Anda di halaman 1dari 19

Criminal Justice Policy Review

http://cjp.sagepub.com

The Influence of Sex Offender Registration on Juvenile Sexual


Recidivism
Elizabeth J. Letourneau, Dipankar Bandyopadhyay, Debajyoti Sinha and Kevin S.
Armstrong
Criminal Justice Policy Review 2009; 20; 136 originally published online Dec 5, 2008;
DOI: 10.1177/0887403408327917
The online version of this article can be found at:
http://cjp.sagepub.com/cgi/content/abstract/20/2/136

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

On behalf of:
Department of Criminology at Indiana University of Pennsylvania

Additional services and information for Criminal Justice Policy Review can be found at:
Email Alerts: http://cjp.sagepub.com/cgi/alerts
Subscriptions: http://cjp.sagepub.com/subscriptions
Reprints: http://www.sagepub.com/journalsReprints.nav
Permissions: http://www.sagepub.com/journalsPermissions.nav
Citations http://cjp.sagepub.com/cgi/content/refs/20/2/136

Downloaded from http://cjp.sagepub.com at FLORIDA STATE UNIV LIBRARY on June 12, 2009

The Influence of Sex Offender


Registration on Juvenile Sexual
Recidivism

Criminal Justice
Policy Review
Volume 20 Number 2
June 2009 136-153
2009 SAGE Publications
10.1177/0887403408327917
http://cjp.sagepub.com
hosted at
http://online.sagepub.com

Elizabeth J. Letourneau
Dipankar Bandyopadhyay
Medical University of South Carolina

Debajyoti Sinha
Florida State University

Kevin S. Armstrong
Medical University of South Carolina

This study examines the influence of South Carolinas sex offender registration policy
on juvenile offender recidivism risk. Juvenile male sexual offender (N = 1,275) risk of
recidivism was examined with an average 9-year follow-up. Survival analysis was used
to examine the influence of covariates, including the primary time-dependent covariate
registration status at time, on risk of new sexual offense charges and adjudications.
A competing risks model was used to explain the effects of covariates on different types
of recidivism events (sexual, assault, and other). Registration status at time had a
significant effect on risk of new other offense charges and a marginal (p < .10) effect
on risk of sexual offense charges (survival analysis) but no statistical evidence of
affecting risk of new adjudications. Results suggest a surveillance effect leading to
unnecessary charges for registered (vs. nonregistered) youth. State and national policy
implications are discussed.
Keywords:

juvenile sexual offender; registration; recidivism

ince 1994, state and federal policies have required long-term and public registration of some adult sexual offenders. Until recently, states could chose whether
to subject juveniles to public registration. In 2006, the Sex Offender Registration and
Notification Act (SORNA), Title I of the Adam Walsh Child Protection and Safety

Authors Note: This research was supported by grants to the first author from the Centers for Disease
Control and Prevention (R49 CE000567) and the National Science Foundation (SES 0455124). The
authors wish to thank Ms. Trudie Trotti, director of research and statistics, and Mr. Errol Campbell, senior
research associate, South Carolina Department of Juvenile Justice (DJJ). We also thank Mr. Charles
Bradberry, senior statistician, and Ms. Diana Tester, statistician, South Carolina State Budget and Control
Board, Office of Research and Statistics. Ms. Janice Rivers originally developed the DJJ database and
provided helpful consultation. Dr. Jill Levenson provided an initial review of this article. The conclusions
expressed in article are those of the authors and do not necessarily represent those of funding agencies or
South Carolina state agencies.

136
Downloaded from http://cjp.sagepub.com at FLORIDA STATE UNIV LIBRARY on June 12, 2009

Letourneau et al. / Registration and Juvenile Recidivism

137

Act of 2006 (P. L. 109-248), was passed and required all states to establish long-term
(25 years or life) public registration of juveniles (14 years and older) adjudicated as
minors for certain sexual offenses. If enacted nationwide, this policy would result in
the public registration of tens of thousands of juveniles (Chaffin, 2008). Many states
already require the public registration of juveniles adjudicated as minors. For
example, South Carolinas registration policy has included lifetime public registration of some juveniles since 1995 (SC Code 23-3-420, chapter 73, Article 6). Thus,
South Carolinas public registration policy corresponds with SORNA requirements
(e.g., long-term public registration of juveniles). As such, examining the effects
of South Carolinas policy might provide insight as to the eventual effects of the
more recently enacted SORNA. As discussed later, the purpose of the present study
was to examine the effects of the South Carolina registration policy on juvenile
recidivism.
Public registration policies were predicated, in part, on the belief that sexual
offenders are at high risk of sexual recidivism and require substantial surveillance
to reduce that risk (U.S. Department of Justice, 2007; Zimring, 2004). Thus, public
registration policies aim to reduce recidivism risk both by deterring new recidivism
events and by reducing the time needed to detect ongoing recidivism (U.S.
Department of Justice, 2007; Zimring, 2004). These twin aims (i.e., deterrence and
improved detection of recidivism) increase the complexity of determining the effectiveness of these policies. For example, study results indicating that registered
youth are less likely than nonregistered youth to reoffend could be interpreted as
supporting a deterrent effect, whereas the opposite results (i.e., greater recidivism
of registered vs. nonregistered youth) could be interpreted as supporting an effective detection (or surveillance) effect. Alternatively, it has been argued that registration policies might be criminogenic in themselves, creating barriers (e.g., to
academic and vocational achievement) that impede successful emancipation from
adolescence into young adulthood and possibly contributing to recidivism
(Letourneau & Miner, 2005). To date, most studies have failed to find support for a
deterrent effect (Adkins, Huff, & Stageberg, 2000; Schram & Milloy, 1995; Zevitz,
2006). Specifically, each of these studies compared groups of (mostly adult male)
sexual offenders subjected to varying degrees of public registration (from no registration requirements to active public notification procedures), and none found significant between-group differences in sexual recidivism rates. One uncontrolled
study has reported lower recidivism of offenders convicted after versus before
enactment of public registration policies, suggesting a deterrent effect (Washington
State Institute for Public Policy, 2005). However, violent crime declined generally
across that same time period, making interpretation of the results difficult. To date,
no published research has examined the effects of registration policies on juvenile
offenders who, despite recidivism rates that typically fail to exceed 10% (cf.,
Fortune & Lambie, 2006), are subject to public registration requirements in many
states and are now a focus of SORNA.

Downloaded from http://cjp.sagepub.com at FLORIDA STATE UNIV LIBRARY on June 12, 2009

138

Criminal Justice Policy Review

The purpose of this study is to examine the effects of South Carolinas registration
policy on sexual offense recidivism by juveniles first adjudicated of a serious sexual
offense as minors. Given the general absence of juvenile-focused research in this area,
specific hypotheses are omitted. Rather, analyses were designed to provide a preliminary examination of the effects of registration on the sexual and nonsexual recidivism
rates of juvenile sexual offenders. The potential influence of other factors on recidivism
(e.g., offender age at offense) was addressed by the inclusion of several covariates.

Method
Sample
The entire population of South Carolina minors adjudicated (i.e., found or pled
guilty) for serious sexual offenses between January 1, 1990, and December 31, 2004
(N = 1,306), was accessed for the present study. Analyses were limited to male
offenders (n = 1,275; 97.6%) between the ages of 6 and 17 (inclusive) at initial sexual offense adjudication. These male youth were White (45.1%), Black (53.8%), or
Other (1.1%). Mean age at initial sexual offense was 14.5 (SD = 1.5 years, range of
7.9 to 17.8). Mean number of nonsexual prior offense adjudications was 0.45 (SD =
1.29, range of 0 to 14). Mean length of follow-up was 9 years (SD = 4.3 years, range
of 0.21 to 15.96 years). Follow-up was 1 or more years for all but four youth.
Additional information on youth and family characteristics could not be included
because these variables were updated upon subsequent arrest.
Many youth (n = 574; 45%) were required to register at some point during followup (either for their index sexual offense or for a subsequent sexual offense).
Demographic and criminal characteristics (i.e., age at index sexual offense adjudication,
race, and number of prior adjudications) were compared for youth required to register
and youth not required to register. No differences reached statistical significance.

Procedures
Data
Prior to obtaining any data, data sets were de-identified and unique identifiers
assigned by a consultant in the South Carolina State Budget and Control Board
Office of Research and Statistics (ORS). All procedures were approved by the
Medical University of South Carolina Institutional Review Board, which waived participant consent procedures. Archival data for this study were extracted from three
South Carolina sources, including (a) sex offender registry records,
(b) Department of Juvenile Justice (DJJ) records, and (c) computerized criminal
history records (CCHR).

Downloaded from http://cjp.sagepub.com at FLORIDA STATE UNIV LIBRARY on June 12, 2009

Letourneau et al. / Registration and Juvenile Recidivism

139

Registry data. South Carolina sex offender registry data were obtained from the
South Carolina Law Enforcement Division (SLED) in collaboration with the South
Carolina Office of Justice Programs Statistical Analysis Center. Registry files
included offenders unique identifier, literal description of the registry offenses, and
initial date of registration for all registered offenders from registry inception
(January 1, 1995) through December 31, 2005.
Juvenile justice data. Juvenile justice data were obtained from the South Carolina
DJJ, in collaboration with ORS. The DJJ data included all charges forwarded to
solicitors from January 1, 1990, through December 31, 2005. Variables of interest
included demographic variables and (a) unique identifiers, (b) literal descriptions of
offenses forward to solicitors, (c) dates of charges, (d) solicitors decisions (and
dates), (e) literal description of disposition offenses, and (f) judicial dispositions and
dates (including start and end dates of incarceration where relevant). This DJJ set did
not capture charges that were never forwarded to solicitors.
Criminal court data. Adult criminal offenses are captured in the CCHR database
and were obtained from SLED in collaboration with ORS. The CCHR database
included all charges and adjudications recorded between January 1, 1990, and
December 31, 2005. Variables of interest included (a) unique identifiers, (b) literal
descriptions of charge (adjudication) offenses, (c) date of charges (adjudication), (d)
literal description of disposition offenses, (e) disposition decisions (e.g., determination of guilt), and (f) date of disposition.
Data preparation steps. South Carolinas registration policy focuses on more
serious forms of sexual offenses for juvenile offenders. Thus, all youth with at least
one guilty adjudication for a felony sexual offense (as defined in chapter 7, Title
24 SC Criminal Justice Commission 1994) were included in this study. The most
common index sexual offenses included sex with a minor (36%), criminal sexual
conduct (23%), and lewd act with a minor (17%). Several additional steps were
taken to prepare data for analysis. (a) When multiple sexual offense charges were
filed on the same date, the most serious charge was selected; (b) when multiple
charges were cleared on the same date, only the earliest charge was retained in
analyses; and (c) records were removed when they indicated the youth was less than
6 years of age or more than 18 years of age at the time of index sexual offense.
Operational definition of index sexual offense. Index sexual offense was defined
as the first felony sexual offense guilty adjudication that occurred for a youth
between 1990 and 2004 (inclusive). Only youth originally prosecuted as minors
were included. Prior to 1990, few juveniles were adjudicated for sexual offenses
(McManus, 2008). Index offenses occurring after 2004 were not included because of
limited follow-up duration.

Downloaded from http://cjp.sagepub.com at FLORIDA STATE UNIV LIBRARY on June 12, 2009

140

Criminal Justice Policy Review

Operational definitions of covariates. The covariate of primary interest was the


time-dependent registration status. Specifically, a youth could either (a) enter the
study as unregistered and remain unregistered for the duration of follow-up or (b)
enter the study as registered and remain registered for the duration of follow-up or
(c) enter the study as unregistered and subsequently be required to register at some
recorded time point within the follow-up period. Option C could occur either
because of a new sexual offense or because registration was applied retroactively to
a youth with a pre-enactment (i.e., pre-1995) sexual offense after that youth was
arrested for a new offense of any kind. Because registration duration is lifetime,
there was no option for a registered youth to revert to nonregistered status.
In addition to registration status at time, six covariates were included in all models. Year of index sexual offense and calendar year of time at risk were included to
account for nonspecific factors (e.g., changes in economic indicators or other legal
changes) that might influence juvenile offending in general (e.g., see Finkelhor &
Jones, 2004). Second, several factors believed to influence juvenile recidivism and
juvenile prosecution (e.g., see Vandiver, 2006; Worling & Lngstrm, 2006) also
were included in the analyses: youth age at index offense, prior criminal history (i.e.,
number of prior guilty adjudications for any offense type), race (dichotomized as
White or minority), and youth age at risk (i.e., time-dependent age at recidivism
event or censoring).
Operational definition of outcome (recidivism) variables. Event of interest,
recidivism, was operationally defined in two ways. First, all postindex offense
charges (hereafter, charges), regardless of adjudication decisions, were included as
recidivism events. In separate analyses, all postindex offense guilty pleas or findings
(hereafter, adjudications) were included as recidivism events. Sexual recidivism
events were specifically interpreted as either any new charge or adjudication for any
type of sexual offenses (i.e., not limited to felony sexual offenses). Assault recidivism events included postindex charges or adjudications for any type of assault
offenses (e.g., assault, domestic violence). Other recidivism events included postindex charges or adjudications for nonperson offenses in three categories: property
offenses (e.g., theft), drug offenses (e.g., possession of a controlled substance), and
public order violations (e.g., driving under suspension). Status offenses (e.g., curfew
violations) were excluded from analyses.
Follow-up. A youth was considered at risk of recidivism either since the date of
index sexual offense adjudication or, if incarcerated for the index offense, since date
of release from incarceration.

Data Analytic Strategy


Survival analysis techniques (J. P. Klein & Moeschberger, 2003) were used to
investigate the effects of covariates on the hazard (instantaneous risk) of committing

Downloaded from http://cjp.sagepub.com at FLORIDA STATE UNIV LIBRARY on June 12, 2009

Letourneau et al. / Registration and Juvenile Recidivism

141

a new sexual offense at any time point. In survival analysis, the response variable is
measured from a time of origin (i.e., date of index sexual offense adjudication or
release from incarceration following index offense) either to the date of recidivism
(i.e., new sexual offense charge or adjudication) or to the end of data collection (i.e.,
right censoring). To model the effects of covariates on the hazard/risk of committing
a sexual offense at any time point, we used the Cox Relative Risk Model (Cox,
1972), which assumes that effect of each covariate on risk remains constant over
time while allowing the baseline risk function of the event to take any shape over
time. This model accommodates time-dependent covariates and allows a youth to
not be at risk of recidivism while incarcerated for a crime unrelated to the recidivism
event of interest (described later). Associated partial likelihood (Cox, 1972) analysis
of Coxs model permits inferences about the effects of covariates (interpreted as relative risk ratio) at any time point via comparing the relative risk of an individual having the event at a time point with the relative risks for the rest of individuals at risk
of the event at that time. In these analyses, each offender contributes two primary
pieces of information: length of follow-up and status (either recidivism event
observed or censored). The censoring mechanism was assumed to be noninformative
(J. P. Klein & Moeschberger, 2003) for these survival models (we address informative censoring next).
The majority of recidivism events for juvenile sexual offenders are nonsexual
offenses (Caldwell, 2002). Thus, these nonsexual recidivism events are of interest in
their own right, and at any time point, the risk of such nonsexual recidivism events
is likely associated with risk of sexual recidivism at that time (e.g., if a new nonsexual offense charge or adjudication results in secure detention, this removes the youth
from being at risk of committing a new community-based sexual offense). In the
case of such informative censoring, all possible types of recidivism events are modeled as competing causes of recidivism event, and the estimated survival for each
cause-specific event is of primary interest (Satagopan et al., 2004; Scrucca, Santucci,
& Aversa, 2007). In competing risks models, the type of endpoint is denoted by the
value 0, 1, 2, or 3, respectively, when the youth had experienced censoring (i.e., no
new offense, cause = 0) or a new sexual offense (cause = 1), a new assault offense
(cause = 2), or a new other offense (cause = 3). The cause-specific hazard (CSH)
model then provides the instantaneous rate/risk of failure from cause j at time t,
given the person is at risk of recidivism because of all causes at time t. The Relative
Risks Model (Cox, 1972) was used to determine the effects of covariates on each of
these three CSH functions at time t (Prentice & Breslow, 1978).
Survival and competing risk analyses were complicated because of the presence
of incarceration periods when the youth was not at risk of recidivism event because,
while incarcerated, youth were not at risk of committing offenses in the community.
Ignoring incarceration periods that occurred during follow-up could result in biased
parameter estimates and standard errors, because the partial likelihood-based inference (Cox, 1972) compares relative risks of different youths at risk of recidivism at

Downloaded from http://cjp.sagepub.com at FLORIDA STATE UNIV LIBRARY on June 12, 2009

142

Criminal Justice Policy Review

any time point (i.e., comparison of relative risks of youths belonging to the risk set
of the time point). The counting process approach (Therneau & Gambsch, 2000)
was used to ensure that the youths incarcerated at any particular time point were
removed from the risk set of that time point (Scrucca et al., 2007). Data on incarceration events were available only from DJJ (i.e., for offenders incarcerated as juveniles) and not from CCHR (i.e., for offenders incarcerated as adults). Last, to verify
the validity of the relative risk assumption of Coxs model and to assess the necessity of further transformation of covariates, plots of martingale and deviance residuals were examined. These residuals are suitable for survival data subjected to
right-censoring and nonmonotonic risk sets over time. The plots indicated adequacy
of modeling assumptions and no necessity of transformation covariates (plots omitted to conserve space). SAS (Version 9.1) software and R (Version 2.6.1) were used
to perform all analyses.

Results
In Table 1, we provide the number of index offenses and recidivism events occurring in each year of follow-up. Consistent with other recidivism studies, sexual
recidivism events were rare, with 95 (7.5%) offenders charged and 32 (2.5%) adjudicated for new sexual offenses during follow-up.

Sexual Recidivism Survival Analyses


Sexual offense charges. Model fit statistics indicated statistically significant
improvement in model fit when covariates were included (see Table 2). Table 2
shows that four of the seven covariates were statistically significantly associated
with instantaneous risk (hazard) of new sexual offense charges. Specifically, index
offense year was positively associated with higher risk, indicating that a youth with
a more recent index sexual offense was at higher relative risk than a comparable
youth with a more distal index offense when both are at an equal time interval away
from their respective index offense dates. Likewise, age at index offense was associated with recidivism risk, indicating that increased age at index offense put a youth
at higher risk of sexual offense charges at any time in the future. Two time-dependent
covariates, calendar year of recidivism event and age at recidivism event, were negatively associated with recidivism. This negative association suggests that the longer
a youth remains in the community offense free, the lower the likelihood of a future
offense. There was moderate statistical evidence for registration status (p = .08) to
support that, following registration, a youth was at higher risk of new sexual offense
charges than risk of a comparable youth who was not registered at that time point
from his index offense. Figure 1 shows the estimated survival probability (KaplanMeier estimate) over time for sexual offense charges adjusting for covariate effects.

Downloaded from http://cjp.sagepub.com at FLORIDA STATE UNIV LIBRARY on June 12, 2009

Letourneau et al. / Registration and Juvenile Recidivism

143

Table 1
Annualized Rate of Index and Recidivist Offenses
Year

Indexa

1990
1991
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005

.05 (67)
.08 (102)
.09 (118)
.10 (132)
.09 (114)
.06 (80)
.06 (77)
.06 (75)
.05 (57)
.05 (61)
.06 (78)
.07 (88)
.06 (82)
.06 (75)
.05 (69)
na

New Sexual
Offense Chargesb

New Sexual
Offense Adjudicationsc

.0 (0)
.0 (0)
.02 (2)
.04 (4)
.02 (2)
.07 (7)
.06 (6)
.11 (10)
.08 (8)
.08 (8)
.07 (7)
.08 (8)
.13 (12)
.07 (7)
.06 (6)
.08 (8)

.0 (0)
.0 (0)
.0 (0)
.03 (1)
.03 (1)
.09 (3)
.06 (2)
.16 (5)
.06 (2)
.13 (4)
.06 (2)
.16 (5)
.13 (4)
.03 (1)
.03 (1)
.03 (1)

Note: Values in parentheses represent the annual count of index and recidivism events.
a. n = 1,275 offenders.
b. n = 95 offenders.
c. n = 32 offenders.

Sexual offense adjudications. Model fit statistics (see Table 2) indicated statistically significant improvement in model fit after including the covariates in the
model. Table 2 shows that four of the seven covariates were statistically significantly
associated with risk of new sexual offense adjudications. Specifically, year of index
offense, age at index offense, calendar year of risk of adjudications, and age at time
of risk were all statistically significantly associated with risk/hazard of event and in
the same direction as the previous (charge) model. Registration status was not statistically significantly associated with risk of new sexual offense adjudications.

Competing Risks Analyses


A second set of analyses was conducted to examine the effects of the same set of
covariates on the cause-specific risks of different types of initial (i.e., first recidivism
event for that youth) recidivism events. The causes included sexual, assault, and
other recidivism events. As with the survival analyses, separate models examined
charge and adjudication events.
Competing risks analysis for new charges. Of 755 initial recidivism charges, 46
were for sexual offenses, 148 were for assault offenses, and 561 were for other

Downloaded from http://cjp.sagepub.com at FLORIDA STATE UNIV LIBRARY on June 12, 2009

144

Downloaded from http://cjp.sagepub.com at FLORIDA STATE UNIV LIBRARY on June 12, 2009

0.79
1.51
0.95
1.61
0.32
0.09
0.43

.23
.24
.24
.24
.22
.07
.25

SE j
11.76***
38.92***
15.42***
46.49***
2.12
1.75
3.01

2
2.21
4.53
0.39
0.20
1.37
1.09
1.54

Hazard
Ratio
1.41, 3.48
2.82, 7.29
0.24, 0.62
0.13, 0.32
0.90, 2.11
0.96, 1.25
0.95, 2.52

95% CI
1.08
1.39
1.30
1.48
0.49
0.08
0.50

j
.40
.36
.42
.35
.37
.12
.43

SE j

**p < .01. ***p < .001.

b. AIC without covariates = 429.05, AIC with covariates = 340.10, Wald (model-based) 2(7) = 45.08, p < .001.

7.42**
14.84***
9.49**
17.54***
1.80
0.45
1.36

2.96
4.00
0.27
0.23
1.64
1.09
1.66

Hazard
Ratio

Sexual Adjudicationb

a. AIC without covariates = 1,259.78, AIC with covariates = 986.62, Wald (model-based) 2(7) = 127.98, p < .001.

Note: CI = confidence interval; AIC = Akaike Information Criterion.

Index year
Index age
Event year
Event age
Race
Priors
Registration

Covariate

Sexual Offense Chargea

Table 2
Sexual Recidivism Risk Survival Analyses Results

1.36, 6.46
1.98, 8.10
0.12, 0.62
0.11, 0.46
0.80, 3.35
0.85, 1.38
0.71, 3.86

95% CI

Letourneau et al. / Registration and Juvenile Recidivism

145

Figure 1
Estimated Survival Functions for Charge Recidivism

Note: The dash-doted line shows the estimated survival of sexual offense charges without accounting
for competing risk events. The solid line shows the estimated survival of sexual offense charges after
adjusting for the competing risks. The dashed line shows the estimated survival of assault offense
charge competing risk events, and the dotted line shows the estimated survival of other offense charge
competing risk events.

offenses. Model fit statistics (see Table 3) indicated statistically significant improvement in model fit after including the covariates. Year of index offense, age at index
offense, calendar year of risk, and age at risk of new charges were statistically significantly associated with the risk of new charges specifically for new sexual offense
charges (see Table 3). Again, the directions of these effects were identical to directions of corresponding effects from the two previous survival models. For cause-specific risk of new charges attributed to assault, all covariates except registration status
had statistically significant effects. Effects of race and prior offenses also were statistically significant. Specifically, minority status and greater number of prior
offenses were statistically significantly associated with increased risk for new
charges attributed to assault offenses. For cause-specific risk of new charges attributed to other offense charges, all covariates except race (for which there was a moderate statistically significant effect, p = .08) and prior offenses were statistically

Downloaded from http://cjp.sagepub.com at FLORIDA STATE UNIV LIBRARY on June 12, 2009

146

Downloaded from http://cjp.sagepub.com at FLORIDA STATE UNIV LIBRARY on June 12, 2009

0.95
1.44
1.09
1.46
0.58
0.17
0.15

1.14
1.64
1.30
1.65
0.16
0.07
0.26

Index year
Index age
Event year
Event age
Race
Priors
Registration

Index year
Index age
Event year
Event age
Race
Priors
Registration

Sexual Offense Chargea


.34
7.22**
.33
16.84***
.35
9.43**
.34
17.83***
.30
1.50
.09
2.72
.39
0.70
Assault Offense Chargec
.19
25.02***
.21
46.65***
.20
29.77***
.21
48.54***
.18
10.77**
.06
7.73**
.25
0.36
Other Offense Chargee
.10
124.46***
.10
295.65***
.11
149.46***
.09
310.08***
.09
3.05
.05
1.76
.11
5.20*

SE j

3.11
5.17
0.27
0.19
0.85
1.07
1.29

2.58
4.20
0.34
0.23
0.56
1.18
0.86

2.49
3.94
0.34
0.24
1.45
1.17
1.38

Hazard
Ratio

2.55, 3.80
4.29, 6.23
0.22, 0.34
0.16, 0.23
0.71, 1.02
0.97, 1.18
1.04, 1.61

1.78, 3.74
2.78, 6.34
0.23, 0.50
0.15, 0.35
0.40, 0.79
1.05, 1.33
0.52, 1.41

1.28, 4.83
2.05, 7.57
0.17, 0.68
0.12, 0.46
0.80, 2.62
0.97, 1.40
0.65, 2.95

95% CI

1.15
1.59
1.33
1.56
0.25
0.09
0.12

1.10
1.30
1.29
1.40
0.63
0.17
0.25

0.64
1.77
0.82
1.65
0.21
0.07
0.80

.11
.10
.12
.10
.10
.04
.12

.23
.22
.24
.23
.23
.08
.29

.49
.52
.53
.57
.49
.14
.67

SE j

Note: CI = confidence interval; AIC = Akaike Information Criterion.


a. AIC without covariates = 576.74, AIC with covariates = 433.95, Wald (model-based) 2(7) = 67.40, p < .001.
b. AIC without covariates = 207.10, AIC with covariates = 169.52, Wald (model-based) 2(7) = 21.65, p < .01.
c. AIC without covariates = 1,869.93, AIC with covariates = 1,370.20, Wald (model-based) 2(7) = 223.06, p < .001.
d. AIC without covariates = 1,209.69, AIC with covariates = 876.35, Wald (model-based) 2(7) = 154.67, p < .001.
e. AIC without covariates = 7043.64, AIC with covariates = 4,900.94, Wald (model-based) 2(7) = 896.35, p < .001.
f. AIC without covariates = 5,929.28, AIC with covariates = 4,216.17, Wald (model-based) 2(7) = 751.63, p < .001.
*p < .05. **p < .01. ***p < .001.

0.91
1.37
1.07
1.44
0.37
0.15
0.32

Index year
Index age
Event year
Event age
Race
Priors
Registration

Covariate

Table 3
Competing Risk Survival Analyses Results
Hazard
Ratio
95% CI

Sexual Adjudicationb
1.71
1.90
0.73, 4.96
11.48***
5.85
2.11, 16.24
2.42
0.44
0.16, 1.24
8.40**
0.19
0.06, 0.59
0.18
1.23
0.47, 3.22
0.26
1.07
0.82, 1.41
1.41
2.23
0.60, 8.34
Assault Adjudicationd
23.11***
3.00
1.92, 4.68
33.89***
3.66
2.36, 5.66
28.14***
0.28
0.17, 0.45
38.25***
0.25
0.16, 0.39
7.64**
0.53
0.34, 0.83
4.27*
1.18
1.01, 1.39
0.74
0.78
0.44, 1.38
Other Adjudicationf
102.31***
3.16
2.53, 3.94
245.77***
4.89
4.01, 5.97
124.23***
0.27
0.21, 0.34
237.19***
0.21
0.17, 0.26
6.48*
0.78
0.65, 0.95
6.10*
1.10
1.02, 1.18
0.98
1.13
0.89, 1.42

Letourneau et al. / Registration and Juvenile Recidivism

147

significant. Being registered at time of risk was statistically significantly associated


with cause-specific risk of other offense charges. Figure 1 gives the estimated causespecific survival functions of events of charges attributed to sexual, assault, and
other charges, adjusting for covariate effects. Each curve attributed to a specific
offense was estimated after accounting for the remaining competing causes of new
events of charges.
Competing risks for adjudications. Of all 573 observed initial adjudication recidivism events, 16 were for sexual offenses, 94 were for assault offenses, and 463 were
for other offenses. Table 3 shows statistically significant improvement in model fit
after including the covariates. Age at index offense, calendar year of risk, and age at
risk of adjudications were statistically significantly associated with the cause-specific risk of adjudications attributed to sexual offenses (see Table 3). For the causespecific risk of adjudications attributed to assault and the cause-specific risk of
adjudications attributed to other offenses, all covariates except registration status at
time of risk were statistically significant. Directions of the effects of covariates on
risks of adjudication events attributed to various causes remained the same as the
corresponding effects described in previous models.

Discussion
The purpose of this study was to examine the influence of registration status at
time of risk on the commission (or detection) of new sexual and nonsexual offenses,
with the aim of disentangling support for deterrence versus surveillance effects.
Consistent with several previous studies focusing on adult offenders (Adkins et al.,
2000; Schram & Milloy, 1995; Zevitz, 2006), our results provided no support for a
deterrent effect on juvenile sex offenders. Thus, we did not find any credible statistical evidence that registration status at time can decrease the risk/hazard of recidivism at that time. On the contrary, for some of our models, we have found moderate
to strong statistical evidence that registration status at time can increase the risk/hazard of new charges. As has been noted recently, crime control policies based on principles of deterrence appear more likely to achieve their intended deterrent effects
when aimed at specific and bounded behaviors (Weisburd, Einat, & Kowalski,
2008, p. 28). Thus, deterrence is context dependent and more likely to occur when
targeted toward highly specified behaviors (Nagin, 2008; Pogarsky, 2008; Weisburd
et al., 2008). It is perhaps not surprising, therefore, that registration fails to reduce
risk of recidivism, given the amorphous quality of sexual behaviors. Additionally,
the effect of deterrence varies by age (Sampson & Cohen, 1988), with juveniles relatively less likely to be deterred, perhaps because of the greater impulsivity of youth.
This further reduces the likelihood of identifying a registration deterrent effect on
juvenile sexual offenders. Determining whether public registration deters would-be

Downloaded from http://cjp.sagepub.com at FLORIDA STATE UNIV LIBRARY on June 12, 2009

148

Criminal Justice Policy Review

juvenile sexual offenders (i.e., primary prevention) is the subject of ongoing


research.
The present study provides some support for a surveillance effect. Specifically,
registration status at time was a significant predictor of other offense charges in
the competing risks model, with a trend for an association between registration status at time and risk of sexual offense charges in the survival analysis. Youth were at
greater risk of being charged with other (or sexual) offenses after registering than
before or in the absence of registration. An important question is whether this
increased relative risk of charges reflected the behavior of the offenders or those providing the surveillance. Had registration status been associated with risk of charges
and adjudications, a surveillance effect in which actual recidivism was identified
an intended goal of SORNA and other public registration policieswould have been
supported. Such a finding might also have indicated that the process of registration
had a criminogenic effect, serving to increase youths risk of re-offense. However,
registration status at time of risk was not associated with increased risk for adjudications in any models. One potential explanation for this pattern of findings is that
the surveillance effect was not a result of increased criminal behavior by the offenders but rather a result of increased policing of these youth. Police officers choose to
make arrests depending on several factors, such as local policies (e.g., Howell,
2003), familiarity with offenders (M. W. Klein, 1974), and global attributions about
classes of offenders (Wortley, 1997). A general belief among police officers that registered juvenile sexual offenders are at high risk of recidivism would be congruent
with lay beliefs about sexual recidivism rates (e.g., Levenson, Brannon, Fortney, &
Baker, 2007) and might lead officers to arrest and charge registered youth for suspicious but ultimately noncriminal behavior (similar to a negative halo error; Balzer &
Sulsky, 1992).
An alternative explanation for the present results is there were too few sexual
offense adjudications to detect what might have been a real registration status effect
(i.e., a Type 2 error). However, the pattern of finding registration status associated
with risk of charges but not adjudications held for the other offense category,
which had the highest rates of occurrence of all charge types. That these other
offenses tended to be less serious than sexual or assault offenses further supports the
hypothesis that the bar for charging registered youth might be significantly lower
than for charging nonregistered youth. Interpreting null effects is perilous, and
numerous other explanations for the current pattern of findings exist. For example,
multiple effects could be operating simultaneously, in which a deterrent effect might
decrease recidivism overall while a surveillance effect might increase the detection
of those crimes that are committed, the end result being no significant change in
observed recidivism rates for registered versus unregistered youth.
With respect to the remaining covariates, several consistent findings occurred
across models. Year of index offense was positively and significantly associated with
relative risk of recidivism in most models. This finding indicates that youth with

Downloaded from http://cjp.sagepub.com at FLORIDA STATE UNIV LIBRARY on June 12, 2009

Letourneau et al. / Registration and Juvenile Recidivism

149

more recent index offenses were more likely to have new recidivism events than
youth with more distal index offenses when compared across the same distance from
the origin (index) event. Delineating the specific causes of this increased risk would
be a useful avenue for future research. One possibility is the influence of other legislation (e.g., the increased bargaining power obtained by prosecutors following passage of South Carolinas 1995 three strikes law). A second consistent finding was
that youth who were older at time of index offense were more likely to reoffend at
any point in the follow-up. This finding corresponds with results from numerous
other studies linking older (youth) age at index offense to recidivism risk (e.g.,
Heilbrun et al., 2000). Furthermore, across all models, two time-dependent covariates (calendar year at time of risk and age at risk of recidivism event) were negatively associated with relative risk of recidivism. Thus, as youth moved further away
from their index offenses, they became less and less likely to obtain a new charge or
adjudication. In other words, a history of nonrecidivism predicted a future of nonrecidivism, as has been reported for adult sex offenders (Hanson, Morton, & Harris,
2003). Lastly, covariates indicating prior offenses and minority status performed as
expected when significant, in that both were associated with increased relative
recidivism risk in some models.
The results of this study should be considered in light of its strengths and limitations. Three limitations deserve comment. First, the percentage of youth with new
sexual offense charges (7.5%) or adjudications (2.5%) was low. This finding is consistent with overwhelming evidence indicating low juvenile sexual recidivism base
rates (e.g., Alexander, 1999; Caldwell, 2002; Fortune & Lambie, 2006) and is a pernicious problem for researchers. Thus, as noted previously, the failure of detecting
effect of registration status at time on risk/hazard of sexual offense adjudication
could be due, in part, to low statistical power for such detection. More specifically,
estimates of model parameters using survival analysis techniques for a particular
data set are dependent on the level of the underlying censoring mechanism. In the
event of high censoring, the parameter estimates as well as the associated standard
errors can be biased. Multistate and national studies with even longer follow-up
durations or studies that oversample higher risk offenders might permit improved
model specification. The second noteworthy limitation of the present study is the
absence of a scientifically rigorous design (i.e., randomization to conditions), which
necessarily limits the extent to which causation can be implied. Some researchers
have been permitted to randomize youth to probation conditions (e.g., Weisburd
et al., 2008), but this is a rare opportunity, indeed. It would be of great scientific and
policy benefit to examine the effects of registration policies in a more scientifically
rigorous manner. In the absence of such, replication of studies is imperative to determine whether consistent results emerge. A third limitation is that neither the juvenile
nor the adult criminal records provided a complete picture of the adjudication
process from initial charge through case disposal, resulting in different definitions of

Downloaded from http://cjp.sagepub.com at FLORIDA STATE UNIV LIBRARY on June 12, 2009

150

Criminal Justice Policy Review

charges for the DJJ data (i.e., charges forwarded to solicitors) versus the CCHR
data (i.e., all charges, regardless of subsequent processing). Given the low base rate
of sexual offense recidivism events, we believe accessing recidivism data from multiple sources was prudent, even in light of these differences. Principal strengths of
this study were the inclusion of all juvenile (male) sexual offenders across a substantial follow-up time period. Both the inclusiveness of the study sample (population) and the length of follow-up increase confidence in the generalizability and
external validity of the results.
In summary, this study provides a reasonable test of the effects of South
Carolinas public registration policy on risk/hazard of juvenile sexual recidivism.
The results have clear policy implications. First, this study found no evidence in
support of a deterrent effect for known offenders. Policy makers should acknowledge the growing body of research that has failed to find evidence of deterrence
and refrain from making deterrence-based arguments to support registration policies. Second, there was no evidence that registration status at time altered the likelihood of new adjudications at that time. This finding suggests that rather than
identifying recidivists, the increased surveillance afforded by public registration
resulted in unnecessary charges. Numerous concerns have been raised regarding
the registration of juvenile offenders, including (a) the deleterious collateral consequences of long-term public registration on youth (e.g., Jones, 2007; Michels,
2007; Trivits & Reppucci, 2002), (b) the fact that meaningful developmental differences between juvenile and adult offenders (e.g., Grisso et al., 2003; Steinberg
& Scott, 2003) are ignored by these policies (Garfinkle, 2003; Trivits & Reppucci,
2002; Zimring, 2004), and (c) these policies ignore emerging and scientifically
rigorous research evidence in support of treatment effectiveness for juveniles who
sexually offend (Borduin, Henggeler, Blaske, & Stein, 1990; Borduin, Schaeffer,
& Heiblum, in press; Hanson et al., 2002; Letourneau et al., in press; Reitzel &
Carbonell, 2006). In light of these concerns and the present studys findings that
registration status at time was not associated with deterrence and was associated
with increased risk of charges (but not adjudications), two final policy recommendations seem warranted. First, state and federal policies should be amended to
remove youth from online registries and to otherwise significantly limit access to
juvenile registry data. In the absence of evidence for effectiveness, the increased
risk of harm posed to publicly registered youth is not justifiable. Second, registration durations should be reduced to reflect the developmental status of youth.
Reduced registration durations acknowledge the substantive differences between
youthful and adult offenders and limit opportunities for unnecessary arrests and
charges. Indeed, many states already limit public access to juvenile offender information and juvenile registration durations. SORNA will require these states to
increase public access and registration durations for many youthful offenders,
changes not in keeping with present results.

Downloaded from http://cjp.sagepub.com at FLORIDA STATE UNIV LIBRARY on June 12, 2009

Letourneau et al. / Registration and Juvenile Recidivism

151

References
Adkins, G., Huff, D., & Stageberg, P. (2000). The Iowa sex offender registry and recidivism. Des Moines:
Iowa Department of Human Rights.
Alexander, M. A. (1999). Sexual offender treatment efficacy revisited. Sexual Abuse: A Journal of
Research and Treatment, 11, 101-116.
Balzer, W. K., & Sulsky, L. M. (1992). Halo performance appraisal research: A critical examination.
Journal of Applied Psychology, 77, 975-985.
Borduin, C. M., Henggeler, S. W., Blaske, D. M., & Stein, R. (1990). Multisystemic treatment of adolescent sexual offenders. International Journal of Offender Therapy and Comparative Criminology, 34,
105-113.
Borduin, C. M., Schaeffer, C. M., & Heiblum, N. (in press). A randomized clinical trial of multisystemic
therapy with juvenile sexual offenders: Effects on youth social ecology and criminal activity. Journal
of Consulting and Clinical Psychology.
Caldwell, M. F. (2002). What we do not know about juvenile sexual reoffense risk. Child Maltreatment,
7(4), 291-302.
Chaffin, M. (2008). Our minds are made up: Dont confuse us with the facts [Special Issue: children with
sexual behavior problems]. Child Maltreatment, 13, 110-121.
Cox, D. R. (1972). Regression models and life-tables, Journal of the Royal Statistical Society- B, 24,
187-220.
Finkelhor, D., & Jones, L. M. (2004). Explanations for the decline in child sexual abuse cases.
Washington, DC: Office of Juvenile Justice and Delinquency Prevention.
Fortune, C., & Lambie, I. (2006). Sexually abusive youth: A review of recidivism studies and methodological issues for future research. Clinical Psychology Review, 26, 1078-1095.
Garfinkle, E. (2003). Coming of age in America: The misapplication of sex-offender registration and
community-notification laws to juveniles. California Law Review, 91, 163-208.
Grisso, T., Steinberg, L., Woolard, J., Cauffman, E., Scott, E., Graham, S., et al. (2003). Juveniles competence to stand trial: A comparison of adolescents and adults capacities as trial defendants. Law and
Human Behavior, 27, 333-363.
Hanson, R. K., Gordon, A., Harris, A. J. R., Marques, J. K., Murphy, W., Quinsey, V. L., et al. (2002). First
report of the collaborative outcome data project on the effectiveness of psychological treatment for
sex offenders. Sexual Abuse: A Journal of Research and Treatment, 14, 169-194.
Hanson, R. K., Morton, K. E., & Harris, A. J. R. (2003). Sex offender recidivism risk: What we know and
what we need to know. Annals of the New York Academy of Science, 989, 154-166.
Heilbrun, K., Brock, W., Waite, D., Lanier, A., Schmid, M., Witte, G., et al. (2000). Risk factors for juvenile criminal recidivism: The postrelease community adjustment of juvenile offenders. Criminal
Justice and Behavior, 27, 275-291.
Howell, J. C. (2003). Preventing and reducing juvenile delinquency: A comprehensive framework.
Thousand Oaks, CA: Sage.
Jones, M. (2007, July 22). The case of the juvenile sex offender: Is he a criminal marked forever or a kid
whose behavior can be changed? New York Times Magazine, pp. 33-39, 56, 58-59.
Klein, J. P., & Moeschberger, M. L. (2003). Survival analysis: Techniques for censored and truncated
data. New York: Springer.
Klein, M. W. (1974). Labeling, deterrence, and recidivism: A study of police dispositions of juvenile
offenders. Social Problems, 22, 292-303.
Letourneau, E. J. (in press). Multisystemic therapy for juvenile sexual offenders: 1-year results from a
randomized effectiveness trial. Journal of Family Psychology.
Letourneau, E. J., & Miner, M. H. (2005). Juvenile sex offenders: A case against the legal and clinical
status quo. Sexual Abuse: A Journal of Research and Treatment, 17, 313-331.

Downloaded from http://cjp.sagepub.com at FLORIDA STATE UNIV LIBRARY on June 12, 2009

152

Criminal Justice Policy Review

Levenson, J. S., Brannon, Y., Fortney, T., & Baker, J. (2007). Public perceptions about sex offenders and
community protection policies. Analyses of Social Issues and Public Policy, 7, 1-25.
McManus, R. (Ed.). (2008). South Carolina criminal and juvenile justice trends: 2007. Blythewood:
South Carolina Department of Public Safety, Office of Justice Programs.
Michels, S. (2007, August 16). Should 14-year-olds have to register as sex offenders? ABC News.
Retrieved December 17, 2007, from http://abcnews.go.com/TheLaw/Story?id=3483364&page=1
Nagin, D. S. (2008). Thoughts on the broader implications of the Miracle of the Cells. Criminology &
Public Policy, 7, 37-42.
Pogarsky, G. (2008). Editorial introduction: Fines, threat of incarceration, and deterrence. Criminology &
Public Policy, 7, 5-8.
Prentice, R. L., & Breslow, N. E. (1978). Retrospective studies and failure time models. Biometrika, 65,
153-158.
Reitzel, L. R., & Carbonell, J. L. (2006). The effectiveness of sexual offender treatment for juveniles as
measured by recidivism: A meta-analysis. Sexual Abuse: A Journal of Research and Treatment, 18,
401-422.
Sampson, R. J., & Cohen, J. (1988). Deterrent effects of police on crime: A replication and theoretical
extension. Law and Society Review, 22, 163-189.
Satagopan, J. J., Ben-Porat, L., Berwick, M., Robson, M., Kutler, D., & Auerbach, A. D. (2004). A note
on competing risks in survival data analyses. British Journal of Cancer, 91, 1229-1235.
Schram, D., & Milloy, C. D. (1995). Community notification: A study of offender characteristics and
recidivism. Olympia: Washington Institute for Public Policy.
Scrucca, L., Santucci, A., & Aversa, F. (2007). Competing risk analysis using R: An easy guide for clinicians.
Bone Marrow Transplantation, 40, 381-387.
Steinberg, L., & Scott, E. S. (2003). Less guilty by reason of adolescence: Developmental immaturity,
diminished responsibility, and the juvenile death penalty. American Psychologist, 58, 1009-1018.
Therneau, T. M., & Gambsch, P. M. (2000). Modeling survival data: Extending the Cox-Model. New
York: Springer.
Trivits, L. C., & Reppucci, N. D. (2002). Application of Megans Law to juveniles. American
Psychologist, 57, 690-704.
U.S. Department of Justice. (2007, May). The national guidelines for sex offender registration and notificationProposed guidelines, May 2007. Retrieved December 17, 2007, from http://www.ojp.usdoj/
smart/proposed.htm
Vandiver, D. M. (2006). A prospective analysis of juvenile male sex offenders: Characteristics and recidivism rates as adults. Journal of Interpersonal Violence, 21, 673-688.
Washington State Institute for Public Policy. (2005). Sex offender sentencing in Washington State: Did
community notification influence recidivism? Olympia, WA: Author.
Weisburd, D., Einat, T., & Kowalski, M. (2008). The miracle of the cells: An experimental study of interventions to increase payment of court-ordered financial obligations. Criminology and Public Policy,
7, 9-36.
Worling, J. R., & Lngstrm, N. (2006). Risk of sexual recidivism in adolescents who offend sexually:
Correlates and assessment. In H. E. Barbaree & W. L. Marshall (Eds.), The juvenile sex offender (2nd
ed., pp. 219-247). New York: Guilford.
Wortley, R. (1997). Attributes as a function of expertise: The case of the police decision to arrest. Journal
of Applied Social Psychology, 27, 525-538.
Zevitz, R. G. (2006). Sex offender community notification: Its role in recidivism and offender reintegration. Criminal Justice Studies, 19, 193-208.
Zimring, F. E. (2004). An American travesty: Legal responses to adolescent sexual offending. Chicago:
University of Chicago Press.

Downloaded from http://cjp.sagepub.com at FLORIDA STATE UNIV LIBRARY on June 12, 2009

Letourneau et al. / Registration and Juvenile Recidivism

153

Elizabeth J. Letourneau, PhD, is an associate professor at the Family Services Research Center, Medical
University of South Carolina. She has conducted research on different aspects of sexual offending for 20
years, most recently focusing on treatment effectiveness and the effects of sex offender legislation. She is
an active member of the Association for the Treatment of Sexual Abusers.
Dipankar Bandyopadhyay, PhD, is an assistant professor of biostatistics at the Medical University of
South Carolina. His main research interests include nonparametric survival analysis, density estimation,
clustered data analysis, Bayesian data analysis and statistics in oral health research. Currently, he is also
interested in the quantitative exploration of the effects of registration/notification laws to deter sexual
violence.
Debajyoti Sinha, PhD, is the Ron and Caroline Hobbs Endowed professor of statistics in the Department
of Statistics, Florida State University. His main research interests include developing and extending
statistical models and associated Bayesian analysis for some common but rather complex examples of
survival data (e.g. data on relapse of cancer, death of the patient).
Kevin S. Armstrong, BS, is a statistical analyst and the data manager at the Family Services Research
Center, Medical University of South Carolina. He has supported faculty directed research since
December, 2003.

Downloaded from http://cjp.sagepub.com at FLORIDA STATE UNIV LIBRARY on June 12, 2009

Anda mungkin juga menyukai