Women offenders: Another look at the evidence

Author: Devon L. L. Polaschek, Professor of Psychology/Crime Science, University of Waikato

Author biography: Devon Polaschek is a clinical psychologist, and professor in the School of Psychology and the New Zealand Institute of Security and Crime Science, University of Waikato. Her research interests include understanding and preventing re-offending in serious violent and sexual offenders, family violence, psychopathy, desistance, reintegration and parole.

One of the few upsides to the increasing numbers of women serving sentences in western correctional systems is the increase in the amount of scholarly writing and empirical research about women offenders. Yet despite this improving evidence base, there still remain relatively few good research studies directed at understanding relevant characteristics of women offenders and designing and implementing effective approaches to exiting from these same systems. Perhaps the biggest increases are seen in articles that, without necessarily offering evidence to support their case, push back on the idea that findings for male offenders automatically apply to women. This concern is legitimate. Crime commission is still overwhelmingly a “man’s game” (Salisbury & Van Voorhis, 2009; p. 541), making it difficult to collect data on large enough samples of women for robust investigations.

But it is also the case that the meta-analytic literature on “what works” with offending is dedicated to providing findings that can be used to allocate scarce resources to reduce reconviction for, say, 10,000 prisoners; an approach sure to cause outrage in anyone committed to the importance of individual differences. In this literature, gender, and often ethnicity, are treated as moderator variables, meaning that the overall findings are simply examined separately for women and men. The studies examined typically include relatively few women, and may themselves not have split out their results by gender. Consequently, we learn, at best, whether findings largely derived from research with men also apply to women. This is actually a very useful question, but it does not enable us to identify whether there are factors relevant for women that have not been considered with men, or often even whether common risk factors are more or less important for women than for men.

Progress may also have been slowed by the growing feminist scholarship in this domain. Prominent critiques are directed at the system itself, and even at whether women should be held responsible for their offending, given the purported effects of social inequality for instance. Kelly Hannah-Moffat, an accomplished sociologist, exemplifies this work. For example, she writes: “The actuarial logic of RNR[1]-inspired assessment gives priority only to those aspects of a woman’s life that are ‘empirically’ [sic] shown to contribute to recidivism. This emphasis is problematic because it locates the problem of crime in the individual and diminishes the role that social and structural contexts play in women’s criminalization” (p. 215, Hannah-Moffat, 2009), and “The foundational RNR model identifies ‘promising targets for change’” (p. 33, Andrews & Bonta, 1994/2010) within the individual, precluding meaningful considerations of how social inequality shapes opportunities and choices, and how it is itself a risk to be managed (p. 37; Hannah-Moffat, 2016). Although important to a free-speaking society, such arguments are of limited value in a correctional setting where the issue of whether individuals will be held responsible, and of whether a particular woman’s behaviour is in fact criminal, have already been settled.

The other major debate that emerges when reading the literature on women offenders centres on the non-criminogenic needs. There is a good argument to be made that addressing issues unrelated to offending, such as physical and mental health[2] is a fundamental human right, and a public good. But again, this can only be done in a system that is tasked with doing so. So, for example, in the US where key case law has dictated that prisoners must be provided with adequate healthcare, prison psychologists spend much of their time providing this care instead of working with offenders to reduce recidivism.

Is “what works” for men relevant to women?

Much literature refuting the male-dominated status quo with regard to rehabilitation is, simply speaking, based on the idea that women are different to men. Given the apparent obviousness of this assertion,  it is, perhaps, somewhat surprising that decades of psychological research shows women to be more similar to than different from men on a wide range of characteristics  (Hyde, 2014), averaging about 84% overlap (Zell, Krizan, & Teeter, 2015). This research suggests that often such differences are exaggerated, or substantially in the eye of the beholder, implying we should be careful not to create or exaggerate differences through stereotyping and expectancy effects. This section examines the case for generalisation from the research literature on (mostly) men, for two of the important issues: risk assessment and treatment targets.

Risk assessment

The difference in the proportions of men and women involved in the criminal justice system as offenders is a very real one. One consequence of this difference is that most high-risk offenders are men, and most women are lower risk offenders. A salient question then, is whether and how to assess women’s risk of future offending, and especially, whether existing instruments, built mainly from data on men are a good fit to women.

A recent review suggests the LSI family of tools (Level of Service Inventory-Revised [LSI-R]; Level of Service Inventory: Ontario Revision [LSI-OR]; Level of Service/Case Management Inventory [LS/CMI]) was found to be as effective with women as it is with men in distinguishing who will be reconvicted from who will not (Geraghty & Woodhams, 2015). The LSI scales are the best researched for women of the risk assessment tools, and are in use in New Zealand Corrections  (see also Olver, Stockdale, & Wormith, 2014; Smith, Cullen, & Latessa, 2009). However, none of the studies in the Geraghty review addressed the issue of calibration: whether a particular score is associated with the same probability of recidivism for men and women. For instance, if a particular score was associated with a lower likelihood of recidivism for women compared to men, and this was not known, the tool is overpredicting women’s risk. So further investigation is needed of this issue if we aren’t to over- or under-manage women’s risk relative to their male counterparts.

Dynamic risk factors for offending

A major concern with applying the RNR model to women has been whether or not dynamic risk factors for offending in men are equally applicable to women. Again, the LSI scales are particularly useful for addressing this question because they were developed by Andrews, Bonta and colleagues and therefore include measurement of the Central Eight criminogenic need areas. As I noted above, when scores on the central eight criminogenic need areas (antisocial history, peers, cognition and personality pattern, substance abuse, family/marital, school/work, and leisure/recreation[3]), are summed together, these “gender-neutral variables and their compilation into a total risk scale LSI-R powerfully predict offense-related outcomes for women” (p. 281, Van Voorhis, Wright, Salisbury, and Bauman, 2010). But the relevance of the individual risk factors has been examined less. Using LS/CMI data from five small studies (total n=354) Andrews et al. (2012) found that each of the eight needs individually predicted recidivism. Interestingly, each was more highly related to recidivism (i.e., more predictive) for women than for men, though this difference was only statistically significant for substance abuse. These results led them to suggest that for women, we should speak of the “Big Five” rather than the “Big Four”: substance abuse is more important than for men, for whom it is only a moderate risk factor.

Is there a psychology of women’s criminal conduct?

To consider the case for a “gender-responsive” (i.e., women-specific) correctional psychology requires an understanding of the strengths and limitations of the RNR approach. As I implied in the opening, the RNR model is at its most useful with resource allocation policy, service development, and broad design decisions. It is frequently misunderstood (see Polaschek, 2012) as much more restrictive and prescriptive than it is, and some of these misunderstandings are pertinent to women offenders.

Today’s RNR model is best thought of as a 15 principle-based empirical guide that tells us about a range of factors that are correlated with reducing reconviction. Successful interventions will contain or address more of these factors than less successful ones. But establishing empirically how the factors relate to recidivism is not part of the RNR model. Consider substance use. For some offenders it may be a risk factor because drug purchases take them into contact with other offenders (Arseneault, Moffitt , Caspi, Taylor, & Silva, 2000). For others, perhaps it reflects binge drinking difficulties in the service of managing painful emotions, or their partner is pushing them to use drugs, leading to an addiction that they pay for by committing crime. We are all familiar with these and other mechanisms that may link substance abuse to crime. The RNR model does not specify which of these is relevant for a particular offender or group of offenders, leaving room for tailoring of programmes to the people they serve. It does not prescribe how change is best achieved with regard to the risk factor of substance abuse per se. All it says is that offenders who have contact with a service that is working to reduce substance use will be more likely to stay conviction free, all other things being equal.

Gender-responsivehybrid models understand women’s offending as partially determined by unique risk factors (Van Voorhis, 2012), and argue that there are unique responsivity issues to address as well. Van Voorhis and colleagues are among those who have investigated whether the addition of women-specific factors to dynamic risk assessments might improve the accuracy of risk prediction (over the LSI scales alone), and by implication, the effectiveness of rehabilitation (Van Voorhis, Wright, Salisbury, & Bauman, 2010). Additional factors investigated for their relationship to recidivism included self-efficacy, parenting stress, housing safety, mental health and adult victimisation. However, though some were predictive with some samples, none was consistent across samples. Most of the predictive factors could be grouped under the Central Eight (Bonta & Andrews, 2016) anyway which may be why their predictive ability was limited or erratic (e.g., education and family support, anger/hostility, relationship dysfunction). In other words, again, gender differences may be more in the detail than in the general nature of the need. And as with more general purported gender differences, there is probably more overlap between genders than we might expect.

A later study (Bell, 2014) tested this idea of overlap by examining hypothesised women-specific and traditional gender-neutral risk factors in samples of women and men. She found that of seven gender-neutral factors, five predicted recidivism for men, and three for women. For nine women-specific risk factors, four predicted recidivism for women and three for men. Women scored higher than men on 12 of the risk factors overall, suggesting higher needs, but most were not related to recidivism. Current substance abuse, so strongly predictive for women in the Andrews et al. (2012) study, did not predict women’s offending here at all.

So what can we make of this research? First, some specific aspects of risk factors may be much more prevalent in women than in men, but much of this detail lies below the surface of the broad categories of criminogenic needs in the RNR model. Research with women can be helpful in identifying and unpacking ths specificity, which in turn may help with tailoring programmes to women. None of this constitutes evidence that women-specific risk factors lie outside the RNR model. Second, the women’s literature remains small enough that we are at risk of making too much of the results of individual studies rather than waiting to see if those results replicate to different samples; very often they won’t. Any single study of men’s dynamic risk predictors is similarly unlikely to come out with the same pattern of results as a big meta-analysis summing many studies. We would not use one or two studies with men to say that the men’s research literature is wrong. We need to be similarly careful with research with women. Third, we do need to know more about the specific forms of major criminogenic needs for women. For example, if women’s criminogenic peers are more often their partners than for men, interventions to reduce this influence will be rather different than if the peers are fellow gang members.

What about trauma? Repeated exposure to traumatic events is common in the childhood lives of offenders, and for women, it is also concerningly prevalent in adulthood (Bell, 2014). The women offenders’ literature on treatment is replete with programmes for “dealing with trauma”, though often the actual processes by which this is done are not specified. This point is important insofar as we know that simply talking about traumatic experiences repeatedly can increase symptomatology for some people, and may not improve adult functioning; and in fact most trauma exposure results in little impairment to adult functioning. Further, traumatic exposure can have many consequences that are not limited to the criteria required for a diagnosis of PTSD. But PTSD is very elevated in prisoner samples (Briere, Agee, & Dietrich, 2016) including in New Zealand.  Seventy-five percent of women in a recent study of New Zealand prisoners reported some form of mental disorder, with 40% meeting the criteria for PTSD in the previous 12 months  (Indig, Gear, & Wilhelm, 2016). Treatment should be made available for these women, but it does not follow that such treatment will necessarily reduce criminal risk. In part this is because PTSD is simply a description of a particular cluster of consequences of traumatic exposure, with no assumption that they are functionally linked to offending.

Beech and Ward (2004; Ward, Polaschek, & Beech, 2006) outlined a model that is useful for linking risk factors to their possible causes, and clarifying different types of risk factors and how they interact. The model shows that major developmental experiences, especially in childhood or adolescence help to shape relatively stable psychological dispositions and it is some of these that constitute the dynamic risk factors clustered under the categories of the Central Eight.

Therefore, developmental factors (e.g., repeated childhood abuse) are only relevant insofar as they have led to current psychological dispositions that are themselves criminogenic needs. The mechanisms by which these experiences translate into criminal propensity are varied. For example, abuse may lead to global anger, and a sense that the world owes one a living, substance abuse, attachment difficulties, and negative cognitions about the self and others (Briere & Scott, 2015). Or it may compromise one’s ability to get out from under the influence of criminal peers. Some of these mechanisms may fuel crime, but since adversity in childhood and adulthood is common for offenders, the influence of these types of factors—especially the historic ones—is already implicitly built into our models of criminal development. The mechanisms are the focus of treatment, regardless of how they were acquired, and treatment of the distress resulting from the events themselves will not necessarily affect these mechanisms, because over time they take on a life of their own (or in behavioural psychology terms, they are maintained by different factors to those that initiated them).


Is it the case that the gender differences are mainly attributable to responsivity issues? Responsivity refers to myriad different issues but is best thought of as an interaction between the person’s ability to engage with the intervention and the intervention’s ability to engage the person. Relatedly, sometimes what is really meant is the ability to engage with any change process.

The Indig study also found that 47% of women prisoners had experienced psychological distress in the past 30 days (cf. 27% of men). Women living in states of psychological distress or with untreated mental disorders simply may not have the energy to contemplate changes on risk factors. Change is hard for all of us. Routine activities in custodial environments will contribute to that stress. Particularly when sexual abuse is part of the history, both women and men may struggle to manage themselves safely when they are routinely strip-searched and subject to other intrusive experiences that trigger flashbacks, and where the presence and behaviour of other prisoners may activate a state of continuous hyperarousal and vigilance. If the therapy or programme occurs in such environments, the offender brings the consequences of this context into the treatment session.

Similarly, if women are worrying about the safety of children, or whether there is enough money to get through the next week, they are unlikely to prioritise “coming to group”, and may not even have a feasible way to do so (petrol, childcare etc.). Ways to address the practical circumstances required to have the “headspace” to engage in personal change may be somewhat different for women, but these ideas mainly come under the category of “common sense”; in common with the state of responsivity research across the board, the research on responsivity factors for women is patchy at best (Bourgon & Bonta, 2014).

Recent interviews with Department of Corrections’ Kowhiritanga facilitators (Kowhiritanga is a group-based rehabilitation programme for women) suggested that women’s groups may need more time to develop a stable working culture, and that women often wanted to talk through issues at some length, as the material was being presented. Contrary to widely held beliefs, women are not generally more talkative than men (Mehl, Vazire, Ramírez-Esparza, Slatcher, & Pennebaker, 2007). But male criminals in group rehabilitation probably talk less than men in other circumstances. Male prisoners are generally mistrustful in groups, worrying that other men will take advantage of their sharing, and not wanting to be perceived as weak or needing help. So women offenders may require us to recognise the need to slow down group process when people are a bit more comfortable with the group environment.


The evidence base for women offenders remains small, but is improving in quality. However, robust evidence is swamped by agenda-driven writing that reveals stereotypic beliefs that crime reflects inherent badness in men and inherent illness and social disadvantage in women, and fundamental errors in logic, in understanding the aetiology of risk and its remediation, and the RNR model itself.

The overall picture suggests that the RNR model applies to women, but that women offenders may be more complex or more diverse to treat (i.e., reduce risk of re-offending) than men. To date, this complexity does not appear to come from greater criminal risk based on personal characteristics, or more numerous or complex, or “different-from-men” criminogenic needs. Rather, it may be coming from (a) higher rates of overall mental disorder (responsivity/non-criminogenic need) (b) higher rates of PTSD (both historic and recent), which may complicate any treatment that draws on past experiences, and may therefore require more individual treatment (c) greater responsibilities in the community (e.g., for children) that make focusing on treatment difficult (d) less control over some of the contextual conditions that facilitate change (e.g., more dependence on criminogenic partners, less ability to achieve financial independence due to childcare, less ability to be physically and emotionally safe due to family harm etc.), which in turn may lead to low self-efficacy for change, and in group settings at least (e) more comfort with expressing emotions and opinions, and greater interest in communality, which may slow down the group process considerably.

Rather than using these potential differences as an opportunity to throw away decades of empirical research on in favour of untested or common-sense ideas, we would make more progress if we took all of the above into account in planning treatment responses to women offenders, and committed resources to investigating the effects. Many of these issues point to the importance of considering the whole person in our responses, and designing services accordingly.


Andrews, D. A., & Bonta, J. (1994). The psychology of criminal conduct. Cincinnati OH: Anderson.

Andrews, D. A., & Bonta, J. (2010). The psychology of criminal conduct (5th ed.). Newark, N.J.: Matthew Bender.

Andrews, D. A., Guzzo, L., Raynor, P., Rowe, R. C., Rettinger, L. J., Brews, A., & Wormith, J. S. (2012). Are the Major Risk/Need Factors Predictive of Both Female and Male Reoffending? International Journal of Offender Therapy and Comparative Criminology, 56, 113-133.

Arseneault, L., Moffitt , T. D., Caspi, A., Taylor, P. J., & Silva, P. A. (2000). Mental disorders and violence in a total birth cohort. Archives of General Psychiatry, 57, 979-986.

Beech, A. R., & Ward, T. (2004). The integration of etiology and risk in sexual offenders: A theoretical framework. Aggression and Violent Behavior, 10, 31-63.

Bell, V. R. (2014). Gender-responsive risk assessment in corrections. El Paso, TX: LFB Scholarly Publishing.

Bonta, J., & Andrews, D. A. (2016). The psychology of criminal conduct (6th ed.). London: Routledge.

Bonta, J., Blais, J., & Wilson, H. A. (2014). A theoretically informed meta-analysis of the risk for general and violent recidivism for mentally disordered offenders. Aggression and Violent Behavior, 19, 278-287.

Bourgon, G., & Bonta, J. (2014). Reconsidering the responsivity principle: A way to move forward. Federal Probation, 78(2), 3-10.

Briere, J., Agee, E., & Dietrich, A. (2016). Cumulative trauma and current posttraumatic stress disorder status in general population and inmate samples. Psychological Trauma: Theory, Research, Practice, and Policy, 8, 439-446.

Briere, J. N., & Scott, C. S. (2015). Principles of trauma therapy: A guide to symptoms, evaluation, and treatment (2nd edition, DSM-5 update ed.). Los Angeles, CA: Sage.

Geraghty, K. A., & Woodhams, J. (2015). The predictive validity of risk assessment tools for female offenders. Aggression & Violent Behavior, 21, 25-38.

Hannah-Moffat, K. (2009). Gridlock or mutability: reconsidering “gender” and risk assessment. Criminology and Public Policy, 8, 209-219.

Hannah-Moffat, K. (2016). A conceptual kaleidoscope: contemplating ‘dynamic structural risk’ and an uncoupling of risk from need. Psychology, Crime and Law, 22, 33-46.

Hyde, J. S. (2014). Gender similarities and differences. Annual Review of Psychology, 65, 373-398.

Indig, D., Gear, C., & Wilhelm, K. (2016). Comorbid substance use disorders and mental health disorders among New Zealand prisoners. Wellington Retrieved from http://www.corrections.govt.nz/__data/assets/pdf_file/0011/846362/Comorbid_substance_use_disorders_and_mental_health_disorders_among_NZ_prisoners_June_2016_final.pdf.

Mehl, M. R., Vazire, S., Ramírez-Esparza, N., Slatcher, R. B., & Pennebaker, J. W. (2007). Are women really more talkative than men? Science, 317(5834), 82.

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.

Polaschek, D. L. L. (2012). An appraisal of the Risk-Need-Responsivity Model of offender rehabilitation and its application in correctional treatment. Legal and Criminological Psychology, 17, 1-17.

Salisbury, E. J., & Van Voorhis, P. (2009). Gendered pathways: A quantitative investigation of women probations’ paths to incarceration. Criminal Justice & Behavior, 36, 541-566.

Smith, P. H., Cullen, F. T., & Latessa, E. J. (2009). Can 14737 women be wrong? A meta-analysis of the LSI-R and recidivism for female offenders. Criminology and Public Policy, 8, 1601-1626.

Van Voorhis, P. (2012). On behalf of women offenders: Women’s place in the science of evidence-based practice. Criminology and Public Policy, 11, 111-145.

Van Voorhis, P., Wright, E. M., Salisbury, E. J., & Bauman, A. (2010). Women’s risk factors and their contributions to existing risk/needs assessment. The current status of a gender-responsive supplement. Criminal Justice & Behavior, 37, 261-288.

Ward, T., Polaschek, D. L. L., & Beech, A. R. (2006). Theories of sexual offending. Chichester: Wiley.

Zell, E., Krizan, Z., & Teeter, S. R. (2015). Evaluating gender similarities and differences using metasynthesis. American Psychologist, 70, 10-20.

[1] Risk Needs Responsivity

[2] Mental health disorders can be risk factors for offending in specific cases. Overall though they are not (Bonta, Blais, & Wilson, 2014), and changeable risk factors for crime for those with mental disorders remain well accounted for by the “central eight”.

[3] The first four are referred to as the “big four” and the second four as the “moderate four”.