Risk factors for violent re-offending
The low base rate (the proportion of a population that exhibits violent recidivism) of violent female offenders affects the accuracy of predictive instruments (Alder & Bazemore, 1979; Brennan, 1998; Odgers, Moretti, & Reppucci, 2005; Quinsey, Harris, Rice, & Cormier,1999). Using the same risk predictors for violent male offenders could easily lead to overrepresentations of females in the violent category (Brennan, 1998). In what follows actuarial and structured clinical guidelines for violent risk assessment are discussed first, followed by identified static and dynamic risk factors relating to violent re-offending by females.
In developing the Violence Risk Appraisal Guide (VRAG, an actuarial tool developed for a male forensic population) Quinsey, Harris, Rice and Cormier (1999) clearly did not test the VRAG on "its ability to predict violent recidivism among female offenders. Generally the predictors of crime among women are the same as those among men, but the base rate of violence in most female population is much, much lower" (p. 248). One study in 2002 by Harris, Rice and Cormier (as cited in Harris, Rice, & Camilleri, 2004) found that the VRAG did not predict violent offending among women. Harris, Rice and Camilleri (2004) used the data from the MacArthur Violence Risk Assessment Study to evaluate the applicability of the (modified) VRAG for nonforensic male and female patients and primarily self-reported violence. The researchers found a stronger accuracy of the VRAG than their previous study which "warrants further research among female patients and offenders" (Harris, Rice, & Camilleri, 2004, p. 1069). The study supports previous findings that "the predictors and causes of violence are quite general rather than specific to particular populations" (Harris, Rice, & Camilleri, 2004, p. 1072); that forensic and nonforensic patients have similar clinical problems and needs; and generality across different measures of violent outcome and across a wide range of follow-up times. The study does not support the suggestion of adjusting actuarial scores in case of psychotic symptoms.
In his discussion of the research-based selection of the static and dynamic factors of the Violence Risk Scale (VRS), a structured guideline "designed specifically to assess the risk of violent recidivism for institutionalized forensic clients who are to be released to the community" (Wong & Gordon, 1999, p. 6), Wong (n.d.) does not refer to how these factors relate to female violent offenders. However, Wong and Gordon (1999) state that the VRS is supposed to be gender and race neutral. However, anger, often considered a dynamic risk factor, appears to differ between male and female offenders. According to Suter and Byrne (2000) women offenders have higher levels of anger and exercise lower levels of control over its expression than male offenders. "Violent female offenders were found to admit to fewer incidents as being anger provoking, and to express lower levels of anger than non-violent females" (Suter & Byrne, 2000, p. 7). The authors link these results to the higher incidence of psychopathology in female offenders.
Motiuk (2000) lists risk factors associated with violent re-offending: history of violence, anger or fear problems, active psychosis, substance abuse, psychopathy, weapon interest, criminal history, childhood problems, lifestyle instability, younger age and being male. However, not much evidence is present to support that these risk factors are applicable to female violent offenders. Blanchette (1997) found that previous attempts at suicide was the strongest indicator of violent re-offending by female offenders.
Odgers and Moretti (2002) note that a list of risk factors is "of limited value in the absence of models that help us understand the differential impact of risk factors and interactions between risk factors across development. The weight of one risk factor always depends on the context in which it occurs" (p. 107). They argue to move to more integrated models of the "processes through which risk factors across multiple domains and levels interact over time" (p. 107).
Weizmann-Henelius, Viemero and Eronen (2004) comment on the minor attention given to risk factors related to violent behaviour in women: "women commit fewer crimes than men, especially violent crimes" (p. 185), often in domestic situations. They review explanations the literature offers regarding violent offending by females. In Finland, violent offenders are referred for an extensive forensic psychiatric assessment. The authors researched a national sample of 61 violent female offenders and found that compared to non-offenders violent offenders reported more problems in their family of origin, more adverse experiences in both childhood and adulthood, more often psychiatric care, substance abuse problems, a history of attempted suicide, a problematic relationship in the year preceding the index offence, personality disorders and cognitive deficits. Comparisons between first-time and repeat violent offenders showed no difference in cognitive abilities but for repeat offenders a younger age at first violent offence, victims often being less emotionally close, a history of non-violent crimes and substance abuse, antisocial personality disorder, borderline personality disorder, witnessed violence in their family of origin, parents divorced, and lived in foster homes more. Substance abuse increased "the risk of violence in antisocial women" (Weizmann-Henelius, Viemero, & Eronen, 2004, p. 193).
Another Finnish study on repeat offending among homicidal female offenders with psychotic and personality disorders (Putkonen, Komulainen, Virkkunen, Eronen, & Lonnqvist, 2003) concludes that re-offending happens within the first two years of the index offence i.e. soon after the offence or soon after release from prison. "Criminal activity before the index offense best predicted repeat offending. … Personality disorders increased and psychotic disorders decreased the risk of recidivism" (p. 949). Being young and addicted to substances made a violent female prone for re-offending. The conclusion is that "when a woman is violent, her recidivism might be similar to that of a violent man" (p. 949). The authors caution that their finding may not be applicable elsewhere, given Finland's racial and social homogeneity.
Odgers, Moretti and Reppucci (2005) reviewed the empirical evidence for the assessment, prediction and management of violence in adolescent girls as "the risk for antisocial behavior in girls is most likely acute during this developmental period" (Odgers & Moretti, 2002, p. 103). Predicting violence in girls faces different issues compared to violence in males or adult females, such as the low base rate of traditional forms of violence among females, the different expression of violence among females as compared to males, the significance of a violent history and an early onset of antisocial and aggressive acts as a predictor of future violence (violent female adolescents tend to disappear when traditional violence measures are used and if they engage in violent behaviour as an adult it happens within the home and has less chance of being detected). "Being female is typically considered a protective factor and is given a negative weighting with actuarial violence assessment models" (Odgers, Moretti, & Reppucci, 2005, p. 9). The authors conclude that "the majority of research on violence prediction in adult women is extremely limited and the findings are unclear at best" (Odgers, Moretti, & Reppucci, 2005, p. 12). There is agreement that psychopathy is an accurate predictor of both violent and general recidivism in adult males, but evidence is lacking to support that psychopathy exists in young people. The authors add that little is known (albeit conflicting information) about the developmental course of aggressive and antisocial behaviour among girls. "The current use of risk assessment instruments to predict future violence in adult females and adolescent males is not widely supported" and "it is clear that the traditional practice of risk assessment for the purpose of violence prediction in adolescent girls is not advisable" (Odgers, Moretti, & Reppucci, 2005, p. 14). However, the authors argue that there is no need for gender-specific risk factors as risk factors appear to be similarly related to violence in boys and girls according to the literature, such as antisocial peers, academic problems and antisocial parental behaviour (Andrews & Bonta, 2003), except for higher rates of sexual abuse and depression for females. However, "clinical experience with forensic populations … presents a very different picture" (Odgers, Moretti, & Reppucci, 2005, p. 16). Some of the gender-related factors that appear important in risk assessment for girls are "exposure to sexual abuse, psychiatric co-morbidity, threat to interpersonal relationships, and insecure attachment" (Odgers, Moretti, & Reppucci, 2005, p. 22). When assessing girls they recommend to include contextual and relational factors in which the violence is embedded and to assess the type, severity and duration of abuse, the relationship with the perpetrator and availability of support and in addition co-morbid disorders. They also suggest taking into consideration factors that could decrease the probability of violence in females as girls tend to internalise problems rather than expressing them in aggressive and violent behaviour. Aggression "may function as a coercive, albeit dysfunctional, strategy to maintain relationships" (Odgers, Moretti, & Reppucci, 2005, p. 19) as girls tend to be aggressive against family members, partners and peers. Research supports the notion that violence and motivation manifests itself differently in females, "we may be measuring a different construct" (Odgers, Moretti, & Reppucci, 2005, p. 21), in line with Shaw and Dubois" (1995) suggested possibility that men and women "experience and use violence for different reasons and under different circumstances" (Problems in understanding women's violence, Gender-blindness, 1). It would be interesting to examine for what kind of violence New Zealand women are sanctioned (i.e., against partner, children or strangers).
In sum, the base rate for violence in women is much lower than in men (except for violence by male and female psychiatric inpatients) affecting the accuracy of predictive instruments. The limited number of studies on the validity of actuarial and structured assessment tools for violent recidivism developed on male offender populations does not recommend their clinical use on female offenders.
In general, the research on violence prediction in adult women is very limited and findings are unclear. However, violence by females appears to happen more in relationships, with different motives and in different circumstances compared to violent males although some argue that risk factors "are generally the same" (Strand & Belfrage, 2001, p. 71) for both groups. Violent female offenders appear incapable of coping with stressful life events hence the development of psychopathologies such as depression, suicide attempts and substance abuse, the last increasing the risk of violence in antisocial women. It is suggested to investigate further the New Zealand violent female offenders population in terms of age, type of offence and relationship with the victim as young women are likely to represent a significant subgroup of violent female offenders.