Juvenile Justice
ctillage
trajectory.pdf
Laws 2014, 3, 651–673; doi:10.3390/laws3040651
laws ISSN 2075-471X
www.mdpi.com/journal/laws/
Article
Trajectories and Risk Factors of Criminal Behavior among Females from Adolescence to Early Adulthood
Julie M. Krupa and Kristina K. Childs *
Department of Criminal Justice, University of Central Florida, 12805 Pegasus Dr., Orlando, FL 32816,
USA; E-Mail: [email protected]
* Author to whom correspondence should be addressed; E-Mail: [email protected];
Tel.: +1-407-823-5981; Fax: +1-407-823-5360.
External Editor: Tamar R. Birckhead
Received: 11 June 2014; in revised form: 28 August 2014 / Accepted: 5 September 2014 /
Published: 29 September 2014
Abstract: Previous research suggests that social environmental and individual-level factors
influence adolescent development and behavior over time. However, little attention has
been devoted to examining how risk factors (i.e., parental support, peer delinquency,
self-control) affect trajectories of criminal behavior among female adolescents. Utilizing
data from the National Longitudinal Study of Adolescent Health (n = 5138 females) and
latent class analysis, three offending trajectories among females from late adolescence to
early adulthood were identified: late escalators, late de-escalators, and stable low/abstainers.
Next, the influence of social environmental and individual-level factors during adolescence
(Wave 1) on these trajectories was assessed. Results identified key differences in the risk
factors related to group placement. The implications of the findings for prevention and
treatment services targeting adolescent females, and directions for future research,
are discussed.
Keywords: offending trajectories; female delinquency; risk factors for criminal behavior
1. Introduction
It is well documented that the proportion of girls involved in the juvenile justice system has steadily
increased over the past two decades. For example, from 1992 to 2011, the proportion of juvenile
OPEN ACCESS
Laws 2014, 3 652
arrests that were girls increased from 23% to 29% [1,2]. In response to these increases, federal and
local initiatives, policy recommendations, and a range of funding opportunities have been created to
address the needs of adolescent female offenders. For example, the Office of Juvenile Justice
Delinquency Prevention (OJJDP) has devoted funding to prevention and intervention services designed
specifically for girls since the late 1990s [3]. Along with these initiatives, a growing body of research
has evolved that has examined gender differences in justice system involvement, theoretical
explanations of delinquency, responsiveness to treatment, and equitable juvenile justice decision-making.
Overall, this body of research suggests that specialized approaches to understanding, preventing,
and responding to juvenile delinquency are warranted due to differences in boys’ and girls’
developmental experiences [4].
However, as Johanssan and Kempf-Leonard argued, “…the research about female risk factors for
general delinquency, as well as research addressing girls’ serious, violent, and chronic offending, has
been insufficient” ([5], p. 217). One area where this argument applies is research devoted to
understanding the developmental pathways of criminal behavior among adolescent girls. Therefore,
using a large sample of female adolescents, this study seeks to expand the body of research on female
offending by examining how individual-level risk and protective factors influence the development of
different trajectories of offending throughout adolescence and early adulthood. Identifying the risk and
protective factors associated with different trajectories of offending among females, specifically, will
assist in the development of gender-specific prevention and intervention programs that are able to:
(a) meet the needs of adolescent girls that are at risk for persistent offending; (b) ensure appropriate
allocation of prevention resources so that agencies’ treatment budgets are going towards programs that
are able to meet the unique needs of adolescent girls (as well as boys); and (c) reduce costly justice
system involvement, persistent criminal behavior, social instability, and poor outcomes among girls
at-risk for delinquent behavior [6,7].
2. Trajectories of Criminal Behavior
In recent years, a large number of group-based trajectory studies have been conducted to describe
the development and continuity of antisocial and delinquent behavior [8–10]. For example, in Moffitt’s
landmark study, she identified two trajectories of adolescent offending [11]. Life-course persistent
offenders displayed an early onset of offending and continued to engage in criminal behavior
throughout the life course. Adolescent-limited offenders engaged in delinquency during adolescence,
but desisted as they got older. Moffitt also identified different risk factors for placement into each
group [11]. Early onset of delinquent behavior, neuropsychological defects, and deprived family
environments were important risk factors among the life-course persisters. Among adolescent-limited
offenders, the imitation of delinquent peers was the strongest risk factor. More recently, a number of
studies have identified more than two groups of offenders, providing support for Moffitt’s concept of
life-course persisters, while identifying greater diversity in offending trajectories [12]. For example,
using data on boys and girls involved in the Seattle Social Development Project, Chung et al.
identified five groups of offending trajectories: non-offenders, late onsetters, desisters, escalators, and
chronic offenders [13]. Similarly, Fergusson and Horwood also identified five groups of offending
trajectories using a New Zealand sample of children (boys and girls) that were followed from birth to
Laws 2014, 3 653
age 21. These groups included low-risk, early onset, intermediate onset, late onset, and chronic
offenders [14]. More recently, Jennings and Reingle conducted a meta-analysis of 105 empirical
studies that examined the number and shape of group-based trajectories of delinquency, violence, and
aggression [15]. The 105 studies included samples of boys only, girls only, but consisted mostly of
samples of boys and girls. Across these studies, 2–7 trajectory groups were identified with the majority
identifying three or four groups. The number of trajectory groups identified in each study varied based
on factors such as geographical location, sample, measurement, and length of observation.
Indeed, when studies rely on samples of both boys and girls, girls can be found in the same
trajectories as boys [14–16]. However, gender differences in the level and duration of criminal
behavior within offending trajectories are commonly identified [8,14,17–19]. For example, Pepler,
Jiang, Craig and Connolly analyzed developmental trajectories of delinquency among adolescent girls
and boys [20]. Adolescents were characterized into five different classes: low, moderate, late onset,
early onset, and chronic. When compared within classes, girls differed from boys in their levels of
delinquency. Girls in the “late onset” group displayed higher levels of delinquency after the age of 15
whereas girls in the “early onset” group tended to desist from delinquency sooner than their male
counterparts and girls in the “moderate” group displayed higher levels of delinquency in early
adolescence and peaked sooner than boys. Using data from the Philadelphia Collaborative Perinatal
Project (CPP), Piquero and Chung found that once relevant controls (i.e., income, family structure,
WISC, mother’s age at child birth, and disciplinary codes) were introduced in their study, the early
onset and seriousness of offending relationship disappeared among the girls (but not the boys), leaving
only a female late-onset group [21]. Using data from the National Longitudinal Study of Youth,
Murphy, Brecht, Huang, and Herbeck found that age at which trajectories changed varied by gender
with females in the “decreased” group declining at an earlier age [22].
In addition, research that has examined gender-specific trajectories has also identified important
differences in the number and shape of offending trajectories [23,24]. Zheng and Cleveland (2013)
examined gender differences in the development of violent and nonviolent delinquent behavior from
ages 12–22 [25]. Females were categorized into three classes: low (60%), desister (30%), and decliner
(11%). Males were categorized into four different classes: low (50%), desister (25%), chronic (13%),
and decliner (12%). Males in the “desister” class reported more violent than nonviolent delinquency
and more overall delinquency than their female counterparts. Also, males in the “decliner” class
reported nonviolent and violent delinquency; while females primarily reported nonviolent behavior.
Finally, the “chronic” class was gender-specific applying only to males.
In sum, previous studies based on boys and girls have documented important gender differences in
the characteristics of offending over time, including the frequency of offending over time, age of onset,
and patterns of desistence and persistence [12,22,25,26]. However, there is a lack of studies that are
dedicated to the examination of female offending from adolescence to early adulthood. Given the
recent increase in girls’ involvement in the juvenile justice system, additional research is needed to
fully understand the nature of offending trajectories among girls so that effective gender-responsive
prevention and intervention programs can be developed. Therefore, the first goal of the current study is
to examine offending trajectories from adolescence to early adulthood among a nationally
representative sample of female adolescents.
Laws 2014, 3 654
Given the lack of research that has been devoted to studying offending trajectories among samples
of female adolescents, there is also scant research on the risk and protective factors related to different
delinquent pathways among girls. Research on the risk and protective factors related to general
delinquency suggests that boys and girls share common risk factors such as parental responsiveness
and acceptance, having delinquent peers, low school commitment, low self-control, low self-esteem,
and abuse [27–30]. However, research also shows that certain risk factors for offending have a greater
influence on girls compared to boys (and vice versa) [18]. For example, parenting practices have been
shown to have a stronger effect on criminal behavior for girls compared to boys [27,30,31] while
males have been found to be more susceptible to peer influences [7,29,32,33]. Individual-level risk
factors such as low self-control, impulsivity, and early aggression have also been found to be stronger
predictors of criminal behavior for boys [29,34–36], whereas mental health problems and suicide
ideation are more prevalent among female offenders [30,37]. Studies examining gender differences in
the impact of low self-esteem on delinquent behavior are mixed [38,39].
Yet, how these factors influence different developmental pathways of offending among female
adolescents, including persistence and desistance, is not well understood. Only a handful of studies
have examined the risk and protective factors associated with different trajectories of criminal behavior
across gender [22,40–42] and these results are rather inconsistent. Based on previous research that has
identified differences in the factors related to different offending trajectories among samples of boys
and girls [12,18,20,24] as well as the extant body of research that documents gender differences in the
risk and protective factors related to general delinquency [30–32], additional research is needed to
fully understand the risk and protective factors related to different patterns of offending among female
adolescents. This information is critical to developing prevention programs that are able to target the
unique risk factors related to different “types” or patterns of offending among female adolescents.
Therefore, the second goal of the current study is to examine the female-specific risk and protective
factors, measured during mid-adolescence (Wave 1), related to different patterns of offending from
adolescence through early adulthood.
In sum, the majority of studies on trajectories of adolescent offending involve samples of boys and
girls [8,12–14,20,21]. This is a limitation to the current body of research on offending trajectories
because a number of studies have revealed important gender differences in the characteristics of
offender trajectories (length of offending, number of different pathways, proportion of individuals that
fall within each trajectory), risk and protective factors related to juvenile delinquency, and
responsiveness to intervention programs [13,17,33,43]. Due to these differences, the lack of research
that focuses specifically on patterns of female offending during adolescence has led to limited
knowledge about the factors that predict different offending trajectories among girls [44,45].
3. Current Study
In general, the current study seeks to add to the growing body of research on female offending.
Specifically, the goals of the current study are to:
1. Identify different trajectories of offending from adolescence through early adulthood among a
large, representative sample of female adolescents;
Laws 2014, 3 655
2. Identify the risk and protective factors that are related to membership in each trajectory
during adolescence.
3.1. Methods
Data from the National Longitudinal Study of Adolescent Health (Add Health), a nationally
representative study of more than 20,000 adolescents, was used in the current study. Add Health used a
multi-stage stratified sampling design to select study respondents from 132 middle and high schools
from 80 communities across the United States. The first wave was collected in 1994–1995 when the
study respondents were in grades 7–12 (for details on the design of the Add Health data, please
see [46]). To date, four waves of data have been collected. Wave 1 interviews were conducted in 1995,
Wave 2 interviews were conducted in 1996, Wave 3 interviews were conducted from 2000 to 2001,
and Wave 4 interviews were conducted from 2008 to 2009 1. Across the four waves, the broader Add
Health study has included a number of data collection programs including in-school surveys, parent
surveys, school administrator surveys, and in-home interviews. The current study uses data from
female respondents that participated in all four waves of the in-home interview.
Given the complexity of the Add Health study design, sampling weights were applied to yield
national population estimates. Therefore, cases that were missing valid Wave 4 weight components
were excluded from the analyses (n = 380, 6.9% of the respondents that were under the age of 18)
(see [47]). These exclusions yielded a final sample size of 5138 females that were involved in all four
waves of the in-home interviews. The average age of the female sample was 15.2 (SD = 1.6) at Wave 1,
16.08 (SD = 1.6) at Wave 2, 21.5 (SD = 1.6) at Wave 3, and 28.1 (SD = 1.6) at Wave 4. Sixty-four
percent of the girls included in the sample were White, 23% were Black, 6% were Asian, and 7% were
an “other” race.
3.1.1. Variables 2
Criminal Behavior. Criminal behavior was measured using 10 self-reported behaviors. The
selection of the items was based on the inclusion of the behavior in all four waves of data collection.
The 10 items included past year involvement in violence, property crime, and public disorder offenses.
The five violence-related items asked respondents how many times in the past year they had gotten
into a serious physical fight, hurt someone so badly they needed medical care, took part in a group
fight, pulled a knife or gun on someone, and used a weapon to steal something. Property crime
offenses included how many times in the past year respondents reported going into a house or building
to steal something, stealing something worth more than $50, and stealing something worth less than
$50. Additional items included how many times in the past year respondents sold marijuana or other
drugs and deliberately damaged property that belonged to another person. All 10 items were originally
coded on an ordinal scale ranging from never (=0) to five or more times (=3). Due to the low number
of females that reported frequent engagement in each of the behaviors, all 10 items were recoded into
1 Response rates for each wave ranged from 77% to 89%. 2 See Harris et al. for a detailed description of each of the study variables [46].
Laws 2014, 3 656
dichotomous indicators representing never (=0) and one or more times in the past year (=1).
The means and standard deviations for each of the dichotomized offenses are reported in Table 1.
Table 1. Means and standard deviations for the offense items.
Wave 1 Wave 2 Wave 3 Wave 4
Mean (SD)
Serious physical fight * 0.24 (0.42) 0.13 (0.34) - 0.12 (0.15)
Hurt someone 0.11 (0.31) 0.04 (0.20) 0.02 (0.15) 0.01 (0.07)
Group fight 0.17 (0.38) 0.14 (0.34) 0.03 (0.18) 0.01 (0.11)
Pulled a knife/weapon out 0.02 (0.15) 0.02 (0.14) 0.01 (0.07) 0.01 (0.11)
Weapon to steal 0.03 (0.17) 0.02 (0.14) 0.01 (0.09) 0.01 (0.06)
Burglary 0.03 (0.18) 0.03 (0.16) 0.01 (0.10) 0.01 (0.06)
Stole something worth more than $50 0.03 (0.18) 0.03 (0.17) 0.02 (0.14) 0.01 (0.10)
Stole something worth less than $50 0.17 (0.36) 0.14 (0.35) 0.05 (0.22) 0.02 (0.15)
Sold drugs 0.04 (0.20) 0.04 (0.19) 0.04 (0.19) 0.02 (0.14)
Deliberately damaged property 0.13 (0.33) 0.09 (0.29) 0.05 (0.21) 0.02 (0.15)
Note: *The Wave 3 interview did not include a measure of past year involvement in a serious physical fight.
At each wave, the 10 items were summated into one overall general offending index ranging
from 0 to 10 3. The average for the offending index at Wave 1 was 0.98 (SD = 1.49, factor loadings
ranged from 0.48 to 0.58, alpha = 0.70), the average at Wave 2 was 0.67 (SD = 1.26, factor loadings
ranged from 0.30 to 0.66, alpha = 0.68), the average at Wave 3 was 0.23 (SD = 0.69, factor loadings
ranged from 0.30 to 0.67, alpha = 0.60), and the average at Wave 4 was 0.16 (SD = 0.54, factor
loadings ranged from 0.30 to 0.60, alpha = 0.58) 4. As can be seen, among the females included in the
sample, the average level as well as the variation in offending revealed a linear decline across the four
waves of data collection. The risk factors used to predict criminal behavior are discussed below. All of
these risk factors were measured at Wave 1 of the in-home interview.
Parental Involvement. Parental involvement was measured using 10 items, five items referred to the
mother figure (e.g., step mother, adoptive mother, foster mother) and five items referred to the father
figure [8,49]. These items asked respondents if they had participated in the following activities with
their mother/father figure in the last four weeks: gone shopping, played a sport, attended a religious
service, worked on a school project, and gone to a movie, play, museum, concert, or sporting event.
Responses were dichotomous representing that they had participated in the activity (=1) or had not
participated in the activity (=0). If a respondent reported living with both a mother and father figure,
the average of the 10 items was used. If a respondent reported living with only a mother-figure or only
3 The Wave 3 interview did not include a measure of past year involvement in a serious physical fight. Therefore, the
wave 3 index ranged from 0 to 9. 4 There were no significant differences in the four delinquency indices across the females included in the sample and
those excluded (due to not participating in all four waves of data or having invalid sampling weights). In addition, the
original Add Health investigators concluded that the bias due to nonresponse was small in magnitude. For the
delinquency and violence indices specifically, they concluded that the bias was not significantly different from zero [48].
Laws 2014, 3 657
a father-figure, the sum of the five items was used (factor loadings for the father ranged from 0.98 to 1.0
and factor loadings for the mother figure ranged from 0.96 to 0.98, alpha = 0.89).
Parental Attachment. Parental attachment was measured using 10 items, five of which referred to
the mother figure and five that referred to the father figure [49,50]. These items asked each respondent
how he or she felt about their mother/father figure including whether the mother/father figure was
warming and loving, if the respondent was satisfied with communication, and if he or she was satisfied
with the overall relationship. Responses were dichotomous representing that they agreed with the
statement (=1) or did not agree (=0). If a respondent reported living with both a mother and father
figure, the average of the 10 items was used. If a respondent reported living with only a mother-figure
or only a father-figure, the sum of the five items used (factor loadings for the father ranged from 0.47
to 0.61 and factor loadings for the mother figure ranged from 0.51 to 0.80, alpha = 0.88).
Parental Control. Parental control was measured using seven items that asked about respondents’
freedom to make their own decisions about curfew on weekend nights, people they could hang around,
what to wear, how much television to watch, what types of television programs to watch, what time to
go to bed on week nights, and what to eat [49]. Each item was coded as a dichotomous variable. These
seven items were summed to create an overall parental control scale (factor loadings ranged from 0.47
to 0.73, alpha = 0.78).
School Attachment. School attachment was measured using five items asking respondents if they
felt close to people at their school, part of the school, happy to be at school, teachers treated them
fairly, and safe at school [51,52]. These items were coded on an ordinal scale ranging from strongly
agree (=1) to strongly disagree (=5). All five items were reverse coded and summed to create an
overall school attachment index (factor loadings ranged from 0.72 to 0.86, alpha = 0.86).
Peer Substance Use. Similar to previous studies using the in-home interviews, peer substance use
was used as a proxy for peer misbehavior [53–55]. Peer substance use was measured using three items
that asked respondents about the substance use activities of their three best friends. Specifically, these
items asked respondents to report how many of their three best friends smoked at least one cigarette a
day, drank alcohol at least once a month, and used marijuana at least once a month. All three items
were summed to create an overall peer substance use scale (factor loadings ranged from 0.81 to 0.85,
alpha = 0.76).
Substance Abuse 5. Substance abuse was measured using four separate items. The first measure
asked respondents how many days, in the past 30 days, each respondent smoked at least one cigarette.
The second measure asked “during the past 12 months, how many days have you gotten drunk or very
high on alcohol”. Responses were coded into five categories ranging from never to at least once per
week. The third item measured the number of times each respondent reported using marijuana in the
past 30 days. The final item measured the use of other, more serious forms of drug use. Three items 5 We chose to include substance use as a predictor of delinquent trajectories instead of a form of delinquent behavior for
many reasons. Indeed, both substance use and delinquent behavior are forms of deviant behavior. However, there are
also important differences in the characteristics of substance use and delinquent behavior that suggest that these
behaviors are conceptually distinct. A number of studies have found substance use and delinquent behavior to be
distinct dimensions of deviant or risk-taking behavior [56,57]. Since the goal of this study was to examine trajectories
of delinquent behavior among girls over time, we felt that is was necessary to include substance use as a risk factor for
delinquent behavior, rather than a form of delinquent behavior itself.
Laws 2014, 3 658
which asked respondents, how many times, in the past 30 days, they used inhalants, cocaine, or “other”
drugs were used to measure use of more serious forms of drug use. Due to the low number of
respondents that reported use, these items were summed and then dichotomized to represent never
using other, more serious forms of substances (=0) or using other, more serious forms of substances
one or more times in the past 30 days (=1).
Self-control. Following previous research using the Add Health data, we measured self-control
using a behavioral measure that included 5 items which pertained to behavior at school [58,59]. These
questions asked how often each respondent had trouble getting along with teachers, paying attention in
school, getting homework done, getting along with other students, and keeping focused. Responses
were measured on an ordinal scale ranging from never (=0) to everyday (=4). These five items were
summed to create an overall proxy measure for self-control (factor loadings ranged from 0.56 to 0.80,
alpha = 0.70). Higher values represent lower levels of self-control.
Depression. Depression was measured using a modified version of the Center for Epidemiological
Studies-Depression Scale (CES-D), which is commonly used to measure depression in adolescents
using the Add Health Data [60–62]. Seventeen items which asked respondents about their emotional
status in the past week were included. Responses ranged from never or rarely (=0) experiencing these
feeling to most of the time or all the time (=3). Some of these items included, having felt “fearful”,
“lonely”, “sad”, “felt people disliked you”, and “felt as though you could not shake off the blues”.
These 17 items were summed to create an overall measure of depression (factor loadings ranged from
0.38 to 0.79, alpha = 0.87).
Self-esteem. Self-esteem was measured using three items which asked respondents about their
feelings about oneself [52]. These items asked whether the respondent felt they had a lot of good qualities,
liked themselves the way they were, and whether they felt loved and wanted. The three items were
summed to create an overall self-esteem scale (factor loadings ranged from 0.80 to 0.88, alpha = 0.78).
3.1.2. Analysis
The analyses proceeded in several steps. First, bivariate correlations among the four self-reported
offending indices were examined. Next, a series of latent class models (LCA) were performed to
identify the best fitting model 6. LCA estimates a model that extracts latent “classes” or categories
based on patterns in observed indicators. The patterns are hypothesized to be related to some
underlying unobserved factor (i.e., trajectory) rather than being causally related [63]. Thus, based on the
offending indices at each wave, identifiable “classes” or trajectories of criminal behavior were extracted 7.
The issue of class enumeration in mixture modeling remains unresolved. Therefore, the reliance on
multiple criteria and theory to aid in selecting the appropriate number of classes is recommended [64].
The criteria used to determine the best-fitting model were the classification table based on class
6 Growth mixture models and latent class growth analyses were considered, but due to the low levels of delinquency
found at each of the four waves, the low number of time points available, and the complex nature of measuring growth
over time, a parsimonious LCA model was chosen. 7 Missing delinquency data ranged from 0.8% in Wave 1 to 10% in Wave 4. All study participants had at least one wave
of valid delinquency data. Mplus uses full information maximum likelihood to estimate the latent classes based on
available information.
Laws 2014, 3 659
probabilities for the most likely latent class membership, the entropy score, the Bayesian information
criterion (BIC), the Lo-Mendell Rubin (LMR), and the Lo-Mendell-Rubin adjusted likelihood ratio test
(LRT) [64,65]. For the classification table, high diagonal values and low off-diagonal values indicate
good classification quality. The values of entropy range from 0 to 1, with scores close to 1 indicating
clear classifications [66]. For LMR and LRT, a significant p-value indicates that the specified model
(with k classes) fits significantly better than a model with one less class (k-1). Furthermore, the
substantive meaning of the classes was also considered. All latent class analyses were conducted in
Mplus 6.0 using maximum likelihood estimation. Also, complex survey commands (i.e., weighting)
were used to account for the complexity of the Add Health survey design [47,67].
Once the most appropriate LCA model was selected, the posterior probabilities for most likely class
membership were used to examine differences in the risk and protective factors (measured at Wave 1)
that were related to each class. Independent analyses of variance (ANOVA) and chi-square tests of
significance (χ²) were used to examine the bivariate relationship between each of the risk/protective
factors and class membership. Finally, the factors that were significantly related to class membership
at the bivariate level were entered into a multinomial logistic regression model to identify which
factors (at Wave 1), controlling for age and race, predicted latent class membership 8. Specifically, the
relative risk ratios (RRR) were used to identify significant relationships. All bivariate and multivariate
models were conducted in Stata 13.0 using survey commands to account for the complexity of the Add
Health study design 9.
3.2. Results
Table 2 provides the bivariate correlations among the four offending indices. Correlations
ranged from 0.17 to 0.54 and were positive and significant (p < 0.01). Wave 1 offending and Wave 2
offending showed the strongest correlations and Wave 1 offending and Wave 4 offending showed the
weakest correlation.
Table 2. Correlations among self-reported offending at Waves 1–4 *.
Wave 1 Wave 2 Wave 3 Wave 4
Wave 1 - - - - Wave 2 0.54 - - - Wave 3 0.19 0.22 - - Wave 4 0.17 0.19 0.26 -
Note: * All correlations were significant (p < 0.01).
The next step of the analyses involved comparing a number of latent class models to identify the
best-fitting model. These results are presented in Table 3. As can be seen, the three-class solution was
the best fitting model across all of the model fit indices. There was a significant decline in the BIC and
8 Age and race were included as control variables. Age was a continuous variable representing age at the time of the
Wave 1 interview. The average age at Wave 1 was 15.2 (SD = 1.6). Race was a categorical variable coded as White
(64%), Black (23%), and Other (13%). 9 The correlations among the risk and protective factors ranged from −0.08 to 0.52 and all variation inflation factors were
less than 2.0.
Laws 2014, 3 660
log likelihood, entropy was close to 1, and the LMR and LRT were significant suggesting that the
two-class model can be rejected in favor of the three-class model. The results of the three-class LCA
model are depicted in Figure 1.
Table 3. Model Fit Indices for the Latent Class Analyses (n = 5138).
BIC Log Likelihood Entropy LMR LRT Average Latent ClassProbabilities
2-Classes 49262.86 −24575.89 0.98 4435.82
(p = 0.40)
4334.36
(p = 0.40) 1.00, 0.95
3-Classes 45833.20 −21896.54 0.99 3472.38
(p = 0.02)
3392.96
(p = 0.02) 1.00, 0.99, 0.99
4-Classes 40083.29 −19943.39 0.98 5189.53
(p = 0.68)
5070.84
(p = 0.68) 0.91, 0.99, 0.99, 0.99
Figure 1. Offending trajectories of the three latent classes (n = 5138).
0
1
2
3
4
5
6
7
8
9
10
1 2 3 4
S el
f- R
ep or
te d
D el
in q
u en
cy
Wave
Stable Low/Abstainer (C1), 92% Late Escalators (C2), 3% Late De-escalators (C3), 5%
The first class, labeled “stable low/abstainers” accounted for the majority of the sample (92%).
The girls in this group reported no to very low levels of criminal behavior across the four time points.
The average level of offending ranged from 0.88 (SE = 0.03) at Wave 1 to 0.06 (SE = 0.01) at Wave 4.
The second group, labeled “late escalators” accounted for 3% of the sample. The girls in this group
reported the highest levels of criminal behavior at Wave 1 and steadily declined during Waves 2 and 3
but increased their criminal behavior at Wave 4 (Wave 1 mean = 2.80, SE = 0.22; Wave 2 mean = 1.71,
SE = 0.23; Wave 3 mean = 0.66, SE = 0.11; Wave 4 mean = 2.52, SE = 0.08). The third group, labeled
“late de-escalators” accounted for 5% of the sample. Girls in this group reported the highest average
levels of criminal behavior at Wave 3 but showed a significant decline in criminal behavior at Wave 4
(Wave 1 mean = 2.09, SE = 0.12; Wave 2 mean = 1.89, SE = 0.12; Wave 3 mean = 2.72, SE = 0.02;
Wave 4 mean = 0.37, SE = 0.01). Importantly, although there are meaningful differences in offending
patterns across the groups, it is clear that none of the groups were characterized by a “high” or
“chronic” pattern of criminal behavior. In fact, all three groups showed relatively low levels of
criminal behavior across the four time points.
Laws 2014, 3 661
The next step in the analyses sought to identify which factors, measured at Wave 1, predicted group
membership. Bivariate analyses, examining the relationship between group membership and each
individual risk/protective factor, are presented in Table 4. Compared to the “late escalators”, the
“stable low/abstainers” showed significantly higher levels of parental involvement, self-control, and
attachment to school and significantly lower levels of depression, peer substance use, marijuana use,
cigarette use, frequency of getting drunk, and “other” drug use. Compared to the “late de-escalators”,
the “stable low/abstainers” reported significantly higher levels of self-control, self-esteem, and
attachment to school and significantly lower levels of depression, parental attachment, and marijuana
use. Compared to the “late-escalators”, the “late de-escalators” reported significantly higher levels of
parental involvement and significantly lower levels of depression, peer substance use, and cigarette
use. Parental control and truancy did not significantly differ across the latent classes. Therefore, these
two factors were not included in the multinomial logistic regression models.
Table 4. Bivariate analyses of the risk/protective factors and latent class trajectory.
Mean (SD) Mean (SD) Mean (SD)
C1 Stable Low/Abstainers
C2 Late Escalators
C3 Late De-Escalators
Self-Control a,b 4.61 (3.17) 5.91 (3.54) 5.83 (3.32) Depression a,b,c 10.50 (7.17) 14.02 (8.43) 11.98 (8.07) Self-Esteem b 5.31 (1.82) 5.27 (1.98) 5.57 (2.03) Parental Attachment b 14.85 (1.81) 15.11 (1.83) 15.14 (1.84) Parental Involvement a,c 1.51 (1.02) 1.29 (0.92) 1.60 (1.06) Parental Control 4.98 (1.66) 4.98 (1.74) 4.95 (1.64) Peer Substance Use a,c 2.24 (2.52) 3.15 (2.77) 2.50 (2.70) Marijuana Use a,b 0.82 (4.79) 2.50 (9.75) 2.15 (9.67) Cigarette Use a,c 3.70 (8.96) 6.82 (11.52) 4.47 (9.21) School Attachment a,b 18.10 (4.81) 17.32 (4.90) 17.43 (5.16) Truancy 8.72 (13.24) 10.10 (16.05) 6.63 (6.99)
Percentage Percentage Percentage
Frequency of Getting Drunk a Never 76.6% 58.8% 69.6% 1–2 times 11.7% 19.4% 14.3% Less than once a month 4.9% 10.3% 6.3% Two or more times per month 3.8% 5.5% 5.5% At least once per week 3.1% 6.1% 4.2%
Use of Other Drugs a No 96.1% 90.9% 89.7% Yes 3.9% 9.1% 10.3%
Notes: a Significant differences between the stable low/abstainers group and the late escalator group (p < 0.05); b Significant differences between the stable low/abstainers group and the de-escalator group (p < 0.05); c Significant differences between the escalator group and the de-escalator group (p < 0.05).
The results of the multinomial logistic regression, controlling for age and race, are presented in
Table 5. Each column in the table represents a comparison of two of three groups. Compared to the
“stable low/abstainers”, the “late escalators” were significantly more likely to report higher levels of
Laws 2014, 3 662
depression and lower levels of self-esteem at Wave 1. For example, a one unit increase in depression
led to 1.1 higher odds of being in the “late-escalators” group compared to the “stable low/abstainers”
group. Compared to the “stable low/abstainers”, the “late de-escalators” reported lower levels of
self-control, higher levels of parental involvement, and the use of other drugs. For instance, use of
other, more serious forms of substances at Wave 1 led to 2.1 higher odds of being in the “late
de-escalators” group compared to the “stable low/abstainers”. Finally, compared to the “late
escalators”, the “late de-escalators” reported significantly lower levels of depression, significantly
higher levels of self-esteem, and higher levels of parental involvement at Wave 1. For example, a one
unit increase in self-esteem was associated with 1.2 higher odds of being in the “late de-escalators”
group compared to the “late escalators” group. Overall, these results suggest that different
risk/protective factors present during mid-adolescence were related to different offending trajectories
from mid-adolescence to early adulthood.
Table 5. Multinomial logistic regression predicting latent class trajectory controlling for
age and race (n = 4814) *.
Stable Low/Abstainers
→ Late Escalators
Stable Low/Abstainers
→ Late De-Escalators
Late Escalators →
Late De-Escalators
RRR (CI) RRR (CI) RRR (CI)
Self-Control 1.03 (0.97–1.09) 1.09 (1.04–1.14) *** 1.06 (0.99–1.13)
Depression 1.06 (1.03–1.08) *** 1.00 (0.98–1.03) 0.95 (0.92–0.98) **
Self-Esteem 0.84 (0.76–0.93) ** 1.01 (0.93–1.11) 1.21 (1.06–1.38) **
Parental Attachment 1.00 (0.92–1.10) 1.07 (0.99–1.16) 1.07 (0.95–1.21)
Parental Involvement 0.87 (0.73–1.03) 1.17 (1.02–1.37) * 1.35 (1.08–1.67) **
Peer Substance Use 1.02 (0.95–1.05) 1.01 (0.94–1.08) 0.99 (0.89–1.09)
Frequency of Getting Drunk 1.18 (1.00–1.39) 1.08 (0.91–1.27) 0.91 (0.73–1.14)
Marijuana Use 1.01 (0.99–1.07) 1.01 (0.99–1.03) 1.00 (0.98–1.03)
Cigarette Use 1.02 (1.00–1.04) 1.00 (0.98–1.02) 0.98 (0.96–1.01)
Use of Other Drugs 1.50 (0.80–2.83) 2.09 (1.22–3.57) ** 1.40 (0.63–3.07)
School Attachment 1.00 (0.95–1.05) 1.00 (0.96–1.04) 1.00 (0.95–1.07)
Note: * p < 0.05, ** p < 0.01, *** p < 0.001.
4. Conclusions
The current study sought to identify different trajectories of offending and to identify the risk and
protective factors related to these patterns among a sample of females from adolescence to early
adulthood. Our results identified three distinct offending trajectories: “late de-escalators”, “late
escalators”, and “stable low/abstainers”. The “stable low/abstainers” reported very low levels of
criminal behavior across the four time points. The “late de-escalators” showed moderate levels of
criminal behavior across the first three time points but substantially decreased offending by the time
they reached their late 20s (Wave 4). The “late escalators” showed a decline in criminal behavior in
their early 20s (Wave 3) but then showed a marked increase in offending by their late 20s (Wave 4).
Laws 2014, 3 663
Interestingly, a group characterized by chronic and stable offending was not identified in the current
analyses. This is an interesting finding because it is inconsistent with a large number of studies
involving samples of boys [11,68] and boys and girls [13,14] that identify a relatively small group of
chronic, persistent offenders. Thus, a subgroup of life-course persistent offenders was not identified
among the girls included in our study. On the other hand, however, the “late de-escalators” may best
represent Moffitt’s adolescent-limited offenders given the decline in criminal behavior around age 22
(average age). To a certain degree, these findings align with previous trajectory studies that relied on
representative samples of girls. A number of studies have also found support for an adolescence-limited
group [69–71]. However, this group accounted for 5% of our sample of girls. In other, nationally
representative samples of girls, this group has been found to comprise 10%–12% of the sample.
Furthermore, a small handful of studies have also identified a delayed or adult-onset group of female
offenders similar to the “late escalators” found in our study [19,72–74]. These studies have found that
the percent of the female sample included in this group ranges from 4% to 24%. Only 3% of the girls
in our sample were included in the “late escalators” group. However, unlike the findings of the current
study, most studies that rely on nationally representative samples do find a small group of life-course
persistent female offenders. In general, the findings of this study highlight the importance of studying
trajectories of behavior among girls separately and the need for additional gender-specific research to
fully understand onset, persistence, and desistance of offending among girls.
Important differences in the risk and protective factors related to each trajectory were also found.
In early adolescence (Wave 1), individual-level factors such as depression and low self-esteem were
found to be significant predictors of placement into the “late escalators” group. In contrast,
both individual- and environmental-level factors were found to be related to placement in the “late
de-escalators” group. Parental involvement was an important predictor of placement in the “late
de-escalators” compared to the “late escalators” and self-control and use of other drugs were
significantly higher among this group compared to the “stable low/abstainers”. Thus, our findings
suggest that different factors spanning both individual and environmental domains are related to
different patterns of offending among female adolescents.
Overall, a number of similarities in the risk and protective factors found in other studies of boys and
girls were identified as important predictors of criminal behavior among the girls included in the
current sample (e.g., parental involvement, self-control). In particular, the importance of individual-level
factors such as depression and self-esteem in predicting different “types” or patterns of offending is an
important finding and is consistent with previous research that shows that adolescent girls’ behavior is
more susceptible to factors such as mental health, self-identity, and suicidal ideation compared to their
male counterparts [75–77]. Nevertheless, in a review of 12 gender-specific programs for female
juvenile offenders to find “what works”, Foley found that many programs focused on risk factors such
as substance abuse or interpersonal skills but did not focus on mental health problems [78]. As a result,
Foley argued that gender-specific programs “could do more to incorporate risk factors that have more
impact on the lives of females than males” ([78], p. 268). Taken together, it is clear that there is a need
to incorporate components of mental health assessment and intervention, confidence-building, and life
skills into prevention and intervention programs targeting at-risk adolescent females.
Given previous research that suggests that parental attachment and control are strong predictors of
female delinquency [28,79], it is surprising that these variables were not found to be significant
Laws 2014, 3 664
protective factors against criminal behavior. It is also surprising that alcohol or marijuana use did not
significantly predict group membership. A strong link between the use of alcohol and marijuana and
delinquency among boys and girls has been documented for quite some time [80–82]. Similarly,
the insignificant relationship between truancy and group membership was also unexpected. Truancy,
and other forms of status offending, are considered primary risk factors for other, more serious forms
of deviant behavior [83]. One possible reason for the insignificant findings regarding parental
attachment, parental control, substance use, truancy, and school attachment is the low levels of
offending found across all three trajectories. Since the variability in criminal behavior across the three
groups was low, the ability to empirically distinguish between these groups is reduced.
A number of additional limitations to the current study should also be mentioned. First, the Add
Health data were collected over 10 years ago. Since this time, there have been a number of shifts in
trends of female offending, substance use, and family functioning. Therefore, future studies should
seek to replicate this study using samples of adolescent females from more recent years. Also, due to
the low levels of self-reported offending reported at each of the four waves, we dichotomized the
offenses. As a result, our offense indices measured versatility, but did not account for variations in the
severity or frequency of criminal behavior at each time point. The inclusion of this information may
provide a more detailed picture of the characteristics of trajectories of criminal behavior among girls
from adolescence to early adulthood. There are also a number of additional risk and protective factors
that have been found to be predictive of girls’ delinquency that were not included in the current study.
These factors include developmental factors such as puberty and genetics [84], neighborhood
context [18,85], and romantic relationships. For example, recent studies have suggested that romantic
partners also play a key role in risk for adolescent girls’ behavior [86]. Therefore, future research
should examine the role of same-sex, mixed-sex, and romantic relationships on trajectories of criminal
behavior among females as well as the impact of developmental and community-level factors.
Moreover, our analyses only accounted for risk and protective factors measured at Wave 1 which
provides information about the factors that influenced initial placement into these groups. Certainly,
understanding which factors relate to group placement during early adolescence is critical to furthering
our understanding of female offending and preventing future behavior. However, it is also important to
understand what factors influence changes in offending at different developmental stages.
For example, identifying what factors led to the de-escalation of criminal behavior among the
“de-escalators” and what factors led to the escalation of criminal behavior among the “escalators” is
also critical information regarding continuity and change in behavior over time. Future research should
seek to explore the risk and protective factors that are related to the continuity, escalation, and
de-escalation of behavior across different developmental periods. Indeed, the impact of different risk
and protective factors, such as parenting, peers, and personality characteristics, has the potential to
change in direction and magnitude as one progresses through different developmental stages [87,88].
It is also important to bring attention to the issue of reification. Reification occurs when an abstract
concept is treated as concrete or regarded as a real entity. Based on the current analyses, one cannot
conclude that the trajectories found in the current analyses represent actual types of female offenders in
the population. Instead, latent subgroups are used to provide a useful heuristic for representing the
heterogeneity in observed items across a sample of the population. They are not assumed to be
empirical realities. Therefore, our findings provide a representation of the variability in offending
Laws 2014, 3 665
among females in the current sample, but do not suggest that these trajectories are concrete realities
that can be used to categorize female offenders. In addition, due to the low levels of criminal behavior
found across the full sample, small latent subgroups, such as a chronic offending group, may get
overlooked [89]. Thus, conclusions about the shape of offending trajectories among female adolescents
based on the current findings should be made with caution. Given the small number of studies using
female only samples, as well as the low variability found among the criminal behavior items, the
findings of this study are meant to be preliminary and to serve as a basis for future research that can
replicate and/or expand upon our findings.
Regardless of these limitations, the present study contributes to the growing body of research on
female offending by identifying the risk and protective factors associated with criminal trajectories
across adolescence to early adulthood. The advantages of the current study are the use of a large,
representative sample and an entirely female sample. A number of previous studies have focused
exclusively on clinical or adjudicated samples of girls. Although these studies provide a significant
contribution to understanding the needs of adolescent offenders, these studies are only able to provide
an accurate understanding of offending trajectories among high-risk offenders. Thus, they are less able
to inform the development of early prevention or intervention strategies that can target delinquency
prior to the onset of serious misbehavior. Often, girls that are already involved in clinical treatment or
the juvenile justice system are exhibiting serious forms of problem behavior and therefore do not
provide an accurate representation of criminal trajectories among the general female adolescent
population. For example, studies based on clinical or adjudicated females typically find that the
percent of the sample that falls into a “life-course persistent” group is relatively high and the percent of
girls that fall into a “low-offending” group is relatively low or nonexistent [90,91]. In contrast, the
findings of this study as well as other nationally representative studies of female offending found a
much higher proportion of girls fell into lower-risk trajectories.
In sum, our findings underscore the importance of gender-specific programming that is able to
address the unique developmental needs of female adolescents and highlight differences in the risk and
protective factors related to different offending trajectories. Thus, it is clear that applying a “one shoe
fits all” approach, both within and across gender subgroups, to the prevention and treatment of
criminal behavior will not be effective due to the unique developmental needs of adolescents.
The findings of our study also accentuate the need for additional research focusing specifically on the
development and continuity of criminal behavior among female adolescents and how this information
can be translated into gender-specific programming based on the unique needs of adolescent girls.
As Foley acknowledged, a great deal of research is needed to understand the development and
continuity of delinquency among adolescent females and to develop, implement, and evaluate
programs that will positively impact the lives of adolescent girls [78]. Future research can build on the
current study by relying on more recent samples of adolescent females and by examining how various
risk and protective factors, including developmental and community-level factors, relate to changes in
offending at different stages of development.
Laws 2014, 3 666
Acknowledgements
This research uses data from Add Health, a program project directed by Kathleen Mullan Harris and
designed by J. Richard Udry, Peter S. Bearman, and Kathleen Mullan Harris at the University of North
Carolina at Chapel Hill, and funded by grant P01-HD31921 from the Eunice Kennedy Shriver National
Institute of Child Health and Human Development, with cooperative funding from 23 other federal
agencies and foundations. Special acknowledgment is due Ronald R. Rindfuss and Barbara Entwisle
for assistance in the original design. Information on how to obtain the Add Health data files is
available on the Add Health website [46]. No direct support was received from grant P01-HD31921
for this analysis.
Author Contributions
Julie M. Krupa was responsible for the development of the research question and the literature
review. She also prepared the data for analysis and conducted preliminary analyses. Kristina K. Childs
performed further analysis and interpreted the results. Both authors contributed equally to the writing,
editing, and revising of the manuscript.
Abbreviations
Add Health: National Longitudinal Study of Adolescent Health;
LCA: Latent Class Analysis;
LMR: Lo-Mendall Rubin;
LRT: Lo-Mendall Rubin Likelihood Ratio Test;
SE: standard error;
SD: standard deviation;
RRR: relative risk ratio.
Conflicts of Interest
The authors declare no conflict of interest.
References
1. Howard N. Snyder. “Juvenile Arrests, 1992.” Office of Juvenile Justice Delinquency and
Prevention, 1994. Available online: https://www.ncjrs.gov/txtfiles/fs-9413.txt (accessed on 19
September 2013).
2. Charles Puzzanchera. “Juvenile Arrests, 2011.” Office of Justice Programs, Office of Juvenile
Justice Delinquency and Prevention, 2013. Available online: http://www.ojjdp.gov/pubs/
244476.pdf (accessed on 19 September 2013).
3. Sarah C. Walker, Ann Muno, and Cheryl Sullivan-Colglazier. “Principles in Practice:
A Multistate Study of Gender-Responsive Reforms in the Juvenile Justice System.” Crime &
Delinquency 58 (2012): 1–25.
Laws 2014, 3 667
4. Meda Chesney-Lind, and Randall G. Shelden. Girls, Delinquency, and Juvenile Justice. Belmont:
Wadsworth, 2004.
5. Pernilla Johansson, and Kimberly Kempf-Leonard. “A Gender-Specific Pathway to Serious,
Violent, and Chronic Offending?: Exploring Howell’s Risk Factors for Serious Delinquency.”
Crime & Delinquency 55 (2009): 216–40.
6. Mark A. Cohen, Alex R. Piquero, and Wesley G. Jennings. “Monetary Costs of Gender and
Ethnicity Disaggregated Group-Based Offending.” American Journal of Criminal Justice 35
(2010): 159–72.
7. Suyeon Park, Merry Morash, and Tia Stevens. “Gender Differences in Predictors of Assaultive
Behavior in Late Adolescence.” Youth Violence & Juvenile Justice 8 (2010): 314–31.
8. Jennifer M. Reingle, Wesley G. Jennings, and Mildred M. Maldonado-Molina. “Risk and
Protective Factors for Trajectories of Violent Delinquency among a Nationally Representative
Sample of Early Adolescents.” Youth Violence & Juvenile Justice 10 (2012): 260–76.
9. Benjamin Aguilar, L. Alan Sroufe, Byron Egeland, and Elizabeth Carlson. “Distinguishing the
Early-onset/Persistent and Adolescence-onset Antisocial Behavior Types: From Birth to 16
Years.” Development and Psychopathology 12 (2000): 109–32.
10. Alex R. Piquero. “Taking Stock of Developmental Trajectories of Criminal Activity over the Life
Course.” In The Long View of Crime: A Synthesis of Longitudinal Research. New York: Springer,
2008, pp. 23–78.
11. Terrie E. Moffitt. “Adolescence-limited and Life-course-persistent Antisocial Behavior:
A Developmental Taxonomy.” Psychological Review 100 (1993): 674–701.
12. Alex R. Piquero. “Assessing the Relationships between Gender, Chronicity, Seriousness, and
Offense Skewness in Criminal Offending.” Journal of Criminal Justice 28 (2000): 103–15.
13. Ick-Joong Chung, Karl G. Hill, J. David Hawkins, Lewayne D. Gilchrist, and Daniel S. Nagin.
“Childhood predictors of Offense Trajectories.” Journal of Research in Crime and Delinquency
39 (2002): 60–89.
14. David M. Fergusson, and John L. Horwood. “Male and Female Offending Trajectories.”
Development and Psychopathology 14 (2002): 159–77.
15. Wesley G. Jennings, and Jennifer M. Reingle. “On the Number and Shape of Developmental/
Life-Course Violence, Aggression, and Delinquency Trajectories: A state-of-the-art Review.”
Journal of Criminal Justice 40 (2012): 472–89. 16. David Huizinga, Shari Miller, and The Conduct Problems Prevention Research Group.
“Developmental Sequences of Girls’ Behavior.” U.S. Department of Justice, Office of Justice
Programs, Office of Juvenile Justice Delinquency and Prevention, December 2013. Available
online: http://www.ojjdp.gov/pubs/238276.pdf (accessed on 15 November 2013). 17. Shari Miller, Patrick S. Malone, and Kenneth A. Dodge. “Developmental Trajectories of Boys’
and Girls’ Delinquency: Sex Differences and Links to Later Adolescent Outcomes.” Journal of
Abnormal Child Psychology 38 (2010): 1021–32.
18. Jennifer M. Reingle, Wesley G. Jennings, and Mildred M. Maldonado-Molina. “The Mediated
Effect of Contextual Risk Factors on Trajectories of Violence: Results from a Nationally
Representative, Longitudinal Sample of Hispanic Adolescents.” American Journal of Criminal
Justice 36 (2011): 327–43.
Laws 2014, 3 668
19. Norman A. White, and Alex R. Piquero. “A Preliminary Empirical Test of Silverthorn and Frick’s
Delayed‐onset Pathway in Girls using an Urban, African‐American, US‐Based Sample.” Criminal Behaviour and Mental Health 14 (2004): 291–309.
20. Debra J. Pepler, Depeng Jiang, Wendy M. Craig, and Jennifer Connolly. “Developmental
Trajectories of Girls’ and Boys’ Delinquency and Associated Problems.” Journal of Abnormal
Child Psychology 38 (2010): 1033–44.
21. Alex R. Piquero, and He Len Chung. “On the Relationships between Gender, Early Onset, and the
Seriousness of Offending.” Journal of Criminal Justice 29 (2001): 189–206.
22. Debra A. Murphy, Mary-Lynn Brecht, David Huang, and Diane M. Herbeck. “Trajectories of
Delinquency from Age 14 to 23 in the National Longitudinal Survey of Youth sample.”
International Journal of Adolescence and Youth 17 (2012): 47–62.
23. Matthew C. Aalsma, and Daniel K. Lapsley. “A Typology of Adolescent Delinquency:
Sex Differences and Implications for Treatment.” Criminal Behaviour & Mental Health 11
(2001): 173–91. 24. Persephanie Silverthorn, Paul J. Frick, and Richard Reynolds. “Timing of Onset and Correlates of
Severe Conduct Problems in Adjudicated Girls and Boys.” Journal of Psychopathology and
Behavioral Assessment 23 (2001): 171–81.
25. Yao Zheng, and H. Harrington Cleveland. “Identifying Gender-Specific Developmental
Trajectories of Nonviolent and Violent Delinquency from Adolescence to Young Adulthood.”
Journal of Adolescence 36 (2013): 371–81.
26. Deborah Gorman-Smith, and Rolf Loeber. “Are Developmental Pathways in Disruptive Behaviors
the Same for Girls and Boys?” Journal of Child and Family Studies 14 (2005): 15–27. 27. Stephen D. Whitney, Lynette M. Renner, and Todd I. Herrenkohl. “Gender Differences in Risk
and Promotive Classifications Associated with Adolescent Delinquency.” The Journal of Genetic
Psychology 171 (2010): 116–38.
28. Machteld Hoeve, Juditch S. Dubas, Veroni I. Eichelsheim, Peter H. van der Laan, Wilma Smeenk,
and Jane R. M. Gerris. “The Relationship between Parenting and Delinquency: A Meta-Analysis.”
Journal of Abnormal Child Psychology 37 (2009): 749–75. 29. Dana Peterson, and Kirstin A. Morgan. “Sex Differences and the Overlap in Youths’ Risk Factors
for Onset of Violence and Gang Involvement.” Journal of Crime & Justice 37 (2014): 129–54.
30. Joanne Belknap, and Kristi Holsinger. “The Gendered Nature of Risk Factors for Delinquency.”
Feminist Criminology 1 (2006): 48–71.
31. Abigail A. Fagan, M. Lee Van Horn, Susan Antaramian, and J. David Hawkins. “How do
Families Matter? Age and Gender Differences in Family Influences on Delinquency and Drug
Use.” Youth Violence and Juvenile Justice 9 (2011): 150–70.
32. Dana J. Hubbard, and Travis C. Pratt. “A Meta-Analysis of the Predictors of Delinquency among
Girls.” Journal of Offender Rehabilitation 34 (2002): 1–13.
33. Abigail A. Fagan, M. Lee Van Horn, J. David Hawkins, and Michael W. Arthur. “Gender
Similarities and Differences in the Association between Risk and Protective Factors and
Self-Reported Serious Delinquency.” Prevention Science 8 (2007): 115–24.
34. Lisa M. Broidy, Daniel S. Nagin, Richard E. Tremblay, John E. Bates, Bobby Brame, Kenneth A.
Dodge, David Fergusson, John L. Horwood, Rolf Loeber, Robert Laird, and et al.
Laws 2014, 3 669
“Developmental Trajectories of Childhood Disruptive Behaviors and Adolescent Delinquency:
A Six-site, Cross-National Study.” Developmental Psychology 39 (2003): 222–45. 35. Teresa C. LaGrange, and Robert A. Silverman. “Low Self-Control and Opportunity: Testing the
General Theory of Crime as an Explanation for Gender Differences in Delinquency.” Criminology
37 (1999): 41–72. 36. Majone Steketee, Marianne Junger, and Josine Junger-Tas. “Sex Differences in the Predictors of
Juvenile Delinquency: Females are more Susceptible to Poor Environments; Males are Influenced
more by Low Self-Control.” Journal of Contemporary Criminal Justice 29 (2013): 88–105.
37. Elizabeth Cauffman, Francis J. Lexcen, Asha Goldweber, Elizabeth P. Shulman, and Thomas
Grisso. “Gender Differences in Mental Health Symptoms among Delinquent and Community
Youth.” Youth Violence and Juvenile Justice 5 (2007): 287–307.
38. Michael K. Ostrowsky. Self-Medication and Violent Behavior. El Paso: LFB Scholarly
Publishing, 2009.
39. Gregory D. Webster, Lee A. Kirkpatrick, John B. Nezlek, Veronica C. Smith, and Layne E.
Paddock. “Different Slopes for Different Folks: Self-esteem Instability and Gender as Moderators
of the Relationship between Self-Esteem and Attitudinal Aggression.” Self & Identity 6 (2007):
74–94.
40. Johannes A. Landsheer, and C. van Dijkum. “Male and Female Delinquency Trajectories from Pre
through Middle Adolescence and their Continuation in Late Adolescence.” Adolescence 40
(2005): 729–48. 41. Bo Wang, Lynette Deveaux, Xiaoming Li, Sharon Marshall, Xinguang Chen, and Bonita Stanton.
“The Impact of Youth, Family, Peer and Neighborhood Risk Factors on Developmental
Trajectories of Risk Involvement from Early through Middle Adolescence.” Social Science &
Medicine 106 (2014): 43–52.
42. Wesley G. Jennings, Mildred M. Maldonado-Molina, Alex R. Piquero, Candice L. Odgers, Hector
Bird, and Glorisa Canino. “Sex Differences in Trajectories of Offending among Puerto Rican
Youth.” Crime & Delinquency 56 (2010): 327–57.
43. David P. Farrington, and Kate A. Painter. Gender Differences in Offending: Implications for
Risk-Focused Prevention. London: Home Office RDA, 2004. Available online:
http://www.crim.cam.ac.uk/people/academic_research/david_farrington/olr0904.pdf (accessed on
15 January 2014).
44. Terrie E. Moffitt. “Juvenile Delinquency and Attention Deficit Disorder: Boys’ Developmental
Trajectories from Age 3 to Age 15.” Child Development 61 (1990): 893–910.
45. David S. Nagin, David P. Farrington, and Terrie E. Moffitt. “Life-course Trajectories of Different
Types of Offenders.” Criminology 33 (1995): 111–39.
46. Kathleen M. Harris. “The Add Health Study: Design and Accomplishments.” Available online:
http://www.cpc.unc.edu/projects/addhealth/data/guides/DesignPaperWIIV.pdf (accessed on 20
November 2013).
47. Kim Chantala, and Joyce Tabor. “Strategies to Perform a Design-Based Analysis using the Add
Health data.” Available online: http://www.cpc.unc.edu/projects/addhealth/data/guides/weight1.pdf
(accessed on 20 November 2013).
Laws 2014, 3 670
48. Naomi Brownstein, William D. Kalsbeck, Joyce Tabor, Pamela Entzel, Eric Daza, and Kathleen
M. Harris. “Non-Response in Wave IV of the National Longitudinal Study of Adolescent Health.”
Available online: http://www.cpc.unc.edu/projects/addhealth/data/guides/W4_nonresponse.pdf
(accessed on 20 November 2013).
49. Martha Gault-Sherman. “It’s a Two-way Street: The Bidirectional Relationship between Parenting
and Delinquency.” Journal of Youth and Adolescence 41 (2012): 121–45.
50. Monica K. Johnson, Robert Crosnoe, and Lyssa L. Thaden. “Gendered Patterns in Adolescents’
School Attachment.” Social Psychology Quarterly 69 (2006): 284–95.
51. Hilary D. Joyce, and Theresa J. Early. “The Impact of School Connectedness and Teacher Support
on Depressive Symptoms in Adolescents: A Multilevel Analysis.” Children and Youth Services
Review 39 (2014): 101–07.
52. Renee V. Galliher, Sharon S. Rostosky, and Hannah K. Hughes. “School Belonging,
Self-Esteem, and Depressive Symptoms in Adolescents: An Examination of Sex, Sexual
Attraction Status, and Urbanicity.” Journal of Youth and Adolescence 33 (2004): 235–45.
53. Gene H. Brody, Xiaojia Ge, Rand Conger, Fredrick X. Gibbons, Velma M. Murry, Meg Gerrard,
and Ronald L. Simons. “The Influence of Neighborhood Disadvantage, Collective Socialization,
and Parenting on African American Children’s Affiliation with Deviant Peers.” Child
Development 72 (2001): 1231–46. doi:10.1111/1467-8624.00344.
54. Scott R. Weaver, and Hazel M. Prelow. “A Mediated-moderation Model of Maternal Parenting
Style, Association with Deviant Peers, and Problem Behaviors in Urban African American and
European American Adolescents.” Journal of Child and Family Studies 14 (2005): 343–56.
doi:10.1007/s10826-005-6847-1.
55. Arielle R. Deutsch, Lisa J. Crockett, Jennifer M. Wolff, and Stephen T. Russell. “Parent and Peer
Pathways to Adolescent Delinquency: Variations by Ethnicity and Neighborhood Context.”
Journal of Youth and Adolescence 41 (2012): 1078–94.
56. Mary R. Gillmore, Sandra S. Butler, Mary J. Lohr, and Lewayne Gilchrist. “Substance Use and
Other Factors Associated with Risky Sexual Behavior among Pregnant Adolescents.” Family
Planning Perspectives 24 (1992): 255–68.
57. Sheryl A. Hemphill, Todd I. Herrenkohl, Andrea N. LaFazia, Barbara J. McMorris, John W.
Toumbourou, Michael W. Arthur, Richard F. Catalano, J. David Hawkins, and Lyndal Bond.
“Comparison of the Structure of Adolescent Problem Behavior in the United States and
Australia.” Crime & Delinquency 53 (2007): 303–21.
58. Dina Perrone, Christopher J. Sullivan, Travis C. Pratt, and Satenik Margaryan. “Parental Efficacy,
Self-Control, and Delinquency: A Test of a General Theory of Crime on a Nationally
Representative Sample of Youth.” International Journal of Offender Therapy and Comparative
Criminology 48 (2004): 298–312.
59. Joseph O. Baker. “The expression of low self-control as problematic drinking in adolescents:
An integrated control perspective.” Journal of Criminal Justice 38 (2011): 237–44.
60. Andrew M. Guest, and Nick Mcree. “A School-Level Analysis of Adolescent Extracurricular
Activity, Delinquency, and Depression: The Importance of Situational Context.” Journal of Youth
and Adolescence 38 (2009): 51–62.
Laws 2014, 3 671
61. R. Barry Ruback, Valerie A. Clark, and Cody Warner. “Why Are Crime Victims at Risk of Being
Victimized Again? Substance Use, Depression, and Offending as Mediators of the
Victimization-Revictimization Link.” Journal of Interpersonal Violence 29 (2013): 157–85.
62. Brianna Remster. “Self-Control and the Depression–Delinquency Link.” Deviant Behavior 35
(2014): 66–84.
63. Allan L. McCutcheon. “Basic Concepts and Procedures in Single and Multiple Group Latent
Class Analysis.” In Applied Latent Class Analysis. Edited by Jacques A. Hagenaars and Allan L.
McCutcheon. Cambridge: Cambridge University Press, 2002, pp. 56–88.
64. Karen L. Nylund, Tihomir Asparouhov, and Bengt O. Muthen. “Deciding on the Number of
Classes in Latent Class Analysis and Growth Mixture Modeling: A Monte Carlo Simulation
Study.” Structural Equation Modeling: A Multidisciplinary Journal 14 (2007): 535–69.
65. Yungtai T. Lo, Nancy R. Mendell, and Donald B. Rubin. “Testing the number of components in a
normal mixture.” Biometrika 88 (2001): 767–78. 66. Jeroen K. Vermunt, and Jay Magidson. “Latent Class Analysis.” In The Sage Encyclopedia of
Social Science Research Methods. Edited by Michael S. Lewis-Beck, Alan E. Bryman and Tim F.
Liao. Newbury Park: Sage, 2003.
67. Bengt O. Muthén, and Linda K. Muthén. Mplus User’s Guide, 3rd ed. Los Angles: Muthén &
Muthén, 2010.
68. Victor van der Geest, Arjan Blokland, and Catrien Bijleveld. “Delinquent Development in a
Sample of High-Risk Youth: Shape, Content, and Predictors of Delinquent Trajectories from Age
12 to 32.” Journal of Research in Crime and Delinquency 46 (2009): 111–43.
69. Steven A. McFadyen‐Ketchum, John E. Bates, Kenneth A. Dodge, and Gregory S. Pettit. “Patterns of Change in Early Childhood Aggressive‐Disruptive Behavior: Gender Differences in Predictions from Early Coercive and Affectionate Mother‐Child Interactions.” Child Development 67 (1996): 2417–33.
70. Amy V. D’Unger, Kenneth C. Land, and Patricia L. McCall. “Sex Differences in Age Patterns of
Delinquent/Criminal Careers: Results from Poisson Latent Class Analyses of the Philadelphia
Cohort Study.” Journal of Quantitative Criminology 18 (2002): 349–75.
71. Benjamin B. Lahey, Carol A. Van Hulle, Irwin D. Waldman, Joseph Lee Rodgers, Brian M.
D’Onofrio, Steven Pedlow, Paul Rathouz, and Kate Keenan. “Testing Descriptive Hypotheses
Regarding Sex Differences in the Development of Conduct Problems and Delinquency.” Journal
of Abnormal Child Psychology 34 (2006): 730–48.
72. Patrick H. Tolan, and Peter Thomas. “The Implications of Age of Onset for Delinquency Risk II:
Longitudinal Data.” Journal of Abnormal Child Psychology 23 (1995): 157–81.
73. Lynn Kratzer, and Sheilagh Hodgins. “A Typology of Offenders: A Test of Moffitt’s Theory
among Males and Females from Childhood to Age 30.” Criminal Behaviour and Mental Health 9
(1999): 57–73.
74. Persephanie Silverthorn, and Paul J. Frick. “Developmental Pathways to Antisocial Behavior:
The Delayed-onset Pathway in Girls.” Development and Psychopathology 11 (1999): 101–26.
75. Sofia Diamantopoulou, Frank C. Verhulst, and Jan van der Ende. “Gender Differences in the
Development and Adult Outcome of Co-occurring Depression and Delinquency in Adolescence.”
Journal of Abnormal Psychology 120 (2011): 644–55.
Laws 2014, 3 672
76. Kristen C. Kling, and Janet Shibley Hyde. “Gender differences in self-esteem: A meta-analysis.”
Psychological Bulletin 125 (1999): 470–500. 77. Susan Nolen-Hoeksema, and Joan S. Girgus. “Emergence of Gender Differences in Depression
during Adolescence.” Psychological Bulletin 115 (1994): 424–43.
78. Allison Foley. “The Current State of Gender-Specific Delinquency Programming.” Journal of
Criminal Justice 36 (2008): 262–69. 79. Dana L. Haynie. “Contexts of Risk? Explaining the Link between Girls’ Pubertal Development
and their Delinquency Involvement.” Social Forces 82 (2003): 355–97.
80. Delbert S. Elliott, David Huizinga, and Scott W. Menard. Multiple Problem Youth: Delinquency,
Substance Abuse, and Mental Health Problems. New York: Springer, 1989.
81. David Huizinga, and Cynthia Jakob-Chien. “The Contemporaneous Co-occurrence of Serious and
Violent Juvenile Offender and Other Problem Behaviors.” In Serious and Violent Juvenile
Offenders: Risk Factors and Successful Interventions. Edited by Roll Loeber and David P.
Farrington. Thousand Oaks: Sage, 1998, pp. 47–67. 82. Scott Menard, Sharon Mihalic, and David Huizinga. “Drugs and Crime Revisited.” Justice
Quarterly 18 (2001): 269–99. 83. Rolf Loeber, and David P. Farrington. “Young Children Who Commit Crime: Epidemiology,
Developmental Origins, Risk Factors, Early Interventions, and Policy Implications.” Development
and Psychopathology 12 (2000): 737–62. 84. Shabnam Javdani, Naomi Sadeh, and Edelyn Verona. “Expanding Our Lens: Female Pathways to
Antisocial Behavior in Adolescence and Adulthood.” Clinical Psychology Review 31 (2011):
1324–48.
85. Leoniek Kroneman, Rolf Loeber, and Alison E. Hipwell. “Is Neighborhood Context Differently
Related to Externalizing Problems and Delinquency for Girls Compared with Boys?” Clinical
Child and Family Psychology Review 7 (2004): 109–22.
86. Shabnam Javdani, Erin M. Rodriguez, Sara R. Nichols, Erin Emerson, and Geri R. Donenberg.
“Risking it for Love: Romantic Relationships and Early Pubertal Development Confer Risk for
Later Disruptive Behavior Disorders in African-American Girls Receiving Psychiatric Care.”
Journal of Abnormal Child Psychology, 2014, in press. doi:10.1007/s10802-014-9875-8.
87. Magda Stouthamer-Loeber, Rolf Loeber, David P. Farrington, Quanwu Zhang, Welmoet van
Kammen, and Eugene Maguin. “The Double Edge of Protective and Risk Factors for
Delinquency: Interrelations and Developmental Patterns.” Development and Psychopathology 5
(1993): 683–701.
88. Edward P. Mulvey, Laurence Steinberg, Jeffrey Fagan, Elizabeth Cauffman, Alex R. Piquero,
Laurie Chassin, George P. Knight, Robert Brame, Carol A. Schubert, Thomas Hecker, and et al.
“Theory and Research on Desistance from Antisocial Activity among Serious Adolescent
Offenders.” Youth Violence and Juvenile Justice 2 (2004): 213–36.
89. Stephanie T. Lanza, and Brittany L. Rhoades. “Latent Class Analysis: An Alternative Perspective
on Subgroup Analysis in Prevention and Treatment.” Prevention Science 14 (2013): 157–68.
Laws 2014, 3 673
90. Charles D. Ayers, James H. Williams, J. David Hawkins, Peggy L. Peterson, Richard F. Catalano,
and Robert D. Abbott. “Assessing Correlates of Onset, Escalation, De-escalation, and Desistance
of Delinquent Behavior.” Journal of Quantitative Criminology 15 (1999): 277–306.
91. Nadine Lanctŏt, and Marc LeBlanc. “Explaining Deviance by Adolescent Females.” Crime and
Justice: A Review of Research 29 (2002): 113–202.
© 2014 by the authors; licensee MDPI, Basel, Switzerland. This article is an open access article
distributed under the terms and conditions of the Creative Commons Attribution license
(http://creativecommons.org/licenses/by/4.0/).
