term risk factors. In our study, long-term risk factors are
also important. If a variety of long-term risk factors is taken into account,
the correlation between serious delinquency and the presence of a sole long-
term risk factor is only weak. This confirms the findings of previous studies.
If we examine several risk factors simultaneously, long-term risk factors
such as limited prosocial behaviour, inadequate parental supervision and
weak school attachment appear to be particularly relevant. In accordance
with the literature, we also conclude that the accumulation of risk factors
Methodological limitations
This research has several methodological limitations. The cross-sectional
character of the study implies that we do not pretend to make any statements
regarding causality, which is similar to much empirical research in the long-
term risk factor tradition. We do claim that the circumstances in which
serious delinquents commit offences differ substantially from those in which
relatively moderate delinquents offend, yet this does not necessarily imply
that these circumstances actually cause serious delinquency. Furthermore,
we relied on self-reported offences with a reference period of 12 months.
This could have caused a retrospective recall bias. However, research has
shown that a reference period of 12 months is still valid with regard to self-
reported delinquency (see e.g. Huizinga and Elliot 1986). Another limitation
is that we have examined only those situations in which actual offences
were committed; we have no information on similar situations in which no
offences were perpetrated. These limitations, however, are inherent aspects
of our research method and are quite similar to the limitations of other
studies. One way to move forward in this field of research would be to
design a longitudinal study that incorporates observations of ‘hotspots’.
This type of study is extremely rare, with the exception of the study by
Wikström and Butterworth (2006).
Wikström and Butterworth (2006) distinguish between three different
groups of adolescents. The first, the non-delinquent group, generally does
not regard delinquency as an option and therefore rarely or never commits
offences, regardless of the situation. The second group commits offences ‘as
a habit’ and is therefore not influenced to a large extent by the situation:
these individuals in fact take advantage of any opportunity that arises.
The third group is most interesting: it does regard crime as an option, yet
situational factors explain whether or not these individuals choose to commit
an offence. The situations in which these adolescents find themselves are
therefore extremely relevant, and the lifestyles of individuals appear to be
a particularly significant factor in this regard (see e.g. Osgood et al. 1996;
Haynie et al. 2006; Pauwels 2007). For this last group, preventive measures
(e.g. by limiting alcohol consumption) could have an impact, whereas such
measures would have little effect on the ‘habitual’ group.
It is interesting to note that our findings indicate that, for the most
serious group of juvenile delinquents, short-term risk factors might play
a role as well. Such factors, most notably having used substances prior to
the offence and the absence of tangible guardians, could be interpreted as
causal factors or as indicators of a different lifestyle that systematically puts
these offenders in risky situations more frequently than others. However,
even if these factors are supposed to be indicators of long-term risks, they
can be triggered only by the features of the actual situation in which the
individual finds himself or herself. Hence, in future research, risk factor
research should be combined more frequently with so-called ‘time budget’
questionnaires in which adolescents reconstruct situations, activities and
companions for a specific period and on an hour-by-hour basis (Wikström
and Butterworth 2006). This type of research could provide a greater insight
into which adolescents commit offences in which situations, thus providing
a starting point for the prevention of delinquency.
The findings of our study are relevant for both theory and empirical
research. The results show that situational factors play an important role in
explaining serious delinquency, if we take into account (an accumulation of)
long-term risk factors. The circumstances in which serious delinquents com-
mit offences differ substantially from those in which moderate delinquents
offend. Hence, the emerging interest in short-term risk factors in theory and
research should produce more integrated theory and more research into
both long-term and short-term risk factors.
Appendix A
Acknowledgements
References
2005 WODC Youth Delinquency Survey]. The Hague: WODC [Research and
Documentation Centre].
Van Widenfelt, B., Goedhart, A., Treffers, P. D. A. and Goodman, R. (2003). Dutch
version of the Strengths and Difficulties Questionnaire (SDQ). European
Child & Adolescent Psychiatry 12, 281–9.
Veenstra, R., Lindenberg, S., Oldehinkel, A. J., De Winter, A. F. and Ormel, J.
(2006). Temperament, environment, and antisocial behavior in a population
sample of preadolescent boys and girls. International Journal of Behavioral
Development 30, 422–32.
Warr, M. (2002). Companions in crime. Cambridge: Cambridge University Press.
Weerman, F. C., Smeenk, W., Slotboom, A., Harland, P., Den Dijker, L., Bijleveld,
C. et al. (2003). De survey van het NSCR-schoolproject. Documentatie en
tabellenboek van de eerste onderzoeksronde [Survey from the NISCALE
school project. Documentation and book of tables from the first study round]
(Rep. No. 2003-6). Leiden: NSCR [Netherlands Institute for the Study of
Crime and Law Enforcement].
Wikström, P.-O. H. (2005) The social origins of pathways in crime: Towards a
developmental ecological action theory of crime involvement and its changes.
In D. P. Farrington (ed.) Integrated developmental and life-course theories of
offending, pp. 211–46. New Brunswick, NJ: Transaction Publishers.
Wikström, P.-O. H. and Butterworth, D. A. (2006). Adolescent crime. Individual
differences and lifestyles. Devon: Willan Publishing.
Wikström, P.-O. H. and Sampson, R. J. (2003). Social mechanisms of community
influences on crime and pathways in criminality. In B. B. Lahey, T. E. Moffitt
and A. Caspi (eds) Causes of conduct disorder and juvenile delinquency, pp.
118–48. New York: Guilford Press.
André M. van der Laan is a Senior Researcher at the Research and Docu-
mentation Centre (WODC) of the Dutch Ministry of Justice. His research
interests are in developmental criminology, the effects of juvenile sanctions
and the adverse effects of sanctions. He coordinates the WODC Youth
Delinquency Survey, which has reported on a regular basis to the Ministry
of Justice on the trends in juvenile crime. He has published on juvenile
crime, risk and protective factors of delinquency, trends in crime, juveniles’
experiences of sanctioning, the effects of sanctions and defiance theory.
a.m.van.der.laan@minjus.nl
Martine Blom
crime and criminal recidivism. She has published on trends in juvenile crime,
recidivism of juvenile offenders and crime groups of different heritage.
m.blom1@minjus.nl
Edward R. Kleemans