Risk factors for (repeated) criminal behaviour

The prevalence of female offenders is between 8% and 18.3% depending on the studies in industrial countries (Bonta, Pang, & Wallace-Capretta, 1995) and is low, relative to male offenders (Stuart & Brice-Baker, 2004). Only little research is available on recidivism by female offenders and "what works" is mainly "what works with male offenders" (Salomone, 2004).

"The general criminology perspective views the factors responsible for female crime as essentially the same as those for male crime" (Bonta, Pang, & Wallace-Capretta, 1995, p. 279). Andrews and Bonta (2003) identified the best-validated risk factors for criminal behaviour and the best predictors of recidivism (Bonta, 2002) as "the Big Four": anti-social attitudes, anti-social associates, history of antisocial behaviour and anti-social personality pattern (including psychopathy, impulsivity, restless aggressive energy, egocentrism, below average intelligence, a taste for risk, poor problem solving and poor self regulation skills). The list continues as "the Big Eight" with problematic circumstances at home (such as low levels of affection, caring and cohesiveness, poor parental supervision, neglect and abuse), problematic circumstances at school or work (low levels of education and achievement and unstable employment history), or with leisure (poor use of recreational time) and substance abuse. The ability to predict criminal behaviour increases with the number and variety of major risk factors assessed and with the number of different sources of information used. The authors admit that "the importance of school/work, personal distress, and noncriminogenic interpersonal targets remains unclear with women and minorities" (Andrews & Bonta, 2003, p. 321). However, they comment that the correlates of criminal behaviour appear "highly similar for males and females" (p. 266) based on research available at the time of their writing. They endorse other findings that risk assessment with the LSI-R has found "parallel results for males and females" (Andrews & Bonta, 2003, p. 266).

Farrington and Painter (2004) researched whether risk factors for offending differed for males and females, by examining the brothers and sisters of males included in the Cambridge Study in Delinquent Development. They concluded that the following most important risk factors are similar for brothers and sisters: low family income, large family size, attending a high delinquency rate school, a convicted father, a convicted mother, a delinquent sibling, parental conflict, separation from a parent, harsh or erratic parental discipline and poor parental supervision. Some factors predicted more strongly for sisters: low social class, low family income, poor housing, low praise by parents, harsh or erratic discipline, parental conflict, low parental interest in education and low parental interest in the children whilst others predicted more strongly for brothers: nervous fathers and mothers and poorly educated fathers and mothers. "In general, risk factors were better predictors of the offending behaviour of sisters than brothers" (Farrington & Painter, 2004, p. 3) and "risk assessment using family factors is likely to be more accurate for females than for males" (Farrington & Painter, 2004, p. 3).

Some studies have researched the originating factors to crime, others the maintaining factors. It is paramount to keep the difference in mind when assessing risk of recidivism. Risk of recidivism is different from risk for violence, escape and misconduct in prison and from classification in prison 1. A gender-specific classification system for women offenders is beyond the scope of this paper but has received international attention (Blanchette & Taylor, 2004; Hardyman & Van Voorhis, 2004).

Holtfreter and Morash (2003) state that identifying risk factors for recidivism is only "identifying predictors within the offender population, not potential influences on crime in the first place" and that evidence supports the notion that "programs attempting to reduce particular risk factors also reduce recidivism" but "this does not mean that all of women's needs, whatever they may be, place them at greater risk for offending and recidivism" (p. 151). Loucks and Zamble (1999) conclude "there are considerable similarities in the factors predicting recidivism in serious offenders, regardless of gender" (p. 30) such as age at first arrest. In particular psychopathy "is as important in predicting general offending in female serious offenders as it is in serious male offenders" (p. 28) and "plays an important role in the prediction of violent behavior and prison maladjustment, as it does for males" (Loucks & Zamble, 2000, p. 31). Measures of personality and current functioning contributed most to the prediction of criminal and violent behaviour and of prison misconduct in female offenders. Loucks and Zamble's findings do not support gender-specific theories of female criminal behaviour (see Arnold (1994) for a theory of crime by Black women in which the process of criminalization is initiated by gender oppression, class oppression and victimization and crime is considered a response to alienation and structural dislocation from family, education and work). Although some experiences such as trauma, victimization and moderate to severe depression may have played an important role in the origins of anti-social behaviour in females, Loucks and Zamble (2000) state these are not significant in explaining serious or repeated offending. The exception is pre-adolescent sexual abuse as a significant predictor for violence (Loucks & Zamble, 2000). Their sample of 100 Canadian federally-sentenced women proved to be representative of the institutional population and of serious female offenders in Canadian prisons (Loucks & Zamble, 2000).

Knowledge of predictors of recidivism for female offenders is important for differentiation between high risk and low risk re-offenders, to ensure public safety and to maximize the effectiveness of treatment and rehabilitation services (Bonta, Pang, & Wallace-Capretta, 1995). The authors examined the predictive validity of the Statistical Information on Recidivism Scale (used in Canada). Only two items proved to be predictive of recidivism for incarcerated female offenders: age at first adult conviction and sentence length. Therefore the SIR scale as a predictive scale for female offenders was not supported (also Dell & Boe, 2000). The authors comment that "risk scales based on criminal history variables may have severe limitations when applied to female offenders" (Bonta, Pang, & Wallace-Capretta, 1995, p. 289). However, many of the risk factors for re-offending with males were equally relevant with females (prior criminal history, certain offence types, sentence length). Further research indicated that a drug/narcotic infraction, life imprisonment and release on full parole were inversely related to recidivism (also found in male offenders). Some variables that predict for males did not for females: history of juvenile delinquency, weapon involved with the offence, offence occurred with an associate and alcohol and drug abuse. Other predictors for female re-offending included committing an unarmed robbery, single-parent mothers, illegal sources of income, depending on welfare, history of physical abuse as an adult, history of self-injury, violence toward staff, and number of incidents in prison.

Moth and Hudson (1999) repeat there have been limited studies on female offender risk (referring to Bonta, Pang, & Wallace-Capretta, 1995 and Loucks & Zamble, 1994). A literature review listed the following static factors predicting recidivism in incarcerated female offenders: younger age at admission to prison, younger age at first conviction, younger age at time of interview, history of committing unarmed robbery, previous drug conviction, physical abuse as an adult, history of self injury, history of violence towards staff and number of incidents towards staff (while incarcerated). Other static factors are: previous incarceration, previous revocation of parole, previous escape, longer aggregated sentence length, prior convictions for violent sexual offence and breaking and entering, history of previous psychiatric hospitalisation, prior suicide attempt, history of serious and repeated antisocial acts, left school prior to 16, early/mid childhood disrupted by adoption, fostering, or institutionalisation. Moth and Hudson (1999) summarise that some static factors may apply to female offenders only: history of physical abuse as an adult, history of self injury, history of psychiatric hospitalisation, prior suicide attempts and a history of early childhood disrupted by adoption, fostering or institutionalisation.

Stuart and Brice-Baker (2004) explored the available theoretical and empirical data in respect of variables that correlate with higher rates of recidivism in adult female prisoners. They concluded that five [static] variables were significantly correlated with recidivism. These are:

  • Age: older offenders have higher recidivism rates than younger offenders, but the authors comment that criminal behaviour in older women, often first time offenders, "may not be a continuation of a pattern originating in young adulthood" (p. 40);
  • Arrests while under legal supervision: positively correlated with higher rates of recidivism;
  • Offence type: drug offences or property offences;
  • Age of first imprisonment: younger age of first offence is correlated to higher rates of recidivism;
  • Not looking forward to release: it is suggested that these women would miss the relationships or friendships within the prison as they are considered sentimental; not being motivated to be released could have higher rates of recidivism. Interestingly, the violent offenders in this study had the lowest recidivism rates, possibly because of longer sentences and subsequently less time spent outside of prison and fewer opportunities to re-offend. "Research shows that extreme violence among women is not typical of those who recidivate, and women who do commit such acts have been found to have significantly shorter non-violent criminal histories" (Stuart & Brice-Baker, 2004, p. 47).

Benda (2005) studied Sampson and Laub's position "that desistance from crime can be explained by social bonding that occurs in adulthood" transitions that represent turning points in people's life-course trajectories" (p. 325) with 300 male and female graduates from a boot camp in the U.S.A. The study findings reveal that "childhood and recent sexual and physical abuse, adverse feelings, living with a criminal partner, and drug use are particularly powerful predictors of women's recidivism" (Benda, 2005, p. 337). In addition, "all life transition, except years of education, are inversely and significantly related to recidivism" and "forming a family with a caring partner serves as a buffer for women" (Benda, 2005, p. 337). Covington (1998) also points to women's "capacity for relatedness and connection" as a particular source of strength. Attachment and relationships are important for women and "focus on female development and mutual, caring, and empowering relationships can be useful tools for correctional programs for women and girls" (p. 6). Odgers & Moretti (2002) add that although aggressive females are more likely to desist from offending during their transition into adulthood, they appear to not function well in other domains of life. For men, Benda's results supported previous research on predictors of recidivism: criminal associates, aggression, carrying a weapon, drug use, younger present age and early age of onset of crime and job satisfaction.

Alder and Bazemore (1979) identified two risk factors with female offenders: the number of prior incarcerations and a history of drug dependency. They conclude that if predictive instruments have not been validated for female offenders separate guidelines should be adopted. The authors queried 25 years ago whether the parole guidelines used with male offenders applied to female offenders. "If male and female subpopulations significantly differ with regard to offense related behaviour, items that predict parole failure accurately for male populations may exhibit serious deficiencies when applied to women offenders' (p. 293). They concluded that there are differences between men and women in type of offences, severity (see also Odgers & Moretti, 2002), number of prior prison sentences, age at admission, prior alcohol use and prior drug use (the last higher for females). They notice that biases could lead to overprediction of recidivism in females when prediction instruments are validated on male data and appear to yield a high percentage of false positives for females (the prediction that a female offender will recidivate while she would have been successful if released) and the fact that cut off scores may differ. For instance, the risk of imprisonment in the RoC*RoI model may be present for a female offender because of breach of conditions rather than because of re-offending. Such statistics would deserve further research and could endorse the predictive accuracy of the RoC*RoI for female offenders.

In summary, the prevalence of female offending is low compared to their male counterparts. Some view the factors responsible for female crime essentially the same as for male offenders. The ability to predict criminal behaviour increases with the number and variety of risk factors assessed and with the number of different sources of information used. Risk of recidivism is not equal to risk for violence, escape, misconduct in prison or classification in prison. For risk assessment purposes one needs to distinguish between originating and maintaining factors to crime. Further, in the present discussion a dichotomy is unfolding between prediction of risk assuming the presence of static, unchangeable risk factors (such as age, criminal history and early family factors) and intervention or reintegration targets focusing on dynamic or changeable factors (criminogenic needs) for which consensus about which factors are to be incorporated in risk assessment models is absent. "The risk principle, which asserts that risk can be predicted, is less viable when applied to women" (Blanchette, 2001, p. 50). Focusing on dynamic, even non-criminogenic factors may prove more effective in terms of reducing re-offending of women, with attention for family factors, connections and relationships, the last two being important strengths for females and in support of a holistic approach of the offending process and offender.


1 Department of Corrections' Policy Department (2002) has investigated the development of a security classification system for women inmates in recognition of the current system being designed for male inmates. It was stated that "women present comparatively lower external risk than male inmates" (p. 20).