Prediction of Recidivism With the Youth Level of Service/Case Management Inventory (Reduced Version) in a Sample of Young Spanish Offenders

Abstract

Intervention in youth recidivism is critical in helping prevent young people from continuing their criminal career into adulthood, on a life-course-persistent trajectory. Andrews and Bonta attempt to provide an explanation of risk and protective factors using a conversion of the Youth Level of Service/Case Management Inventory (YLS/CMI), which predicts recidivism. In this study, scores have been obtained from 382 adolescents (M age = 16.33 years) from the juvenile court, to check the ability of a reduced version of the YLS/CMI, to predict recidivism. The outcome variables for recidivism were examined in the 2-year follow-up period, after their first assessment in the court. The risk factors showed good levels of recidivism prediction. Recidivists obtained significant higher mean total risk scores than nonrecidivists in the reduced (M = 6.54, SD = 2.44; M = 3.66, SD = 2.85), with areas under the curve (AUCs) ranging from .601 to .857. The factors that emerged as the most discriminative were education/employment, criminal friends, and personality. All the protective factors differentiated between recidivists and nonrecidivists. The results, therefore, showed that this reduced version would be capable of predicting youth recidivism in a reliable way.

Keywords

recidivism Spanish youth offenders YLS/CMI reduced version

Predicting criminal behavior, and in particular juvenile crime, is an issue of major concern in today’s society. Although the general level of youth offending does not appear to have increased, there has been a rise in some violent crimes in recent years (Benavente, 2009; Capdevila, Ferrer, & Luque, 2005; Pérez, 2010; Puzzanchera & Adams, 2011; Smit & Bijleveld, 2015). Similarly, a large percentage of crimes are committed by a small percentage of juvenile offenders (Cottle, Lee, & Heilbrun, 2001). Furthermore, in the Spanish juvenile system, rates of recidivism for general youth offenders, during follow-up periods of 2 to 4 years, stand at between 26% and 40% (Acosta, Muñoz de Bustillo, Martín, Aragón, & Betancort, 2012; Bravo, Sierra, & Del Valle, 2009; Cuervo, Villanueva, & Prado-Gascó, 2017; García-España, García Pérez, Benítez Jiménez, & Pérez Jiménez, 2011; San Juan & Ocáriz, 2009). In this context, intervention in youth recidivism is critical to help prevent young offenders from continuing their criminal career into adulthood, on a life-course-persistent trajectory (Moffitt, 2006).

Detecting risk and protective factors has become crucial in preventing and reducing crime. A risk factor for offending is a variable that predicts a high probability of later offending (Farrington, Loeber, & Ttofi, 2012; Ribeaud & Eisner, 2010). Meanwhile, protective factors can be considered variables that predict a low probability of offending among persons exposed to risk factors. They have been related to desistance in different prospective longitudinal studies (Farrington et al., 2012; Hartman, Turner, Daigle, Exum, & Cullen, 2009; Ttofi, Farrington, Piquero, & DeLisi, 2016). Social learning theories (Catalano & Hawkins, 1996) try to structure the wide range of risk and protective factors in accordance with their theoretical assumptions. These theories are mainly based on the fact that behavior is interiorized through interaction with the environment, so criminal conduct will be more likely in youth who perceive more rewards for performing an antisocial activity than a prosocial one. One perspective of social learning theories attempts to provide an in-depth explanation of the theoretical frame of risk and protective factors through Andrews and Bonta’s (2010) general personality and social psychological model of criminal conduct. This model considers the individual as an agent that interacts with his or her environment, and who cannot be explained without this interactive, dynamic context. At the same time, in a wider context, the youth is also influenced by sex, age, or race (Shepherd, 2015; Zhang, 2016).

Besides these contextual variables, the model includes some individual factors, which are considered the best predictors of recidivism. These variables or factors were antisocial attitudes, antisocial friendships, an antisocial personality pattern, a history of previous offences (considered “the Big Four”), plus deficient family circumstances, education and employment, substance abuse, and free time for leisure and recreation. Taken together, these factors are referred as “the Central Eight” and are the same as those put forward by Hoge and Andrews (2006) in the Youth Level of Service/Case Management Inventory (YLS/CMI). This inventory enables young people to be classified as being at low, moderate, high, or very high risk of recidivism. Several studies show the validity of the inventory (Anderson et al., 2016; Catchpole & Gretton, 2003; Flores, Travis, & Latessa, 2004; Rennie & Dolan, 2010). Results from different meta-analyses support the predictive accuracy of the YLS/CMI even with female and distinct ethnic groups (Olver, Stockdale, & Wormith, 2009; Schwalbe, 2007). However, there is very little work carried out on Spanish justice populations and systems. There are some studies on the relation between offending and acculturation processes of Hispanics living in the United States (Jennings et al., 2016; Jennings, Zgoba, Piquero, & Reingle, 2013; Piquero, 2015), but not with Spanish youth offenders living in Spain.

Various Spanish studies have included this instrument to examine the risk factors that successfully discriminate between juvenile reoffenders and those who do not reoffend. For example, significant differences were found between recidivist and nonrecidivist youth in all the factors of the YLS/CMI, with the exception of education/employment and leisure (Garrido, López, Silva, López, & Molina, 2006). In other studies, the two factors that best predicted recidivism were past and current offences and substance abuse (Cuervo & Villanueva, 2015; Graña, Garrido, & González, 2006).

Given the evidence of the good results in discriminating between offenders and nonoffenders and in predictive validity (Chu et al., 2015; Schmidt, Hoge, & Gomes, 2005; Thompson & Putnins, 2003), a screening version of the instrument has been developed—the Youth Level of Service/Case Management Inventory (YLS/CMI; Hoge & Andrews, 2011). To our knowledge, there are few studies that use the shortened version of the Youth Level of Service/Case Management Inventory (YLS/CMI). In the first, the total YLS/CMI-SV score was significantly correlated with police contacts, the number of new criminal charges, and the number of new criminal convictions (Van de Ven, 2004). This study also showed that the total YLS/CMI-SV score was significantly correlated with the total YLS/CMI score, and concluded that the instrument was a reliable and valid assessment of risk and need factors. Significant differences according to sex were also found, males scoring significantly higher than females. Another more recent study carried out in Singapore revealed that the reduced version was significantly predictive of general, violent, and nonviolent recidivism for male young offenders, and a useful measure for assessing the levels of risk for male young offenders; whereas, the results for females were less clear (Chu, Yu, Lee, & Zeng, 2014). In 2010, the Irish Youth Justice Service approved the use of this instrument to identify the needs that young people present and to determine which young people may benefit most from project intervention (Irish Youth Justice Service, 2011).

Given that research on the YLS/CMI-SV and on Spanish youth offenders is currently limited, it would be very useful to have a brief Spanish version of this validated instrument. Information gathered using this tool could be used for screening out low-risk youth from others at greater risk, and affording more accuracy for case management planning in intervention (Hoge, 2005). Furthermore, to our knowledge, research on the use of the screening, specifically with serious young offenders groups, is also limited (Olver, Stockdale, & Wong, 2012). In this study, we, therefore, examine the predictive accuracy of a reduced version instrument for different recidivism outcomes, including offences both against people and against property.

The aim of this study is, therefore, to analyze the predictive validity of the conversion of the YLS/CMI in relation to youth recidivism, by examining its possible differential contribution over a follow-up period of 2 years, when most recidivist acts take place (Mulder, Brand, Bullens, & van Marle, 2011). The second objective is to explore the relationship between the specific risk and protective factors as assessed on the reduced version of YLS/CMI, and recidivistic outcomes as a function of gender, age, and nationality.

The objectives of the study led to the formulation of the following hypotheses.

Hypothesis 1: The reduced version of the YLS/CMI will predict recidivism and the number of criminal charges.

Hypothesis 2: The most effective risk predictors will be criminal history, antisocial peer group, antisocial attitudes, and antisocial personality, identified by Andrews and Bonta (2010) as the Big Four; and in the case of protective factors, these same factors would be related to lower rates of recidivism.

Method

Participants

The participants were all the juveniles with a disciplinary record in the juvenile court of a Spanish province in the period from March 2008 to May 2011. Data were gathered for all the charges in 135 municipalities, covering a total of 604,344 inhabitants. The study, therefore, included a wide range of young offenders, from those occasionally committing minor crimes, such as shoplifting, to those persistently committing serious crimes, such as sexual assaults. In total, there were 382 juveniles, aged 14.27 to 17.99 years. The average age was 16.33 years (SD = 1.04 years); 311 were male (81.4%) and 71 (18.6%) were female. In terms of nationality, the largest percentage (78.7%) were Spanish, followed by 8.4% of Romanian or other Eastern European nationalities, 6.5% from South American countries, and 6.3% from Arab countries. The distribution of crime on the sample was as follows: person-related offences, 52.1% (n = 184); and property-related offences, 47.9% (n = 169).

Instrument

The YLS/CMI (Garrido et al., 2006; Hoge & Andrews, 2006) is a recidivism risk inventory, which is compiled by the assessor through different sources of information such as interviews with parents, reports from the youth, observations. The inventory consists of 42 items grouped into eight risk factors. Each of the subscales is assessed at low, moderate, and high risk level, according to the YLS/CMI administration guidelines. The sum score of the eight factors provides a recidivism total risk level for each young person. The inventory also allows factors of strength (protective factors) to be recorded. Protective factors are considered not only when there is an absence of risk in a factor but also when there is an explicit presence of a positive factor. For example, illegal car racing must be a risk factor in the leisure area, and being part of a sports club could be considered a protective factor in the same area. It is possible to assess the minor with a protective factor on each scale, except for prior and current offences, because the positive factor here would be normative for all participants instead of protective. Hence, the maximum score of protective factors is seven.

The YLS/CMI-SRV (Hoge & Andrews, 2011) contains eight items for the eight risk/need domains of the YLS/CMI (Hoge & Andrews, 2006). This screening version was designed to identify young people at risk and conduct a preliminary assessment. It can be used for assisting in decisions about the level of supervision or intervention appropriate for the youth, and when there is a need of evaluating a large number of youth. It is primarily designed to be used in systems related to juvenile justice and correctional, where decisions must be made about further judicial processing or interventions. High-risk youth should be referred for a comprehensive risk assessment.

As for internal consistency, good results were found in previous studies (Cronbach’s α = .77 and .85; Van de Ven, 2004). In this study, the alpha value of the reduced version was .81. Construct validity was evaluated by correlating reduced version scores with parallel scores from the YLS/CMI. Satisfactory correlations were obtained between parallel scores from the two instruments. The total score from the two instruments were significantly correlated (r = .946, p < .001).

Procedure

The initial individual interviews to obtain a profile of the young person and information to complete the YLS/CMI were carried out by the juvenile court technical team. The interviews took place in the juvenile court offices around 3 to 6 months after charging. The technical team in the juvenile court was trained for 1 month in the use of the YLS/CMI, by an expert on the instrument. This team is administrating the YLS/CMI instrument since 2008 in their daily routine work with continuous supervision by the expert. The scores for the reduced version were obtained from the complete YLS/CMI, taking into account the risk level of each scale from the original inventory. Risk/need levels (low, moderate, or high) from each subscale are provided by the instrument (YLS/CMI), as stated before. It is important to point out that the screening version was not administered. Scores from the original YLS/CMI were converted to the reduced as follows: (a) history of conduct disorder (0 = 0, 1-5 = 1), (b) current school or employment problems (0 = 0, 1-7 = 1), (c) some criminal friends (0-1 = 0, 2-4 = 1), (d) alcohol/drug problems (0 = 0, 1-5 = 1), (e) leisure/recreation (0 = 0; 1-3 = 1), (f) personality/behavior (0 = 0, 1-7 = 1), (g) family circumstances/parenting (0 = 0, 1-2 = 1, 3-4 = 2, 5-6 = 3), (h) attitudes/orientation (0 = 0, 1 = 1, 2-3 = 2, 4-5 = 3). Ratings are then added for a total score, which ranges from 0 to 12 (1 point for each “yes” response; Hoge & Andrews, 2011; Van de Ven, 2004). Although the screening does not take into account protective factors, in this study, they have been scored in the same way as that of the original version.

The outcome variables for recidivism were measured in two different ways: dichotomously (reoffender/nonreoffender) and quantitatively (the number of subsequent charges). Recidivism was considered when the minor presented any type of charge after the first evaluation with the inventory. Both variables were examined in the 2-year follow-up period, after the initial assessment of each minor. There was a fixed follow-up period for each participant but this period started in different time lines.

Data Analysis

First, the descriptive analysis of gender and recidivism, related to the reduced version scores are presented. Next, predictive analyses are calculated. The area under the curve (AUC) statistic was selected due to its accuracy for predicting recidivism (Rice & Harris, 2005; Swets, Dawes, & Monahan, 2000). In this study, forward stepwise logistic regression (for recidivists/nonrecidivists) and negative binomial regression (for the number of criminal court contacts) were used to examine the predictive validity of the reduced version. These multivariate analyses attempted to determine whether recidivism differences persist when controlling for demographic characteristics and protective factors in the follow-up period.

Results

The ANOVA analysis with the total reduced version score and gender of the offenders revealed gender to be a significant factor; in other words, boys presented a higher risk of recidivism than girls (boys: M = 4.70, SD = 3.07; girls: M = 3.47, SD = 2.70), F(1, 375) = 9.57, p = .002. Similarly, significant differences were found when gender was related to each item. The percentages of risk factors for each reduced version item depending on sex are presented in Table 1. Males were found to be significantly more likely than females to present risk in their history of conduct disorders and current school or employment problems, and to have some criminal friends and alcohol/drug problems.

Table 1.

Distribution of Risk Reduced Version Inventory Percentages in Boys and Girls (N = 382).

Factors	Male %			Female %			Total	χ²	Significant
1. History of conduct disorder (0-1)	14.9			1.6			16.5	4.095	.043
2. Current school or employment problems (0-1)	66			13.1			79.1	3.928	.048
3. Some criminal friends (0-1)	44.2			6.5			50.8	8.463	.004
4. Alcohol/drug problems (0-1)	30.6			3.4			34%	9.601	.002
5. Leisure/recreation (0-1)	63.9			13.6			77.5	0.902	.342
6. Personality/behavior (0-1)	41.4			7.6			49%	2.294	.130
	1	2	3	1	2	3
7. Family circumstances/parenting (0-3)	27.5	16.5	1.8	5.2	2.6	0	53.7	5.754	.124
8. Attitudes/orientation (0-3)	14.1	17	2.6	2.6	2.9	0	39.3	4.852	.183

Regarding males and females on recidivism, statistically significant differences are observed with a value of χ²(1) = 4.011, p = .045. Boys recidivated in 28.5% of the cases, whereas girls did it on 16.9% of the cases. Boys also obtained higher mean of charges than girls (M = 0.71, SD = 1.59; M = 0.23, SD = 0.56), F(1, 375) = 6.31, p = .012.

The total mean score for risk factors assessed by the reduced version for the entire sample was 4.47 (SD = 3.04), ranging from 0 to 11 points. Meanwhile, reoffenders’ risk scores had a mean of 6.50 (SD = 2.43).

The results for recidivism showed that 99 out of the 376 minors had a further disciplinary charge in the juvenile court during the 2-year follow-up period. As a result, 26.3% (N = 99) were reoffenders (16.9% female and 28.5% male). Differences on the total risk scales depending on recidivism were observed. Recidivists obtained higher mean total risk scores than nonrecidivists (M = 6.54, SD = 2.44; M = 3.66, SD = 2.85), F(1, 37) = 80.38, p = .000.

The reduced version total risk score was also significantly correlated to the number of criminal charges during the follow-up period for the overall sample (N = 383; r = .339, p < .001) and the male subgroup (r = .337, p < .001), but the total risk score was not associated with future recidivism in the female subgroup (r = .162, n.s.). The correlation between the number of protective domains and the total reduced version risk score was significant (r = –.615, p < .001; N = 382).

According to the protective factors in each subscale, the total mean for juveniles was 1.05 (SD = 3.04), ranging from 0 to 7 points. Table 2 shows the distribution of protective factors in the sample (the protective factor in Area 1 is not considered by the inventory, because not having a criminal record does not denote strength). Boys had a higher percentage of protective factors than girls in some factors. The significant factors were good performance in school, having positive friends, and having productive leisure/recreation activities.

Table 2.

Percentage of Protective Factors in Boys and Girls (N = 383).

	Male %	Female %	Total%	χ²	Significant
Current school or employment problems	7.3	3.1	10.5	3.846	.050
Some criminal friends	6	3.7	9.7	10.034	.002
Alcohol/drug problems	10.2	3.7	13.9	2.492	.114
Leisure/recreation	16.8	7.6	24.3	12.896	.000
Personality/behavior	7.3	2.9	10.2	2.656	.103
Family circumstances/parenting	13.4	4.2	17.5	1.505	.220
Attitudes/orientation	14.7	4.7	19.4	1.997	.158

The fact of having a lower rate of protective factors was significantly related to recidivism. Recidivists had a mean score of 0.28; whereas, for nonrecidivists, it was 1.35 (F = 27.314, p = .000).

Predictive Statistics

Table 3 shows the AUC values for various recidivistic outcomes. For the overall sample, the AUC score predicting general recidivism using the reduced version total score was .775 (SE = 0.025, 95% confidence interval [CI] = [0.72, 0.83]). In the male group, recidivism was significantly predicted for offences against persons and property offences, whereas among girls, only property offences were predicted.

Table 3.

AUC Values for Recidivistic Outcomes Across the Entire Sample, as Well as Sex Subgroup and Type of Violent Offence (N = 383).

	Recidivism
	AUC (SE)	95% CI	p
Overall sample	.775 (0.025)	[0.726, 0.824]	.000
Male subgroup	.783 (0.027)	[0.731, 0.836]	.000
Property offences	.812 (0.035)	[0.744, 0.881]	.000
Against person offence	.772 (0.042)	[0.689, 0.855]	.000
Female subgroup	.666 (0.075)	[0.519, 0.813]	.071
Property offences	.857 (0.075)	[0.710, 1]	.015
Against person offence	.601 (0.099)	[0.406, 0.795]	.432

Note. AUC = area under the curve; CI = confidence interval.

As can be seen in Table 4, in the logistic regression analysis, demographic variables, type of crime, and reduced version total score were included in Model 1. The reduced version total score and age were significant in the prediction of recidivism, whereas sex, the type of crime committed, and nationality were not significant (Nagelkerke R² = .325). Model 2 also found that protective factors were significant in predicting recidivism (Nagelkerke R² = .228). However, the best model was obtained by introducing both variables—risks and protective factors—in Model 3 (Nagelkerke R² = .338). However, protective factors were no longer significant when analyzed with risk factors.

Table 4.

Logistic Regression Analysis of Recidivism in a Follow-Up Period of 2 Years.

	Model 1		Model 2		Model 3
	OR	CI 95% (B)	OR	CI 95% (B)	OR	CI 95% (B)
	OR	[LL, UL]	OR	[LL, UL]	OR	[LL, UL]
Age	0.61**	[0.46, 0.80]	0.65**	[0.50, 0.83]	0.60**	[0.46, 0.79]
Boy	1.57	[0.72, 3.43]	2.057	[0.96, 4.41]	1.59	[0.72, 3.51]
Spanish	0.62	[0.33, 1.18]	0.822	[0.45, 1.48]	0.67	[0.35, 1.27]
Against person offence	0.73	[0.42, 1.28]	0.811	[0.48, 1.36]	0.73	[0.42, 1.28]
Total risk score	1.48**	[1.33, 1.64]	—	—	1.38**	[1.22, 1.55]
Protective factors	—	—	0.437**	[0.30, 0.63]	0.711	[0.49, 1.02]
Constant	127.62		347.07		220.56

Note. Model 1: N = 348, −2 log likelihood = 311.952, Cox and Snell R² = .232, Nagelkerke R² = .338. Model 2: N = 348, −2 log likelihood = 316.087^a, Cox and Snell R² = .223, Nagelkerke R² = .325. Model 3: N = 348, −2 log likelihood = 344.732, Cox and Snell R² = .157, Nagelkerke R² = .228. OR = odds ratio; CI = confidence interval; LL = lower limit; UL = upper limit.

p ≤ .05. **p ≤ .01.

The negative binomial regression predicting the number of new criminal records during the 2-year follow-up period is presented in Table 5. In Model 1, the results show that the total risk score was significant, age was also significantly related to the number of criminal contacts (p < .001), indicating that young people at younger ages engaged in more reoffending. Males had significantly more criminal contacts than females (p < .05). Nationality was also significant; Spaniards were more likely than foreigners to present more reoffend in the follow-up period. The type of crime committed was not significant.

Table 5.

Negative Binomial Regression Effects of Total Risk Score and Their Association With Number of Criminal Files.

	Model 1		Model 2		Model 3
	IRR	CI 95% Wald	IRR	CI 95% Wald	IRR	CI 95% Wald
	IRR	[LL, UL]	IRR	[LL, UL]	IRR	[LL, UL]
Intercept	243.03	[11.31, 5,702.56]	739.06	[35.52, 15,375.8]	422.25	[18.23, 9,780.05]
Age	0.58**	[0.48, 0.71]	0.61**	[0.50, 0.74]	0.58**	[0.48, 0.718]
Boy	2.33*	[1.21, 1.44]	3.35**	[1.74, 6.44]	2.41**	[1.24, 4.69]
Spanish	1.38*	[1.28, 1.49]	0.87	[0.57, 1.33]	0.69	[0.44, 1.09]
Against person offence	0.93	[0.63, 1.37]	0.96	[0.66, 1.40]	0.93	[0.63, 1.38]
Total risk score	1.33**	[1.28, 1.49]	—	—	1.28**	[1.17, 1.40]
Protective factors	—	—	0.41**	[0.30, 0.57]	0.64**	[0.47, 0.88]

Note. Model 1: N = 347, log likelihood = −304.98, AIC = 622.22, BIC = 645.09. Model 2: N = 347, log likelihood = −316.73, AIC = 645.47, BIC = 668.58. Model 3: N = 347, log likelihood = −300.12, AIC = 614.25, BIC = 641.21. IRR = incidence rate ratio; CI = confidence interval; LL = lower limit; UL = upper limit; AIC = Akaike information criterion; BIC = Bayesian information criterion.

p ≤ .05. **p ≤ .01.

Protective factors were introduced in Model 2. Protective factors, age, and sex were significant. Interestingly, the effect of nationality declined and was no longer statistically significant. This may be an indication that protective factors partially mediate the effect of nationality on criminal behavior. In Model 3, when risk and protective factors were included in the model, they were both significant (p = .000, .012). Similarly, the variables of sex and age were once again significant.

Tables 6 and 7 show results of logistic and negative binomial regression. The factors that best predicted general recidivism were current school or employment problems, criminal friends, and personality/behavior. These factors predicted both recidivism and number of offences.

Table 6.

Logistic Regression Analysis of Risk Factors and Their Association With Recidivism (N = 377).

	OR	CI 95% (B)
	OR	LL	UL
Step 1
Criminal friends	0.11**	0.066	0.21
Constant	0.79
Step 2
Criminal friends	0.17**	0.09	0.33
Personality	0.38**	0.21	0.68
Constant	1.01
Step 3
School problems	0.23	0.07	0.68
Criminal friends	0.21**	0.11	0.40
Personality	0.40*	0.23	0.72
Constant	1.09

Note. −2 log likelihood = 349.882, Cox and Snell R² = .205, Nagelkerke R² = .298, N = 377. OR = odds ratio; CI = confidence interval; LL = lower limit; UL = upper limit.

p ≤ .05. **p ≤ .01.

Table 7.

Negative Binomial Regression Effects of Risk Factors and Their Association With Number of Criminal Files.

	IRR	95% CI
Intercept	1.67	[0.89, 3.15]
History of conduct disorder	0.78	[0.50, 1.20]
Current school or employment problems	0.21**	[0.07, 0.57]
Some criminal friends	0.24**	[0.13, 0.43]
Alcohol/drug problems	1.03	[0.69, 1.54]
Leisure/recreation	0.48	[0.18, 1.27]
Personality/behavior	0.51**	[0.31, 0.83]
Family circumstances/parenting	0.98	[0.76, 1.26]
Attitudes/orientation	0.99	[0.77, 1.26]

Note. N = 384, log likelihood = −338.11, AIC = 694.23, BIC = 729.62. IRR = incidence rate ratio; CI = confidence interval; AIC = Akaike information criterion; BIC = Bayesian information criterion.

p ≤ .05. **p ≤ .01.

Finally, logistic regression (Table 8) and negative binomial regression were also conducted with protective factors. Both analyses yielded similar results: Rejection of drugs or alcohol, appropriate leisure, and good performance in school or employment were the significant protective factors that best predict the number of future criminal records.

Table 8.

Logistic Regression Analysis of Protective Factors and Their Association With Recidivism (N = 377).

	OR	CI 95% (B)
	OR	LL	UL
Step 1
Alcohol/drug problems	22.88**	3.11	167.84
Constant	0.019
Step 2
Alcohol/drug problems	14.21*	1.88	107.06
Leisure/recreation	2.76**	1.29	5.90
Constant	0.013
Step 3
Current school or employment problems	6.84	0.88	52.76
Alcohol/drug problems	10.973*	1.44	83.24
Leisure/recreation	2.399*	1.11	5.17
Constant	0.003

Note. N = 377, −2 log likelihood = 394.824^a, Cox and Snell R² = .104, Nagelkerke R² = .152. OR = odds ratio; CI = confidence interval; LL = lower limit; UL = upper limit.

p ≤ .05. **p ≤ .01.

Discussion

The main aim of this study was to analyze the predictive validity of the reduced version of the YLS/CMI for youth recidivism by examining its possible differential contribution over a follow-up period of 2 years. It was hypothesized that the reduced version of the YLS/CMI would predict recidivism. This hypothesis was confirmed for the two outcome measures (general recidivism and the number of criminal charges). We will first discuss some of the data for age, gender, risk, and protective factors, and we will, subsequently, present the predictive results. Finally, limitations and conclusions of the study will be commented.

Results from this study support previous findings on the influence of gender, age, and nationality on risk (Cuervo & Villanueva, 2015; Flores et al., 2004; Upperton & Thompson, 2007). Boys presented a higher risk of recidivism and higher scores for protective factors than girls. Furthermore, minors have lower recidivism and number of charges as their age increases. Interestingly, the effect of nationality was significant in the risk model, but was not significant when it was introduced with protective factors or with risk and protective factors together. This could be an indication that protective factors partially mediate the effect of nationality on delinquent behavior. The correlation of the protective factors and being foreign is significant, even though the effect of being Spanish is reduced when protective factors are taken into account. It seems that when a minor has protective factors, the nationality is no longer important for the reoffending. Therefore, these results reinforce the influence of variables such as sex, age, or race in the final decision of the minor about his or her offending activity, as defended in the Andrews and Bonta’s (2010) general personality and social psychological model of criminal conduct.

According to the first hypothesis, the total score of the reduced version has a strong predictive validity for recidivism. When gender and type of crime were taken into account, recidivism was significantly predicted for crimes against persons and for property offences in the male group. Meanwhile, for girls, only property offences were predicted. This phenomenon was also observed by Chu et al. (2014), who attributed it to the low rate of violent recidivism among females.

The total risk score and the total protective factors were analyzed to check their predictive value. The risk and protective factors together yielded the most predictive value compared with other combinations of variables.

The fact that boys had a high number of risk and protective factors may seem counterintuitive, even though it could denote a qualitative aspect of the characteristic of protective factors. Because protective factors are not merely the absence or the contrary of risk factors, but a more complex concept, the criteria in the assessment of these factors may be more subjective and less clear than risk factors. Further research in the assessment of youth strengths is needed.

However, the risk score was more powerful in predicting recidivism than protective factors. In the final model, protective factors increased the predictive validity. Hence, having prosocial attitudes and conducts would distance minors from crime. This distance could be either symbolic or real, preventing the minor from critical situations, possible fights, drugs, and bad habits in general. In fact, risk scores decrease over time, indicating that these youth have lower levels of risk as they move away from the legal system. Furthermore, it seems that the worst situation related to risk occurs near the time of disposition (Mulvey et al., 2016).

This research, therefore, confirms that this total score is a predictor of general recidivism and of the number of later offences committed. Same effects can be found in the original version of the instrument, showing the reduced even higher Nagerkelke values (Cuervo & Villanueva, 2015) and yielding a significant increment in the amount of variance explained by dynamic risk factors (Lodewijks, de Ruiter, & Doreleijers, 2010). This study, therefore, lends support to previous studies that significantly predicted general recidivism and the number of criminal charges using the Screening Inventory (Chu et al., 2014).

The second objective was to explore the relationship between the specific risk and protective factors as assessed using the reduced version and recidivistic outcomes as a function of gender, age, and nationality. According to the hypothesis, the best predictors of juvenile recidivism would be a previous criminal record, an antisocial peer group, antisocial attitudes, and an antisocial personality, identified by Andrews and Bonta (2010) as the Big Four. Our results partially confirm this hypothesis.

The factors that best predict recidivism were current school or employment problems, some criminal friends, and personality/behavior. Our findings support the influence of peer relations and personality/behavior on recidivism, in Andrews and Bonta’s (2010) Big Four factors. These factors partially support those found in the study using the screening version carried out in Singapore, in which the best predictors were a history of conduct problems, school problems, leisure deficits, antisocial personality or behavior, and negative family circumstances (Chu et al., 2014). The area that emerged as the most significant risk/need factors assessed in Ireland using this screening version was having antisocial peers (86.1%; Irish Youth Justice Service, 2011). In the Canadian study, the predominant risk factors were attitudes/orientation, family circumstances, and criminal friends (Van de Ven, 2004). These results regarding education and personality factors match with the ones found in previous studies with the original inventory (Rennie & Dolan, 2010; Viljoen, Elkovitch, Scalora, & Ullman, 2009; Weerman, 2010) even showing higher explained variance values (Cuervo & Villanueva, 2015).

School problems, criminal friends, and an antisocial personality are, therefore, important factors predicting recidivism in various studies. The school environment and friends form part of the young person’s immediate context. The costs and rewards that juveniles obtain from their environment and from interactions in their school or leisure environments will lead them to a more prosocial or antisocial life.

A poor performance at school is a major precursor of recidivism (Garrido, 2009; Rennie & Dolan, 2010; Viljoen et al., 2009; Weerman, 2010). Truancy from school and bad behavior are related to recidivism in other studies (Bravo et al., 2009; Iborra, Rodríguez, Serrano, & Martínez, 2011; San Juan, Ocáriz, & De la Cuesta, 2007). Some authors maintain that students with problems at school should be identified, because this poor performance could lead them to engage in delinquent behavior (Hart, Toole, Price-Sharps, & Shaffer, 2007). Focusing the intervention on truancy and behavioral management will, therefore, prevent a possible escalation to more severe problems.

An antisocial personality is another commonly accepted risk factor for delinquency and recidivism, and is a significant factor in the onset and persistence of the offender’s development (Cuervo & Villanueva, 2015; Graña et al., 2006; Viljoen et al., 2009). Aggressive and other antisocial behaviors are left unchecked in the close environment/family, whereas prosocial values and behavior are unrewarded. As the child grows up and starts to spend more time outside, opportunities for involvement with antisocial peers increase. Furthermore, if the antisocial behavior is established within the family context, this will influence the social network that the child develops. Well-socialized children are less likely to accept the friendship of antisocial children (Andrews & Bonta, 2010; Cuervo & Villanueva, 2015; Warr, 2005).

In summary, the most significant risk factors in relation to recidivism seem to be factors associated to relational abilities from the minors’ social context, related to their education, the association with peers, and antisocial personality or behavior. Hence, these are dynamic and modifiable factors that can be modulated by presenting the minors with the real cost or consequences of their negative behavior and with the consequences of a positive lifestyle. These findings raise questions about the most predictive factors in the model of Andrews and Bonta (2010): Which of the Central Eight actually comprise the Big Four? Do the Big Four really exist internationally? It seems that as even Bonta and Andrews (2017) recognize, the Big Four are not present in other types of samples as general offenders, youthful offenders, mentally disordered, racial minorities, and drugs offenders. The social context and nature of each culture must be taken into account when analyzing predicting factors of recidivism.

The protective factors that best predict fewer future criminal charges were good performance in school, rejection of drugs, and positive leisure. These last predicted both recidivism and number of offences. For example, a minor who not only says that he or she would not use any drugs but also states that he or she openly rejected them, and has good performance in school, and positive and structured leisure time will be significantly less likely to reoffend (Serin, Chadwick, & Lloyd, 2016). The school environment is one of the most closely related to a prosocial adaptation. Having good neighbourhood socioeconomic status as well as parents and family was also found by other studies to be factors influencing the young people not to reoffend (Stouthamer-Loeber, Loeber, Wei, Farrington, & Wikstrom, 2002; Van et al., 2009). It is important to notice that two of the most risk-predictive factors (criminal friends and personality/behavior) do not coincide with the most predictive protective factors (rejection of alcohol/drugs and appropriate leisure). This could be related to the fact that protective factors are not merely the opposite of the risk factors. This may explain why, in the final model, the addition of protective factors increased the predictive validity as mentioned before.

Having positive aspects in the juvenile’s environment predicts good integration into society. The positive effect of protective factors has been proven in other studies (Cuervo & Villanueva, 2015; Farrington et al., 2012; Hoge & Andrews, 2006; Lodewijks et al., 2010). Juveniles who do reoffend have fewer protective factors than their nonrecidivist counterparts. It is, therefore, essential to remember the importance of the strength factors in providing the juvenile with tools for rejecting involvement in antisocial activities, and enhancing his or her values and capabilities (Andrews, Bonta, & Wormith, 2011; Bravo et al., 2009; Onifade et al., 2008). The inclusion of these factors would, thus, be beneficial not only for prediction but also for intervention. Intervention focused on separating youth from every antisocial aspect of their lives would be principal, because the risk decreases over time (Mulvey et al., 2016).

Limitations

This study, therefore, analyzed reoffending from the juvenile system records. This research may, consequently, have underestimated recidivism rates for youth who reoffended when above 18 years old. However, our findings are consistent with other Spanish, English, and Australian studies in terms of the percentages of recidivism rates (García-España et al., 2011; Garrido et al., 2006; Jennings, 2002). Moreover, because the reduced version was scored from the full YLS/CMI, results must differ from the ones that would have been obtained if the screening version would have been administrated. In this sense, we are aware that this study is only a first approach to the analysis of a brief screening version. Next step should have to include the real administration of the Spanish version of the YLS/CMI-SV. Finally, because the sampling was not probabilistic and was obtained in a Spanish community, the generalization of the results is necessarily limited by these cohort restrictions. In this sense, it would be interesting to extend this study to other populations in Spain and Spanish-speaking countries.

Conclusion

According to these data, the young person’s environment in the broadest sense is crucial in predicting recidivism. Intervention strategies should focus on the immediate context with which the juvenile interacts. The data show that shortcomings in education or work, antisocial conduct with antisocial friends, and a tendency toward an antisocial personality lead the minor toward antisocial conducts. The results of this study partly support the Andrews and Bonta model and the contribution of risk factors to recidivism, emphasizing school education, peers, and an antisocial personality, as well as the importance of strength factors in the juvenile, focusing on the relational abilities to reject any implication in any antisocial activity.

It also confirms that the reduced version is able to predict general recidivism and the number of criminal charges in a Spanish sample prospectively over 2 years, with similar values to the complete version (Cuervo & Villanueva, 2015). The reduced inventory requires considerably less time to be administrated and, as such, could be used in settings with a high flow of youth offenders and limited resources for complete evaluation (Chu et al., 2014). However, it is essential that practitioners are prepared and qualified to administer this tool to obtain an objective assessment of each minor and to be able to evaluate all of them using the same procedure (Bosker, Witteman, Hermanns, & Heij, 2015).

The reduced version could, therefore, be a positive contribution to those working with juveniles when identifying minors with higher risk of additional intervention and with low risk of minimal or no further processing, as well as a validated and accurate instrument according to the European Economic and Social Committee (2006), and for the use of evidence-based assessment.

Footnotes

Acknowledgements

We would like to thank the professionals from the Castellón Juvenile Court and the Davalos Fletcher Foundation for their support. We are also grateful to all the young people who participated in this project.

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) received no financial support for the research, authorship, and/or publication of this article.

References

Acosta

Bustillo de Muñoz

M. C.

Martín

Aragón

Betancort

(2012). Evaluation of the effectiveness of minimum intervention measures on young offenders. The Spanish Journal of Psychology, 15, 702-709.

Anderson

Davidson

Barnes

Campbell

Petersen

Onifade

(2016). The differential predictive validity of the Youth Level of Service/Case Management Inventory: The role of gender. Psychology, Crime & Law, 22, 666-677.

Andrews

D. A.

Bonta

(2010). The psychology of criminal conduct (5th ed.). Cincinnati, OH: Anderson Publishing.

Andrews

D. A.

Bonta

Wormith

S. J.

(2011). The Risk-Need-Responsivity (RNR) model. Criminal Justice and Behaviour, 38, 735-755.

Benavente

(2009, January). Conductas antisociales de la juventud, desde una perspectiva juridical. Communication presented at the VIII Jornadas de Orientación, Universidad Politécnica de Valencia, Spain.

Bonta

Andrews

D. A.

(2017). The psychology of criminal conduct. New York, NY: Routledge.

Bosker

Witteman

Hermanns

Heij

(2015). Improving agreement about intervention plans in probation by decision support. International Journal of Offender Therapy and Comparative Criminology, 59, 1459-1473.

Bravo

Sierra

M. J.

Del Valle

J. F.

(2009). Evaluación de resultados de la nueva ley de responsabilidad penal de menores. Reincidencia y factores asociados. [Results analyses of the new law of criminal responsibility in minors. Recidivism and associated factors] Psicothema, 21, 615-621.

Capdevila

Ferrer

Luque

(2005). La reincidencia en el delito en la justicia de menores [Criminal recidivism in youth justice]. Barcelona, Spain: Generalitat de Catalunya.

10.

Catalano

R. F.

Hawkins

J. D.

(1996). The social development model: A theory of antisocial behavior. In Hawkins

J. D.

(Ed.), Delinquency and crime: Current theories (pp. 149-197). New York, NY: Cambridge University Press.

11.

Catchpole

R. E. H.

Gretton

H. M.

(2003). The predictive validity of risk assessment with violent young offenders: A 1-year examination of criminal outcome. Criminal Justice and Behavior, 30, 688-708.

12.

Chu

Lee

Zeng

Yim

Tan

Ang

. . . Ruby

(2015). Assessing youth offenders in a non-Western context: The predictive validity of the YLS/CMI ratings. Psychological Assessment, 27, 1013-1021.

13.

Chu

C. M.

Lee

Zeng

(2014). The utility of the YLS/CMI-SV for assessing youth offenders in Singapore. Criminal Justice and Behavior, 41, 1437-1457.

14.

Cottle

Lee

Heilbrun

(2001). The prediction of criminal recidivism in juveniles: A meta-analysis. Criminal Justice and Behavior, 28, 367-394.

15.

Cuervo

Villanueva

(2015). Analysis of risk and protective factors for recidivism in Spanish youth offenders. International Journal of Offender Therapy and Comparative Criminology, 59, 1149-1165.

16.

Cuervo

Villanueva

Prado-Gascó

(2017). Predicción de la reincidencia juvenil mediante el Inventario YLS/CMI y baremos para su valoración [Youth recidivism prediction using the YLS/CMI Inventory and norms for assessment]. Revista Mexicana De Psicología, 34, 24-36.

17.

European Economic and Social Committee. (2006). Opinion on the prevention of juvenile delinquency. Ways of dealing with juvenile delinquency and the role of the Juvenile Justice System in the European Union. Brussels, Belgium: Official Journal of the European Union.

18.

Farrington

Loeber

Ttofi

(2012). Risk and protective factors for offending. In Welsh

B. C.

Farrington

D. P.

(Eds.), The Oxford handbook of crime prevention (pp. 46-69). Oxford, UK: Oxford University Press.

19.

Flores

Travis

Latessa

(2004). Case classification for juvenile corrections: An assessment of the Youth Level of Service/Case Management Inventory (YLS/CMI), executive summary. Washington, DC: U.S. Department of Justice.

20.

García-España

García Pérez

Benítez Jiménez

M. J.

Pérez Jiménez

(2011). Menores reincidentes y no reincidentes en el sistema de justicia juvenil andaluz [Recidivist and non-recidivist minors in the Andalusian juvenile justice system]. Alternativas, 18, 35-55. Retrieved from http://hdl.handle.net/10045/18709

21.

Garrido

(2009). La predicción y la intervención con los menores infractores: Un estudio en Cantabria (Estadísticas, instrumentos, protocolos y evaluación). Gobierno de Cantabria, Consejería de Empleo y Bienestar Social) [Prediction and intervention with youth offenders: A study in Cantabria Statistics, instruments, protocols and assessment]. Santa Cruz, Spain: Dirección General de Políticas Sociales.

22.

Garrido

López

Silva

López

M. J.

Molina

(2006). El modelo de la competencia social de la ley de menores [The social competence model in the minors law]. Valencia, Spain: Tirant Lo Blanc.

23.

Graña

J. L.

Garrido

González

(2006). Reincidencia delictiva en menores infractores de la Comunidad de Madrid: Evaluación, características y propuestas de intervención [Criminal recidivism in youth offenders in Madrid’s community: Assessment, characteristics and intervention proposals]. Madrid, Spain: Agencia para la Reeducación y Reinserción del Menor Infractor (ARRMI) de la Comunidad de Madrid.

24.

Hart

J. O.

Toole

Price-Sharps

Shaffer

(2007). The risk and protective factors of violent juvenile offending: An examination of gender differences. Youth Violence and Juvenile Justice, 5, 367-384.

25.

Hartman

J. L.

Turner

M. G.

Daigle

L. E.

Exum

M. L.

Cullen

F. T.

(2009). Exploring the gender differences in protective factors implications for understanding resiliency. International Journal of Offender Therapy and Comparative Criminology, 53, 249-277.

26.

Hoge

Andrews

(2006). Youth Level of Service/Case Management Inventory (YLS/CMI). Toronto, Ontario, Canada: Multi-Health Systems.

27.

Hoge

R. D.

(2005). Youth Level of Service/Case Management Inventory. In Grisso

Vincent

Seagrave

(Eds.), Mental health screening and assessment in juvenile justice (pp. 283-294). New York, NY: Guilford Press.

28.

Hoge

R. D.

Andrews

D. A.

(2011). Youth Level of Service/Case Management Inventory–Screening Version. Toronto, Ontario, Canada: Multi-Health Systems.

29.

Iborra

Rodríguez

Serrano

Martínez

(2011). Situación del menor en la Comunitat Valenciana: víctima e infractor [Overview of youth in Valencian community: victim and offender]. Valencia, Spain: Generalitat Valenciana.

30.

Irish Youth Justice Service. (2011). Progress Report on Garda Youth Diversion Project Development 2009-2011 (Report compiled by: Community Programmes Unit, Irish Youth Justice Service Department of Justice and Equality. Co-funded by the Irish Government and the European Union). Dublin: Governments Publications.

31.

Jennings

(2002). One year juvenile reconviction rates: July 2000 cohort. London, England: Home Office National Statistics.

32.

Jennings

W. G.

Gonzalez

J. R.

Piquero

A. R.

Bird

Canino

Maldonado-Molina

(2016). The nature and relevance of risk and protective factors for violence among Hispanic children and adolescents: Results from the Boricua Youth Study. Journal of Criminal Justice, 45, 41-47.

33.

Jennings

W. G.

Zgoba

K. M.

Piquero

A. R.

Reingle

J. M.

(2013). Offending trajectories among native-born and foreign-born Hispanics to late middle age. Sociological Inquiry, 83, 622-647.

34.

Lodewijks

de Ruiter

Doreleijers

(2010). The impact of protective factors in desistance from violent reoffending: A study in three samples of adolescent offenders. Journal of Interpersonal Violence, 25, 568-587.

35.

Moffitt

(2006). Life-course persistent versus adolescence-limited antisocial behavior. In Cicchetti

Cohen

D. J.

(Eds.), Developmental psychopathology (pp. 570-598). New York, NY: John Wiley.

36.

Mulder

Brand

Bullens

van Marle

(2011). Risk factors for overall recidivism and severity of recidivism in serious juvenile offenders. International Journal of Offender Therapy and Comparative Criminology, 55, 118-135.

37.

Mulvey

E. P.

Schubert

C. A.

Pitzer

Hawes

Piquero

Cardwell

(2016). An examination of change in dynamic risk of offending over time among serious juvenile offenders. Journal of Criminal Justice, 45, 48-53.

38.

Olver

M. E.

Stockdale

K. C.

Wong

(2012). Short and long-term prediction of recidivism using the Youth Level of Service/Case Management Inventory in a sample of serious young offenders. Law and Human Behavior, 36, 331-344.

39.

Olver

M. E.

Stockdale

K. C.

Wormith

J. S.

(2009). Risk assessment with young offenders: A meta-analysis of three assessment measures. Criminal Justice and Behavior, 36, 329-353.

40.

Onifade

Davidson

Campbell

Turke

Malinowski

Turner

(2008). Predicting the who, what and when of delinquency with the Youth Level of Service Case Management Inventory. Criminal Justice and Behaviour, 35, 474-483.

41.

Pérez

(2010). Los infractores menores de edad en el ámbito judicial [Youth offenders in criminal context]. Revista De Derecho Penal y Criminología, 4, 401-417.

42.

Piquero

A. R.

(2015). Understanding race/ethnicity differences in offending across the life course: Gaps and opportunities. Journal of Developmental and Life-Course Criminology, 1, 21-32.

43.

Puzzanchera

Adams

(2011). Juvenile arrests 2009. Washington, DC: U.S. Department of Justice, Office of Justice Programs, Office of Juvenile Justice and Delinquency Prevention.

44.

Rennie

Dolan

(2010). Predictive validity of the Youth Level of Service/Case management. Journal of Forensic Psychiatry and Psychology, 1, 1-19.

45.

Ribeaud

Eisner

(2010). Risk factors for aggression in pre-adolescence: Risk domains, cumulative risk and gender differences—Results from a prospective longitudinal study in a multi-ethnic urban sample. European Journal of Criminology, 7, 460-498.

46.

Rice

M. E.

Harris

G. T.

(2005). Comparing effect sizes in follow-up studies: ROC area, Cohen’s d, and r. Law and Human Behavior, 29, 615-620.

47.

San Juan

Ocáriz

(2009). Evaluación de la intervención educativa y análisis de la reincidencia en la Justicia de Menores en la CAPV [Assessment of educational intervention and recidivism analysis in Juvenile Justice in Basque Country]. Vitoria-Gasteiz, Spain: Servicio Central de publicaciones del Gobierno Vasco.

48.

San Juan

Ocáriz

De la Cuesta

J. L.

(2007). Evaluación de las medidas en medio abierto del Plan de Justicia Juvenil de la Comunidad Autónoma del País Vasco [Assessment of community measures in the Juvenile Justice Plan in Basque Country community]. Boletín Criminológico, 96, 1-4.

49.

Schmidt

Hoge

R. D.

Gomes

(2005). Reliability and validity analyses of the youth level of service/case management inventory. Criminal Justice and Behavior, 32, 329-344.

50.

Schwalbe

C. S.

(2007). Risk assessment for juvenile justice: A meta-analysis. Law and Human Behavior, 31, 449-462.

51.

Serin

Chadwick

Lloyd

(2016). Dynamic risk and protective factors. Psychology, Crime & Law, 22, 151-170.

52.

Shepherd

(2015). Finding color in conformity: A commentary on culturally specific risk factors for violence in Australia. International Journal of Offender Therapy and Comparative Criminology, 59, 1297-1307.

53.

Smit

P. R.

Bijleveld

C. C.

(2015). Juvenile delinquency and juvenile justice trends in Europe. In Krohn

M. D.

Lane

(Eds.), The handbook of juvenile delinquency and juvenile justice (pp. 15-26). Chichester, UK: John Wiley.

54.

Stouthamer-Loeber

Loeber

Wei

Farrington

Wikstrom

(2002). Risk and promotive effects in the explanation of persistent serious delinquency in the explanation of persistent serious delinquency in boys. Journal of Consulting and Clinical Psychology, 70, 111-123.

55.

Swets

J. A.

Dawes

R. M.

Monahan

(2000). Better decisions through science. Scientific American, 283, 82-87.

56.

Thompson

A. P.

Putnins

A. L.

(2003). Risk-need assessment inventories for juvenile offenders in Australia. Psychiatry, Psychology and Law, 10, 324-333.

57.

Ttofi

M. M.

Farrington

D. P.

Piquero

A. R.

DeLisi

(2016). Protective factors against offending and violence: Results from prospective longitudinal studies. Journal of Criminal Justice, 45, 1-3.

58.

Upperton

Thompson

(2007). Predicting juvenile offender recidivism: Risk-need assessment and juvenile justice officers. Psychiatry, Psychology and Law, 14, 138-146.

59.

Van de Ven

(2004). Assessment of risk and need factors and service use in diverted youth (Unpublished doctoral dissertation). Department of Psychology, Carleton University, Ottawa, Ontario, Canada.

60.

Van Domburgh

Loeber

Bezemer

Stallings

Stouthamer-Loeber

(2009). Childhood predictors of desistance and level of persistence in offending in early onset offenders. Journal of Abnormal Child Psychology, 37, 967-980.

61.

Viljoen

J. L.

Elkovitch

Scalora

Ullman

(2009). Assessing risk for violence in adolescents who have sexually offended: An examination of the predictive validity of the ERASOR, YLS/CMI, PCL-YV. Criminal Justice and Behaviour, 36, 981-1000.

62.

Warr

(2005). Making delinquent friends: Adult supervision and children’s affiliations. Criminology, 43, 77-106.

63.

Weerman

(2010). Delinquency after secondary school: Exploring the consequences of schooling, working and dropout. European Journal of Criminology, 7, 339-355.

64.

Zhang

(2016). Testing the predictive validity of the LSI-R using a sample of young male offenders on probation in Guangzhou, China. International Journal of Offender Therapy and Comparative Criminology, 60, 456-468.