Predictive Validity of the YLS/CMI in a Sample of Spanish Young Offenders of Arab Descent

Abstract

This study was conducted to assess the predictive validity of the Youth Level of Service/Case Management Inventory (YLS/CMI) in young offenders of Arab descent, living in Spain. To address this subject, the Inventory was administered to a sample of Arab minor offenders (N = 116), and results were compared to a sample of non-Arab minor offenders (N = 140), who were all aged between 14 and 17 years. The charges filed after the date of the first assessment carried out by the Youth Offending Team were coded during the follow-up period (2012-2017). The Inventory showed a similar predictive validity for both groups. However, the values were always slightly higher in the non-Arab group than in the Arab group. With subtle cultural differences, the YLS/CMI seems to be a risk instrument capable of predicting recidivism among Arab young offenders.

Keywords

Arab young offenders predictive validity YLS/CMI recidivism risk factors

Introduction

The prediction of criminal behaviour has become a key point in the criminological field due to its capacity to prevent continuance in the short term (youth reoffending), and in the long term (adult offending; Bersani & Doherty, 2018; Sampson & Laub, 2003). One of the most well-known risk instruments which is in widespread use in many countries is the Youth Level of Service/Case Management Inventory (YLS/CMI), by Andrews and Bonta (1995). The theoretical framework of this Inventory states that for the prediction of criminal recidivism (Bonta, Law, & Hanson, 1998), the factors with the greatest predictive value are antisocial attitudes, antisocial friendships, an antisocial personality pattern, and a history of previous offenses, which are known as the “Big Four” (Andrews, Bonta, & Wormith, 2006). These four factors are followed by another group of factors with moderate correlations, which are deficient family circumstances, education and employment, substance abuse, and leisure and free time. Together, these factors are referred to as the “Central Eight” (Andrews & Bonta, 2010). However, as even Bonta and Andrews (2017) acknowledge, the Big Four are not present in some types of samples, such as offenders with mental health disorders, racial minorities, or drugs offenders. The social context and nature of each culture must consequently be taken into account when analyzing the factors predicting recidivism.

With regard to the validity of the YLS/CMI, several studies show the discriminant capacity of the Inventory for reoffenders and non-reoffenders (Anderson, Hawes, & Snow, 2016; Cuervo & Villanueva, 2015; Flores, Travis, & Latessa, 2004; Rennie & Dolan, 2010). As for the accuracy of the predictive validity of the Inventory, some studies have showed scores for area under the curve (AUC) values ranging from .57 to .75 (Marshall, Egan, English, & Jones, 2006; Schwalbe, 2007; Shepherd, Singh, & Fullam, 2015). This analysis assesses the capability of the total eight-factor score in predicting recidivism where a score of .50 indicates a chance prediction, and a value of 1 a perfect prediction.

However, there are also some critical studies about the general application of risk assessment instruments to different races or cultures (Martel, Brassard, & Jaccoud, 2011; Wilson & Gutierrez, 2014). The fact that most instruments were developed and validated using Caucasian male offender populations creates some doubt about their capacity to deal with the unique characteristics of demographically different offender groups, such as ethnic minorities (Olver, Stockdale, & Wormith, 2009; Wormith & Bonta, 2018). Risk assessment instruments may fail to capture accurately the full range of aspects (language, custom, and religion) that are specific to the Arab culture which is the focus of this study. In fact, some authors have called for new risk assessments, including “culturally-specific risk factors that provide a more accurate measure of risk for groups of minority offenders” (Wilson & Gutierrez, 2014, p. 197).

Despite the limited research in this area, the few studies that have addressed this issue in adults have consistently showed a better prediction of recidivism for nonminority offenders than for minority offenders (Gutiérrez, Wilson, Rugge, & Bonta, 2013; Wormith, Hogg, & Guzzo, 2015). For example, in the previously quoted studies, the Aboriginal offenders yielded higher risk scores and higher rates of recidivism than their non-Aboriginal peers. In the few studies we know of containing a sample of low-risk adult offenders from a Muslim country, for example, Bhutta and Wormith (2016) found that the predictive validities of the LS/CMI (the adult version of the YLS/CMI) were comparable with those in Western countries. AUC values ranged from .55 to .82, and Cronbach’s alpha for the total score of the Inventory was .75. However, as the authors point out, these results may be confined only to low-risk probationers and LS/CMI and not to all kind of offenders and risk assessment tools.

On the contrary, Schmidt, van der Meer, Tydecks, and Bliesener (2018) found that the predictive power of the LS/CMI was reduced for adult offenders with an Arab migration background compared with German offenders without this background.

Even fewer studies have been undertaken on minors belonging to ethnic groups (Rembert, Henderson, & Pirtle, 2014), especially from European countries and in Spanish justice populations and systems. Although some studies with young offenders show that, in general, the YLS/CMI significantly predicts reoffending for minority youths as well as for White youths (Barnes et al., 2016; Olver, Stockdale, & Wormith, 2014), other studies indicate that the minority group may present significantly higher risk scores and recidivism (Liddell, Blake, & Singh, 2016; Perrault, Vincent, & Guy, 2017; Thompson & McGrath, 2012). Moreover, false positives are more common in minority ethnic groups (Shepherd & Lewis-Fernandez, 2016) while false negatives are usually much lower (Douglas, Pugh, Singh, Savulescu, & Savell, 2017).

These results have been consistently found in Australian studies with Aboriginal young offenders (Shepherd et al., 2015; Thompson & McGrath, 2012), and in American studies with African American young offenders (Onifade, Davidson, & Campbell, 2009; Perrault et al., 2017). Some subtle differences have also been found, such as Black youth scoring significantly higher than White youth on the prior/current offense scale, a static factor in the Inventory (Perrault et al., 2017), and the instrument being able to accurately predict recidivism for high-risk youth in all ethnic groups, but not so accurately for the general sample (Shepherd et al., 2015). In spite of these findings, no studies in Europe with young offenders of Arab descent living in Spain has been carried out.

One of the most prominent immigrant groups in Spain is the Arab population, being the Moroccan, the most frequent nationality in obtaining the Spanish citizenship (N = 24,247; Informe del Instituto Nacional de Estadística [INE], 2016). The second and later generations have been brought up in Spain, and immersed into an acculturation process. However, this group still maintains some key features of Arab culture, such as the centrality of the family and religion (Erickson, Al-Timimi, 2001; Haboush, 2007; Soriano & Santos, 2002). In fact, most Arabs in Spain consider themselves Muslims.

Within the criminology field, the number of Moroccan young adults in Spanish prisons has increased in the last decade (García-España, 2016). Among minors, Arab young offenders account for 6% to 11% of the total youth offender population in Spain, but double the Spanish recidivism rate: 40% versus 20% (Capdevila, Ferrer, & Luque, 2005; Cuervo, Villanueva, Prado-Gascó, 2017). One study carried out in Catalonia (Spain) even argues that young offenders from the Maghreb are the group with the most risk factors and the fewest protective factors, that is, the group with the hardest criminal and criminological profile (Capdevila et al., 2005). However, to our knowledge, no study has analyzed the predictive validity of the YLS/CMI in the Arab youth population living in Spain. This question is of critical importance due to the discriminatory processes that this cultural group experiences.

Even in countries with a long tradition of integration, such as the Netherlands, Moroccan-Dutch individuals report high levels of perceived discrimination (Schrier et al., 2014). In fact, social integration is especially difficult in the case of Arab population (in contrast to other Christian and European groups; Awad, 2010). The Spanish population feels a great religious, cultural, and linguistic distance toward this group, with a higher level of rejection of Arabs compared to other immigrant groups (Maya & Puertas, 2008; Navas, Tejada, & Fernández, 2011). The possibility of a poor predictive validity in risk assessment instruments would be an institutional type of discrimination for this population. This would lead to assume wrong decisions in the custody process, based in biased risk assessment. In this line, the principle of fairness should drive the use of risk assessment: First, any risk instrument must be shown to predict recidivism with similar accuracy across different groups (predictive fairness) and second, the use of the instrument in itself must not yield average score differences between racial groups (minimize disparate impact), (Monahan & Skeem, 2016; Skeem & Lowenkamp, 2016). In other words, risk assessments should provide similar ability to discriminate between risk classifications for different racial groups, regardless of the base rate of offending in each group (Campbell, Papp, Barnes, Onifade, & Anderson, 2018).

The study was therefore of interest due to its exploration of the applicability and validity of a Western risk/need instrument in a previously unexamined Arab youth minority in Spain. It was hypothesized that the predictive validity of the YLS/CMI would be more accurate for the non-Arab group than for the Arab group (predictive bias). Second, Arab young offenders were expected to present more recidivism, more risk factors, and fewer protective factors than the non-Arab group (disparate impact).

Method

Participants

The entire sample consisted of 256 minors, aged 14 to 17 years (M = 15.82 years; SD = 1.05), from a province in the Valencian Region in Spain, who had committed a crime and had consequently been interviewed by the Technical Team in the Juvenile Court during the period from 2012 to 2017. Of this sample, 59 were women and 197 were men, accounting for 23% and 77% respectively. This sample was subdivided into two subgroups: Arab-Spanish participants and non-Arab–Spanish participants.

The Arab group consisted of 116 subjects, of whom 14 were females and 102 males (12.1% and 87.9%). The mean age was 15.76 years (SD = 1.09). All were Arab-Spanish adolescents who had been born in Spain or moved there in very early childhood. The criterion of classification as Arab or non-Arab was based on the culture and customs inculcated in the minors.

Meanwhile, the non-Arab group comprised 140 participants, of whom 45 were women and 95 men (32.1% and 67.9%), with an average age of 15.88 years (SD = 1.01). Significant differences between the two groups could be found for the variable gender, χ²(1) = 14.41, p = .000*, due to the large proportion of men in the Arab group (87.9% men and 12.1% women), compared with the non-Arab group (67.9% men and 32.1% women). The age variable showed no differences between the two groups (t = −0.82, p = .41).

Instrument

The instrument used in this study was the YLS/CMI by Hoge and Andrews (2006), which was translated into Spanish by Garrido, López, Silva, López, and Molina (2006) as the “Inventario de Gestión e Intervención para Jóvenes” (IGI-J). This instrument is completed by the Technical Team in the Juvenile Court to evaluate the general risk in a minor’s life, with data from a range of information sources (interviews with the young offender and his or her family, previous court records, contact with school and social centers, etc.). During 2 months, the members of the Technical Team were trained by an expert to understand the protocol of the Inventory and obtain common criteria.

The inventory consists of 42 items, which can be classified according to eight risk factors. In each factor, the evaluator marks the risk items that can be applied to the youth (1 = presence; 0 = absence), with each variable factor having between three and seven items. The factors included in the questionnaire are as follows: (a) prior and current offenses and dispositions, (b) family circumstances/parenting, (c) education/employment, (d) peer relations, (e) substance abuse, (f) leisure/recreation, (g) personality/behaviour, and (h) attitudes, values, and beliefs.

The total general risk score of the YLS/CMI is obtained from the sum of each of the areas that constitute the Inventory. This score obtained indicates a minor’s level of risk of recidivism. The score can in turn be classified into different ranges according to the manual: low (from 0 to 8 points), moderate (from 9 to 22 points), high (from 23 to 32 points), and very high (from 33 to 42 points). According to the overall score obtained on the Inventory, the Youth Offending Team proposes which kind of measure should be adopted with the juvenile.

The instrument also includes the protective factors. These 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. It is possible to assess the minor with a protective factor on each scale except for prior and current offenses, as the positive factor here would be normative for all participants instead of protective. The maximum score for protective factors is therefore 7.

The Spanish version of the Inventory has shown adequate psychometric properties in previous studies, obtaining a Cronbach’s alpha ranging from .87 (Cuervo & Villanueva, 2013) to .91 (Cuervo et al., 2017) for all the items on the Inventory. In this study, the Cronbach’s alpha was .85.

Procedure

The study was conducted under the cooperation agreement established between the Justice Department and the Psychology Department of the University. The data for this study were obtained from an analysis of records of the Juvenile Court of a province in Spain’s Valencian region. The analysis included the number of offenses for each minor in a follow-up period from 2012 to 2017. Demographic data related to ethnicity, nationality and gender, and the risk of youth recidivism obtained by the YLS/CMI were collected.

The participant’s selection procedure was the following one: First, all Arab juvenile offenders who had committed a crime during the period from 2012 to 2017 were selected according to the minor self-identification as Arab, providing a total of 116 subjects. The sample from the non-Arab comparison group was subsequently collected by a random selection procedure.

The recidivism variable was coded in binary format (0 = no recidivism, 1 = recidivism), but also in a continuous way (number of new offenses). The variable “criminal recidivism” refers to charges filed after the date of the first assessment carried out on the minor by the Youth Offending Team, which will be referred to as the baseline. Each minor therefore has a different baseline, considered from 2012.

Data Analyses

The analyses presented here have been mainly structured in two components: if the instrument predicts recidivism with similar accuracy across different groups (predictive validity) and if the use of the instrument yields average score differences between racial groups (disparate impact). For the first component, predictive validity of the YLS/CMI, point-biserial correlations, AUC analyses, hierarchical logistic and negative binomial regression analyses were conducted. In addition, the reliability analysis focused on the assessment of internal consistency using Cronbach’s (1951) alpha was also performed.

As commented before, the outcome variables for youth recidivism were measured dichotomously and quantitatively. In the first case, logistic regression was performed, as this strategy allows to predict a certain behaviour when the response variable is dichotomous (Flores, Holsinger, Lowenkamp, & Cohen, 2017). In the second case, binomial negative regression was carried out. Generalized linear regression with negative binomial distribution applies to count variables and appears to be quite appropriate for the non-normal distribution of the dependent variable under study (Weerman & Hoeve, 2012).

For the second component, disparate impact, a series of t tests for independent samples and chi-square tests were conducted to examine possible variations in the YLS/CMI scores for the two offender groups. The effect size was also calculated in accordance with Cohen (1988), and the confidence interval was 95% in all the analyses.

Results

Predictive Validity of Risk Scores for Recidivism

The internal consistency analyses (Cronbach’s alpha) showed a reliability of .848 for the Arab sample and .855 for the non-Arab group. A contingency table was constructed to practically assess classification errors between the values predicted and those obtained for recidivism in each group (Table 1). The YLS/CMI predicted the correct outcomes (true positives and true negatives) for 73.3% in Arab minor offenders and 75.9% in non-Arab minor offenders. As can be seen, overclassification errors were slightly higher in the non-Arab group (6.79%), while underclassification errors were slightly higher for Arab minors (20%). Arab youth tend to be more classified as false negatives, that is, they engaged in a criminal act but they were judged to be at low risk.

Table 1.

Classification Table of Recidivism.

	Predicted risk level
	True positive (positive hit)		False negative (underclassification error)
Recidivism	%	n	%	n
Arab group	11.43	12	20.00	21
Non-Arab group	14.72	19	17.05	22
	False positive (overclassification error)		True negative (negative hit)
Recidivism	%	n	%	n
Arab group	6.66	7	61.90	65
Non-Arab group	6.79	9	61.24	79

In addition, point-biserial correlations were run to determine the relationship between recidivism and the total YLS/CMI score (rpb = .396, n = 256, p < .001). In the Arab sample, there was a positive and significant correlation between recidivism and the total risk score (rpb = .355, p < .001). The same is true for the non-Arab sample (rpb = .426, p < .001).

An AUC analysis was performed to assess the capability of the total eight-factor score to predict recidivism. In this case, an AUC of .73 (SE = .05) was observed for the Arab group, and it was therefore significant (p < .001). In this case, the confidence interval for the AUC value lay between .63 and .83. On the contrary, an AUC of .76 (SE = .04) was obtained for the non-Arab group (p = .000). In this other case, the confidence interval for the AUC value ranges from .67 to .84. For both groups, the instrument correctly discriminated that a minor who engaged in a criminal act received a higher risk classification than an individual who did not.

The results of a logistic regression for the non-Arab group are presented below (Table 2), including the total risk score and demographic variables (gender and age). In the first model, which contains risk score along with demographics, the total risk score emerged as the most significant predictor of recidivism. However, gender was the most important variable (inverse) in this second model. Taken together, these two variables (being male and presenting a high risk) account for 32.5% of the variance in the prediction of recidivism.

Table 2.

Logistic Regression Analysis of Recidivism for Non-Arab Offender Minors.

	B	SE	Wald	df	p	Exp(B)	95% CI (B)
	B	SE	Wald	df	p	Exp(B)	LL	UL
Model 1: Risk scores
Women(1)	−1.22	0.54	5.03	1	.025*	0.293	0.10	0.85
Age	−0.04	0.21	0.04	1	.826	0.95	0.62	1.45
Total risk	0.16	0.03	21.1	1	.000*	1.17	1.09	1.25
Constant	−1.54	3.39	0.21	1	.648	0.213
Nagelkerke R² = .325; *p < .05; log likelihood = 127.289
Model 2: Protective scores
Women(1)	−0.92	0.48	3.63	1	.057	0.400	0.16	1.02
Age	−0.02	0.19	0.01	1	.933	0.984	0.67	1.44
Total protective	−0.68	0.34	4.03	1	.045*	0.506	0.26	0.984
Constant	−0.050	3.13	0.00	1	.987	0.951
Nagelkerke R² = .120; *p < .05; log likelihood = 149.746
Model 3: Risk and protective scores
Women(1)	−1.21	0.55	4.81	1	.028*	0.299	0.10	1.02
Age	−0.06	0.22	0.07	1	.785	0.943	0.62	1.44
Total risk	0.16	0.04	17.53	1	.000*	1.17	1.08	0.984
Total protective	−0.15	0.35	0.18	1	.668	0.859	0.43	1.72
Constant	−1.26	3.13	0.00	1	.715	0.951
Nagelkerke R² = .326; *p < .05; log likelihood = 127.093

Note. CI = confidence interval; LL = lower limit; UL = upper limit.

In the second model, containing protective score along with demographics, the total score for protective factors was the only variable predicting recidivism. Nevertheless, the percentage of variance in the prediction of recidivism explained by this variable is much lower than in the previous case (12%).

In the third model, including risk and protective scores and demographic variables, the total risk score was the variable that contributed significantly to the final model. The final model was statistically significant, accounting for 32.6% of the variance in the prediction of recidivism. In turn, gender was also a significant variable in the total model, and the protective total score was not significant in this model, which shows that when the two variables (protective and risk scores) are introduced, the significance of protective score disappears.

The regression analyses obtained for the sample of Arab minor offenders are presented in Table 3. As in the non-Arab group, in the first model the total risk score was the most significant variable predicting recidivism. However, unlike the previous case, neither gender nor age is a significant variable. This model accounted for 20.2% of the variance in the prediction of recidivism.

Table 3.

Logistic Regression Analysis of Recidivism for Arab Offender Minors.

	B	SE	Wald	df	p	Exp(B)	95% CI (B)
	B	SE	Wald	df	p	Exp(B)	LL	UL
Model 1: Risk scores
Women(1)	−0.66	0.85	0.60	1	.437	.515	0.09	2.74
Age	−0.30	0.21	1.98	1	.159	.737	0.48	1.13
Total risk	0.13	0.38	12.36	1	.000*	1.14	1.06	1.22
Constant	2.83	3.39	0.71	1	.400	16.89
Nagelkerke R² = .202; *p < .05; log likelihood = 114.385
Model 2: Protective scores
Women(1)	−.956	0.82	1.36	1	.224	.384	0.077	1.92
Age	−.217	0.20	1.15	1	.283	.805	0.542	1.19
Total protective	−.374	0.28	1.77	1	.184	.688	0.396	1.19
Constant	2.93	3.23	0.83	1	.363	18.84
Nagelkerke R² = .053; *p < .05; log likelihood = 126.657
Model 3: Risk and protective scores
Women(1)	−0.63	0.85	0.55	1	.457	.532	0.10	2.81
Age	−0.28	0.22	1.62	1	.203	.755	0.49	1.16
Total risk	0.15	0.47	11.07	1	.001*	1.17	1.07	1.28
Total protective	0.30	0.33	0.81	1	.369	1.35	0.69	2.62
Constant	2.04	3.49	0.34	1	.559	7.72
Nagelkerke R² = .211; *p < .05; log likelihood = 113.599

Note. CI = confidence interval; LL = lower limit; UL = upper limit.

In the second model, containing protective score, none of the variables analyzed were shown to be significant, unlike the case of the non-Arab group. This model explained 5.3% of the variance in the prediction of recidivism. In the third model, including risk and protective scores and demographic variables, the total risk score was again the only variable that significantly contributed to the final model. The model explains 21.1% of the variance on the prediction of recidivism.

Finally, the results from the negative binomial regression are presented for both groups. Similar results to the logistic regression models were found. The total risk score for the Inventory also predicted the number of charges in the follow-up period in both cases.

In the first case, referring to the non-Arab group (Table 4), the total risk score emerged as the most significant predictor of subsequent criminal charges. Likewise, the effect of the gender was also significant. If the minor is a male, he therefore is more likely to have a greater number of subsequent criminal charges than if he were a female. In the second case, referring to the Arab group (Table 5), the only variable that showed a predictive value for the number of subsequent criminal charges was the total risk score. On this occasion, gender was not a significant variable.

Table 4.

Negative Binomial Regression of Number of Criminal Files for Non-Arab Offender Minors.

						95% CI Wald
	B	SE	Wald χ²	df	Significance	LL	UL
Intercept	−2.37	2.60	0.83	1	0.362	−7.471	2.730
Women	−0.78	0.39	3.98	1	0.046*	−1.548	−0.014
Age	0.03	0.16	0.04	1	0.824	−0.282	0.354
Total risk	0.11	0.02	22.17	1	0.000*	0.067	0.163
Total protective	−0.21	0.27	0.61	1	0.433	−0.755	0.324

Note. n = 129; log likelihood = −118.532; AIC = 247.063; BIC = 261.362. CI = confidence interval; LL = lower limit; UL = upper limit; AIC = Akaike information criterion; BIC = Bayesian information criterion.

p < .05.

Table 5.

Negative Binomial Regression of Number of Criminal Files for Arab Offender Minors.

							95% CI Wald
	B	SE	χ² Wald	df	Significance	LL	UL
Intercept	−4.15	2.48	2.80	1	0.094	−9.013	0.710
Women	−0.70	0.70	0.99	1	0.320	−2.081	0.679
Age	0.16	0.14	1.20	1	0.273	−0.128	0.453
Total risk	0.13	0.03	18.99	1	0.000*	0.076	0.200
Total protective	0.09	0.26	0.13	1	0.712	−0.421	0.617

Note. n = 105; log likelihood = −126.865; AIC = 263.730; BIC = 277.000. CI = confidence interval; LL = lower limit; UL = upper limit; AIC = Akaike information criterion; BIC = Bayesian information criterion.

p < .05.

Risk Scores and Recidivism

The results for the risk factors in the different areas of the Inventory are presented in Table 6. As can be seen, there are no significant differences in the total risk score, and only some differences in the subscales of family circumstances and peer relations. The average risk score on the scales was higher for the non-Arab group in both cases. The effect size of the significant risk factors analyzed was low-medium, ranging from −0.49 to 0.14.

Table 6.

Descriptive Statistics for the Arab Group and Non-Arab Group (Risk Factors).

Inventory factors	Arab group	Non-Arab group	t	p	Cohen’s d
Inventory factors	M (SD)	M (SD)	t	p	Cohen’s d
Prior and current offenses	0.24 (0.54)	0.17 (0.43)	1.06	.287	0.14
Family circumstances	1.19 (1.35)	1.80 (1.62)	−3.20	.002*	−0.41
Education	2.04 (1.43)	1.79 (1.63)	1.33	.185	0.16
Peer relations	1.47 (1.32)	2.31 (1.52)	−3.63	.000*	−0.59
Substance abuse	0.48 (1.00)	0.74 (1.18)	−1.83	.069	−0.23
Leisure/recreation	1.89 (1.08)	1.94 (1.10)	−1.86	.064	−0.04
Personality/behaviour	0.91 (1.34)	1.11 (1.38)	−1.17	.240	−0.14
Attitudes/orientation	0.47 (0.89)	0.49 (.91)	−1.77	.860	−0.02
Total score	8.70 (6.44)	10.15 (7.24)	−1.67	.095	−0.21

p < .05.

For protective factors, significant differences were found in the subscales of peer relations, χ²(1) = 5.27, p = .002, and substance abuse, χ²(1) = 24.29, p < .001. These differences showed a higher mean for the non-Arab group on the Peer Relations subscale and a higher mean for the Arab group on the Substance Abuse subscale. No differences were found for the total protective score.

When the absence or presence of recidivism was assessed, χ²(1) = 0.15, p = .069, no significant differences were found. Likewise, no differences were found between the two groups with regard to subsequent criminal charges (t = 1.62; p = .105). That is, both groups recidivated in a similar way, regardless of the type of recidivism variable used.

Conclusion

The current investigation examined the use of a common risk/need assessment tool with young offenders of Arab descent. First, we hypothesized that the predictive validity of the YLS/CMI would be more accurate for the non-Arab group than for the Arab group. This was not fully supported by the results. With very subtle differences, the Inventory showed a similar predictive validity for both groups. However, the values of the AUC analyses and the regression models for the YLS/CMI were always slightly higher in the non-Arab group than in the Arab group (AUC = .76; .73; R² = .33; .21, respectively). For both groups, the AUC values obtained in this study were in the upper range in comparison to previous studies (Shepherd et al., 2015).

In the different regression models, gender was never a significant predictor variable of recidivism in the group of Arab youths. However, this may be due to the small number of girls in the sample of Arab young offenders. This is consistent with previous studies that showed a low criminal involvement in Arab girls (Junger-Tas, Ribeaud, & Cruyff, 2004). On the contrary, the variable risk score was consistently a significant predictor for recidivism in both groups. For all the participants in this study, and regardless of culture, the risk score in the YLS/CMI was able to predict subsequent reoffending during the follow-up period.

If we turn to look at the classification errors in the prediction, we can find the following results that are not consistent with previous literature (Rembert et al., 2014; Shepherd & Lewis-Fernandez, 2016). While the overclassification error (false positive) was similar in both groups, the underclassification error (false negative) was higher in the Arab group. A kind of “positive discrimination bias” when estimating the reoffending risk of Arab youth offenders in comparison to the non-Arab group seems to exist. This is quite paradoxical if we take into account previous studies analyzing classification errors of the YLS/CMI in cultural minorities. Does it have to be with the specific minority studied? Further research is needed to clarify this issue. In spite of this bias, Arab youth offenders present the same recidivism rate as non-Arab youth. In general, the results obtained therefore support the predictive validity of the YLS/CMI for recidivism among Arab young offenders, as reported by Bhutta and Wormith (2016) in low-risk adult offenders. That is, the Inventory presented predictive fairness, contributing to the objective premise of assessment and equality before the law (Rembert et al., 2014).

Second, Arab young offenders were expected to present higher levels of recidivism, more risk factors, and fewer protective factors than the non-Arab group. This hypothesis was not supported. Arab young offenders did not present more risk factors or fewer protective factors than their non-Arab peers. Furthermore, contrary to the predictions, the differences in the recidivism rate between both groups were not significant. They reoffended to a similar extent, both in relation to presence/absence of recidivism, and in relation to number of subsequent criminal charges. In sum, regarding the disparate impact question, the answer was negative: The use of the instrument did not yield average score differences between racial groups.

In relation to risk factors, and contrary to the hypothesis posited, non-Arab young offenders presented significant higher levels of risk than the Arab group for two factors: family circumstances/parenting and peer relations. This is logical considering the importance of the family in Arab culture, and the emphasis on preserving family honor (Al-Krenawi & Graham, 2000; Schmidt et al., 2018).

Moreover, the Arab group presented a significant protective factor compared to their non-Arab peers: substance abuse. In Arab communities, alcohol consumption is frowned upon (Baron-Epel et al., 2015). However, due to acculturation processes, descendants of Arabs are exposed to Western alcohol consumption norms (Arfken & Ahmed, 2016). In spite of the pressure to drink because of social norms, in this study they seem to choose not to drink, and the active rejection of alcohol and drugs is a protective factor for reoffending, which is not present in the non-Arab group. Even with increasing acculturation, they wish to maintain their ethnic identity (Haboush, 2007). In fact, the existence of a strong cultural identity or engagement in adult offenders has even proven to be a protective factor against violent reoffending (Shepherd, Delgado, Sherwood, & Paradies, 2018). In youth, Rojas-Gaona, Hong, and Peguero (2016) defend that the need to balance the demands of his or her original culture represented by his or her first-generation immigrant parents and the demands of the new host culture seems to protect them from criminogenic influences. Moreover, religiosity (as a general attitude to guide behaviour in life) maybe also a protective factor for reducing risk and recidivism, as found by Bhutta and Wormith (2016).

The overall results obtained for risk and protective factors are not consistent with previous studies that establish a high risk profile for Arab young offenders in Catalonia, Spain (Capdevila et al., 2005): substance abuse, with no permanent address, living in the street, no school attendance, a traumatic family background, and so on. A possible explanation for these apparently contradictory results may be that the minors discussed in the study cited (Capdevila et al., 2005) seem to be unaccompanied minors, that is, minors coming from the Maghreb to Spain alone in search of a better future for them and their families. However, this is not the profile of the participants in this study. The participants here are mainly second- and later-generation Arab-Spanish. They were born in Spain or moved there in very early childhood, accompanied by their families. This characteristic seems to define a clearly distinctive profile of the minors, that is consistent with the results obtained in this study and with the more common adolescent-limited trajectory of offending, as suggested by Moffitt (1993, 2006).

Finally, several limitations of this research must be outlined. The first limitation is that this study analyzed recidivism only with reference to juvenile system records. Therefore, an underestimation of recidivism rates for youth who were 18 years old at the time of the offense may have occurred. However, the results from this study are consistent with previous recidivism rates (Cuervo & Villanueva, 2015; Hilterman, Nicholls, & van Nieuwenhuizen, 2014). Second, as the sample was obtained in a Spanish province, the generalization of the data is limited by this factor. It would be advisable to extrapolate these analyses to other Spanish culture minorities (as, for example, Romanian populations), as well as to other countries with populations of Arab origin. Likewise, although both groups were mostly made up of male minor offenders, it would be advisable to have a more balanced sample of women and men, in both intragroup and intergroup terms, to compare the differences more adequately.

Despite these limitations, this study presents important implications for risk assessment professionals. Although the YLS/CMI seems to provide a good prediction of recidivism in Spanish young offenders of Arab descent, professionals must be aware of cultural differences, particularly the value of protective factors in this population, and the errors of classification that may affect the decision-taking custody process. It is our duty to translate this awareness into the provision of culturally sensitive services.

Footnotes

Acknowledgements

We would like to thank the professionals at the Castellón Juvenile Court for their support, and we are also grateful to all the young participants in this research.

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

Al-Krenawi

Graham

J. R.

(2000). Culturally sensitive social work practice with Arab clients in mental health settings. Health & Social Work, 25(1), 9-22. doi:10.1093/hsw/25.1.9

Anderson

S. A.

Hawes

D. J.

Snow

P. C.

(2016). Language impairments among youth offenders: A systematic review. Children and Youth Services Review, 65, 195-203. doi:10.1016/j.childyouth.2016.04.004

Andrews

D. A.

Bonta

(1995). The level of supervision inventory-revised. Toronto: Multi-Health Systems, 106, 19-52.

Andrews

D. A.

Bonta

(2010). The psychology of criminal conduct (5th ed.). New Providence, NJ: LexisNexis.

Andrews

D. A.

Bonta

Wormith

J. S.

(2006). The recent past and near future of risk and/or need assessment. Crime & Delinquency, 52, 7-27. doi:10.1177/0011128705281756

Arfken

C. L.

Ahmed

(2016). Ten years of substance use research in Muslim populations: Where do we go from here? Journal of Muslim Mental Health, 10(1), 13-24. doi:10.3998/jmmh.10381607.0010.103

Awad

G. H.

(2010). The impact of acculturation and religious identification on perceived discrimination for Arab/Middle Eastern Americans. Cultural Diversity & Ethnic Minority Psychology, 16(1), 59-67. doi:10.1037/a0016675

Barnes

A. R.

Campbell

N. A.

Anderson

V. R.

Campbell

C. A.

Onifade

Davidson

W. S.

(2016). Validity of initial, exit, and dynamic juvenile risk assessment: An examination across gender and race/ethnicity. Journal of Offender Rehabilitation, 55, 21-38. doi:10.1080/10509674.2015.1107004

Baron-Epel

Bord

Elias

Zarecki

Shiftan

Gesser-Edelsburg

(2015). Alcohol consumption among Arabs in Israel: A qualitative study. Substance Use & Misuse, 50, 268-273. doi:10.3109/10826084.2014.962051

10.

Bersani

B. E.

Doherty

E. E.

(2018). Desistance from offending in the twenty-first century. Annual Review of Criminology, 1, 311-334. doi:032317-092112

11.

Bhutta

M. H.

Wormith

J. S.

(2016). An examination of a risk/needs assessment instrument and its relation to religiosity and recidivism among probationers in a Muslim culture. Criminal Justice and Behavior, 43, 204-229. doi:10.1177/0093854815604011

12.

Bonta

Andrews

D. A.

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

13.

Bonta

Law

Hanson

(1998). The prediction of criminal and violent recidivism among mentally disordered offenders: A meta-analysis. Psychological Bulletin, 123, 123-142. doi:10.1037/0033-2909.123.2.123

14.

Campbell

Papp

Barnes

Onifade

Anderson

(2018). Risk assessment and juvenile justice: An interaction between risk, race, and gender. Criminology & Public Policy, 17, 525-545. doi:10.1111/1745-9133.12377

15.

Capdevila

Ferrer

Luque

(2005). La reincidencia en el delito en la justicia de menores [Criminal recidivism in youth justice]. Barcelona, Spain: Centre d’estudis Jurídics i Formació Especialitzada, Generalitat de Catalunya.

16.

Cohen

M. A.

(1988). Some new evidence on the seriousness of crime. Criminology, 26, 343-353. doi:10.1111/j.1745-9125.1988.tb00845.x

17.

Cronbach

L. J.

(1951). Coefficient alpha and the internal structure of tests. Psychometrika, 16, 297-334. doi:10.1007/BF02310555

18.

Cuervo

Villanueva

(2013). Reiteración y reincidencia delictivas en menores españoles con expediente judicial [Reiteration and criminal recidivism in Spanish youth with criminal files]. Revista Mexicana De Psicología, 30, 61-68.

19.

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. doi:10.1177/0306624X14557917

20.

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.

21.

Douglas

Pugh

Singh

Savulescu

Fazel

(2017). Risk assessment tools in criminal justice and forensic psychiatry: The need for better data. European Psychiatry, 42, 134-137. doi:10.1016/j.eurpsy.2016.12.009

22.

Erickson

C. D.

Al-Timimi

N. R.

(2001). Providing mental health services to Arab Americans: Recommendations and considerations. Cultural Diversity & Ethnic Minority Psychology, 7, 308-327. doi:10.1037/1099-9809.7.4.308

23.

Flores

A. W.

Holsinger

A. M.

Lowenkamp

C. T.

Cohen

T. H.

(2017). Time-free effects in predicting recidivism using both fixed and variable follow-up periods: Do different methods produce different results. Criminal Justice and Behavior, 44, 121-137. doi:10.1177/0093854816678649

24.

Flores

A. W.

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.

25.

García-España

(2016). De menores inmigrantes en protección a jóvenes extranjeros en prisión [From immigrant minors in protection services to young foreigners in prison]. InDret, 3, 1-27.

26.

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.

27.

Gutiérrez

Wilson

H. A.

Rugge

Bonta

(2013). The prediction of recidivism with Aboriginal offenders: A theoretically informed meta-analysis. Canadian Journal of Criminology and Criminal Justice, 55, 55-99. doi:10.3138/cjccj.2011.E.51

28.

Haboush

K. L.

(2007). Working with Arab American families: Culturally competent practice for school psychologists. Psychology in the Schools, 44, 183-198. doi:10.1002/pits.20215

29.

Hilterman

E. L.

Nicholls

T. L.

van Nieuwenhuizen

(2014). Predictive validity of risk assessments in juvenile offenders: Comparing the SAVRY, PCL: YV, and YLS/CMI with unstructured clinical assessments. Assessment, 21, 324-339. doi:10.1177/1073191113498113

30.

Hoge

R. D.

Andrews

D. A.

(2006). Youth Level of Service/Case Management Inventory (YLS/CMI). Multi-Heath Systems. Retrieved from https://www.mhs.com/MHS-Publicsafety?prodname=yls-cmi

31.

Informe del Instituto Nacional de Estadística. (2016). Estadística de migraciones 2015 [Migration statistics 2015]. Madrid, Spain: Spanish Statistical Institute.

32.

Junger-Tas

Ribeaud

Cruyff

M. J.

(2004). Juvenile delinquency and gender. European Journal of Criminology, 1, 333-375. doi:10.1177/1477370804044007

33.

Liddell

Blake

Singh

(2016). Over-represented and misunderstood: Pacific young people and juvenile justice in NSW. Australian & New Zealand Journal of Criminology, 49, 1-19. doi:10.1177/0004865816666614

34.

Marshall

Egan

English

Jones

R. M.

(2006). The relative validity of psychopathy versus risk/needs-based assessments in the prediction of adolescent offending behaviour. Legal and Criminological Psychology, 11, 197-210. doi:10.1348/135532505X68719

35.

Martel

Brassard

Jaccoud

(2011). When two worlds collide: Aboriginal risk management in Canadian corrections. The British Journal of Criminology, 51, 235-255. doi:10.1093/bjc/azr003

36.

Maya

Puertas

(2008). Estereotipos, inmigración y trabajo [Stereotypes, immigration and work]. Papeles del Psicólogo, 29, 6-15.

37.

Moffitt

T. E.

(1993). Adolescence-limited and life-course-persistent antisocial behavior: A developmental taxonomy. Psychological Review, 100, 674-701.

38.

Moffitt

T. E.

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

Cohen

D. J.

(Eds.), Developmental psychopathology: Risk, disorder, and adaptation (pp. 570-598). Hoboken, NJ: John Wiley.

39.

Monahan

Skeem

J. L.

(2016). Risk assessment in criminal sentencing. Annual Review of Clinical Psychology, 12, 489-513. doi:10.1146/annurev-clinpsy-021815-092945

40.

Navas

Tejada

Fernández

(2011). Concordance between acculturation attitudes and perceptions of the Spaniards toward Maghreb immigrants: Relationship to prejudice. Anales de Psicología/Annals of Psychology, 27(1), 186-194.

41.

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. doi:10.1177/0093854809331457

42.

Olver

M. E.

Stockdale

K. C.

Wormith

J. S.

(2014). Thirty years of research on the level of service scales: A meta-analytic examination of predictive accuracy and sources of variability. Psychological Assessment, 26, 156-176. doi:10.1037/a0035080

43.

Onifade

Davidson

Campbell

(2009). Risk assessment: The predictive validity of the Youth Level of Service Case Management Inventory with African Americans and girls. Journal of Ethnicity in Criminal Justice, 7, 205-221. doi:10.1080/15377930903143544

44.

Perrault

R. T.

Vincent

G. M.

Guy

L. S.

(2017). Are risk assessments racially biased? Field study of the SAVRY and YLS/CMI in probation. Psychological Assessment, 29, 664-678. doi:10.1037/pas0000445

45.

Rembert

D. A.

Henderson

Pirtle

(2014). Differential racial/ethnic predictive validity. Youth Violence and Juvenile Justice, 12, 152-166. doi:10.1177/1541204013485606

46.

Rennie

C. E.

Dolan

M. C.

(2010). The significance of protective factors in the assessment of risk. Criminal Behaviour and Mental Health, 20, 8-22. doi:10.1002/cbm.750

47.

Rojas-Gaona

C. E.

Hong

J. S.

Peguero

A. A.

(2016). The significance of race/ethnicity in adolescent violence: A decade of review, 2005–2015. Journal of Criminal Justice, 46, 137-147. doi:10.1016/j.jcrimjus.2016.05.001

48.

Sampson

R. J.

Laub

J. H.

(2003). Life-course desisters? Trajectories of crime among delinquent boys followed to age 70. Criminology, 41, 555-592. doi:10.1111/j.1745-9125.2003.tb00997.x

49.

Schmidt

van der Meer

Tydecks

Bliesener

(2018). How culture and migration affect risk assessment. European Journal of Psychology Applied to Legal Context, 10, 65-78. doi:10.5093/ejpalc2018a7

50.

Schrier

A. C.

Peen

de Wit

M. A.

van Ameijden

E. J. C.

Erdem

Verhoe

A. P.

. . . Beekman

A. T. F.

(2014). Ethnic density is not associated with psychological distress in Turkish-Dutch, Moroccan-Dutch and Surinamese-Dutch ethnic minorities in the Netherlands. Social Psychiatry and Psychiatric Epidemiology, 49, 1557-1567. doi:10.1007/s00127-014-0852-x

51.

Schwalbe

C. S.

(2007). Risk assessment for juvenile justice: A meta-analysis. Law and Human Behavior, 31, 449-462. doi:10.1007/s10979-006-9071-7

52.

Shepherd

S. M.

Delgado

R. H.

Sherwood

Paradies

(2018). The impact of indigenous cultural identity and cultural engagement on violent offending. BMC Public Health, 18, Article 50. doi:10.1186/s12889-017-4603-2

53.

Shepherd

S. M.

Lewis-Fernandez

(2016). Forensic risk assessment and cultural diversity: Contemporary challenges and future directions. Psychology, Public Policy, and Law, 22, 427-438. doi:10.1037/law0000102

54.

Shepherd

S. M.

Singh

J. P.

Fullam

(2015). Does the Youth Level of Service/Case Management Inventory generalize across ethnicity? International Journal of Forensic Mental Health, 14, 193-204. doi:10.1080/14999013.2015.1086450

55.

Skeem

J. L.

Lowenkamp

C. T.

(2016). Risk, race, and recidivism: Predictive bias and disparate impact. Criminology, 54, 680-712. doi:10.1111/1745-9125.12123

56.

Soriano

Santos

(2002). El perfil social de la mujer inmigrante marroquí en España y su incidencia en la relación intercultural [The social profile of Moroccan immigrant woman in Spain and its incidence in the intercultural relation]. Papeles de Geografía, 36, 171-184.

57.

Thompson

A. P.

McGrath

(2012). Subgroup differences and implications for contemporary risk-need assessment with juvenile offenders. Law and Human Behavior, 36, 345-355. doi:10.1037/ h0093930

58.

Weerman

F. M.

Hoeve

(2012). Peers and delinquency among girls and boys: Are sex differences in delinquency explained by peer factors? European Journal of Criminology, 9, 228-244. doi:10.1177/1477370811435736

59.

Wilson

H. A.

Gutierrez

(2014). Does one size fit all? A meta-analysis examining the predictive ability of the Level of Service Inventory (LSI) with aboriginal offenders. Criminal Justice and Behavior, 41, 196-219. doi:10.1177/0093854813500958

60.

Wormith

J. S.

Bonta

(2018). The Level of Service (LS) instruments. In Singh

Jay P.

Kroner

D. G.

Wormith

J. S.

Desmarais

S. L.

Hamilton

(Eds.), Handbook of recidivism risk/needs assessment tools (pp. 117-145). West Sussex, UK: John Wiley.

61.

Wormith

J. S.

Hogg

S. M.

Guzzo

(2015). The predictive validity of the LS/CMI with Aboriginal offenders in Canada. Criminal Justice and Behavior, 42, 481-508. doi:10.1177/0093854814552843