Psychopathy Scores Predict Recidivism in High-risk Youth: A Five-year Follow-up Study

The construct of psychopathy is characterized by a constellation of traits, including interpersonal and affective deficits, poor decision-making, and impulsivity. The relationship between psychopathic traits and antisocial behaviors, including criminal behavior, has been well-established among adults (Allen et al., 2022; Edwards et al., 2023; Leistico et al., 2008) and adolescents (Braga et al., 2023; Caldwell, 2011; Cauffman et al., 2009; Dolan & Rennie, 2008; Edens & Cahill, 2007; Geerling et al., 2020; Gretton et al., 2004; Hilterman et al., 2014; Khanna et al., 2014; McCuish et al., 2018; Schmidt et al., 2011; Shepherd & Strand, 2016; Sitney et al., 2016; Stockdale et al., 2010; Vincent et al., 2008). Based on this strong relationship between psychopathic traits and antisocial behaviors, it is estimated that crimes stemming from individuals scoring high on psychopathy are responsible for nearly $460 billion in social and economic costs per year, with $56.7 billion being accounted for by juveniles with elevated psychopathic traits (Anderson et al., 1999; Kiehl & Hoffman, 2011). Given the extraordinary social and financial cost of psychopathy to society, there has emerged a heightened interest in identifying individuals scoring high on psychopathic traits during adolescence to predict their likelihood of reoffending over time. Understanding this temporal dimension of antisocial risk carries the potential to identify and disrupt maladaptive and life-course-persistent trajectories.

Though there are a variety of assessments designed to measure psychopathic traits (da Silva et al., 2020; Gillespie et al., 2023), the Hare Psychopathy Checklist-Revised (PCL-R; Hare, 2003) has garnered the most support within the forensic community as a risk assessment tool for predicting violent and general recidivism (Dolan & Doyle, 2000; Khiroya et al., 2009; Walters, 2003). An adolescent version—the Hare Psychopathy Checklist: Youth Version (PCL:YV; Forth & Kosson, 2003)—was later developed to identify youth characterized by elevated psychopathic traits at an early stage in development. Assessment protocols for both the PCL-R and PCL:YV include semi-structured interviews with the participant as well as auxiliary reviews of available participant institutional files (e.g., criminal records). The PCL-R and PCL:YV are commonly used to inform decision-making throughout the criminal legal pipeline—such as sentencing, treatment assessment, and parole—for adults and adolescents, respectively (DeMatteo & Olver, 2022; Salekin et al., 1996).

The predictive utility of the PCL:YV for rearrest has strong empirical support (Caldwell, 2011; Cauffman et al., 2009; Dolan & Rennie, 2008; Edens & Cahill, 2007; Gretton et al., 2004; Hilterman et al., 2014; Khanna et al., 2014; McCuish et al., 2018; Schmidt et al., 2011; Shepherd & Strand, 2016; Sitney et al., 2016; Stockdale et al., 2010; Vincent et al., 2008). These studies report that lifestyle and antisocial psychopathic traits (e.g., impulsivity and early behavioral problems) are most useful in predicting subsequent rearrest. In some samples, affective psychopathic traits (e.g., a lack of empathy) also predict violent rearrest (Braga et al., 2023). However, most studies conducted thus far have focused predominantly on non-Hispanic/Latino samples (see Braga et al., 2023 for a review).

Given the fact that Hispanic youth in the U.S. are at disproportionate risk of being detained or committed to juvenile correctional facilities compared to their White and non-Hispanic/Latino peers (Sickmund et al., 2023), there is an impetus to carefully assess the nature of this relationship across cultures and minorities. While Braga and colleagues’ (2023) results suggest that the percentage of White participants in a study had no significant effect on the relationship between psychopathic traits and offense outcomes, this same result does not specifically suggest the stability of these results within specific ethnic groups (i.e., Hispanic/Latino samples). Studies with more ethnically diverse samples suggest weaker or non-existent relationships between the PCL:YV and antisocial outcomes (Edens & Cahill, 2007). Additionally, though some studies (see Salekin, 2008) were conducted with high portions of Hispanic/Latino participants (47%), analyses were solely conducted on the whole sample rather than constrained to this specific ethnicity. This raises the question of whether previously observed relationships between psychopathic traits and antisocial outcomes would extend to other ethnicities, such as Hispanic/Latino samples.

Another issue is how to characterize the nature of the relationship between psychopathic traits and recidivism risk. Research suggests that the risk associated with psychopathic traits may be mediated by the ability to reason or solve problems, which is commonly operationalized through measures of estimated intelligence (intelligence quotient: IQ). While the majority of the literature suggests a negative relationship between IQ and recidivism (Beaver et al., 2013; Loeber et al., 2012; Lynam et al., 1993; Moffitt et al., 1981; Salekin et al., 2010; Schwartz & Beaver, 2019; Schwartz et al., 2015), IQ has also been shown to interact with psychopathic traits and subsequent outcomes in a variety of ways. Some research suggests that a combination of high IQ and high psychopathic traits increases the risk of reoffending in juveniles (compared to either measure alone) (Hampton et al., 2014), while other research suggests non-interactive negative relationships between IQ and rearrest (Beaver et al., 2013; Loeber et al., 2012; Lynam et al., 1993; Moffitt et al., 1981; Salekin et al., 2010; Schwartz & Beaver, 2019; Schwartz et al., 2015), complex curvilinear relationships (Mears & Cochran, 2013), or no relationship (Salekin et al., 2010) between the measure and outcomes. Thus, the extent to which measurements of psychopathy interact with other constructs, namely, IQ, or predict antisocial outcomes over and beyond other measures comorbid with psychopathic traits (i.e., history of substance dependence [SUD]) in adolescents to predict early adulthood offending and violence in high-risk samples remains unclear.

The current study serves as an important conceptual replication and extension of previous work, evaluating the forensic utility of the PCL:YV, and its underlying factors and facets for assessing risk of general and violent felony rearrest in a predominantly Hispanic/Latino adolescent sample. In addition to assessing risk related to PCL:YV Total scores, we examined a two-factor model of psychopathic traits (Hare, 2003; Harpur et al., 1989; Kennealy et al., 2007). This specific two-factor model of psychopathic traits was established via factor analysis (Harpur et al., 1989; Hare et al., 1990), yielding two correlated overarching factors that held explanatory value for the underlying individual items. Later confirmatory factor analyses suggest each of these factors can be further accounted for by two underlying facets (Vitacco et al., 2005; Kosson et al., 2013). Factor 1 is composed of interpersonal (Facet 1) and affective (Facet 2) facets (e.g., grandiosity and a lack of empathy), and Factor 2 is composed of antisocial (Facet 3) and developmental (Facet 4) facets (e.g., impulsivity and early behavioral problems). Rearrest risk was assessed into early adulthood (i.e., across a five-year follow-up period) while controlling for other variables of interest suggested to be associated with rearrest risk (i.e., IQ and SUD).¹ Additionally, we tested the potential for interactive effects between these variables (i.e., PCL:YV factor interactions and PCL:YV x IQ interactions) to confer synergistic risk above the variables themselves, as tested in prior literature (Corrado et al., 2004; Hampton et al., 2014; Vincent et al., 2008). The current study serves as one of the largest longitudinal analysis of the relationship between psychopathic traits and rearrest outcomes in a Hispanic/Latino sample. We hypothesized that those who scored higher on the PCL:YV (Braga et al., 2023; Edens et al., 2007; Kennealy et al., 2010; Leistico et al., 2008; Salekin, 2008; Walters, 2003), would be characterized by higher rates of general and violent recidivism, including a shorter time to rearrest for both categories, even when controlling for SUD (Allen et al., 2022; Edwards et al., 2023; Tolou-Shams et al., 2023), and estimated IQ (Schwartz & Beaver, 2019). We hypothesized PCL:YV Total scores, and both Factor 1 and Factor 2 scores to be predictive of felony/violent re-arrest, with Factor 2 scores being more strongly related to the rearrest outcomes compared to Factor 1 scores. Furthermore, while Hispanic/Latino youth are at disproportionate risk of rearrest compared to their White and non-Hispanic/Latino peers, we hypothesize that the aforementioned risk factors will generalize when tested in the Hispanic/Latino subsample.

The purpose of this study was to assess the forensic utility of the PCL:YV for predicting general and violent recidivism in a large sample of predominantly Hispanic/Latino high-risk male adolescents up to five years post-release. Analyses controlled for additional variables of interest (i.e., IQ and SUD), and also considered the interactive effects of psychopathic traits and IQ. As hypothesized, and consistent with prior literature, we observed that higher juvenile psychopathic traits, lower estimated IQ, and meeting criteria for a higher number of substance dependencies (at a bivariate level), were predictive of shorter time to and higher rates of rearrest into adulthood. Psychopathic traits were the strongest predictor (see Fig. 2), even while controlling for the aforementioned measures (see Tables 3 and 4). These effects generalized strongly to participants identifying as Hispanic/Latino (see "Bivariate Analyses: Hispanic/Latino Sub-Sample", "Multivariate Analyses: Hispanic/Latino Sub-Sample" sections). This study illustrates the utility and generalizability of the PCL:YV in a high-risk sample.

PCL:YV Factor 1 and Factor 2 scores were associated with both felony and violent felony rearrest outcomes. Although PCL:YV Factor 2, and its facets (Facets 3 and 4)—measuring specifically lifestyle/antisocial psychopathic traits, respectively—demonstrated the strongest predictive effects for both outcome categories, it is important to note that PCL:YV Total scores, a summary metric most commonly used for risk assessment, showed comparable effects, largely consistent with previous findings (Braga et al., 2023; Caldwell, 2011; Cauffman et al., 2009; Dolan & Rennie, 2008; Edens & Cahill, 2007; Gretton et al., 2004; Hilterman et al., 2014; Khanna et al., 2014; McCuish et al., 2018; Schmidt et al., 2011; Shepherd & Strand, 2016; Sitney et al., 2016; Stockdale et al., 2010; Vincent et al., 2008: see Tables 3 and 4). The predictive effects of psychopathic traits on felony rearrest persisted even while controlling for the number of SUDs, a variable comorbid with psychopathy (Hemphill et al., 1994). Models including all Facet measures (compared to Total scores or Factor scores) largely performed the best in predicting subsequent rearrest (as assessed by R²). The robustness of these effects (i.e., the strong positive relationship between psychopathic traits and felony and violent felony rearrest) carries importance not only for risk assessments aimed at prediction, but also risk assessments focused on amenability to treatment.

While psychopathy has traditionally been viewed as a “treatment-resistant” cluster of personality traits, emerging research suggests that interventions tailored to individual traits are more successful in curbing antisocial behaviors as compared to non-tailored treatments—a concept commonly referred to as risk-needs-responsivity (Anderson & Kiehl, 2014; Andrews et al., 1990a, b; Simourd & Hoge, 2000; Taxman & Smith, 2021). Indeed, research suggests that adolescents scoring high on the PCL:YV who received trait-tailored treatment—instead of receiving “treatment as usual”—were nearly three times less likely to violently reoffend upon release than those who received standard treatment (Caldwell et al., 2006). A necessary prerequisite to trait-tailored interventions, though, is a comprehensive and accurate assessment of treatment relevant traits (i.e., psychopathic traits). While our effects suggest the importance of antisocial developmental behavioral patterns for the prediction of outcomes, we also find interpersonal/affective traits (i.e., Factor 1 and Facet 2: see Fig. 3) may also have utility in predicting specific outcomes, such as rearrest for violent felony offenses. While these effects were limited to bivariate analyses, they are largely in line with effects observed in prior literature (Braga et al., 2023). Additionally, literature suggests that assessing interpersonal/affective traits has added practical utility when considering treatment amenability (Caldwell et al., 2006; Caldwell et al., 2011), strengthening the arguments behind careful and complete administration of the PCL:YV for its use in the criminal legal services.

The present results contribute to the literature about the generalizability of the PCL:YV. The PCL:YV has emerged as a robust estimate of psychopathic traits, and it has strong predictive utility. Here we found that the PCL:YV generalized to Hispanic youth from a very high-risk adolescent sample. Hispanic/Latino youth are more likely to be detained or committed to juvenile correctional facilities than their White and non-Hispanic/Latino peers (Sickmund et al., 2023), leaving them at disproportionate risk of being at the receiving end of criminal legal decisions. For this very same reason, Hispanic/Latino youth can also disproportionately benefit from accurate assessments of risks in cases of mitigating circumstances, and the same case can be made regarding risk-needs-responsivity assessments. But, importantly, these potentially beneficial instances rely on the strength of relationships between traits of interest (e.g., psychopathic traits) and outcomes of interest (e.g., risk of reoffense), for which our present analyses provide novel support.

Our results also observed a negative relationship between estimated IQ and felony and violent felony rearrest, largely consistent with prior literature (Beaver et al., 2013; Loeber et al., 2012; Lynam et al., 1993; Moffitt et al., 1981; Schwartz et al., 2015; Schwartz & Beaver, 2019, though see: Salekin et al., 2010). There are multiple possible interpretations of these effects. On one hand, intellectual functioning may play a protective role against risky decision making. On the other hand, it could decrease individuals’ likelihood of being caught in risky situations, an explanation supported by previous findings of positive associations between IQ and criminality more generally (Schwartz & Beaver, 2019).

While psychopathic traits and estimated IQ demonstrated positive and negative associations, respectively, between felony and violent felony rearrest, we note that interactions of the two measures also emerged as significant predictors for felony rearrest in the full sample while controlling for each measure individually. In the literature, theory derived from Cleckley (1941) suggests that those high in both measures demonstrate an elevated risk for subsequent antisocial behavior (Hampton et al., 2014; Munoz et al., 2008), specifically noting that if an individual is both intelligent and conniving or manipulative (i.e., a high PCL:YV Factor 1 × IQ interaction term) one may expect that individual to be immersed more fully in self-serving antisocial behaviors. Notably, our effects do not completely mirror this theory. Instead we observe that a lower IQ exacerbates the rearrest risk afforded by psychopathic traits (see Fig. 3 for a visualization of this interaction effect for Felony Rearrest across PCL:YV Factor 2, Facet 1, and Facet 3 traits). Thus, it may be the case that high psychopathic traits, in combination with low IQ may denote a special rearrest risk that is not captured by either measure individually. That said, it is important to note that the independent effects of the individual factors (e.g., PCL:YV Factor 2 and IQ) are much stronger predictors, suggesting that the parts may be more important than their interactions when considering risk of future rearrest. Likewise, though the effect was significant for the interaction terms themselves (albeit a small effect sizes), the steps including the interaction in the multivariate Total and Factor models did not significantly improve the R² of the multivariate regression, and the interaction step in the Facet model only moderately improved the variance accounted for, reducing any strong conclusions to be drawn from the effect. Accordingly, these effects were also not observed in the Hispanic/Latino subsample. Regardless, these results potentially suggest the value of a more nuanced view regarding the relationships between IQ, psychopathy, and antisocial outcomes, one that considers not only linear relationships between the measures and outcomes, but also considers whether there are certain combinatorial subsets of the measures that may confer additional risk outside of the linear relationships.