Using Latent Class Analysis to Explore Subtypes of Youth Who Have Committed Sexual Offenses

Date01 October 2019
DOI10.1177/1541204018820578
Published date01 October 2019
Subject MatterArticles
Article
Using Latent Class Analysis to
Explore Subtypes of Youth
Who Have Committed
Sexual Offenses
Adam Brown
1
Abstract
Attempts to identify typologies of youth who have committed sexual offenses have been challenged
by their overlapping characteristics with youth who have committed nonsexual crimes, as well as
methodological limitations that make the results difficult to translate into direct practice. In the
current study, a technical new way of identifying subtypes of these young people was proposed using
latent class analysis, a person-centered approach that allows categorical subtypes to be revealed by
the data rather than hypothesized differences based on individual factors. The indicators included in
this analysis were sexual behaviors only, thereby eliminating any overlap with general delinquents. In
a sample of 573 male youth between the ages of 11 and 20 (M¼16.75, SD ¼1.72), four unique
classes were identified. Research implications are offered.
Keywords
juvenile sexual offender, sexual offending, typologies, latent class analysis
In 2014, law enforcement agencies in the United States reported that individuals under the age of 18
accounted for approximately 16%of the 16,473 arrests made for rape and 17%of the 43,422 arrests
for all other sexual crimes except prostitution (U.S. Department of Justice, 2014). Gi ven these
statistics, there is concern among researchers, treatment providers, and policy makers with regard
to male youth who commit acts of sexual abuse (Malin, Saleh, & Grudzinskas, 2014). A growing
body of research has revealed that youth who have committed sexual offenses (YSOs) are a hetero-
geneous population across a variety of domains (e.g., Christiansen & Vincent, 2013). However,
complicating this research is that many indicators found in YSO etiology overlap with those found
among general delinquents (Couwenbergh et al., 2006; DeLisi, Neppl, Lohman, Vaughn, & Shook,
2013). This presents a problem for both treatment providers and those who target sexual abuse
prevention, as factors associated with sexual offending by youth are easily obscured by factors also
associated with nonsexual delinquency. Further blurring this line is that YSOs are likely to have
committed acts of general delinquency in addition to their sexual offenses (Brown & Burton, 2010;
1
Silberman School of Social Work at Hunter College, CUNY, New York, NY, USA
Corresponding Author:
Adam Brown, Silberman School of Social Work at Hunter College, CUNY, 2180 3rd Avenue, New York, NY 10035, USA.
Email: adam.brown@hunter.cuny.edu
Youth Violence and JuvenileJustice
2019, Vol. 17(4) 413-430
ªThe Author(s) 2018
Article reuse guidelines:
sagepub.com/journals-permissions
DOI: 10.1177/1541204018820578
journals.sagepub.com/home/yvj
Ronis & Borduin, 2007) and more likely to later reoffend with an act of nonsexual delinquency than
with a sexual offense (Caldwell, 2016; Christiansen & Vincent, 2013; Hendriks, van den Berg, &
Bijleveld, 2015). With such an overlap in behaviors and etiologies, parsing risk factors for sexual
offending from those for general delinquency is tricky. This presents a challenge to policy makers,
researchers, and treatment providers aiming to understand sexual abuse committed by youth. In the
following, a technical new way of identifying subtypes of YSOs with greater parsimony is proposed
using indicators that do not overlap with general delinquency.
Pathways to Sexual Offending and YSO Typologies
Attempts have been made to create typologies of YSOs that are rooted in theoretical assumptions of
criminality and largely driven by menta l health and personality (Fox & DeLisi, 20 17; Hunter,
Figueredo, Malamuth, & Becker, 2003; O’Brian & Bera, 1986; Worling, 2001). The authors of one
study found distinct typologies between YSOs without concurrent criminality (sex only) and YSOs
with concurrent criminality (sex plus; Butler & Seto, 2002). In it, the authors found that sex-only
YSOs had significantly fewer conduct problems and risks of future delinquency than their sex-plus
YSO counterparts who resembled criminally versatile youths who scored higher on risk indicators
for general delinquency. Does this suggest that sex-plus YSOs are general delinquents who happen
to commit a sexual crime among a constellation of delinquent acts, whereas sex-only YSOs repre-
sent a subtype that commits sexual abuse for reasons unrelated to general delinquency?
To date, theoretically informed typologies of YSOs such as these have yet to be empirically
validated. Aebi, Vogt, Plattner, Steinhausen, and Bessler (2012) suggested a reason for this might
be that subgroups of YSOs tend to have similar treatment needs. It is also possible that translating
extant research into practice is often challenging. As the number of individual risk factors
increases, the number of risk profiles increases exponentially, rendering factor-based profiles of
little or no use in terms of practical application (Lanza, Rhoades, Nix, & Greenberg, 2010),
particularly when the most common modality for YSO treatment is group based (see Schmucker,
Lo¨sel, & Schmucker, 2017, for review).
Furthermore, a methodological disadvantage to these approaches is that all risk factors are
weighted equally and assumed to be interchangeable (Lanza et al., 2010), meaning that the exposure
to any risk factor is equal regardless of how well related they are. In other words, such methods
require a priori assumptions of implicit meaning attached to the outcome (e.g., child victims vs. no
child victims, sex only vs. sex plus, use of force vs. no use of force) and the authors’ interpretations
of YSO subtypes are explained by how the variables hold together in relationship to one another.
By contrast, the use of a person-centered approac h to analysis, such as latent class analysis
(LCA), assumes that each factor derives meaning in relationship with all other factors (Laursen
& Hoff, 2006; Magnusson, 1998) while offering an opportunity to create typologies that confer a
higher degree of interpretability for clinicians and policy makers. LCA is similar to cluster analysis
conceptually but based upon a measurement model of factor analysis. Despite innumerable possi-
bilities of configurations, circumstances tend to organize into a relatively small number of patterns,
thereby placing emphasis on regularities and configurations of interactive indicators which distin-
guish qualitatively different groups of individuals (Bergman & El Khouri, 2003; Keller, Cusick, &
Courtney, 2007). Latent class analyses model heterogeneity by using latent categorical variables to
represent subpopulations which differ according to their patterns on a variety of indicators
(McCutcheon, 1987), resulting in an empirical classification of individuals who share a common
profile (Keller et al., 2007).
An additional feature of extant YSO studies is that certain sexual offense variables (e.g., victim
age) have been used as isolated indicators to test subtypes (Aebi, Vogt, Plattner, Steinhausen, &
Bessler, 2012; Fanniff & Kolko, 2012). A probabilistic perspective, rather than deterministic,
414 Youth Violence and Juvenile Justice 17(4)

To continue reading

Request your trial

VLEX uses login cookies to provide you with a better browsing experience. If you click on 'Accept' or continue browsing this site we consider that you accept our cookie policy. ACCEPT