Journal Article Critque

See discussions, stats, and author profiles for this publication at: https://www.researchgate.net/publication/240281495

The Prediction of Criminal Recidivism in JuvenilesA Meta-Analysis

Article in Criminal Justice and Behavior · June 2001

DOI: 10.1177/0093854801028003005

CITATIONS

402

READS

3,840

3 authors, including:

Some of the authors of this publication are also working on these related projects:

DeMatteo, D., Heilbrun, K., Arnold, S., & Thornewill, A. (in preparation). Problem-solving courts and the

criminal justice system. New York: Oxford. View project

Kirk Heilbrun

Drexel University

139 PUBLICATIONS 3,400 CITATIONS

SEE PROFILE

All content following this page was uploaded by Kirk Heilbrun on 31 March 2015.

The user has requested enhancement of the downloaded file.

CRIMINAL JUSTICE AND BEHAVIOR Cottle et al./ JUVENILE RECIDIVISM

THE PREDICTION OF CRIMINAL RECIDIVISM IN JUVENILES

A Meta-Analysis

CINDY C. COTTLE

RIA J. LEE

KIRK HEILBRUN MCP Hahnemann University

A meta-analysis was conducted to identify risk factors that best predict juvenile recidivism, defined as rearrest for offending of any kind. Twenty-three published studies, representing 15,265 juveniles, met inclusion criteria. Effect sizes were calculated for 30 predictors of recidi- vism. Eight groups of predictors were compared: (a) demographic information, (b) offense his- tory, (c) family and social factors, (d) educational factors, (e) intellectual and achievement scores, (f ) substance use history, (g) clinical factors, and (h) formal risk assessment. The domain of offense history was the strongest predictor of reoffending. Other relatively strong predictors included family problems, ineffective use of leisure time, delinquent peers, conduct problems, and nonsevere pathology.

Juvenile offending is a serious concern in our society today. Menand women younger than 18 years old comprise approximately 19% of the population (Census Bureau, 2000). However, recent crime statistics show that juvenile offenders were responsible for approxi- mately 29% of criminal arrests committed in the United States in

367

AUTHORS’ NOTE: We would like to thank Jack L. Vevea, Ph.D., University of North Carolina at Chapel Hill, and Joseph C. McClintock, Ph.D, Columbia Assess- ment Services, Inc., for their assistance in data analysis throughout this project. Cor- respondence concerning this article should be addressed to Kirk Heilbrun, Depart- ment of Clinical and Health Psychology, MCP Hahnemann University, Mail Stop 626, 245 N. 15th Street, Philadelphia, PA 19102-1192; e-mail: Kirk.Heilbrun@ drexel.edu.

CRIMINAL JUSTICE AND BEHAVIOR, Vol. 28 No. 3, June 2001 367-394 © 2001 American Association for Correctional Psychology

1998, accounted for 18% of all persons arrested in 1998, and are among the fastest growing groups of offenders (FBI, 1998). The per- centage of offenses committed by children and adolescents increased by 24% from 1989 to 1998, whereas the percentage of offenses com- mitted by adults over the same time period increased by only 3.8% (FBI, 1998).

A number of important arrest trends concerning juvenile offenders can be observed between 1989 and 1998. Among the total offender population, including offenders of all age groups, arrest trends show that the total crime index decreased by 14%, arrests for property crimes decreased by 19%, and arrests for violent crimes increased by 4% (FBI, 1998). However, a comparison of juvenile (under 18 years old) and adult (age 18 and older) offenders with regard to these arrest statistics shows that for the total crime index, juvenile arrests decreased by 9% and adult arrests by 16%. Similarly, arrests for prop- erty crimes decreased by 12% for juveniles and 23% for adults. For violent offending, juvenile arrest trends show a 15% increase, com- pared with a 3% increase in adult offenders (FBI, 1998). During this period, several important trends emerged in the arrest rates of juvenile offenders regarding specific types of crime: arrest rates decreased for homicide (23%), rape (3%), car theft (39%), burglary (22%), and lar- ceny (4%). By contrast, there were increases in arrests for aggravated assault (21%), arson (10%), and robbery (9%) (FBI, 1998). With the exception of homicide and car theft, these rates reflect a less favorable pattern of arrests for juveniles when compared with adults. Spe- cifically, for crimes in which there was an increase in arrests among both juveniles and adults, juvenile offenders showed a greater increase than adult offenders. Similarly, for offenses in which there was a decrease in the number of total arrests among both juveniles and adults, juvenile offenders showed a much smaller decrease in arrest rates than did adult offenders. This may be particularly problematic, considering that first-time juvenile offenders have a longer period at risk for reoffending over the course of a lifetime, and younger age at first offense has been associated with higher risk of reoffending (Dur- ham, 1996; Walters, 1996).

To be effective, the secondary and tertiary prevention of such offending must incorporate empirically supported risk factors for reoffending in adolescence and early adulthood. Moffitt (1993) has

368 CRIMINAL JUSTICE AND BEHAVIOR

suggested that there are two distinct and qualitatively different groups of juveniles who behave antisocially: adolescent-limited and life- course-persistent. Those with adolescent-limited antisocial behavior tend to begin during mid-adolescence and desist in young adulthood, and may be strongly influenced by situational factors. The second group begins antisocial behavior at an earlier age, persists in such behavior throughout the life course, and shows a pattern of progres- sively more serious offending with age. Although the latter group accounts for only 5% to 6% of juvenile offenders, it has been sug- gested that they are responsible for the majority of crimes committed by juveniles, and later, adult offenders (Moffitt, 1993).

A number of factors have been associated with reoffending in juve- niles. Male gender is a relatively stable predictor of recidivism (e.g., Dembo, et al., 1998; Hoge, Andrews, & Leschied, 1996), although offending by females is increasing (in 1998, females accounted for 29% of all juvenile arrests for property offenses, compared with 19% in 1981; Office of Juvenile Justice and Delinquency Prevention, 1999). Similarly, the type of crime committed by juvenile offenders (e.g., assault, auto theft) appears to have been linked reliably to recidi- vism by a number of studies (e.g., Archwamety & Katsiyannis, 1998; Dembo et al., 1998; Myner, Santman, Cappelletty, & Perlmutter, 1998).

However, there are also inconsistencies regarding risk factors across different studies. For example, substance use has been found to be a strong predictor of recidivism in some studies (e.g., Dembo, Turner, Sue, Schmeidler, Bordon, & Manning, 1995; Roy, 1995) but not others (e.g., Wierson & Forehand, 1995), or has been predictive only for a specific substance, such as cocaine (e.g., Dembo, Williams, Schmeidler, Getreu, & Berry, 1991). According to some investigators (e.g., Dembo et al., 1998; Minor, Hartman, & Terry, 1997; Putnins, 1984), living in a single-parent family is associated with increased risk for reoffending, but this has not been replicated in other studies (e.g., Myner et al., 1998; Niarhos & Routh, 1992).

Intellectual functioning has also been inconsistent across studies in its capacity to function as a risk factor for juvenile reoffending. Duncan, Kennedy, and Patrick (1995) reported that full-scale IQ scores were significant predictors of juvenile recidivism, for example, but Katsiyannis and Archwamety (1997) did not. Such inconsisten-

Cottle et al. / JUVENILE RECIDIVISM 369

cies may be at least partly attributed to the use of different measures of intellectual functioning. To better determine the predictive impact of intellectual functioning on recidivism, it may be useful to consider only studies that use standardized and validated measures of intellec- tual ability, such as the Wechsler scales (e. g., WISC-III) (Weschler, 1991).

Yet another domain in which inconsistency has been observed is history of conduct problems. Some (e.g., Niarhos & Routh, 1992; Wierson & Forehand, 1995) have reported it to be predictive of reoffending, others (e.g., Duncan et al., 1995; Repo & Virkkunen, 1997), have found no relationship between previous conduct prob- lems and reoffense risk, and one study (Myner et al., 1998) actually found a negative relationship. Such inconsistencies underscore the importance of summarizing empirical findings across studies of juve- nile recidivism.

PREVIOUS REVIEWS

Loeber and Dishion (1983) reviewed studies from 1962 to 1980 that considered variables related to delinquency (defined as first-time offending) and recidivism (operationalized as having two or more arrests) in juveniles. To be included in the analysis, the authors required that a study provide sufficient data to allow its incorporation into prediction tables and use objective predictor variables obtained at least a year before the outcome. Outcome measures included police contact, arrest, conviction, and self-reported delinquency; studies using only parole violations as an outcome were excluded. Both retro- spective and prospective studies were included. Variables were grouped into four categories: behavioral (e.g., lying, truancy), family (e.g., socioeconomic status, criminality in family members), educa- tional (e.g., grade point average, low vocabulary), and aggregated (e.g., two or more variables).

A relative improvement over chance (RIOC) ratio was calculated for each variable. Results of this review suggested that composite measures of parental family management techniques were best in improving the accuracy of prediction over chance. The child’s prob- lem behaviors and a history of stealing, lying, or truancy were also

370 CRIMINAL JUSTICE AND BEHAVIOR

found to be strong predictors of delinquency. The weakest predictive variables were separation from parents and socioeconomic status.

When reoffending was considered as an outcome, these studies suggested that a history of stealing, lying, and/or truancy was the strongest predictor. Self-reported child problem behavior (e.g., aggressiveness), prior delinquency, and criminality among family members also improved the accuracy of prediction of recidivism. Socioeconomic status was again the weakest in predicting recidivism.

The Loeber and Dishion (1983) study was a timely review of the lit- erature pertaining to juvenile delinquents and recidivists. The authors provided a quantitative review of the variables related to delinquency and recidivism, and devised a ratio (RIOC) that allowed a meaningful appraisal of the strength of predictive variables. There were also sig- nificant limitations, however. Only 11 unique samples were available for review. This limited number is particularly problematic for vari- ables such as socioeconomic status, which was included in only two studies. More generally, however, this relatively small number of unique samples limits the confidence with which other variables, even those addressed more frequently, may be generalized as accurate in predicting recidivism.

A meta-analysis conducted by Simourd and Andrews (1994) focused only on juvenile delinquency, although it did not distinguish first-time offenders from recidivists or consider adult criminal behav- ior. Simourd and Andrews included 60 studies, with risk factors cate- gorized as follows: (a) social class, (b) family structure or parental problems, (c) personal distress, (d) minor personality variables, (e) parent-child relations, (f ) educational difficulties, (g) temperament or conduct problems, and (h) antisocial peers or attitudes. Although this meta-analysis made no distinction between outcomes of initial offending and reoffending, it does provide some useful findings. First, no gender differences were reported; male and female juveniles were found to have similar risk factors for offending. Second, socioeco- nomic status, family structure, parental problems, and personal dis- tress were not predictive of offending for either male or female juve- niles. Finally, the risk factors most strongly predictive of offending were (a) antisocial peers or attitudes (53 studies); (b) temperament or conduct problems, such as psychopathy, impulsivity, and substance use (45 studies); (c) educational difficulties, such as poor grades,

Cottle et al. / JUVENILE RECIDIVISM 371

dropout status (34 studies); (d) poor parent-child relations, such as problems in attachment or supervision (41 studies); and (e) minor per- sonality variables, such as empathy, moral reasoning (9 studies).

The present meta-analysis will focus on research conducted between 1983 and 2000, whereas the Simourd and Andrews (1994) review included studies published between 1964 and 1994. In addition to this difference in time periods, the present study will focus on reoffending rather than both initial offending and recidivism. The pur- pose of this distinction lies in the comparability of the two offender populations. It is not feasible to make meaningful assumptions about predictors of reoffending behavior based on predictors found to be associated with first-time delinquency, because the two offender pop- ulations differ in terms of their homogeneity. Whereas studies of risk factors for first-time delinquents tend to be based on broad samples of juveniles prior to their identification as delinquent (e.g., Bennett & Basiotis, 1991; Loeber & Stouthamer-Loeber, 1986), studies examin- ing recidivism risk factors typically are based on more homogenous samples of adolescents already identified as delinquent. Therefore, variables significantly associated with reoffending behavior in juve- niles are not necessarily useful in initially distinguishing between adolescents who will or will not become delinquents. Differences between our results and those of Simourd and Andrews should be con- sidered accordingly.

All of the studies included in the present meta-analysis focus on determining factors associated with reoffending behavior in juvenile offender populations. Some differences can be seen among the included studies concerning specific areas of focus. For instance, a number of studies examine associations between recidivism rates and psychological assessment data (e.g., Anderson & Walsh, 1998; Bleker, 1983), as well as risk assessment measures (e.g., Ashford & LeCroy, 1990; Dembo, Williams, Fagan, & Schmeidler, 1994). Some of the studies focus on specific differences among juvenile offender populations, such as gender (e.g., Archwamety & Katsiyannis, 1998) or age (e.g., Stattin & Magnusson, 1989). Other research examines reoffending behavior as it relates to sentencing decisions. For instance, a number of the articles included in this analysis can be dis- tinguished on the basis of placement of the participants at the time of the investigation. The studies considered in the present review typi-

372 CRIMINAL JUSTICE AND BEHAVIOR

cally focus on juveniles placed in correctional facilities (e.g., Katsiyannis & Archwamety, 1997), within the community (e.g., Anderson & Walsh, 1998), or on probation (e.g., Funk, 1999). Simi- larly, some studies examine potential relationships between court dis- positions and recidivism rates (e.g., Niarhos & Routh, 1992). In gen- eral, however, all of the research studies included in this meta-analysis share the primary objective of identifying factors related to recidivism in juveniles.

It is important to note that this analysis does not include predictor variables pertaining to the various interventions available for this pop- ulation of juvenile offenders. The research on interventions for juve- nile delinquents has previously been reviewed using meta-analytic techniques (e.g., Lipsey, 1992; Whitehead & Lab, 1989). In general, studies examining the effects of interventions on juvenile recidivism tend to show reductions in reoffending behavior following specific treatments such as multisystemic therapy (e.g., Henggeler, Melton, Brondino, Scherer, & Hanley, 1997), as compared with interventions such as wilderness or challenge programs, or probation (Lipsey, 1999). In light of the existing empirical literature focusing on inter- vention effects on juvenile recidivism, this particular area of predic- tors is not reviewed in the present analysis.

METHODOLOGICAL CONSIDERATIONS

Several considerations are noteworthy in conducting a meta-analy- sis. First, some researchers may only report statistically significant findings. As a result, those conducting a meta-analysis are forced to omit unreported findings or to use predetermined values, either of which is likely to yield skewed effect sizes (Rosenthal, 1991). Second, it is difficult to locate unpublished studies, often the source of non- significant findings. This phenomenon has been called the file-drawer problem (Rosenthal, 1991) and increases meta-analytic bias in favor of studies reporting statistically significant relationships. Two approaches to minimizing such a bias can be used: (a) locating as many unpublished studies as possible by contacting researchers known to investigate the area and (b) estimating the number of unretrieved studies reporting null results that would be needed to bring the overall p value to a given level by using a formula that takes

Cottle et al. / JUVENILE RECIDIVISM 373

into account the number of studies used and the exact p values of those studies (Rosenthal, 1991).1

The comparability of studies is also an important concern in meta-analysis. In the area of juvenile recidivism, studies may differ in how the investigators measure particular risk factors or define out- comes (e.g., rearrest vs. reconviction) and in the length of time partici- pants are at risk for reoffending. The process of determining inclusion criteria and of evaluating the reliability of those criteria across studies is one of the most important steps in a meta-analysis (Stock, 1994), as the inadequate specification of reliable inclusion criteria would raise questions about the overall reliability and validity of the meta-analytic findings.

Defining recidivism poses a particular problem in considering the outcomes to be used in the present meta-analysis. Criminal recidivism is generally defined by variables such as rearrest (e.g., Ashford & LeCroy, 1990), reconviction (e.g., Hoge et al., 1996), probation viola- tion (e.g., Hoge et al., 1996), or recommitment to an institution (e.g., Dembo et al., 1998). Two issues are relevant. First, the comparability of these various forms of recidivism is not known. Second, it is likely that these measures substantially underestimate the prevalence of antisocial behavior during a given period, judging from studies with adults that have considered self-reported and collateral-described vio- lent behavior as well as that reflected in official records such as rearrest (see, e.g., Lidz, Mulvey, & Gardner, 1993, Steadman et al., 1998).

The definition of juvenile is also important in the present meta- analysis. Studies investigating juveniles often vary in the age range considered, with some studies including young adults (up to 21 years old; e.g., Repo & Virkkunen, 1997; Steiner, Cauffman, & Duxbury, 1999) as juveniles and others limiting juveniles to a given age (most frequently the law’s definition of younger than 18 years old; e.g., Hoge et al., 1996; Jung & Rawana, 1999).

Also important is the duration of the follow-up periods that are used in different studies. Because the overall effect size of a given predictor for recidivism may vary as a function of follow-up duration, the influ- ence of duration must also be considered.

Although several meta-analyses have considered the predictors of criminal recidivism in adults (e.g., Bonta, Law, & Hanson, 1998;

374 CRIMINAL JUSTICE AND BEHAVIOR

R. Hanson & Bussiere, 1998), there has been only one meta-analytic study (Simourd & Andrews, 1994) on the prediction of juvenile reoffending. The present meta-analysis will seek to identify risk fac- tors most strongly associated with such offending by juveniles in light of the methodological concerns discussed in this section. We will focus on “general recidivism,” as opposed to “violent recidivism” or “sexual recidivism,” because only eight studies were identified that pertained to violent or sexual reoffending exclusively.2

METHOD

PARTICIPANTS

Computer searches of the databases PsychLit, PsychInfo, and MedLine were conducted to locate published articles from 1983 to 2000 using the following key words: delinquency, juvenile crime, recidivism, criminal behavior, sexual offenses, sexual assault, aggres- sion, violence, violent behavior, prediction of violence, psychopathy, court transfers, and rearrest. Investigators known to conduct research in the area of juvenile criminal behaviors were contacted to locate additional articles. To be selected for this meta-analysis, we required that a study consider juveniles between the ages of 12 and 21 years who had at least one prior arrest, and provide data on subsequent offending, as defined by either official records or self-report reflecting reincarceration, rearrest, or violation of probation or parole.

A total of 23 published studies, representing 22 unique samples, were identified as meeting criteria for inclusion in the meta-analysis. One group of researchers produced two reports (Dembo et al., 1991, 1998) using essentially the same sample. To avoid double counting, we selected the version with the longer follow-up period for use in the meta-analysis (Dembo et al., 1998). The remaining 22 studies used in the meta-analysis are identified in the References section by an accompanying asterisk (*).

Most of the studies (n = 22) used a single source (i.e., official records) to obtain recidivism data. Only one study (Jung & Rawana, 1999) used official records, self-report, and information from collat- eral sources to obtain recidivism data, although this study did not com-

Cottle et al. / JUVENILE RECIDIVISM 375

pare the recidivism rates obtained through these different sources. The mean sample size of these studies was 688.4 with a range of 45 to 9,176. The mean outcome period was 45.26 months, with a wide range in the duration of outcome periods (1-192 months). The overall mean recidivism rate was 48% (see Table 1).

The majority of the 15,265 participants were male (83.31%) and the mean age was 14.7 years. A total of 47.9% of the participants were Caucasian, 38.18% were African American, and the remaining 18% were classified as “other” (see Table 2).

PREDICTOR VARIABLES

A review of the 22 unique samples resulted in the identification of 30 predictor variables. Using the meta-analysis conducted with adult criminal offenders by Bonta et al. (1998) as a model, these predictors were divided into eight domains. These domains were as follows: (a) demographic information, (b) offense history, (c) family and social factors, (d) educational factors, (e) standardized tests scores, (f ) sub- stance use history, (g) clinical factors, and (h) formal risk assessment.

376 CRIMINAL JUSTICE AND BEHAVIOR

TABLE 1: Characteristics of Studies (n = 25) Included in the Meta-Analysis

Variable M SD Range k

Samples 688.4 1,914.0 45-9,176 22 Follow-up (months) 45.26 58.81 1-192 10a,b

Recidivism base rates (%) 48.0 14.7 22-75 17c

NOTE: k = number of unique samples a. Thirteen studies failed to report a mean follow-up period: Archwamety & Katsiyannis (1998); Ashford & LeCroy (1988); Bleker (1983); Dembo et al. (1998); Funk (1999); Grenier & Roundtree (1987); C. Hanson, Henggeler, Haefele, & Rodick (1984); Katsiyannis & Archwamety (1997); Minor, Hartmann, & Terry (1997); Myner, Santman, Capelletty, & Perlmutter (1998); Niarhos & Routh (1992); Putnins (1984); Towberman (1994). b. Ten studies failed to report a follow-up range: Archwamety & Katisyannis (1998); Ash- ford & LeCroy (1990); Ashford & LeCroy (1988); Bleker (1983); Dembo, Williams, Fagan, & Schmeidler (1994); Funk (1999); Grenier & Roundtree (1987); C. Hanson et al. (1984); Myner et al. (1998); Towberman (1994). c. Six studies failed to report comparable recidivism rates: Dembo et al. (1994); Grenier & Roundtree (1987); C. Hanson et al. (1984); Hoge, Andrews, & Leschied (1996); Myner et al. (1998); Towberman (1994).

Demographic information. Gender, race, and socioeconomic status made up the demographic information domain. The limited age range of the samples precluded the use of current age as a meaningful pre- dictor of recidivism. That is, because this meta-analysis involves juve- niles, the samples included have an artifactually limited age range (typically ranging from 14 to 18 years old). Because the age of the par- ticipants in these studies tends to be truncated around the time at which the risk for offending is highest (i.e., 18 years old; Moffitt, 1993), genuine differences between the current ages of the partici- pants in predicting recidivism would not be found.

Cottle et al. / JUVENILE RECIDIVISM 377

TABLE 2: Age, Race, and Gender of Participants (n = 15,265) Included in Meta- Analysis

Variable M SD Range k

Age (years) 14.71 .91 6-21 11a, b

Race (%) 13c

Caucasian 47.85 African American 38.18 Otherd 18.00

Gender (%) Male 83.31 21e

Female 16.69 21e

NOTE: k = number of unique samples a. Twelve studies failed to report a mean age: Archwamety & Katsiyannis (1998); Ashford & LeCroy (1990); Ashford & LeCroy (1988); Bleker (1983); Grenier & Roundtree (1987); Hoge, Andrews, & Leschied (1996); Katsiyannis & Archwamety (1997); Myner, Santmann, Capelletty, & Perlmutter (1998); Putnins (1984); Repo & Virkkunen (1997); Towberman (1994); Wierson & Forehand (1995). b. Five studies failed to report an age range: Ashford & LeCroy (1990); Ashford & LeCroy (1988); Dembo, Williams, Fagan, & Schmeidler (1994); Grenier & Roundtree (1987); C. Hanson, Henggeller, Haefele, & Rodick (1984); Stattin & Magnusson (1989). c. Nine studies failed to report racial breakdown of the samples: Grenier & Roundtree (1987); C. Hanson et al. (1984); Hoge et al. (1996); Jung (1999); Minor, Hartmann, & Terry (1997); Putnins (1984); Repo & Virkkunen (1997); Stattin & Magnusson (1989); Wierson & Forehand (1995). d. Includes Hispanic, Native American, Asian American, and minority participants not specified further. e. Two studies failed to report a gender breakdown of the samples: Grenier & Roundtree (1987); and Minor et al. (1997).

Offense history. Six predictors were included in the offense history domain. These included age at first contact with the law, age at first commitment, number of prior arrests, number of prior commitments, type of crime committed, and length of first incarceration.

Family and social factors. Seven predictors were included in the family and social factors categories. These predictors were as follows: having been a victim of physical or sexual abuse, living with a single parent, parent pathology, number of out-of-home placements, family problems (e.g., poor relationships within the family), effective use of leisure time, and having delinquent peers.

Educational factors. Three predictors made up the educational fac- tors domain: history of special education, school attendance, and school achievement (e.g., teacher reports, grade point average, grade placement).

Intellectual and achievement scores. Four types of test scores were included in this domain. They were standardized achievement scores, verbal IQ scores, performance IQ scores, and full-scale IQ scores.

Substance use history. Substance use history was composed of two predictors: substance use and substance abuse.

Clinical factors. Four predictors were grouped in this domain. They included severe pathology (e.g., psychosis, suicidality), conduct problems (e.g., presence of conduct-disordered symptoms), non- severe pathology (e.g., stress, anxiety), and history of treatment. Most of the studies investigating diagnostic predictors of recidivism (Arch- wamety & Katsiyannis, 1998; Jung & Rawana, 1999; Myner et al., 1998; Niarhos & Routh, 1992; Towberman, 1994) used clinical judg- ment to code pathology variables. Other studies used rating scales (Duncan et al., 1995; C. Hanson, Henggeler, Haefele, & Rodick, 1984; Steiner et al., 1999), teacher ratings (Stattin & Magnusson, 1989), and the Diagnostic Interview Scale for Children-2 (Wierson & Forehand, 1995).

378 CRIMINAL JUSTICE AND BEHAVIOR

Formal risk assessment. There were not enough studies investigat- ing formal measures of risk assessment to divide this domain into sep- arate variables. Thus, a composite variable—formal risk assessment— was created to represent any study investigating combinations of vari- ables to predict recidivism.

PROCEDURE

A meta-analysis including all 30 predictor variables was con- ducted, with recidivism as the outcome. Raw statistics from each study were converted to correlation coefficients using formulas pro- vided by Rosenthal (1991). These coefficients were normalized using Fisher’s transformation formula:

Zr = ½ loge [1 + r/1 – r].

The effect sizes were then used to calculate an overall weighted effect size (wtd. Zr) for each variable:

Weighted Zr = (Σ wjzr)/(Σwj).

The mean levels of significance were calculated by converting each p value to a normal deviate (Z ) corresponding to each, and averaging the weighted Zs:

Weighted Z = (Σwjzj)/(Σwj 2 )

½ .

To address the file-drawer problem, we used a variation of Rosen- thal’s (1991) method for estimating the number of unretrieved studies reporting null results that would be needed to bring the overall p value to a nonsignificant level. Rosenthal’s formula incorporates the num- ber of studies used and the exact p values of those studies (Rosenthal, 1991). However, in the current meta-analysis, weighted Z values were used to compensate for the variability in the number of participants per study. Thus, to provide a better understanding of the power of the included studies and of the relative size of samples needed in future research, a formula was derived to calculate the total number of partic- ipants needed to yield nonsignificant results:

N× = [(Σ(Z*df ) /1.645) 2

– Σ(df ) 2]½ + 3.

Cottle et al. / JUVENILE RECIDIVISM 379

RESULTS

All variables within the demographic information domain were significantly associated with recidivism (see Table 3). Being male (Zr = .11, p < .001) and of a minority race (Zr = .07, p < .001) were posi- tively related to recidivism. Juveniles from a low socioeconomic background were also at increased risk of reoffending (Zr = –.07, p < .001). A hierarchical regression analysis was conducted to determine whether race remained significant after controlling for socioeconomic status (SES) by entering both gender and SES before race in the equa- tion. Results indicated that race did not remain significantly associ- ated with recidivism after the influence of SES was controlled (F change = 1.80, p = .25).

Each of the offense history variables was also significantly associ- ated with recidivism. Juveniles with earlier age of first contact with the law (Zr = –.34, p < .001), earlier age at first commitment (Zr = –.35, p < .001), more prior arrests (Zr = .06, p < .001), more previous com- mitments (Zr = .17, p < .001), longer incarcerations (Zr = .19, p < .001), and those who committed more serious crimes (Zr = .16, p < .001) were at higher risk for recidivism.

Five of the seven family and social variables were significant pre- dictors of recidivism as well. Juveniles with a history of having been physically or sexually abused (Zr = .11, p < .001), raised in a sin- gle-parent home (Zr = .07, p < .001), having a greater number of out-of-home placements (Zr = .18, p < .001), or significant family problems (Zr = .23, p < .001) were at increased risk of recidivism. Juveniles who did not use their leisure time effectively (Zr = .23, p < .001) and those with delinquent peers (Zr = .20, p < .001) were also at increased risk. However, the presence of pathology in the parents was not found to be a significant predictor (Zr = .05, p = ns).

Among the educational factors, only a history of special education was significantly associated with recidivism (Zr = .13, p < .01). Nei- ther school attendance (Zr = –.05, p = ns) nor the schools’ reports of academic achievement (Zr = –.03, p = ns) were significant predictors.

Three of the four standardized test score variables were significant predictors of recidivism. Juveniles with lower standardized achieve-

380 CRIMINAL JUSTICE AND BEHAVIOR

(text continues on p. 384)

Cottle et al. / JUVENILE RECIDIVISM 381

TABLE 3: Predictors of Recidivism in Juveniles (n = 15,265) by Domain

Variable Zr n k Study

Demographic information

Gender (male) .111** 9,671 3 Dembo et al. (1998); Hoge et al. (1996); Minor et al. (1997)

Race (minority) .067** 10,121 6 Archwamety & Katsiyannis (1998); Dembo, Schmeidler, Nini-Gough, Sue, Borden & Manning (1998); Katsiyannis & Archwamety (1997); Minor, Hartmann, & Terry (1997); Myner, Santman, Cappelletty, & Perlmutter (1998); Niarhos & Routh (1992)

Socioeconomic status

–.065** 10,363 3 Dembo et al. (1998); Myner et al. (1998); Stattin & Magnussin (1989)

Offense History

Age at first contact with law

–.341** 1,225 8 Archwamety & Katsiyannis (1998); Duncan, Kennedy, & Patrick (1995); C. Hanson, Henggeler, Haefel, & Rodick (1984); Katsiyannis & Archwamety (1997); Minor et al. (1997); Myner et al. (1998); Niarhos & Routh (1992); Wierson & Forehand (1995)

Age at first commitment

–.346** 720 3 Archwamety & Katsiyannis (1998); Katsiyannis & Archwamety (1997); Towberman (1994)

Number of prior arrests

.058** 10,155 7 Archwamety & Katsiyannis (1998); Dembo et al. (1998); Duncan et al. (1995); Jung & Rawana (1999); Katsiyannis & Archwamety (1997); Niarhos & Routh (1992); Wierson & Forehand (1995)

Number of prior commitments

.174** 585 3 Archwamety & Katsiyannis (1998); Katsiyannis & Archwamety (1997); Duncan et al. (1995)

(continued)

382 CRIMINAL JUSTICE AND BEHAVIOR

Type of crime .159** 10,267 7 Archwamety & Katsiyannis (1998); Dembo et al. (1998); Katsiyannis & Archwamety (1997); Minor et al. (1997); Myner et al. (1998); Niarhos & Routh (1992); Wierson & Forehand (1995)

Length of first incarceration

.187** 641 3 Archwamety & Katsiyannis (1996); Katsiyannis & Archwamety (1997); Myner et al. (1998)

Family and social factors

Victim of abuse .112** 9,949 5 Archwamety & Katsiyannis (1998); Dembo et al. (1998); Katsiyannis & Archwamety (1997); Myner et al. (1998); Towberman (1994)

Single parent .070** 10,501 5 Dembo et al. (1998); Minor et al. (1997); Myner et al. (1998); Niarhos & Routh (1992); Putnins (1984)

Parent pathology

.047 529 3 Hoge et al. (1996); Myner et al. (1998); Niarhos & Routh (1992)

Number of out-of-home placements

.184** 424 2 Myner et al. (1998); Towberman (1994)

Family problems

.227** 1,054 5 C. Hanson et al. (1984); Hoge et al. (1996); Jung & Rawana (1999); Niarhos & Routh (1992); Towberman (1984)

Effective use of leisure time

–.233** 588 2 Hoge et al. (1996); Jung & Rawana (1999)

Delinquent peers

.204** 1,525 7 Archwamety & Katsiyannis (1998); Hoge et al. (1996); Jung & Rawana (1999); Katsiyannis & Archwamety (1997); Myner et al. (1998); Niarhos & Routh (1992); Towberman (1994)

Educational factors

History of special education

.130** 432 2 Archwamety & Katsiyannis (1998); Katsiyannis & Archwamety (1997)

School attendance

–.048 299 2 Myner et al. (1998); Towberman (1994)

TABLE 3 Continued

Variable Zr n k Study

Cottle et al. / JUVENILE RECIDIVISM 383

School report of achievement

–.028 10,025 6 Dembo et al. (1998); Duncan et al. (1995); Hoge et al. (1996); Jung & Rawana (1999); Myner et al. (1998); Niarhos & Routh (1992)

Intellectual and achievement scores

Standardized achievement score

–.153** 506 3 Archwamety & Katsiyannis (1998); Duncan et al. (1995); Katsiyannis & Archwamety (1997)

Verbal IQ score –.111* 716 4 Archwamety & Katsiyannis (1998); Bleker (1983); Hanson et al. (1984); Katsiyannis & Archwamety (1997)

Performance IQ score

–.031 491 2 Archwamety & Katsiyannis (1998); Katsiyannis & Archwamety (1997)

Full scale IQ score

–.142** 1,756 5 Archwamety & Katsiyannis (1998); Duncan et al. (1995); Katsiyannis & Archwamety (1997); Niarhos & Routh (1992); Stattin & Magnusson (1989)

Substance use history

Substance use .014 9,366 2 Dembo et al. (1998); Towberman (1994)

Substance abuse

.149** 1,111 6 Archwamety & Katsiyannis (1998); Duncan et al. (1995); Katsiyannis & Archwamety (1997); Myner et al. (1998); Niarhos & Routh (1992); Wierson & Forehand (1995)

Clinical factors

Severe pathology

.069 346 2 Archwamety & Katsiyannis (1998); Niarhos & Routh (1992)

Conduct problems

.255** 1,667 7 Duncan et al. (1995); Myner et al. (1998); Niarhos & Routh (1992); Repo & Virkkunen (1997); Stattin & Magnusson (1989); Towberman (1994); Wierson & Forehand (1995)

TABLE 3 Continued

Variable Zr n k Study

(continued)

ment test scores (Zr = –.11, p <.001), lower full-scale IQ scores (Zr = –.14, p < .001), and lower verbal IQ scores (Zr = –.11, p < .01) were at increased risk of recidivism. Performance IQ scores (Zr = –.03, p = ns) were not significantly related to recidivism.

Substance abuse was significantly associated with recidivism (Zr = .15, p < .001). Substance use, however, was not (Zr = .01, p = ns).

Two of the four clinical factors were significant predictors of recidi- vism. Juveniles with a history of conduct problems (Zr = .26, p < .001) or nonsevere pathology (Zr = .31, p < .001) were at increased risk. However, neither history of severe pathology (Zr = .07, p = ns) nor his- tory of psychiatric treatment (Zr = .02, p = ns) were significant.

The composite variable encompassing any kind of formal assess- ment of risk was significant in predicting recidivism (Zr = .12, p < .001).

We also considered all risk factors separately, rank-ordered them according to the magnitude of their Z scores, and calculated the num- ber of additional participants with null results needed to alter the sig- nificance of the finding (see Table 4). As may be seen, the strongest

384 CRIMINAL JUSTICE AND BEHAVIOR

Nonsevere pathology

.305** 953 7 Bleker (1983); Duncan et al. (1995); C. Hanson et al. (1984); Jung & Rawana (1999); Niarhos & Routh (1992); Steiner, Cauffman, & Duxbury (1999); Wierson & Forehand (1995)

History of treatment

.019 9,366 2 Dembo et al. (1998); Towberman (1994)

Formal risk assessment

Risk assessment instruments

.118** 10,353 6 Archwamety & Katsiyannis (1996); Ashford & LeCroy (1988); Dembo et al. (1994, 1998); Jung & Rawana (1999); Katsiyannis & Archwamety (1997)

NOTE: Zr = weighted mean effect size, k = number of unique samples. *p < .01. **p <.001.

TABLE 3 Continued

Variable Zr n k Study

predictors were age at first commitment, age at first contact with the law, and history of nonsevere pathology. The minimum number of participants in an unretrieved study with a Z value of zero needed to nullify the significance ranged from 473 to 81,345, thus reflecting the variability in the power of the studies.

Cottle et al. / JUVENILE RECIDIVISM 385

TABLE 4: Predictors of Recidivism in Juveniles (n = 15,265) by Predictive Strength

Variable Zr n k nx

Age at first commitment –.346** 720 3 2,273 Age at first contact with the law –.341** 1,225 8 3,298 Nonsevere pathology .305** 953 7 2,244 Family problems .277** 1,054 5 2,165 Conduct problems .255** 1,667 7 6,949 Effective use of leisure time –.233** 588 2 1,343 Delinquent peers .204** 1,525 7 2,842 Length of first incarceration .187** 641 3 1,022 Number of out-of-home placements .184** 424 2 617 Number of prior commitments .174** 585 3 699 Type of crime .159** 10,267 7 81,345 Standardized achievement score –.153** 506 3 599 Substance abuse .149** 1,111 6 1,273 Full scale IQ score –.142** 1,756 5 5,014 History of special education .130* 432 2 473 Risk assessment instruments .118** 10,353 6 60,117 History of abuse .112** 9,949 5 59,436 Gender (male) .111** 9,671 3 58,698 Verbal IQ score –.111* 716 4 522 Single parent .070** 10,501 5 37,930 Severe pathology .069 346 2 ns Race (minority) .067** 10,121 6 30,018 Socioeconomic status .065** 10,363 3 36,703 Number of prior arrests .058** 10,155 7 26,145 School attendance –.048 299 2 ns Parent pathology .047 529 3 ns Performance IQ score –.031 491 2 ns School report of achievement –.028 10,025 6 ns History of treatment .019 9,366 2 ns Substance use .014 9,366 2 ns

NOTE: Zr = weighted mean effect size, k = number of unique samples, nx = the number of participants with null results needed to make finding nonsignificant. *p < .01. **p < .001.

DISCUSSION

The major goal of this meta-analysis was to identify the risk factors most strongly associated with reoffense risk in juveniles. The term predictor variables in this context is used in a methodological sense. When conducting meta-analytic research and regression-based statis- tical analyses, researchers typically examine the relationships between predictor and outcome variables. Using the term predictor variable does not necessarily imply that the variables used in these studies, either singly or in combination, will accurately predict which juvenile delinquents will reoffend.

The results show the strongest individual predictors to be a younger age at first commitment, younger age at first contact with the law, and history of nonsevere pathology. In a broader sense, the domains of offense history and family and social factors were consistently associ- ated with recidivism, whereas other domains contained certain vari- ables that were significant but were less consistent across the entire domain.

As noted earlier, the sample of participants in this analysis is con- siderably more homogeneous than it tends to be in delinquency research with first-time or nonoffenders. The present meta-analysis sample consisted entirely of adolescents who had already been adjudi- cated delinquent at least once. This may account for some of the results, including the low correlations between recidivism and vari- ables such as substance use, school attendance and achievement, and history of treatment. Given the nature of the sample in this analysis, it is likely that a substantial proportion of the participants have a history of substance use and academic problems, as well as prior experience with treatment programs. Therefore, such variables may have dimin- ished discriminative power because of restricted range.

Some of the variables identified as significant predictors of recidi- vism can be considered static because they are not subject to change through planned intervention. Nonetheless, static variables are poten- tially useful in the a priori identification of juvenile recidivism risk. Demographic information provides one domain of static predictors considered in this study. Among these, male gender and lower socio- economic status were positively related to recidivism. Additional static risk factors for recidivism include an earlier age of onset of

386 CRIMINAL JUSTICE AND BEHAVIOR

offending, more arrests and commitments, longer incarcerations, and more serious types of offenses. Static variables from the family and social domain include a history of physical or sexual abuse, being raised in a single-parent home, and having a greater number of out-of-home placements. Finally, static predictors from among those in the educational and testing domains include a history of being in special education classes, and lower standardized achievement, full- scale IQ, and verbal scale IQ scores.

We also identified variables that can be considered dynamic, with a potential to change through planned intervention. Assuming such potential would suggest that such dynamic variables could be targeted in risk-reduction intervention planning. Among the predictors identi- fied in this analysis, variables in the family and social domain may be considered dynamic risk factors, including family instability and problematic interactions, association with delinquent peers, and poor use of leisure time. Conduct problems, nonsevere pathologies, and substance abuse (but not necessarily substance use) can also be con- sidered dynamic. Similarly, achievement test scores, one marker of educational performance, also presented as potential dynamic risk factors that are inversely related to recidivism.

However promising this may appear theoretically, it is important to note that it has not been well-tested empirically. This relative absence of empirical support may be partly attributed to the difficulty in imple- menting such a study, which would involve controlling a potentially large number of dynamic risk factors that would vary across subjects, and providing interventions that would also vary widely. Such prob- lems are not insurmountable—they may be addressed through a com- bination of studies targeting single, well-controlled variables and specific interventions with other broader studies addressing the effective-ness of a programmatic approach that targets and addresses all identified dynamic risk factors.

Considering this need, the demarcation of both static and dynamic risk factors for this population ( j uveniles who have already been arrested at least once) may promote the development of risk assess- ment tools that would allow both a priori classification of risk, and the targeting of dynamic areas for risk-reducing intervention planning. This is the structure that has been used with the HCR-20 (Webster, Douglas, Eaves, & Hart, 1997), a risk assessment tool for adults that is

Cottle et al. / JUVENILE RECIDIVISM 387

beginning to be validated for predictive accuracy (Douglas, Ogloff, Nicholls, & Grant, 1999; Douglas & Webster, 1999). It is also similar to the structure of a combined risk/needs assessment tool for juve- niles, such as the Youth Level of Service/Case Management Inventory (YSL/CMI) (Hoge & Andrews, 1994), an adaptation of the Level of Service Inventory for adults (Andrews & Bonta, 1995). Finally, it is comparable to another risk tool that is under development for assess- ing violence risk for boys younger than the age of 12 (the Early Assessment Risk List for Boys, or EARL-20B) (Augimeri, Webster, Keogl, & Levene, 1998). The present findings underscore the value of this structure and the consistency of risk-relevant domains supported in the present study with the domains of the YLS/CMI (offense his- tory, family, education, peers, substance abuse, leisure and recreation, and personality/behavior).

Some of the dynamic risk factors found in this study have implica- tions for the rehabilitation of juvenile offenders who have been arrested for at least one prior offense. The family variables are consis- tent with a growing body of evidence suggesting that multisystemic treatment, delivered in a family context, is effective in reducing juve- nile recidivism (Henggeler et al., 1997; Henggeler, Schoenwald, Borduin, Rowland, & Cunningham, 1998). Interventions delivered in residential placement should be structured around the risk factors presently supported (family, conduct problems and nonsevere pathol- ogies, substance abuse, and educational achievement) for maximum risk-reduction impact. It might also be helpful to distinguish between high-risk versus low-risk juveniles using such a tool because the inten- sity and areas of rehabilitation might be different.

The present results may also have implications for public policy. The accurate identification of higher risk individuals and the ongoing assessment of changing risk status could be useful for decision makers in program planning, resource allocation, and legislation and policy affecting juveniles. In particular, an improved understanding of the impact of interventions through the pathways of juvenile careers could provide an important means for optimally effective allocation of resources for intervention efforts. Such results may help to identify strategies that are most likely to be involved in the deterrence of per- sistent criminal conduct, as well as those that may have less impact on juvenile recidivism.

388 CRIMINAL JUSTICE AND BEHAVIOR

There are several important limitations to the present findings. First, our review identified a relatively small number of prospective studies that assessed predictors of recidivism among juveniles. This was particularly true for studies using specific outcomes, such as vio- lent or sexual reoffending. A recent meta-analysis of the predictors of recidivism in adult offenders (Bonta et al., 1998) distinguished between violent, sexual, and general recidivism. In the present study, however, this distinction could not be made because of the limited number of studies using violent and sexual recidivism. More research is needed using these outcomes before a meta-analysis can be per- formed. To accomplish this, it would be useful for researchers to dis- tinguish between the outcomes of violent offending (nonsexual crimes against persons), sexual offending (sexual crimes against per- sons), and other offenses. This would allow specific outcomes to be assessed and would also permit their combination into larger units, such as all violent offenses (including sexual), or all offenses (all categories).

Another problem with the limited number of prospective, empirical studies in this area concerns the commonality of the predictors employed across studies. Some variables may have failed to achieve statistical significance because of the small number of studies that had investigated them or the limited power resulting from the low number of total participants across studies. For example, only two studies reported the relationship between effective use of leisure time and recidivism. If more studies representing a more diverse population of offenders had been performed using this variable, it would have been possible to provide a more accurate estimate of the strength of this predictor.

A related concern is the limited information provided in published studies. Journals may fail to publish nonsignificant results or report results in a way that does not allow meta-analyses to compare their findings with those of other studies (e.g., reporting beta weights). Although every attempt was made to locate the actual effect sizes in nonsignificant findings and to calculate effect sizes in comparable form, there were some studies that could not be included in the present analysis for these reasons. Similarly, despite vigorous attempts to locate relevant dissertations and unpublished studies (including care- ful searches of various databases and attempts to obtain unpublished

Cottle et al. / JUVENILE RECIDIVISM 389

data from a number of researchers), such data could not be located. Although the file-drawer problem was assessed, these omissions may have affected the present findings.

Several inconsistencies regarding substance use and substance abuse have not been resolved. Only two studies examining substance use as a predictor variable were included in the analysis (Dembo et al., 1998; Towbermann, 1994). However, only one of the two (Dembo et al., 1998) actually reported that substance use was not a significant predictor. Our finding of overall nonsignificance is a consequence of weighting the respective effect sizes of the two studies, resulting in the study with the larger number of participants (Dembo et al., 1998) out- weighing the results of the other study (Towberman, 1994). One explanation for the inconsistent findings between these two studies may involve the differing methodologies used. Towberman (1994) focused on whether juveniles used drugs at the time of their offenses. Others (e.g., Dembo et al., 1994), in contrast, considered lifetime use of various substances. Further research may clarify these discrepan- cies, particularly if substance use and substance abuse can be defined more consistently across investigations.

Some of these limitations may affect the overall significance of the results in this meta-analysis. The small number of studies examining predictors such as history of treatment, out-of-home placements, and substance use places substantial weight on sample size, methodologi- cal differences, and similar factors that may vary across studies. Inclu- sion of unpublished data and relevant future research would likely increase the power of this analysis.

A better understanding of factors associated with juvenile recidi- vism may improve efforts to distinguish between juveniles whose offending is limited to adolescence and those for whom such behavior is more persistent. The current findings suggest that age of onset is an important temporal variable that increases the risk for subsequent offending among juveniles. Two age-related variables—age at first arrest and age at first contact with the law—were among those with the strongest predictive relationship to recidivism. The integration of such data with tools designed to assess risk among younger children may allow us to identify more precisely the nature and level of risk associ- ated with preadolescent children with conduct problems.

390 CRIMINAL JUSTICE AND BEHAVIOR

Finally, the potential usefulness of longitudinal research is clear. Previous meta-analyses with adult offenders have yielded separate risk factors for general, violent, and sexual recidivism. It would be useful to be able to compare the risk factors of these groups across the life span, but this cannot be done until further research is performed with children and adolescents in these areas. Such research would also allow a better understanding and more accurate identification of chil- dren and adolescents whose antisocial behavior persists through adulthood, distinguished from those who desist from such behavior in early adulthood.

NOTES

1. A more easily computed alternative, based on counting rather than on adding Z scores, is also available. This method is more versatile, because it can be used when the exact p values are not known, but may also be less powerful. The number of unpublished studies needed to alter effect size estimates can also be computed (Rosenthal, 1991).

2. Although there may be a number of studies of sexual and violent recidivism, in many cases they were treatment studies and could not be used to identify risk factors for recidivism. Also, due to the heterogeneity of predictor variables between the articles, no meta-analysis could be conducted for violent or sexual recidivism.

REFERENCES

Anderson, L. & Walsh, J. (1998). Level of service inventory–revised. Toronto, Canada: Multi Health Systems.

Andrews, D., & Bonta, J. (1995). Level of Service Inventory-Revised. Toronto, Canada: Multi Health Systems.

*Archwamety, T., & Katsiyannis, A. (1998). Factors related to recidivism among delinquent females at a state correctional facility. Journal of Child and Family Studies, 7, 59-67.

*Ashford, J. B., & LeCroy, C. W. (1988). Predicting recidivism: An evaluation of the Wisconsin juvenile probation and aftercare risk instrument. Criminal Justice and Behavior, 15, 141-151.

*Ashford, J. B., & LeCroy, C. W. (1990). Juvenile recidivism: A comparison of three prediction instruments. Adolescence, 25, 441-450.

Augimeri, L., Webster, C., Keogl, C., & Levene, K. (1998). Early Assessment Risk List for Boys (EARL-20B), Version 1. Toronto, Canada: Earlscourt Child and Family Centre.

Bennett, R., & Basiotis, P. (1991). Structural correlates of juvenile property crime: A cross- national, time-series analysis. Journal of Research in Crime and Delinquency, 28, 262-288.

*Bleker, E. (1983). Cognitive defense style and WISC-R PV sign in juvenile recidivists. Journal of Clinical Psychology, 39, 1030-1032.

Cottle et al. / JUVENILE RECIDIVISM 391

Bonta, J., Law, M., & Hanson, K. (1998). The prediction of criminal and violent recidivism among mentally disordered offenders: A meta-analysis. Psychological Bulletin, 123, 123-142.

Census Bureau (2000). Resident population estimates of the United States by age and sex: April 1, 1990 to July 1, 1999, with short-term projection to July 1, 2000. Washington, DC: Author. Available: http://www.census.gov/.

*Dembo, R., Schmeidler, J., Nini-Gough, B., Sue, C. C., Borden, P., & Manning, D. (1998). Pre- dictors of recidivism to a juvenile assessment center: A three year study. Journal of Child and Adolescent Substance Abuse, 7, 57-77.

Dembo, R., Turner, G., Sue, C. C., Schmeidler, J., Bordon, P., & Manning, D. (1995). Predictors of recidivism to a juvenile assessment center. International Journal of the Addictions, 30, 1425-1452.

*Dembo, R., Williams, L., Fagan, J., & Schmeidler, J. (1994). Development and assessment of a classification of high risk youths. Journal of Drug Issues, 24, 25-54.

Dembo, R., Williams, L., Schmeidler, J., Getreu, A., & Berry, E. (1991). Recidivism among high risk youths: A 2 1/2-year follow-up of a cohort of juvenile detainees. International Journal of the Addictions, 26, 1197-1221.

Douglas, K., Ogloff, J.R.P., Nicholls, T., & Grant, I. (1999). Assessing risk for violence among psychiatric patients: The HCR-20 violence risk assessment scheme and the psychopathy checklist: Screening version. Journal of Consulting and Clinical Psychology, 67, 917-930.

Douglas, K., & Webster, C. (1999). The HCR-20 violence risk assessment scheme: Concurrent validity in a sample of incarcerated offenders. Criminal Justice and Behavior, 26, 3-19.

*Duncan, R., Kennedy, W., & Patrick, C. (1995). Four-factor model of recidivism in male juve- nile offenders. Journal of Clinical Child Psychology, 24, 250-257.

Durham, J. L. (1996). Crime in America. Santa Barbara, CA: ABC-CLIO. Federal Bureau of Investigation. (1998). Crime in the United States 1998. Washington, DC:

Department of Justice, Federal Bureau of Investigation. *Funk, S. (1999). Risk assessment for juveniles on probation. Criminal Justice and Behavior, 26,

44-68. *Grenier, C. E., & Roundtree, G. A. (1987). Predicting recidivism among adjudicated delin-

quents: A model to identify high risk offenders. Journal of Offender Counseling, Services, and Rehabilitation, 12, 101-112.

*Hanson, C. L., Henggeler, S. W., Haefele, W. F., & Rodick, J. D. (1984). Demographic, individ- ual, and family relationship correlates of serious and repeated crime among adolescents and their siblings. Journal of Consulting and Clinical Psychology, 52, 528-538.

Hanson, R. K., & Bussiere, M. T. (1998). Predicting relapse: A meta-analysis of sexual offender recidivism studies. Journal of Consulting and Clinical Psychology, 66, 348-362.

Henggeler, S., Melton, G., Brondino, M., Scherer, D., & Hanley, J. (1997). Multisystemic ther- apy with violent and chronic juvenile offenders and their families: The role of treatment fidelity in successful dissemination. Journal of Consulting and Clinical Psychology, 65, 821-833.

Henggeler, S. W., Schoenwald, S. K., Borduin, C. M., Rowland, M. D., & Cunningham, P. B. (1998). Multisystemic treatment of antisocial behavior in youth. New York: Guilford.

Hoge, R. D., & Andrews, D. A. (1994). The youth level of service/case management: Inventory and manual. Ottawa, Ontario: Department of Psychology, Carleton University.

*Hoge, R. D., Andrews, D. A., & Leschied, A. W. (1996). An investigation of risk and protective factors in a sample of youthful offenders. Journal of Child Psychology and Psychiatry, 37, 419-424.

392 CRIMINAL JUSTICE AND BEHAVIOR

*Jung, S., & Rawana, E. P. (1999). Risk and need assessment of juvenile offenders. Criminal Jus- tice and Behavior, 26, 69-90.

*Katsiyannis, A., & Archwamety, T. (1997). Factors related to recidivism among delinquent youths in a state correctional facility. Journal of Child and Family Studies, 6, 43-55.

Lidz, C. W., Mulvey, E. P., & Gardner, W. (1993). The accuracy of predictions of violence to oth- ers. Journal of the American Medical Association, 269, 1007-1111.

Lipsey, M. (1992). Juvenile delinquency treatment: A meta-analytic inquiry into the variability of effects. In T. Cook, H. Cooper, D. Cordray, H. Hartmann, L. Hedges, R. Light, T. Louis, & F. Mosteller (Eds.), Meta-analysis for explanation: A casebook (pp. 83-127). New York: Russell Sage Foundation.

Lipsey, M. (1999). Can interventions rehabilitate serious delinquents? Annals of the American Academy Political and Social Sciences, 564, 142-166.

Loeber, R., & Dishion, T. (1983). Early predictors of male delinquency: A review. Psychological Bulletin, 94, 68-99.

Loeber, R., & Stouthamer-Loeber, M. (1986). Family factors as correlates and predictors of juvenile conduct problems and delinquency. In M. Tonry & N. Morris (Eds.), Crime and jus- tice: An annual review of research (Vol. 7, pp. 29-149). Chicago: University of Chicago Press.

*Minor, K. I., Hartmann, D. J., & Terry, S. (1997). Predictors of juvenile court actions and recidi- vism. Crime and Delinquency, 43, 328-344.

Moffitt, T. E. (1993). Adolescent-limited and life-course-persistent antisocial behavior: A devel- opmental taxonomy. Psychological Review, 100, 674-701.

*Myner, J., Santman, J., Cappelletty, G., & Perlmutter, B. (1998). Variables related to recidivism among juvenile offenders. International Journal of Offender Therapy and Comparative Criminology, 42, 65-80.

*Niarhos, F. J., & Routh, D. K. (1992). The role of clinical assessment in the juvenile court: Pre- dictors of juvenile dispositions and recidivism. Journal of Clinical Child Psychology, 21, 151-159.

Office of Juvenile Justice and Delinquency Prevention Statistical Briefing Book. (1999). Data Source: Arrest data for 1981-1994 from unpublished data from the Federal Bureau of Inves- tigation and Crime in the United States reports for 1995, 1996, 1997, and 1998. Washington, DC: Government Printing Office, 1996, 1997, 1998, and 1999, respectively [Online]. Avail- able: http://ojjdp.ncjrs.org/ojstatbb/quw54.html.

*Putnins, A. (1984). Family structure and juvenile recidivism. Family Therapy, 11, 61-64. *Repo, E., & Virkkunen, M. (1997). Young arsonists: History of conduct disorder, psychiatric

diagnoses and criminal recidivism. Journal of Forensic Psychiatry, 8, 311-320. Rosenthal, R. (1991). Meta-analytic procedures for social research. Newbury Park, CA: Sage. Roy, S. (1995). Juvenile restitution and recidivism in a midwestern county. Federal Probation,

59, 55-62. Simourd, L., & Andrews, D. (1994). Correlates of delinquency: A look at gender differences.

Forum on Corrections Research, 6, 26-31. *Stattin, H., & Magnusson, D. (1989). The role of early aggressive behavior in the frequency,

seriousness, and types of later crime. Journal of Consulting and Clinical Psychology, 57, 710-718.

Steadman, H. J., Mulvey, E., Monahan, J., Robbins, P., Appelbaum, P., Grisso, T., Roth, L., & Silver, E. (1998). Violence by people discharged from acute psychiatric inpatient facilities and by others in the same neighborhoods. Archives of General Psychiatry, 55, 1-9.

Cottle et al. / JUVENILE RECIDIVISM 393

*Steiner, H., Cauffman, E., & Duxbury, E. (1999). Personality traits in juvenile delinquents: Relation to criminal behavior and recidivism. Journal of the American Academy of Child and Adolescent Psychiatry, 38, 256-262.

Stock, W. A. (1994). Systematic coding for research synthesis. In H. Cooper & L. Hedges (Eds.), The handbook of research synthesis (pp. 125-138). New York: Russell Sage Foundation.

*Towberman, D. (1994). Psychosocial antecedents of chronic delinquency. In N. J. Pallone (Ed.), Young victims, young offenders: Current issues in policy and treatment (pp. 151-164). New York: Haworth.

Walters, J. P. (1996). The state of violent crime in America. Washington, DC: National Crime Prevention.

Webster, C., Douglas, K., Eaves, D., & Hart, S. (1997). HCR-20: Assessing risk for violence (Version 2). Burnaby, Canada: Mental Health, Law, and Policy Institute, Simon Fraser University.

Wechsler, D. (1981). Manual for the Wechsler Intelligence Scale for Children (3rd ed.) San Antonio, TX: The Psychological Corporation.

Whitehead, J. T., & Lab, S. P. (1989). A meta-analysis of juvenile correctional treatment. Jour- nal of Research in Crime and Delinquency, 26, 276-295.

*Wierson, M., & Forehand, R. (1995). Predicting recidivism in juvenile delinquents: The role of mental health diagnoses and the qualification of conclusions by race. Behaviour Research and Therapy, 33, 63-67.

394 CRIMINAL JUSTICE AND BEHAVIOR

View publication statsView publication stats