Most Juveniles Who Commit Violent Offenses Do Not Reoffend in Adulthood

Most Juveniles Who Commit Violent Offenses Do Not Reoffend in Adulthood

These findings may help inform new delinquency interventions that target the needs of a small group of violent adolescent offenders accounting for a large amount of violent crimes. This is the bottom line of a recently published article in Law and Human Behavior. Below is a summary of the research and findings as well as a translation of this research into practice.

Featured Article | Law and Human Behavior | 2017, Vol. 41, No. 3, 273-283

Violent Offending Among Juveniles: A 7-Year Longitudinal Study of Recidivism, Desistance, and Associations With Mental Health


Sascha Hein, University of Houston
Baptiste Barbot, Pace University and Yale University
Amanda Square and John Chapman, Yale University
Catherine Foley Geib, Connecticut Judicial Branch, Court Support Services Division
Elena L. Grigorenko, University of Houston


Serious and violent offending among juveniles is a consistent concern of researchers, practitioners, and policymakers, yet the development of violent offending remains poorly understood because of limited availability of relevant data, small sample sizes, and shortage of longitudinal data sets. This study analyzed developmental patterns of violent offending over 7 years in the complete population of court-referred youth in Connecticut between 2006 and 2012 (N 58,678; mean age at first offense 14.7 years). This unique dataset provided several key findings: First, results from a latent class growth analysis showed that violent crimes peaked at age 14–15, with high-rate adolescent offenders (3.7% of the sample) accounting for 31.9% of all violent offenses. Further, 74.2% of this group desisted from violent crimes in adulthood. Higher levels of self-reported anger/irritability slightly increased the odds of violent recidivism (odds ratio, OR 1.09), where higher levels of depression/anxiety depressed the odds (OR 0.89). The overrepresentation of males, non-Hispanic Black, and Hispanic youth among high-rate adolescence offenders were traceable through adolescence but not beyond the age of 18. Together, these finding may help to inform new delinquency interventions that target the needs of this proportionally small group of violent adolescent offenders accounting for a large amount of violent crimes.


juvenile delinquency, gender differences, ethnic differences, violent crimes, latent class growth analysis (LCGA)

Summary of the Research


“Juvenile delinquency is a common and troublesome phenomenon worldwide. In the United States, about 1.47 million juveniles were arrested in 2011, accounting for 12.7% of all violent crimes nationwide. Regarding the developmental patterns of criminal acts, the existence of an age-crime curve that peaks during adolescence has been well documented. The rate and severity of offenses occurring during this developmental period are particularly predictive of later offenses. Developmental criminological research has contributed to our understanding of many forms of acting-out and delinquent behaviors along the life-course, helping to describe their onset, patterns of continuity and extinction. From this literature, it is evident that not all forms of juvenile delinquency lead to an adult criminal career and that the adolescent-crime “peak” represents partly a developmentally grounded phenomenon. This peak is characterized by an overall increase of all types of crimes, including violent crimes, as aggressive and antisocial behaviors may be manifested in multiple new contexts in adolescence. The increase in violent crime in adolescence results from both an increase of crimes for youth who have had a history of violence and aggression before adolescence, and the onset of violent crimes in youth without a history of violent behavior” (p. 273-274).
“Despite this evident increase in adolescence, violent offenses among juveniles are rare and mainly committed by a small group of frequent and chronic offenders. Violent offenses are particularly concerning because they represent an overt pathway toward serious delinquency including violent and chronic offending during adolescence, and increase the likelihood of later adult offenses and arrests. Violent and serious offending in adulthood is associated with an earlier age of onset of criminal behavior, extensive criminal careers, and male gender and African American race, among other factors” (p. 274).
“Mental health problems are highly prevalent in youth involved in the juvenile justice system, with rates of any psychiatric disorders of 60–70% compared with 20% in community samples. Externalizing (e.g., conduct disorder, oppositional defiant disorder, and attention deficit hyperactivity disorder) and internalizing (e.g., major depression, anxiety) problems, as well as posttraumatic stress disorder symptom severity and psychotic symptoms are among the disorders found to be more prevalent. The substantial mental health needs in this group suggests that there is an association between certain mental health disorders and criminality, as well as between psychopathology and the seriousness of recidivism” (p. 274).

Current Study

“The current study had three aims: (a) to investigate the frequency of violent criminal offenses in a complete population of court-involved juveniles over the course of 7 years; (b) to examine whether gender, race and ethnicity, age at first offense, and mental health factors explain differences between one-time and repeated violent offenders; (c) to ascertain the developmental course of violent offending over time, identifying subgroups of juveniles with varying trajectories, and to link these trajectories to juveniles’ gender, race and ethnicity, and mental health” (p. 274).
“We analyzed a longitudinal statewide dataset of every criminal case record of all youth (N = 58,678) who committed at least one offense between January 1, 2006, and December 31, 2012. To count and compare the number of offenses across different offense types, we included all 58,678 individuals in the analysis. However, to model age trends, we focused on youth who committed their first offense between age 10 and 17 years (n = 57,847) because of the rarity of violent offenses in young children” (p. 274).
“Broadly defined, violent offenses included all crimes against persons, specifically, criminal homicide, robbery, assault, violent sex offenses, and other person offenses (e.g., reckless endangerment, kidnapping, and strangulation). Accordingly, we first grouped offenses committed between the age of 10 and 17 into five categories to derive a count of five types of violent offenses: homicide, robbery, assault, violent sexual offenses, and other crimes against persons (e.g., reckless endangerment, strangulation, and threatening). For comparison, we counted all offenses in the categories of status offenses, property offenses, drug law violations, and crimes against public order” (p. 274-275).
“The Massachusetts Youth Screening Instrument—Version 2 (MAYSI-2) data were available for a subsample of 10,105 juveniles. It consists of seven scales including Alcohol/ Drug Use, Angry-Irritable, Depressed-Anxious, Somatic Com- plaints, Suicide Ideation, Thought Disturbance, and Traumatic Experiences. Threshold scores (with score ranges designated as “Caution” and “Warning”) assist in the identification of youth with potential mental health problems in need of additional psychiatric evaluation, immediate attention, and possible intervention” (p. 275).


“Assault was the most frequently committed violent offense, while homicide was the least prevalent violent offense. Among other offense types, crimes against public order were the most common offense and drug law violations were the least common. Regarding the patterns of repeated offending, the highest percentage of juveniles committing two or more offenses was observed for crimes against public order, while 13.95% of the sample committed two or more violent offenses. Overall, there were 16,809 juveniles who committed a total of 25,508 violent offenses. In this sample, the prevalence of offenses increases with age across all types with a peak at age 15, followed by a steep decrease at age 16 and 17. Together, 14- and 15-year-olds account for about half of all offenses committed by 10- to 17-year old juveniles” (p. 276).
“The offender profiles across the five offense types were analyzed with a zero-inflated Poisson LCA. Results showed that a three-class model fit best to the data. On average, youth in Class 1 (“property offenders”) showed elevated levels of violent offenses and crimes against public order compared with Class 2, which had the highest mean level for property offenses and drug law violations among all three classes, and a similarly low mean level of status offenses as the other classes. Juveniles in Class 2 (“low-level offenders”) represent the largest group, characterized by overall low mean levels across the five offense types. Juveniles in Class 3 (“violent offenders”) showed the highest mean levels of violent offenses and crimes against public order in the sample, had a higher level of property offenses than low-level offenders, and showed overall low levels of status offenses and drug law violations. Males were overrepresented among property offenders as well as among violent offenders whereas gender was more balanced in low-level offenders. Race and ethnicity differences were observed across classes with Black non-Hispanic juveniles being overrepresented among violent offenders and property offenders. Hispanic youth were also overrepresented among violent offenders” (p. 276-277).
“Additional analyses presented here were conducted on the subsample of individuals with available MAYSI data. The sample was characterized by an overall low level of alcohol and drug use, compared with higher levels on the angry-irritable, depressed-anxious and somatic complaints subscales. Zero-inflated Poisson regression models were utilized to predict the number of violent offenses using demographic variables as covariates (gender, race/ethnicity, and age at first offense) and the MAYSI scores as the focal variables. One-unit increases in scores on the alcohol/drug use subscale of the MAYSI increases the count of violent offenses by 1.06 units. Somatic complaints decreased the number of violent offenses. Regarding the demographic variables, results showed that the count of violent offenses for Black, non-Hispanic and Hispanic juveniles increases by 1.63 and 1.34, respectively, compared with White, non-Hispanic juveniles. Moreover, for each unit (i.e., 2.4 months) decrease in the age at first offense, the log count of violent offenses increases by 1.18” (p. 277-278).

“Finally, a logistic regression analysis was conducted to identify factors that discriminate one-time violent offenders from recidivists. Regarding the MAYSI subscales, one unit increase in scores on the angry-irritable scale increase the log odds of reoffending by 0.10, whereas one unit increases in the depressed-anxious scale and the somatic complaints scale decreased the log odds of recidivism by 0.10 and 0.04 units. Race and ethnicity also contributed significantly to the prediction of group probabilities. Specifically, the odds of Black, non-Hispanic juveniles recidivating with a violent offense were 2.14 times higher as compared with White, non-Hispanic youth. Similarly, the odds of Hispanic juveniles recidivating were 1.50 times higher as compared with White, non-Hispanic youth. Moreover, juveniles who were younger at their first offense were more likely to be violent reoffenders. One unit decrease in the age at first offense increased the log odds of violent reoffending by 0.317” (p. 278).
“The next set of analyses identified two classes of juveniles who qualitatively differed in their mean growth of violent offending between the age of 10 and 17. Juveniles in Class 1 are characterized by an evident peak of violent offending in adolescence (age 14–15), followed by a decline in violent offending at the age of 16 and 17. Although the class of high-rate adolescence peak juveniles constitutes a small proportion of the overall sample, they accounted for 31.9% of all violent offenses recorded for juveniles in the age of 10 to 17. The majority of youth followed a pattern that can be described with a low peak of violent offending in adolescence (age 14–15)” (p. 279).

“Small differences between both classes were observed regarding the levels of mental health factors in the subsample of juveniles with available data. High-rate adolescence peak offenders had higher scores on the alcohol/drug use subscale compared with low-rate adolescence peak offenders. However, compared with high-rate adolescence peak offenders, low-rate adolescence peak offenders had higher scores on the depressed-anxious subscale, higher scores on the somatic complaints subscale, and higher scores on the suicide ideation subscale. Males were overrepresented among high-rate adolescence peak offenders compared with low-rate adolescence peak offenders. Regarding race and ethnicity, Black, non-Hispanic juveniles were overrepresented among high-rate adolescence peak offenders compared to low-rate adolescence peak offenders, as were Hispanic youth” (p. 279-280).
“As apparent from the decline in the age-crime curve after age 15, not all high-rate adolescence peak offenders continue to commit violent offenses in adulthood. Although these juveniles were almost three times as likely to commit a violent offense beyond age 18 compared with low-rate adolescence peak offenders, this finding highlights that the majority of youth did not reoffend as an adult. To further elucidate the differences between violent reoffenders based on these trajectories, we compared two subgroups of high-rate adolescence peak offenders with regard to the MAYSI-2 variables, gender, race and ethnicity, age at first offense, and age at first violent offense: youth who committed at least one violent offense as an adult and youth who did not. Results showed no differences between high-rate adolescence peak offenders who recidivate as an adult and those who do not with regard to gender and mental health. Results showed no significant race and ethnicity differences between violent reoffenders in adulthood and desisting juveniles. Taken together, the overrepresentation of males, Black, non-Hispanic, and Hispanic youth in the class with a peak of violent offending in adolescence were traceable up to adulthood but not beyond the age of 18 among high-peak adolescence offenders” (p. 280).

Translating Research into Practice

“Consistent with prior evidences in the literature, the descriptive analysis of the violent crimes committed by juveniles in CT during the years 2006–2012 clearly confirms a “peak” of offenses around age 15. Moreover, the frequency of all types of crime appears to increase noticeably from early adolescence to age 15, and concerns all types of crimes including violent crimes. After the peak at age 15, the prevalence of all offenses reduces at or below the level of 14-year-olds, except for the drug offenses, which is the only offense type that is not associated with a decrease in prevalence after age 15” (p. 280).
“A closer look into the types of violent offenses further elucidates these developmental trends: Assault accounts for most of the increase of violent offenses during the adolescence “peak,” while sex offenses and homicides remain extremely rare with a relatively stable prevalence across age groups. This result mirrors longitudinal findings showing that trajectories of physical aggression are particularly unstable, with a common temporary elevation during “normative” adolescence turmoil. Together, these results are consistent with the fact that some of the developmental changes in the peak of the adolescence “crisis” place youth at risk for externalizing behaviors. These changes include neurobiological changes, psychosocial changes, contextual changes, as well as increased susceptibility to deviant peers influence” (p. 280).

“Adding to the body of research showing the relationship between mental health and juvenile delinquency, the present study elicited differences between high-rate adolescence peak offenders characterized with higher anger-irritability and alcohol and drug use, and low-rate adolescence peak offenders, showing higher depression-anxious, somatic complaints and suicidal ideation. Similarly, anger-irritability increased the odds of reoffending with another violent crime, whereas depression/anxiety and somatic complaints decreased such odds. The literature suggests that clinical depression is associated with an increased risk of violent crime. However, the literature examining the association between psychiatric traits and violent crime is limited. It is plausible that an individual’s ability to feel remorse, guilt, and worry about his or her performance might reflect normative reactions to the committed offense and, therefore, decrease the odds of reoffending. Another explanation for this difference is that violent crimes may be underreported. Therefore, the difference may pertain only to those charged with violent reoffending rather than those who have engaged in repeated violent criminal behaviors. Furthermore, although individuals with higher levels of mental health needs may commit crimes at disproportionate rates, the offenses may not be directly related to disorder symptoms but stemming from third factors and needs such as homelessness, poverty and unemployment that result in less severe crimes” (p. 280).

“Regarding gender, racial, and ethnicity differences, it is well acknowledged that males commit violent crime at higher rates than do females. In this sample, males were overrepresented among violent and property offenders, as well as in the class characterized by the peak of violent offending in adolescence. However, the count of violent offenses was not significantly associated with gender. A possible explanation for this finding is that the gender gap in violent offending has narrowed over time in youths, except for high-level adolescent offenders. Another explanation is that, although arrests for both male and female youth have decreased in recent years, the rates of decreases are lower for females than for males. In this sample, Black, non-Hispanic, and Hispanic juveniles were overrepresented among violent and property offenders as compared with White, non-Hispanic juveniles. This overrepresentation of minority youth, particularly for violent crimes, is in accordance with national data” (p. 280-281).

“Additionally, in this analysis, Black and Hispanic racial and ethnic backgrounds were found to be among the significant factors in discriminating one-time violent offenders from recidivists, which is in direct contrast with the findings from a recent meta-analysis. Reasons for this discrepancy include the use of adult offender samples in most of the studies, different indicators of violent behavior than the ones used in the present study, and follow-up periods to measure recidivism that ranged from 6 months to 10 years. Yet, the link between violent reoffending and race and ethnicity may be attributable to factors not considered in this analysis. For instance, other research studies have found that the association between juvenile recidivism in general and non-White race does not remain significant after addressing variables such as SES. In our analyses, we controlled for an imperfect measure of SES as operationalized by eligibility for free and reduced lunch in school. Arguably, other indicators of SES (e.g., family income) should be taken into account. Moreover, the associations between race/ethnicity and violent offending might represent an accumulation of risks located at multiple contextual levels, as differences in offending between Whites and minority groups have been linked to community disadvantage, residential inequality, and social cohesion in community contexts” (p. 280-281).

Other Interesting Tidbits for Researchers and Clinicians

“Violent offenses constitute an evident behavior among other types of delinquent and deviant behavior. However, repeated violent offending remains a rare phenomenon, hampering the robust inference of the developmental course of violent offending across the formative years of adolescence. Despite the potential underestimation of violent offending because of the use of administrative data, the study presented here is unique in that it used a statewide longitudinal dataset that encompasses the entire population of court-involved youth, allowing for the investigation of the magnitude and the age-patterns of violent offending in conjunction with mental health needs of more than 10,000 individuals. Contrary to studies that have used selected samples of juvenile and outlined a great diversity of (nonviolent) offense trajectories in adolescence, this study elicited for the first time the clear distinction between two classes of violent offenders (high-rate vs. low-rate)” (p. 281).

“The existence of a group of few violent offenders (emerging as early as 12 years of age, and peaking between the age of 14 –15 years) that accounts for a large proportion of violent offenses highlights the implications for appropriate delinquency interventions for different subgroups of youth. This is particularly important given that many high-risk juvenile offenders do not receive an evidence-based intervention such as multisystemic therapy. Thus, there is a need to maximize the effort to differentiate and tailor interventions to meet specific needs of violent offenders. Moreover, given that some youth persist with violent behavior into adulthood, it is important to examine the long-term effects of such interventions and to determine potential mechanisms of change. The present study has offered some preliminary insights toward this endeavor, by eliciting key demographics, mental health conditions and other individual factors associated with heightened risk for the most serious forms of criminal careers” (p. 281).

Join the Discussion

As always, please join the discussion below if you have thoughts or comments to add!

Authored by Becca Cheiffetz

Becca Cheiffetz is a master’s student in the Forensic Psychology program at John Jay College of Criminal Justice. She graduated in 2015 from Sam Houston State University with a BS in Psychology and plans to continue her studies in a Clinical/Forensic Psychology PhD program in the near future. Her professional interests include providing clinical evaluations and treatment for individuals in prison as a prison psychologist and conducting forensic assessments for defendants in criminal court.

This site uses Akismet to reduce spam. Learn how your comment data is processed.

Leave a Reply

Your email address will not be published. Required fields are marked *

This site uses Akismet to reduce spam. Learn how your comment data is processed.