Risk Factors for Criminal Recidivism in Female Offenders [PDF]

1

Risk Factors for Criminal Recidivism in Female Offenders A Thesis Submitted to the Faculty of Drexel University by Jacey R. Erickson in partial fulfillment of the requirements for the degree of Master of Science May 2014

2

© Copyright 2014 Jacey Erickson. All Rights Reserved.

ii Dedications To my greatest love, my daughter Grace.

iii

Acknowledgements I am extremely grateful to my advisor, Dr. Kirk Heilbrun, for his continued support, guidance, and patience throughout this process. I am thankful that he has continued to encourage and have faith in me, despite how long this project has taken. Without his support, this project would not have been possible. I would also like to thank my thesis committee. Dr. Ralph Fretz, thank you for all of your guidance and support, especially in assisting me in obtaining follow-up data. Dr. DeMatteo, thank you for your time, patience and constructive feedback. A special thank you to Tracy Fass for creating the original male database upon which the female database was based, and thank you for all your advice. Additionally, I would like to thank Linda Nwoga and Natalie Anumba for spending hours entering follow-up data.

iv

Table of Contents LIST OF TABLES ......................................................................................................................... vi ABSTRACT.................................................................................................................................. vii 1. BACKGROUND AND LITERATURE REVIEW ...................................................................1 1.1 Female Offender Overview ................................................................................................1 1.2 The Gender Neutral Perspective .......................................................................................2 1.3 The Level of Service Inventory-Revised (LSI-R) .............................................................3 1.4 The Gender Specific Theory ..............................................................................................8 1.5 Rationale ..........................................................................................................................18 1.6 Hypotheses .......................................................................................................................18 1.6.1 Gender Specific Variables ......................................................................................18 1.6.2 LSI-R Variables......................................................................................................19 2. METHODS ..............................................................................................................................19 2.1 Participants .......................................................................................................................19 2.2 Materials ..........................................................................................................................21 2.3 Design & Procedure .........................................................................................................22 3. METHOD OF ANALYSIS......................................................................................................24 4. RESULTS ................................................................................................................................25 4.1 Descriptive Statistics ........................................................................................................25

v

4.2 Inferential Statistics .........................................................................................................28 5. DISCUSSION ..........................................................................................................................29 5.1 Limitations ......................................................................................................................34 LIST OF REFERENCES ...............................................................................................................38 TABLES ........................................................................................................................................42

vi

List of Tables 1. Characteristics of Participants..................................................................................................46 2. Race of Participants .................................................................................................................48 3. History of Abuse ......................................................................................................................49 4. Participant Rearrest Within Two Years After Release ............................................................50 5. Types of Rearrest Offenses ......................................................................................................51 6. Relationship Between Abuse History and Rearrest .................................................................52 7. PAI Scores and Rearrest ..........................................................................................................53 8. Relationship Between LSI-R Scores and Rearrest...................................................................54

vii

Abstract Risk Factors for Criminal Recidivism in Females Offenders Jacey R. Erickson Kirk Heilbrun, Ph.D.

A substantial rise in the number of females involved in the criminal justice system has resulted in increased interest in females’ risk factors for crime and the emergence of two competing perspectives on criminal offending: the gender specific theory and the gender neutral theory. The gender specific theory posits that females have unique, gender specific risk factors for crime, while the gender neutral theory asserts that the risk factors are the same, or very similar, for males and females. Recent research provides support for the gender neutral model, as traditional risk assessment instruments, which were created based on research with male offenders, have been found to predict recidivism similarly in females and males. However, there has been little research on gender specific variables. This study examined the validity of the Level of Service Inventory-Revised (LSI-R), a traditional risk assessment instrument, and several gender specific variables in predicting recidivism in 300 female offenders being released from state custody after incarceration. The LSI-R Overall score and Criminal History, Alcohol/Drug and Financial subscores significantly predicted recidivism, but the other subscores did not. Of the gender specific variables tested, only the Borderline scale of the Personality Assessment Inventory (PAI) was found to be significantly correlated with recidivism. The Anxiety, Depression, Nonsupport, and Suicidal Ideation subscores of the PAI were not significantly associated with recidivism. These results indicate that the LSI-R may not be as accurate in predicting

viii

recidivism in females as in males. Furthermore, Borderline personality traits should be further investigated as potential risk factors for recidivism in females.

1 CHAPTER 1: BACKGROUND AND LITERATURE SURVEY 1.1 Female Offender Overview A substantial increase in the number of crimes committed by females over the past thirty years has resulted in significant growth in the population of incarcerated females and the number of females re-entering society after incarceration. According to the General Accounting Office, the number of females imprisoned in federal and state correctional facilities increased over 500 percent from 1980 to 1999 (1999). From 1977 to 2004, there was a 757% increase in the number of females serving sentences longer than a year (Frost, Greene & Pranis, 2006). The population of incarcerated female offenders is growing at a faster rate than that of male offenders (Frost et al., 2006; Sabol, West, & Cooper, 2009) as the number of incarcerated female offenders increased 3.2% per year from 2000-2006, compared with 1.9% for male offenders (Bureau of Justice Statistics, 2008). Furthermore, the population of female offenders on probation and parole is growing more rapidly than male offenders. In 2011, 1 in 4 probationers and 1 in 10 parolees were female (Maruschak & Parks, 2012). Despite the rapid increase in female offenders, females still constitute only a small percentage of the overall arrests in the United States. Females comprised 7% of all prisoners in the United States in 2007 (Sabol, West, & Cooper, 2009) and also in 2011 (Maruschak & Parks, 2012). A meta-analysis of 77 studies from around the world found all studies reported that females committed fewer crimes than males (Ellis, 1998). Because males make up the largest percentage of offenders in the United States and countries such as Canada and England, most empirical studies on criminal offending have been conducted solely with male offenders. There has been a lack of research on female

2

offenders until recently (Rettinger & Andrews, 2010). Consequently, most criminological theories were created using research on male offenders (Blanchette & Brown, 2006; Bonta, Pang, & Wallace-Carpretta, 1995). Furthermore, “the majority of traditional criminological explanations of criminal conduct have either ignored females completely or assumed generalizability across gender without female-specific empirical support” (Blanchette & Brown, p. 29). Scholars have begun to focus more attention on female offenders as crime and incarcerations rates of females continue to rise (Heilbrun et. al, 2008). The aim has been to discover whether females have distinctive risk factors for offending and if so, to use those specific risk factors to accurately identify those females at high risk of reoffending and to tailor treatment to address those risks. The ultimate goal is to decrease risk of recidivism upon re-entry into the community. In the quest to understand and ultimately prevent female criminality, two competing perspectives of female criminality have emerged: the gender neutral perspective and the gender-specific perspective. There has been ongoing controversy over which perspective most accurately predicts female criminality (Rettinger & Andrews, 2010). 1.2 The Gender Neutral Perspective The most accepted and best-supported model of assessment and treatment of criminal offenders is the risk-needs-responsivity model (RNR), which was originally created based on research with male offenders in Canada. The risk principle of RNR, which extensive research supports (Gendreau, 1996; Lowenkamp, Latessa, & Holsinger, 2006), involves treating offenders based on their level of risk of recidivism, and focusing

3

treatment on the criminological needs that are specifically related to criminal behavior. The risk principle posits that the individuals with the highest risk of criminal offending should receive the most intensive level of services and those with the lowest levels should have less intensive (or in some cases, very limited) services. The need principle involves identifying and treating certain dynamic risk factors that are associated with criminal behavior (termed criminogenic needs) (Andrews & Bonta, 2003a). The RNR model enumerates the “central 8” risk/need factors that are the best predictors of criminal behavior. These eight factors, which were also based on research with male offenders, can be further categorized into the “big 4” (antisocial personality pattern; history of antisocial behavior; antisocial attitudes, beliefs, and cognitiveemotional states; and antisocial peers) and the “modest 4,” (substance abuse, family/marital situation, school/work, and leisure recreation) which are less predictive of, or closely related to criminal behavior (Andrews & Bonta, 2003a). 1.3 The Level of Service Inventory-Revised (LSI-R) The RNR model is the theoretical basis for the Level of Supervision Inventory (LSI), an assessment instrument widely used to identify an offender’s individual risks and need (Andrews, 1982). Since it was created in 1982, it has been used extensively to predict risk of recidivism, parole violation, and halfway house failure, and has been used in making placement decisions for both male and female offenders (Coulson, Ilacqua, Nutbrown, Giulekas, & Cudjoe, 1996). There have been several revisions of the LSI over the years; the most recent are the Level of Service Inventory-Revised (LSI-R) (Andrews & Bonta, 1995) and the Level of Service/Case Management Inventory (LS/CMI)

4

(Andrews, Bonta & Wormith, 2004). The current study focuses on the LSI-R because of the data that was available. However, both the LSI-R and LS/CMI are reviewed here, as the LS/CMI is the updated version of the LSI-R, based on the same principles (RNR, Central Eight Factors) as the LSI-R. Therefore, the studies on the LS/CMI are still meaningful for the current study. The LSI-R, which is perhaps the best known and extensively researched risk assessment instrument, has consistently been shown to be a valid predictor of recidivism in males across offender types, sentence lengths and settings (Simourd, 2004; Flores, Lowenkamp, Holsinger, & Latessa 2006a). For example, the LSI-R has been shown to have strong predictive validity in male offenders serving long-term sentences in Canada (Simourd, 2004) and in the United States (Manchak, Skeem, & Douglas, 2008); in male probationers and parolees in the United States (Lowenkamp & Bechtel 2007); and in federal probationers in the United States (Flores et al. 2006a). Although the LSI-R was created based on Canadian norms, the LSI-R and its most recent edition, the LS/CMI (Andrews et al., 2004), are used to assess criminological risks and needs in a number of countries, including the United States, Canada, Australia, and the United Kingdom. US norms have been developed for the LSI-R and were released in 2003 (Andrews & Bonta, 2003b). Proponents of the gender neutral perspective on female criminality assert that the risk factors predictive of crime are the same, or very similar, for males and females; accordingly, the RNR model is equally valid for males and females. If this is true, then

5

traditional risk assessment instruments based on the RNR model, such as the LSI-R, should predict crime as accurately in females as they do in males. The validity of the LSI-R in predicting criminal behavior in females has been studied, but much less extensively than in males. A 2009 meta-analysis of 25 data sets which collectively included data for 14,737 female offenders found a mean correlation of .35 between LSI-R score and criminal behavior (Smith, Cullen, & Latessa, 2009). Smith ultimately concluded, based on this meta-analysis, that “the LSI-R performs virtually the same for female offenders as it does for male offenders” (pg. 198). There are several factors to consider, however, when interpreting the results of this study. First, half of the effect sizes used in the meta-analysis were calculated from unpublished data and manuscripts and were therefore not subject to peer review. Second, the outcome measures used varied from study to study. Thirty percent of the studies used reincarceration, 30% used reconviction, and the remaining studies used other measures of recidivism (selfreport, rearrest, or community supervision violation). Third, the study does not specify the racial/ethnic makeup of the 14,737 females. This is important because there is concern that, like gender, criminological risk factors are not generalizable across ethnicities (Holsinger, Lowenkamp, & Latessa, 2006). At least two studies have found that the LSI-R does not predict recidivism as well in African American males as it does with Caucasian males (Fass et al., 2008; Schlager, 2007). Fass et al.’s 2008 study found that African American males were more likely to be overclassified as a risk for recidivism than Hispanic and Caucasian males, and Hispanic and Caucasian males were more likely to be underclasssified.

6

A review of eight of the most recent studies from 1996 to present investigating the predictive validity of the LSI-R in females reveals that the LSI-R seems to predict recidivism in females almost as well as in males in all of the studies (Coulson et al., 1996; Flores, Lowenkamp, Smith, & Latessa, 2006b; Folsom & Atkinson, 2007; Holsinger et al., 2006; Lowenkamp, Holsinger, & Latessa, 2001; Vose, Lowenkamp, Smith, & Cullen, 2009). In fact, at least two studies have found a stronger correlation between LSI-R score and reincarceration for females than for males (Flores, et. al, 2006b; Lowenkamp, et al., 2001). The strength of the relationship between LSI-R score and recidivism in females varies from study to study, ranging from r as low as .11 (Vose et al., 2009) to as high as .55 (Coulson et al., 1996). Studies on the predictive validity of the LS/CMI, the most recent version of the LSI-R, also indicate that the LS/CMI predicts recidivism comparatively well in females and males. Rettinger and Andrews’ 2010 study of 411 female offenders in Canada showed strong correlations between the Total LS/CMI score and general recidivism (r=.63), violent recidivism (r=.45) and total number of new offenses (r=.54). In that study, all eight of the LS/CMI’s domains were correlated with recidivism in females (Rettinger & Andrews, 2010). A later 2012 meta-analysis by Andrews et al. (2012) found a stronger correlation between LS/CMI score and recidivism in females (r=.53) than in males (r=.39), which was explained by the superior predictive validity of the substance abuse domain of the LS/CMI in females (r=.46) relative to males (r=.17). This difference was statistically significant. Evidence of over-prediction of recidivism in females, particularly low risk

7

females, was noted, however. Of the five data sets included in the 2012 meta-analysis, four were Canadian samples and there is no information provided on the racial makeup of the samples. It is also important to note that the female sample sizes of the studies included in the meta-analysis were relatively small, ranging from 43 females offenders (Andrews & Robinson, 1984) to 133 female offenders (Raynor, 2007), for a total number of 354 female offenders included in the meta-analysis of five data sets. There were almost six times more males (2,069) included in the meta-analysis than females. Furthermore, Andrews acknowledges that the authors of the LSI-R and LS/CMI were involved in 6 of the 7 studies included in the meta-analysis (Andrews et al., 2012). The aforementioned studies showing the LSI-R and LS/CMI’s ability to predict recidivism comparably in males and females have been used as support for the gender neutral model of female criminality. However, several limitations should be noted. All of the studies on the predictive validity of the LSI-R in females used predominantly Caucasian samples with limited ethnic diversity and a majority of the studies used Canadian samples. For example, the samples included in Andrew’s 2010 meta-analysis were 74% Caucasian, 12% Aboriginal, 10% black and 4% other. Vose’s 2009 study was 80% Causcasian and 20% black. The racial makeup of female inmates in the United States appears to be much different than the samples used in the existing studies investigating the predictive validity of the LSI-R in females. The most recent data on female inmates in state and federal prisons in the United States reflect the following racial/ethnic distribution of females sentenced to more than one year of incarceration: 49% White, 25% Black, 18% Hispanic and 8% categorized in the ‘other’ category, which

8

included American Indians, Alaska natives, Asians, Native Hawaiians, other Pacific Islanders and persons identifying two or more races (Carson & Sabol, 2012). Furthermore, the existing United States studies used samples from very different regions of the United States, which differ both geographically and, demographically. Two of the studies were conducted in a “large Midwestern state” (Flores et al., 2006b; Lowenkamp et al., 2001), one was conducted in “a northern Midwestern state” (Holsinger et al., 2006), one was conducted in “Southwestern US” (Flores et al, 2006a) and another was conducted in Iowa (Vose et al., 2009). None of the existing studies on the predictive validity of the LSI-R or LS/CMI in females have included samples from states in the northeastern United States, nor have they included samples from larger urban areas of the United States. There are other important factors to consider when interpreting the current studies on the predictive validity of the LSI-R and LS/CMI in females. Andrew’s 2010 and 2012 studies used data from as long ago as 1984 and 1990, which required the original LSI-R score to be converted to an LS/CMI score, which may have affected the data. Additionally, the authors of the LSI-R and LS/CMI were involved in many of the studies investigating the predictive validity of the LSI-R and LS/CMI in females. 1.4 The Gender Specific Theory According to the gender-specific perspective of female criminality, gender plays an important role in female offending, a role that is not considered in traditional criminological theories, which were based primarily on male offenders (Blanchette & Brown, 2006). Proponents of the gender-specific theory contend that females and males

9

may share some of the same risk factors for criminal behavior, but those risk factors are qualitatively and quantitatively different, and are related to criminal behavior in females in unique ways (Salisbury & Van Voorhis, 2009). They also argue that females’ “pathways to crime” are different than men, although there is no consensus among feminist theorists regarding the nature of these pathways. DeHart (2008) asserts that criminal behavior in females is often as result of “survival strategies,” while Salisbury and Van Voorhis (2009) maintain that there are three pathways to crime in females: a childhood victimization pathway, a relational pathway, and a social and human capital pathway. A major emphasis of the gender-specific theory is the role that history of victimization and trauma plays in female criminality. It is undisputed that females often have significant histories of abuse and victimization, especially sexual victimization, relative to males (Covington, 1998; Lowenkamp, Holsinger, & Latessa, 2001; McClellan, Farabee, & Crouch, 1997; Rettinger & Andrews, 2010). Greenfeld and Snell (1999) found that “nearly 6 in 10 women in State prisons had experienced physical or sexual abuse in the past” (pg. 1) and a 2008 study found that 70% of the incarcerated women sampled reported sexual victimization so severe in their past that it constituted rape or serious sexual assault (McDaniels-Wilson & Belknap, 2008). Females are not only at high risk of sexual victimization while in the community, but also during periods of incarceration. A recent Bureau of Justice Statistics investigation on sexual abuse in state prisons found that females had a much higher rate of sexual victimization during incarceration than males. In fact, the rate of inmate-on-

10

inmate sexual victimization was three times higher for females (13.7%) than males (4.2%) and the rate was particularly high for bisexual female inmates, as 18% of bisexual inmates surveyed reported being sexually victimized during incarceration in state prison (2012). Advocates of the gender-specific theory argue that this trauma history is related to two other risk factors for criminal behavior and recidivism: substance abuse and mental illness, especially Post-Traumatic Stress Disorder (Covington, 1998). The feminist pathways perspective goes further and asserts that victimization in childhood is the “primary causal factor” leading to criminal behavior in females (Blanchette & Brown, 2006, p. 35). Although it is clear that the prevalence of childhood abuse and victimization is more common (or at least reported more frequently) among female offenders than male offenders, and female offenders report that they believe their history of sexual victimization is related to their criminal behavior (McDaniels-Wilson & Belknap, 2008), it is unclear whether a history of childhood victimization is related to offending risk, or is a better predictor of criminal offending in females than in males. The few studies investigating the relationship between childhood abuse and recidivism have shown mixed results (Lowenkamp, Holsinger, & Latessa, 2001; Wright, Salisbury, & Van Voorhis, 2007). At least one study has shown that childhood abuse is related to prison misconduct in females (Wright, et al., 2007), but an earlier 2001 study found it was not predictive of recidivism (Lowenkamp et al., 2001). The most recent study investigating the role of victimization in female offending and mental health functioning, conducted on the same data set as the current study, found that victimization history (history of sexual abuse, violence, and/or violence in one’s

11

family of origin) was not a significant predictor of recidivism in the first year following release from custody, but was associated with higher levels of stress and mental health problems (Anumba, DeMatteo, & Heilbrun, 2012). Female offenders have a higher prevalence of mental health problems than do male offenders (Covington & Bloom, 1999; Ditton, 1999; Sacks et al., 2008). Specifically, rates of depression, anxiety, and self-injurious behavior are much higher in female offenders than male offenders (Belknap & Holsinger, 2006; Bloom, Owen, & Covington, 2003). Furthermore, incarcerated females also have more severe mental health problems than females in community based treatment (Sacks, 2004). However, there is a lack of research on the relationship between mental health and recidivism in females. Blanchette and Brown (2006) concluded that “personal distress, mental ability and mental health variables are not strongly associated with women’s likelihood of recidivism” (pg. 105). However, other researchers have found a correlation between self-injury and violent recidivism (Bonta, 1995). It is important to note that the existing research in this area often aggregates mental illness into broad categories, potentially missing any relationships between the specific mental illnesses and symptoms more prevalent in females and recidivism (Salisbury, Van Voorhis & Spiropoulos, 2008). Female offenders with mental health disorders often have high rates of cooccurring substance abuse problems (Abram et. al, 2003; Sacks, 2004). The most recent data on drug dependence in female inmates nationwide shows that in 2004, 60% of all female state prisoners and 43% of female federal prisoners met criteria for having a drug dependence or abuse problem during the year prior to their incarceration (Mumola &

12

Karberg, 2006). Although males also have high rates of substance abuse, gender specific theorists assert that substance abuse plays a different role in criminal behavior in females than in males (Van Voorhis, et al. 2010). Greenfeld and Snell (1999) found that 40% of female inmates in state prisons reported being under the influence of drugs at the time of their offense, as compared to 32% of males. Men were more likely to have been using alcohol at the time of their offense (38% of males were under the influence of alcohol at the time of the offense, as compared to 29% of females) (Greenfeld & Snell, 1999). Abram, Teplin and McClelland’s 2003 study of 1,272 female offenders found that “72% with a severe mental disorder also had a substance use disorder” (pg. 1009), and Sacks et al. found the rate of co-occurring disorders higher among female offenders than male offenders (2008). Furthermore, while male and female offenders both have high rates of substance abuse prior to incarceration, “evidence suggests that male and female inmates differ in preincarceration substance abuse” (Phillips, Nixon, & Pfefferbaum, 2002). For example, one study found males more likely to be problem drinkers than females (Nunes-Dinis & Weisner, 1997). Another study found that male inmates reported more problems with alcohol and females reported more problems with cocaine than males (Peters, et. al, 1997). Incarcerated females frequently have health issues, such as sexually transmitted infections and related disorders (e.g., Hepatitis B & C, cervical cancer), hypertension, asthma and disorders related to poverty and poor nutrition (Braithwaite, Treadwell, & Arriola, 2005; Covington, 2007; Stein, Caviness, & Anderson, 2012). Approximately 5% of females are pregnant upon entering prison, and often these pregnancies are high-risk as

13

a result of substance abuse and inadequate prenatal care (Bloom, et al., 2003; Covington, 2007). Unintended pregnancies are an added stress on female offenders, and are not uncommon occurrences in the lifetimes of female offenders (LaRochelle, 2012). There are high rates of HIV infection among female offenders, as well (Abram, Teplin & McClelland, 2003), with HIV infection more prevalent among female offenders than in male offenders (Greenfeld & Snell, 1999; Maruschak, 2009). It appears that there is no existing research on the question of whether there is a relationship between health problems and risk of recidivism. Female offenders are more likely than male offenders to have needs related to the care of their children (Holtfreter & Cupp, 2007). A 1999 study by the Bureau of Justice Statistics found that almost 71% of women under correctional supervision had “at least one child under the age of 18, with an average of 2.11 children” (Greenfeld & Snell, 1999 pg. 1). Furthermore, “72% of women on probation, 70% of women held in local jails, 65% of women in State prisons, and 59% of women in Federal prisons have young children (Greenfeld & Snell). A majority of the women lived with their children prior to incarceration (Greenfeld & Snell). Although these differences are often highlighted by gender-specific theorists as factors that are important in female criminality, research suggests that childcare issues are not significantly correlated with recidivism. One study (Bonta, 1995) found that merely having children was not associated with recidivism, but also found that single parents had higher recidivism rates. Similar studies have not found significant correlations between recidivism and parenthood, or recidivism and single

14

parenthood, even when single mothers have serious financial difficulties related to single parenting (Rettinger, 1998; Rettinger & Andrews, 2010). Female offenders, on average, are more economically disadvantaged than male offenders and are less likely to have been employed prior to incarceration. Greenfeld and Snell’s 1999 special report on female offenders indicated that 4 in 10 female offenders were employed full-time prior to incarceration, compared with 6 in 10 male offenders. Nearly four times more female offenders than male offenders were receiving welfare prior to incarceration and 37% of women had income of less than $600 per month (Greenfeld & Snell). Female offenders scored significantly higher on the Financial domain of the LSI-R than males, indicating that female offenders have more severe financial hardships than males (Heilbrun, et al., 2008). As is the case with many of the other gender specific variables, there is a lack of research investigating whether poverty is a stronger predictor of offending in women than men (Blanchette & Brown, 2006). At least one study failed to find any differences between male and females in the relationship between education and employment and recidivism (Makarios, Steiner, & Travis, 2010). Advocates for the gender-specific theory of female criminality emphasize the role of relationships in female criminality. In the gender specific theory, relationships are viewed as uniquely important to the development and functioning of women; when they are negative, this may lead to substance abuse, risky behaviors, and ultimately, criminal offending (Staton-Tindall, et al., 2011). Some gender specific theorists assert that codependent relationships frequently facilitate criminal behavior (Koons, et al., 1997).

15

Recent research on the role of relationships shows that for females, a positive relationship with parents is associated with decreased likelihood of using drugs and fewer HIV risk behaviors. Furthermore, negative peer influence was correlated with high rates of drug use (Staton-Tindall). Only one study to date could be found on the effect of relationships on criminal behavior. The non-support (NON) scale of the Personality Assessment Inventory (PAI), which measures social support and quality of existing relationships based on self-report, was not found to be a significant predictor of arrest, but did significantly predict stress level, measured by the stress score on the PAI (Anumba, DeMatteo, & Heilbrun, 2012). Proponents of the gender-specific theory of criminality contend that the failure to consider the “well-established gender differences in the prevalence, incidence, and developmental course of antisocial behavior” between male and female offenders in riskneeds assessment results in overclassification of women as high risk and failure to address important gender-specific needs in treatment planning (Rettinger & Andrews, 2010, p. 31). However, despite numerous studies identifying important differences between male and female offenders, there is a lack of research on whether these “genderspecific” differences are actually causally related, or even correlated with recidivism. Furthermore, recent research has indicated that there may not be gender-related differences in risk factors for crime (Blanchette & Brown, 2006; Heilbrun et al., 2008). Salisbury, Van Voorhis, and Spiropoulos (2009), leading proponents of the gender specific theory of female criminality, investigated the role of several gender specific factors and found that a number of them were not related to outcome in the

16

community. Specifically, mental health status, self-esteem, relationships, and childhood victimization were not related to community outcome. Adult emotional abuse, physical abuse, and harassment, and LSI-R scores were correlated with community outcome, however. Parental stress was found to be related to technical violations and self-efficacy was found to be a protective factor. It is important to note that the addition of genderspecific variables did not increase the predictive validity of the LSI-R with respect to community outcomes, but did with respect to serious prison misconducts. Like the LSI-R studies, the sample in this study was also predominantly white (53.3% white, 28.8% black and 16% Hispanic). The pilot study was conducted in Western US state on a relatively small sample (n=134 for the release sample). A subsequent study (Van Voorhis, 2010) further investigated the relationship between gender-specific variables and community outcome in 8 female samples from Maui, Colorado, Minnesota, and Missouri, reporting that of fifteen gender specific factors, only five were significantly associated with outcome (parental stress, self-esteem and self-efficacy, family support and educational assets). However, the study also provided support for the predictive validity of gender neutral variables, as most of the gender-neutral factors from the LSI-R were significantly correlated with rearrests and incarcerations. The gender-neutral variables with the strongest correlations were criminal history, substance abuse, financial problems, education and employment, and housing. Again, the samples investigated were predominantly Caucasian. It is also important to note that there were differences across samples in the tests used and gender-responsive

17

data collected—six assessment tools were used. There were also differences in follow-up periods. In addition to investigating the predictive validity of the LSI-R in females, the Rettinger and Andrews study (2010) included gender-specific variables, and found that no indicators of emotional distress (including depression, phobias, eating disorders, anxiety and nervousness, sleeping difficulties and mental health system involvement) were related to reoffending. However, minority status, and history of abuse were significantly correlated with reoffending. Self-abuse (suicide attempts and self-injury) was reported at high rates (30-38%) and was found to predict future violence, but not general reoffending. Factors related to parenting were not correlated with recidivism (single parenthood status, stress from parenting responsibilities). The relationship between financial difficulties and reoffending proved interesting, as current use of welfare was not correlated with recidivism, but reliance on welfare was associated with general recidivism--and not wanting legitimate employment was correlated with general and violent reoffending. Although there were important relationships between certain gender-specific variables and reoffending, the only gender specific variables with incremental validity relative to the LS/CMI general risk/need score were official history of violence and measure of anger or violence as a problematic condition. The most recent meta-analysis of LSI-R and LS/CMI studies by Andrews et al. (2012), discussed at length above, found that substance abuse was the only domain of the LS/CMI that was more strongly related to recidivism in females than males. Importantly,

18

it also showed evidence of overprediction of recidivism in lower risk females for the first time, providing support for the “overclassification” theory of gender specific theorists. 1.5 Rationale At present, there are conflicting studies on the predictive validity of “gender specific” variables. Although the LSI-R and LS/CMI, both based on the RNR model, seem to predict recidivism in female offenders generally, there is a disagreement within the existing literature on whether adding gender-specific variables enhances the predictive validity of the LSI-R. Research is quite consistent on the predictive validity of the LSI-R in females, but there are no studies on the predictive validity of the LSI-R in females from the Northeastern United States or in samples that are predominantly African American. This is important because demographics can vary greatly by region--and race, culture and ethnicity may play a role in risk factors for recidivism. The aim of the current study is to assess the predictive validity of the LSI-R in female offenders from New Jersey, from a sample that includes a large percentage of Caucasian, African American and Hispanic females. The predictive validity of several gender specific variables will also be investigated to determine whether they are related to recidivism, and if so, whether they add to the predictive validity of the LSI-R. 1.6 Hypotheses 1.6.1 Gender-Specific Variables It is hypothesized that the anxiety, suicidal ideation, depression, non-support and borderline features full scale scores of the PAI will be significant predictors of

19

recidivism. It is further hypothesized that a history of sexual abuse and a history of domestic abuse will not be significant predictors of recidivism. 1.6.2 LSI-R Variables Consistent with previous studies, it is hypothesized that LSI-R total score and all sub-scales will be a significant predictor of recidivism. Also consistent with previous studies, it is hypothesized that the Substance Abuse subscale of the LSI-R will be the strongest predictor of recidivism. The bulk of the existing research on the LSI-R supports its predictive validity in females, while results are still mixed on the gender-specific variables. This is why the hypotheses only predict a significant relationship between LSI-R scores and recidivism. CHAPTER TWO: METHODS 2.1 Participants The archival records of 300 female offenders were reviewed as part of this study. The 300 offenderwere released from a minimum security private assessment and rehabilitation center in New Jersey between 2004 and 2007. Of the 300 females included in the sample, 191 had been transferred to the assessment and rehabilitation center directly from New Jersey state prisons to aid in a gradual transition from prison to the community. These females were still in custody of the New Jersey Department of Corrections (DOC) while residing at the center and are thus deemed “DOC females.” The remaining 109 females (the “halfway back” females) were sent to the assessment and rehabilitation centers from the community after violating a parole condition, which is usually related to substance abuse relapse. These halfway back females were under the

20

supervision of the New Jersey Parole Board while residing at the assessment and treatment center and were sent to the center to receive treatment for a set period of time rather than return to prison (thus, they are “halfway back” to prison). The typical length of stay at the assessment and treatment centers ranges from 60 to 90 days. Only a small number of the DOC females who were released from the private assessment and treatment centers were released directly into the community, as most of the females were released from the assessment and treatment centers into halfway houses. However, there are no data on the number of females who were released directly into the community rather than to a halfway house. Generally, most DOC females go to halfway houses upon leaving the assessment and treatment center, where they stay an average of 6-9 months. Some of the halfway back females may stay in halfway houses, depending on need. The purpose of the assessment and treatment centers, operated by Community Education Centers, is to conduct thorough assessment to tailor rehabilitation to each offender’s specific needs and to assist the New Jersey Department of Corrections in making informed placement decisions for the DOC females. For those offenders who enter the facilities directly from prison, another purpose of the minimum security centers is to help offenders gradually transition to life in the community. Only 20% of inmates in New Jersey state prisons are transitioned from prison to the private assessment and treatment centers in New Jersey prior to their eventual release. To be considered for placement in these facilities from prison, inmates must be 18 months from parole eligibility date (this requirement has since changed to 24 months

21

from parole eligibility date), be classified as minimum security status, and have no prior adult arson or sex offenses. The facilities require a requisite level of participation and inmates may be returned to prison for disciplinary issues or failure to fulfill the requirements of the program. The sample of inmates who attend Community Education Centers across the state of New Jersey are similar in age, ethnicity/race, criminal history, and substance abuse history to those of the entire New Jersey Department of Corrections population (Heilbrun et al., 2008). 2.2 Materials All individuals entering the assessment and rehabilitation center were assessed by master’s level assessment counselors shortly after arrival at the facility. The standard assessment process at the assessment and rehabilitation center includes a thorough and structured interview, a review of each individual’s records from the Department of Corrections, and a standard battery of assessment tests. The assessment counselors were trained to administer all of the assessment tests properly by a PhD-level psychologist who is also a LSI-R master trainer. Based on the self-report information obtained during the interview and a detailed review of each individual’s file, a detailed assessment report was written for each individual that includes a biographical history, criminal history, test results and explanation of test results, and treatment and placement recommendations based on all information collected during the assessment process. The LSI-R was administered to all females entering from August 2003 to March 2006. After 2006, the LS/CMI was administered. Because there were fewer females who had been administered the LS/CMI at the time of data collection for this study, and fewer

22

of the LS/CMI had been released for more than one year at the time of collecting followup data, only females who had been administered the LSI-R were included in the present study. The LSI-R has satisfactory inter-rater and test-retest reliability, as Andrews (1982) reports inter-rater and test-retest reliability over several trials that ranged from r=.80 to .99. The validity of the LSI-R in predicting recidivism in both male and female offenders was discussed previously. In addition to the risk assessment tools, individuals entering the assessment and rehabilitation center were administered the Texas Christian University Drug Screen (TCU-DS-II), a measure used to detect alcohol and drug use disorders (Simpson, Knight & Broome, 1997). In a sample of 400 male inmates, the TCU-DS-II’s accuracy in detecting alcohol or drug dependence was 82.1% and the test-retest reliability was .95 (Peters, et al., 2000). All females were also administered the Personality Assessment Inventory (PAI) (Morey, 1991), a test of personality and psychopathology similar to the Minnesota Multiphasic Personality Inventory (MMPI). The PAI manual reports internal consistency coefficients from .45 to .90 and the test-retest reliability ranges from .31 to .91 (Morey, 1991). The PAI is comprised of four types of scales: validity scales, clinical scales, interpersonal scales, and treatment consideration scales (Morey, 1991). 2.3 Design and Procedure Upon initial assessment at the assessment and treatment center, a file is created for all individuals entering the center that includes their assessment interview and report, the records from their time at the assessment and treatment center, and their raw test scores. A database with 241 variables was created in SPSS based on a typical file. A

23

corresponding coding manual with operational definitions for each variable was created to assist in data entry. The information from the files of all 912 females who attended the center from 2004-2006 were entered in the database by one researcher, who created the database and coding manual. Also included in the database are the total and subscale scores for all of the tests administered, and variables related to mental health history, family history, criminal history, medical, education, employment, religion, and substance abuse. Some of the variables are self-reported variables (i.e., number of children and history of sexual and domestic abuse) and other are from the official Department of Corrections file (i.e., number of prior arrests, convictions, and probation and parole violations). After all of the files were entered into the database, a random sample of 300 individuals was sent to the New Jersey Department of Corrections to obtain outcome data. Three research assistants trained on the database and coding manual entered the outcome data into the database and immediately thereafter, all identifying information was deleted from the database. The file from the assessment and rehabilitation center included the date each individual was discharged from the facility, but as stated previously, did not state where each individual was discharged to, as a few individuals were released directly back into the community rather than being discharged to a halfway house. The New Jersey Department of Corrections did not have the ability to provide the date each individual was released from the halfway house attended, which will be discussed further in the limitations section.

24

The variables to be examined in this study include: the LSI-R total score and eight subscale scores, reported history of sexual abuse and history of domestic abuse, and the anxiety, depression, borderline features, and non-support scales of the PAI. Although the relationship between non-support and re-arrest has already been investigated in this sample by Anumba, DeMatteo, and Heilbrun (2012), that study had only a one year follow-up period with a rather low rate of re-arrest. The dependent variable is whether the individual was arrested in the two years after release from the center. Data on the number of days from release to first arrest is available, but will not be investigated because of the issues regarding halfway house attendance discussed above. Whether an individual attended a halfway house immediately after release from the assessment and rehabilitation center may affect how soon they are re-arrested. There is less concern over the effect of halfway house attendance on re-arrest over the two year follow-up period. Since the follow up period is two years, all individuals who attended halfway houses will have had an ample amount of opportunity and “street time” to be arrested. The issues regarding halfway house attendance will be discussed further in the limitations section. CHAPTER 3: METHOD OF ANALYSIS A point biserial correlation was run to examine the relationship between the independent and quasi-independent variables (LSI-R total score and subscores, and the scores on the anxiety, depression, borderline features, and non-support scales of the PAI) and the categorical dependent variable (whether the individual was arrested after two years). A chi-square was run to examine the relationship between the categorical

25

variables (whether the subject has a history of domestic abuse and/or a history of sexual abuse) and dependent variable (whether the individual was arrested in the two years after release into the community). CHAPTER 4: RESULTS 4.1 Descriptive Statistics The mean age of the sample was 35.75 years (SD=.816), with subjects ranging in age from 18 to 66 years (See Table 1). As stated previously, the racial makeup of the sample varied from that of previous female studies and the most recent national statistics on female inmates (Carson & Sabol, 2012), as the current sample was predominantly African American. As can be seen in Table 2, the racial makeup was as follows: African American (n=166, 55.3%), Caucasian (n=96, 32.0%), Hispanic (n=35, 11.7%), Native American (n=2, .7%) and Other (n=1, .3%). The most recent national statistics on females inmates serving more than one year in state and federal prison had the following racial breakdown: 49.3% White, 25.1% Black, 17.7% Hispanic, and 8.0% Other (See Table 2). It is interesting to note that the LSI-R norms for female offenders in the United States, created by the authors of the LSI-R, used a sample that had a racial makeup that was predominantly Caucasian, also. The racial makeup of that sample was: 65.79% Caucasian, 14.15% African American, 1.83% Hispanic, 3.98% Native American, .27% American Indian, .11% Unknown, and 11.89% Other (See Table 2). The sample size of female inmates was 1,859, of which only 263 were African American. Similar to the

26

present study, the ages ranged from 16-67, with a mean of 34.65 (SD= 9.00) (Andrews & Bonta, 2003b). At the time of assessment, 36% of the sample were incarcerated for a technical violation (n=108), 12.7% were incarcerated for a violent offense (n=38), 16.7% were incarcerated for a property offense (n=50), and 22.3% were incarcerated for a drug offense (n=67). (The numbers do not add up to 100% because inmates were frequently incarcerated for more than one offense and it was unclear what the current offense was for a small number of inmates). Seven inmates were currently incarcerated for murder or manslaughter (2.3%). The mean number of previous adult arrests was 11.52 (SD= 11.02). The mean LSI-R score of the sample was similar to that of the US norms published by Andrews and Bonta in 2003. The mean of our sample was 29.75 (SD= 6.65) and the mean of the sample used to create the US norms was 28.44 (SD= 7.75). As can be seen in Table 1, the means of the individual LSI-R subscores ranged from 1.01 (SD= 1.08) on the Attitudes and Orientations sub-scale, to 2.67 (SD= 1.11) on the Companions sub-scale. PAI subscale means ranged from a low of 38.67 (SD= 9.87) for the Treatment Rejection (RXR) scale to a high of 74.04 (SD= 18.94) on the Drug scale (See Table 1). Three participants’ PAI scores were not included because their PAI profile indicated that the PAI scores were invalid. The low mean Treatment Rejection score indicates that the sample reported having attitudes indicating amenability to treatment. Although the sample reported high rates of drug usage and issues related to drug usage, as evidenced by the clinically elevated Drug scale mean of 74.04 (SD= 18.94), the Alcohol subscale

27

mean for the sample was much lower, at 53.38 (SD= 14.55). As can be seen in Table 1, none of the subscales that this study focuses on (Anxiety, Depression, Borderline Features, and Non-support scales) were in the clinically elevated range (PAI sub-scores over 70 are considered clinically elevated). The only mean score in the clinically elevated range was the Drug scale mean. Approximately one in four inmates (26%) were rearrested in the 2 years following their release from the assessment and treatment center (n=78) (See Table 4). African American and Hispanic participants had the exact same arrest rate at 22.9%, while Caucasian subjects had a much higher re-arrest rate at 32.3%. A chi-square, however, revealed no statistically significant difference in re-arrest rate between African American, Caucasian, and Hispanic subjects χ2 (2, n=295) = 3.10, p= .212, V=.102. Those subjects who identified themselves as Native American (n=2) or Other (n=1) were not included in the chi-square because there were only 3 subjects in those groups. Of the 78 females who were rearrested in the 2 years post-release, 37.2% (n=29) were arrested for drug offenses, 15.4% (n=12) were arrested for theft offenses, 12.8% (n=10) were arrested for assault, 7.7% (n=6) were arrested for prostitution, 1.3% (n=1) were arrested for a parole or probation violation, 2.6% (n=2) were arrested for an “other violent offense,” 5.1% (n=4) were arrested for an “other property offense,” and 17.9% (n=14) were arrested for an “other offense” which is neither a property offense nor a violent offense (See Table 5.)

28

4.2 Inferential Statistics Just under half of the sample (47%) reported a history of domestic abuse and approximately one-third of the sample (37.3%) reported a history of sexual abuse (See Table 3). A chi-square test found no significant difference in rearrest rate between those inmates with a history of domestic abuse compared with those who did not have a history of domestic abuse, χ2 (1, n=296) = .093, p=.760, φ= -.018. Similarly, a chi-square test found no significant difference in re-arrest rate in the 2 years post release between inmates with a history of sexual abuse and inmates without a history of sexual abuse χ2 (1, n=297) = .254, p=.614, φ= .029. One-fourth of the sample (24.7%, n= 74) had a history of both sexual abuse and domestic abuse. There was no significant difference found in re-arrest rate for inmates who had a history of both sexual and domestic abuse compared to those who had only a history of one type of abuse or no abuse history χ2 (1, n=296) = .580, p= .446, φ= .044. Nearly 60% (59.5%, n=178) of the sample had a history of either sexual abuse or domestic abuse. A chi-square revealed no significant difference in re-arrest rate between those who had a history of either type of abuse and those who had no history of abuse χ2 (1, n=296) = .580, p= .446, φ= -.027 (See Table 5). A point biserial correlation was used to investigate the relationship between the PAI variables and LSI-R scores and re-arrest. The anxiety, depression, non-support, and suicidal ideation subscales were not significantly correlated with recidivism (see Table 6). However, a significant relationship was detected between the borderline subscale and re-arrest, r (297) = .133, p = .023. It was a significant positive relationship, as a higher

29

score on the borderline scale was significantly correlated with a higher risk of re-arrrest. The effect size, however, was small, as only 1.8% of the variance in rearrest was explained by the borderline score (r2= .018). Although it was hypothesized that that the overall score and all of the sub-scores of the LSI-R would be significantly correlated with recidivism, only the overall score r (298) = .097, p = .047, the Criminal History subscore r (298) = .106, p = .034, the Financial subscore r (298) = -.159, p = .003, and the Alcohol/Drug subscore r (298) = .120, p = .019, were significantly correlated with re-arrest. It should be noted that the relationship between overall score and rearrest was barely significant (p =. 047). Furthermore, the effect sizes were small for all scores that were significantly correlated with recidivism, with r2 values ranging from .009 for Overall score to .025 for the Financial subscore (See Table 8). The Financial subscore had the higher r2 value, but still only 2.5% of the variance in rearrest could be explained by the Financial subscore. Interestingly, the Financial subscore showed the strongest relationship with rearrest . Furthermore, the relationship was a negative relationship, as the lower the score on the Financial subscore, the higher the risk of re-arrest. The subscores of the LSI-R are scored from 0-4; the higher the score, the higher the risk (0= very low risk, 4= very high risk). Therefore, according to our results, the better a female’s financial situation was, the more likely she was to be re-arrested in the two years after release into the community. CHAPTER 5: DISCUSSION Previous studies have investigated the predictive validity of the LSI-R in female offenders, but there are no published studies to date on the predictive validity of the LSI-

30

R in female offenders in the Northeastern United States. This is the first study to investigate the predictive validity of the LSI-R in New Jersey female offenders, using a predominantly minority sample. Furthermore, this study explored the predictive validity of variables that have been considered “gender specific” factors in recidivism. Like the LSI-R, gender specific factors have not been researched in female offenders from New Jersey prior to the current study. With respect to the LSI-R, the results indicate that only the Overall score and the Criminal History, Financial, and Alcohol/Drug subscores were significantly correlated with recidivism (see Table 8). It is noteworthy that the Overall score was barely significant at the .05 level (p=.047). The remaining subscores were not correlated with recidivism (Education/ Employment, Accommodations, Leisure/Recreation, Companions, Emotional/Personal, and Attitudes/Orientations) (see Table 8). The LSI-R results are surprising for a number of reasons. First, the existing research on the predictive validity of the LSI-R in females has consistently found significant relationships between not only the Overall score and recidivism (Smith, Cullen & Latessa, 2009), but between all subscores and recidivism (Rettinger & Andrews, 2010). The most recent meta-analysis conducted by Andrews et. al, even found a stronger relationship between LSI-R Overall score and recidivism in females (r = .59) than in males (r = .39) (2012). Furthermore, the subscores that were significantly correlated with rearrest had very small effect sizes; each subscore significantly correlated with recidivism only explained 1-2% of the variance in rearrest.

31

The inaccuracy of the LSI-R in predicting recidivism in the present sample may be due to the racial diversity of the sample. As mentioned previously, the existing studies investigating the relationship between LSI-R scores and recidivism used predominantly Caucasian samples. Furthermore, the most recent meta-analyses were based on Canadian samples (Rettinger & Andrews, 2010; Andrews, et. Al, 2012) that may differ both culturally and racially from the current sample. At least two previous studies have shown that the LSI-R to be less accurate in predicting recidivism in African American males than in Caucasian males (Fass, et al., 2008; Schlager, 2007), indicating that race may affect the accuracy of prediction in female offenders, as well. Interestingly, our results reveal a negative relationship between the Financial subscore of the LSI-R and recidivism (see Table 8). This was the strongest relationship of all of the statistically significant correlations (r = -.159, p = .003) but still only accounted for 2.5% of the variance in rearrest. Ultimately, as a female’s finances improved, so did her risk of recidivism. This result is contrary to the hypotheses of the present study and the existing LSI-R studies with females and males, and is also contrary to the gender specific theory, which highlights financial problems as an important factor in female criminal behavior and recidivism (Van Voorhis, et al., 2010). It is unclear why there is a negative relationship between Financial subscore and recidivism, although there are a few possibilities. The negative relationship between the Financial subscore and recidivism may be related to one of the limitations of the study, discussed below, regarding the unknown halfway house status of our sample. A large percentage of our sample attended halfway houses after release from the assessment and

32

treatment center, as most females recently discharged from the assessment and treatment center attend halfway houses. However, we do not have information about which females attended halfway houses and how long they remained there prior to being released directly into the community. The halfway house likely acts as a protective factor for recidivism—theoretically it should be easier to remain law abiding in a structured facility with supervision and ongoing treatment/education than to be living independently. Those females who were in a better financial position at the time of discharge may have been more likely to be released to their home, or have a shorter stay at a halfway house, consequently increasing their risk of recidivism. Another possibility is that having access to money increases the ability to purchase drugs or alcohol, making relapse and recidivism more likely. Furthermore, a female with a better financial position may have better job skills, and a higher level of education than other females, meaning they will be more likely to find a job after discharge into the community. While this may appear to be a protective factor, the stress of having a job may increase the risk of substance abuse and recidivism. The lack of a significant relationship between the Education/Employment and Companion subscores is relevant not only to assessing the validity of the LSI-R in females, but also to assessing the validity of the gender specific theory. Gender specific theorists emphasize the role of relationships/companions and lack of education and employment in female criminal behavior (Van Voorhis, et al., 2010). Therefore, even gender specific theorists would expect that the Education/Employment and Companion subscores would be significantly related to recidivism. In fact, if the gender specific

33

theory is correct in its emphasis on finances, substance abuse, and relationships in recidivism, we would expect stronger relationships between those subscores and recidivism relative to the other LSI-R subscores that are not considered gender specific. That was not the case in the present study. Also inconsistent with the gender specific theory of criminality is the lack of a significant relationship between the Anxiety, Depression, Non-Support and Suicidal ideation subscores of the PAI and recidivism. The only PAI subscale tested that was significantly correlated with recidivism was the Borderline subscale. This may be the most important finding of the current study, and the only finding that lends support to the gender specific theory of criminality. The Borderline subscale of the PAI has been found to be a reliable screening instrument for assessing traits and characteristics of Borderline Personality Disorder (Gardner & Qualter, 2009). Of all of the variables investigated in the present study, the Borderline subscore of the PAI is the only variable that is truly a variable specific to females, as 75% of all individuals diagnosed with BPD are females (APA, 2000). Furthermore, BPD is associated with many of the variables that gender specific theorists classify as gender specific. For example, Borderline Personality Disorder is highly co-morbid with substance abuse and Axis I disorders, including Depression and Anxiety (APA, 2000; van den Bosch, Hysaj & Jacobs, 2012). Also, individuals who suffer from BPD often have high rates of sexual and physical abuse in childhood and adulthood, and high rates of co-morbid PTSD (Wingenfeld, et al., 2011; Yen, et al., 2002). BPD has been characterized as a “relationship disorder” (Linehan, 1993), as

34

interpersonal impairments and “unstable and intense interpersonal relationships” are a hallmark of the disorder (American Psychiatric Association, 2013). Also, many of the features or symptoms of BPD “tend to occur in interpersonal contexts” (Wright, et al., 2013). BPD is also associated with lower education levels and underemployment (Bland et al., 2004; Gardner & Qualter, year), and with criminality, as rates of BPD are much higher in prison populations than in the community (Black et al., 2007; Sansone & Sansone, 2009). Even in prisons, however, the rate of BPD in females is two to three times that of males (Black et al.; Zlotnick et al., 2008). A recent study of female and male offenders with Borderline Personality Disorder found that over half (54.5%) of all females met criteria for BPD. Furthermore, BPD was associated with higher LSI-R scores, suggesting a link between BPD and recidivism (Black et al.). Further research is necessary to investigate the role of BPD in recidivism in females. BPD may be a “super variable” for predicting recidivism among females, as it is a variable that affects females 2-3 times more than males and also encompasses many gender specific variables. 5.1 Limitations The present study has a number of limitations that must be considered when analyzing the results. As stated previously, the date which began the two-year follow up period was the date of release from the assessment and treatment center, not the precise date that the individual was released into the community. Many females were released to halfway houses upon completion in the assessment and treatment center, where they

35

stayed an average of 6-9 months. Although it is entirely possible to be re-arrested while in a halfway house, it is likely that a halfway house has some protective influence over re-arrest that may have played a role in the current study’s low re-arrest rate. Another limitation related to the halfway house issue and the low re-arrest rate is that of technical violations. Data on technical violations was not available for the current study. Therefore, there may have been individuals in the sample who were sent back to prison after a technical violation for which they were never arrested or criminally charged. Since the current study used only re-arrest as the only outcome measure, those individuals would have been represented as not re-arrested during the two-year follow up, even though they were not truly successful upon release and were not re-arrested only because they were incarcerated. The fact that many females attended halfway houses plays a role as well, as those females who attended halfway houses may have been more likely to be returned to prison for violations without being charged with a crime (i.e. being sent back to prison for failing to follow the rules of the halfway house). It should also be noted that the current study only includes arrest data from the state of New Jersey. Participants may have been arrested in other jurisdictions. Furthermore, this study was not able to identify criminal behavior(s) that occurred that did not result in arrest. Future studies should use a variety of recidivism variables, including not only re-arrest, but also convictions, technical violations, failed drug tests, and self- report measures of criminal behavior and/or substance abuse. Future studies should also include arrest and conviction data from all states, if possible.

36

The participants in the current study may systematically differ from those of other studies and may also differ from the female New Jersey Department of Corrections population, despite being similar in age, ethnicity/race, criminal history and substance abuse history (Heilbrun, et al., 2008). The results of the PAI testing indicated that the study sample had very low scores on the treatment rejection scale, indicating that they were amenable to treatment. This may have been due to a number of factors. First, to be transferred into the assessment and treatment center, and to successfully complete it, inmates must comply with the rules and requirements of the center. Those inmates who are successfully discharged from the facility may differ from those who are sent back to prison for failure to follow the rules of the center. The current study only includes individuals who were successfully discharged from the facility. Second, the assessment and treatment facility’s programming may actually increase acceptance to treatment. The victimization data used in the current study was based on each participant’s self-report of whether they had been a victim of sexual abuse or domestic abuse. No information was obtained on the severity of the abuse, the recency of the abuse, or the perpetrator. It is important for future research to investigate these variables to determine whether they are associated with recidivism. Someone who had suffered from less severe abuse at a young age may have a much different risk of recidivism than someone who recently suffered severe abuse. Furthermore, abuse data was collected for the present study in the process of routine assessment at the assessment and treatment center. It was not conducted for the purpose of the study, so at the time of collection there was no operational definition of abuse. It was based on both the interviewer and the participant’s

37

opinion of what constituted abuse. Future research should operationally define the abuse variable prior to obtaining data and use that definition when interviewing participants. As noted previously, the racial makeup of the current study differs from the population of all female offenders in the United States. Therefore, the study may only be generalizable to African American, Caucasian, and Hispanic females being released from prison in New Jersey. Further research should be conducted in New Jersey and across the country with samples that differ not only racially, but also geographically and culturally.

38

References

Abram, K. M., Teplin, L. A., & McClelland, G. M. (2003). Comorbidity of severe psychiatric disorders and substance use disorders among women in jail. American Journal of Psychiatry, 160, 1007-1010. American Psychiatric Association (2000). Diagnostic and statistical manual of mental disorders: DSM-IV-TR (4th ed. text revision). Washington, D.C.: American Psychiatric Association. American Psychiatric Association (2013). Diagnostic and statistical manual of mental disorders (5th ed.). Washington, DC: Author. Andrews, D. A. (1982). The Level of Service Inventory (LSI): The First Follow Up. Toronto, Ontario Ministry of Correction Services. Andrews, D. A., & Bonta, J. (1995). LSI-R: The Level of Service Inventory-Revised. Toronto, Ontario: Multi-Health Systems. Andrews, D. A., & Bonta, J. (2003a). The Psychology of Criminal Conduct. 3d ed. Cincinnati, OH: Anderson Publishing Company. Andrews, D. A., & Bonta, J. (2003b). LSI-R Norms Manual Supplement. Toronto, Ontario: Multi-Health Systems. Andrews, D. A., & Bonta, J. & Wormith, J. S. (2004). Level of service/case management inventory: LS/CMI manual. Toronto, Ontario: Multi-Health Systems. 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 offending? A test with the eight domains of the Level of Service/Case Management Inventory. International Journal of Offender Therapy and Comparative Criminology, 56, 113-133. Andrews, D. A. & Robinson, D. (1984). The Level of Supervision Inventory: Second report. Toronto, Ontario, Canada: Ontario Ministry of Correctional Services. Anumba, N., DeMatteo, D., & Heilbrun, K. (2012). Social functioning, victimization, and mental health among female offenders. Criminal Justice and Behavior, 39, 12051219.

39

Belknap, J., & Holsinger, K. (2006). The gendered nature of risk factors for delinquency. Feminist Criminology, 1, 48-71. Blanchette, K., & Brown, S. L. (2006). The assessment and treatment of women offenders: An integrative perspective. West Sussex, UK: Wiley. Black, D. W., Gunter, T., Allen, J., Blum, N., Arndt, S. Wenman, G., & Sieleni, B. (2007). Borderline personality disorder in male and female offenders newly committed to prison. Comprehensive Psychiatry, 48, 400-405. Bland, A. R., Williams, C.A., Scharer, K., & Manning, S. (2004). Emotion processing in borderline personality disorders. Issues in Mental Health Nursing, 25, 655-672. Bloom, S. B, Owen, B., & Covington, S. (2003). Gender responsive strategies: research, practice, and guiding principles for women offenders. Washington, DC: U.S. Department of Justice, National Institute of Corrections. Bonta, J., Pang, B., & Wallace-Capretta, S. (1995). Predictors of recidivism among incarcerated female offenders. The Prison Journal, 75, 277-294. Braithwaite, R., Treadwell, H., & Arriola, K. (2005). Health disparities and incarcerated women: A population ignored. American Journal of Public Health, 95, 16791681. Bureau of Justice Statistics (2008). Sexual Victimization Reported by Former State Prisoners. Washington, D.C. US Department of Justice, Office of Justice Programs. Retrieved December 7, 2012 from http://bjs.ojp.usdoj.gov/content/pub/pdf/svrfsp08.pdf. Canadian Centre for Justice Statistics (2001). Women in Canada (Catalogue No. 85F0033MIE). Ottawa, Ontario, Canada: Statistics Canada. Carson, E. A., & Sabol, W. J. (2012). Prisoners in 2011. (NCJ 239808). Washington, D. C.: US Department of Justice, Office of Justice Programs, Bureau of Justice Statistics. Retrieved February 23, 2013 from http://bjs.ojp.usdoj.gov/index.cfm?ty=pbdetail&iid=4559 Coulson, G., Ilacqua, G., Nutbrown, V., Giulekas, D. & Cudjoe, F. (1996). Predictive utility of the LSI for incarcerated female offenders. Criminal Justice and Behavior, 23, 427-439. Covington, S. (1998). Women in prison: Approaches in the treatment of our most invisible population. Women and Therapy, 21, 141-155.

40

Covington, S. S. (2007). Women and the criminal justice system. Women’s Health Issues, 17, 180-182. DeHart, D. D. (2008). Pathways to prison: Impact of victimization in the lives of incarcerated women. Violence Against women, 14, 1362-1381. Ditton, P. M. (1999). Mental health and treatment of inmates and probationers (Special Report NCJ 174463). Bureau of Justice Statistic, US Department of Justice. Retrieved December 16, 2012 from http://bjs.ojp.usdoj.gov/index.cfm?ty=pbdetail&iid=787 Ellis, L. (1988). The victimful-victimless crime distinction, and seven universal demographic correlates of victimful criminal behavior. Personality and Individual Differences, 9, 525-548. Fass, T., Heilbrun, K., DeMatteo, D. & Fretz, R. (2008). The LSI-R and Compas: Validation data on two risk-needs tools, 35, 1095-1108. Federal Bureau of Investigation (2002). Crime in the United States in 2002. Uniform Crime Reporting Program. Retrieved November 14, 2009 from http://www.fbi.gov/ucr/cius_02/html/web/. Flores, A. W., Lowenkamp, C. T., Holsinger, A. M., & Latessa, E. J. (2006a). Predicting outcome with the Level of Service Inventory-Revised: The importance of implementation integrity. Journal of Criminal Justice, 34, 523-529. Flores, A. W., Lowenkamp, C. T., Smith, P., & Latessa, E. J. (2006b). Validating the Level of Service Inventory-Revised on a Sample of Federal Probationers. Federal Probation, 70, 44-48. Frost, N. A., Greene, J., & Pranis, K. (2006). Hard hit: The growth in the imprisonment of women, 1977-2004. Institute on Women & Criminal Justice: The Punitiveness Report. Gardner, K. & Qualter, P. (2009). Reliability and validity of three screening measures of BPD in a nonclinical population. Personality and Individual Differences, 46, 636641. Gendreau, P., Little T., & Goggin, C. (1996). A meta-analysis of the predictors of adult offender recidivism: What works! Criminology, 34, 575-607. General Accounting Office. (1999). Women in Prison: Issues and challenges confronting US correctional systems, Washington, D.C.: Author.

41

Glaze, L. E., & Palla, S. (2005). Bureau of Justice Statistics Bulletin: Probation and Parole in the United States, 2004 (NCJ 210676). Washington, D. C.: U.S. Department of Justice, Office of Justice Programs, Bureau of Justice Statistics. Retrieved December 3, 2012, from http://bjs.ojp.usdoj.gov/index.cfm?ty=pbdetail&iid=1108 Greenfeld, L. A., & Snell, T. L. (1999). Bureau of Justice Statistics Special Report: Women offenders (NCJ 175688). Washington, D.C.: U.S. Department of Justice, Office of Justice Programs, Bureau of Justice Statistics. Retrieved December 3, 2012 from http://bjs.ojp.usdoj.gov/index.cfm?ty=pbdetail&iid=568 Heilbrun, K., DeMatteo, D., Fretz, R., Erickson, J., Girardi, D., & Halpern, C. (2008). Criminal recidivism of female offenders: The importance of structured, community based aftercare. Corrections Compendium, 33, 1-2, 30-32. Heilbrun, K., DeMatteo, D., Fretz, R., Erickson, J., Yasuhara, K., & Anumba, N. (2008). How “specific” are gender-specific rehabilitation needs? An empirical analysis. Criminal Justice and Behavior, 35, 1382-1397. Holsinger, A. M., Lowenkamp, C. T., & Latessa, E. J. (2006). Exploring the validity of the Level of Service Inventory-Revised with Native American offenders. Journal of Criminal Justice, 34, 331-337. Knight, K., Simpson, D. D., Morey, J. T. (2002). Evaluation of the TCU Drug Screen. Retrieved December 14, 2012 from https://www.ncjrs.gov/App/Publications/abstract.aspx?ID=196682 LaRochelle, F., Castro, C., Goldenson, J. Tulsky, J. P., Cohan, D. L., Blumenthal, P. D., & Sufrin, C. B. (2012). Contraceptive use and barriers to access among newly arrested women. Journal of Correctional Health Care, 18, 111-119. Linehan, M. M. (1993). Cognitive-behavioral treatment of borderline personality disorder. New York, NY: Guilford Press. Lowenkamp, C. T., Holsinger, A. M., & Latessa, E. J. (2001). Risk/need assessment, offender classification, and the role of childhood abuse. Criminal Justice and Behavior, 28, 543-563. Lowenkamp, C. T., Latessa, E. J., & Holsinger, A. M. (2006). The risk principle in action: What have we learned from 13,676 offenders and 97 correctional programs? Crime and Delinquency, 52, 77-93.

42

Lowenkamp, C. T., & Bechtel, K. (2007). The predictive validity of the LSI-R on a sample of offenders drawn from the records of the Iowa Department of Corrections data management system. Federal Probation, 71, 25-29. Makarios, M., Steiner, B., Travis, L. F. (2010). Examining the predictors of recidivism of men and women released from prison in Ohio. Criminal Justice and Behavior, 37, 1377-1392. Manchak, S. M., Skeem, J. L., & Douglas, K. S. (2008). Utility of the Revised Level of Service Inventory (LSI-R) in predicting recidivism after long-term incarceration. Law and Human Behavior, 32, 477-488. Maruschak, L. M. (2009). Bureau of Justice Statistics Special Report: HIV in prisons (NCJ 228307). Washington, D.C.: U.S. Department of Justice, Office of Justice Programs, Bureau of Justice Statistics. Retrieved December 7, 2012 from http://bjs.ojp.usdoj.gov/index.cfm?ty=pbdetail&iid=1747 Maruschak, L. M., & Parks, E. (2012). Probation and Parole in the United States, 2011. Washington, D. C.: US Department of Justice, Office of Justice Programs, Bureau of Justice Statistics. Retrieved February 22, 2013 from http://bjs.ojp.usdoj.gov/index.cfm?ty=pbdetail&iid=4538 McClellan, D. S., Farabee, D. & Crouch, B. M. (1997). Early victimization, drug use, and criminality: A comparison of male and female prisoners. Criminal Justice and Behavior, 24, 455-476. McDaniels-Wilson, C., & Belknap, J. (2008). The extensive sexual violation and sexual abuse histories of incarcerated women. Violence Against Women, 14, 1090-1127. Morey, L. C. (1991). The Personality Assessment Inventory professional manual. Odessa, FL: Psychological Assessment Resources. Mumola, C. J., & Karberg, J. C. (2006). Bureau of Justice Statistics Special Report: Drug Use and Dependence, State and Federal Prisoners, 2004. (NCJ 213530). Washington, D. C.: U.S. Department of Justice, Office of Justice Programs, Bureau of Justice Statistics. Retrieved February 24, 2013 from http://bjs.ojp.usdoj.gov/index.cfm?ty=pbdetail&iid=778 Nunes-Dinis, M. C., Weisner, C. (1997). Gender differences in the relationship of alcohol and drug use to criminal behavior in a sample of arrestees. American Journal of Drug and Alcohol Abuse, 23, 129-41. Peters, R. H., Strozier, A. L., Murrin, M. R., & Kearns, W. D (1997). Treatment of substance-abusing jail inmates. Journal of Substance Abuse Treatment, 14, 339349.

43

Peters, R. H., Greenbaum, P. E., Steinberg, M. L., Carter, C. R., Ortiz. M. M., Fry, B. C., et al. (2000). Effectiveness of screening instruments in detecting substance abuse disorders among prisoners. Journal of Substance Abuse Treatment, 18¸349-358. Phillips, J. A., Nixon, S. J., & Pfefferbaum, B. (2002). A comparison of substance abuse among female offender subtypes. Journal of the American Academy of Psychiatry and Law, 30, 513-519. Raynor, P. (2007). Risk and need assessment in British Probation: The contribution of the LSI-R. Psychology, Crime and Law, 13, 125-138. Rettinger, L. J., & Andrews, D. A. (2010). General risk and need, gender specificity, and the recidivism of female offenders. Criminal Justice and Behavior, 37, 29-46. Rettinger, L. J. & Andrews, D. A. (2012). Are the major risk/need factors predictive of both female and male offending?: A test with the eight domains of the level of Service/Case Management Inventory. International Journal of Offender Therapy and Comparative Criminology, 56, 113-133. Sabol, W. J., West, H. C., & Cooper, M. (2009). Prisoners in 2008 (NCJ 228417). Washington, D. C.: US Department of Justice, Office of Justice Programs, Bureau of Justice Statistics. Retrieved December 3, 2012 from http://bjs.ojp.usdoj.gov/index.cfm?ty=pbdetail&iid=1763 Sacks, J. Y. (2004). Women with co-occurring substance use and mental disorders (COD) in the criminal justice system: A research review. Behavioral Sciences and the Law, 22, 477-501. Sacks, J. Y., McKendrick, K., Hamilton, Z., Cleland, C. M., Pearson, F. S., & Banks, S. (2008). Treatment outcomes for female offenders: Relationship to number of axis I diagnoses. Behavioral Sciences and the Law, 26, 413-434.’ Salisbury, E. J., & VanVoorhis, P. (2009). Gendered pathways: An empirical investigation of women probationers’ paths to incarceration. Criminal Justice and Behavior, 36, 541-566. Salisbury, E. J., Van Voorhis, P., & Spiropoulos, G. V. (2009). The predictive validity of a gender-responsive needs assessment: An exploratory study. Crime and Delinquency, 55, 550-585. Sansone, R. A., Sansone, L. A. (2009). The interface: Borderline personality and criminality. Psychiatry, 6, 16-20.

44

Schlager, M. D., Simourd, D. J. (2007). Validity of the Level of Service InventoryRevised (LSI-R) among African American and Hispanic male offenders. Criminal Justice and Behavior, 34, 545-554. Simourd, D. J. (2004). Use of dynamic risk/need assessment instruments among longterm incarcerated offenders. Criminal Justice and Behavior, 31, 306-323. Simpson, D. D., Knight, K., & Broome, K. M. (1997). TCU/CJ Forms Manual: Drug dependence screen and initial assessment. Fort Worth: Texas Christian University, Institute of Behavioral Research. Smith, P., Cullen, F. T., & Latessa, E. J. (2009). Can 14,737 women be wrong? A metaanalysis of the LSI-R and recidivism for female offenders. Criminology and Public Policy, 8, 183-208. Staton-Tindall, M., Frisman, L., Lin, H., Leukefeld, C., Oser, C. Havens, J. R., Prendergast, M. Surratt, H., & Clarke, J. (2011). Relationship influence and health risk behavior among re-entering women offenders. Women’s Health Issues, 21, 230-238. Stein, M. D., Caviness, C. M., & Anderson, B. J. (2012). Incidence of sexually transmitted infections among hazardously drinking women after incarceration, 22, 1-7. Van den bosch, L. M. C., Hysaj, M., & Jacobs, P. (2012). DBT in an outpatient forensic setting. International Journal of Law and Psychiatry, 35, 311-316. Van Voorhis, P., Wright, E. M., Salisbury, E., & Bauman, A. (2010). Women’s risk factors and their contributions to existing risk/needs assessment. Criminal Justice and Behavior, 37, 261-288. Vose, B. Lowenkamp, C. T., Smith, P. & Cullen, F. T. (2009). Gender and the predictive validity of the LSI-R: A study of Parolees and probationers. Journal of Contemporary Criminal Justice, 25, 459-471. West, H. C., & Sabol, W. J. (2008). Prisoners in 2007. Washington, D.C.: US Department of Justice, Office of Justice Programs, Bureau of Justice Statistics. Wingenfeld, K., Schafrath, C., Rullkoetter, N., Mensebach, C., Schlosser, N., Beblo, T., Driessen, M., Meyer, B. (2011). Associations of childhood trauma, trauma in adulthood and previous-year stress with psychopathology in patients with major depression and borderline personality disorder. Child Abuse and Neglect, 35, 647654.

45

Wright, E. M., Salisbury, E. J., & Van Voorhis, P. (2007). Predicting the misconducts of women offenders: The importance of gender-responsive needs. Journal of Contemporary Criminal Justice, 23, 310-340. Wright, A. G., Hallquist, M. N., Beeney, J. E., & Pilknois, P. A. (2013). Borderline personality pathology and the stability of interpersonal problems. Journal of Abnormal Psychology, 122, 1094-1100. Yen, S., Shea, Mt. T., Battle, C. L. (2002). Traumatic exposure and posttraumatic stress disorder in borderline, schizotypal, avoidant, and obsessive-compulsive personality disorders: Findings from the collaborative longitudinal personality disorders study. Journal of Nervous Mental Disorders, 190, 510-518. Zlotnick, C., Clarke, J. G., Friedman, P.D., Roberts, M. B., Sacks, S., & Melnick, G. (2008). Gender differences in comorbid disorders among offenders in prison substance abuse programs. Behavioral Sciences and the Law, 26, 403-4

46

Table 1: Characteristics of Participants ________________________________________________________________________ M

SD

Age

35.75

.82

Number of previous arrests

11.52

11.02

Overall score

29.75

6.65

Criminal History

2.39

.92

Education/Employment

2.46

.78

Financial

2.03

.85

Family/Marital

1.89

.93

Accommodations

1.26

.59

Leisure/Recreation

2.28

.92

Companions

2.67

1.11

Alcohol/Drug

2.40

1.28

Emotional/Personal

2.27

.47

Attitudes/Orientations

1.01

1.08

Anxiety

52.41

10.87

Depression

51.92

10.58

Suicidal Ideation

48.16

7.84

Borderline

58.55

11.44

Non-support

50.21

11.03

Alcohol*

53.38

14.55

Drug*

74.04

18.94

Treatment Rejection*

38.67

9.88

LSI-R

PAI

47

________________________________________________________________________ *These PAI items were not investigated in the current study, but were included to provide a description of the sample.

48

TABLE 2: Race of Participants ________________________________________________________________________ Current Study

National Statistics

LSI-R Norms

________________________________________________________________________ Freq.

%

Freq.

%

Freq.

%

African-American

166

55.3

26,000

25.1

263

14.2

Caucasian

96

32.0

51,100

49.3

1,223

65.8

Hispanic

35

11.7

95,400

17.7

34

1.8

Native American

2

.7

---------

----

74

4.0

American Indian*

-----

-----

---------

-----

5

.3

Asian*

-----

-----

---------

-----

11

.6

Other

1

.3

8,274

8.0

221

11.9

Unknown*

---

-----

---------

---

2

.1

_______________________________________________________________________ *The LSI-R norms included three racial categories that were not included in the other samples. Those categories were American Indian, Asian, and Unknown. The LSI-R norms were also missing information for 26 females (1.4%)

49

TABLE 3: History of Abuse ________________________________________________________________________ Frequency Percentage ________________________________________________________________________ History of Sexual Abuse Yes

112

37.3

No

186

62.4

Yes

141

47.0

No

156

52.0

Yes

74

24.7

No

226

75.3

History of Domestic Abuse

History of Both Types of Abuse

________________________________________________________________________ *Data on history of sexual or domestic abuse was missing for 3 individuals.

50

Table 4: Participant Rearrest Within 2 years After Release into the Community

Race

# Rearrested

# Not Rearrested

% Rearrested

African American

38

127

22.9

Caucasian

31

64

32.3

Hispanic

8

27

22.9

*Other

1

3

25.0

78

221

Total

90

51

TABLE 5: Types of Rearrest Offenses

Type of Offense

n

%

Assault

10

12.8

Theft

12

15.4

Drug

29

37.2

Prostitution

6

7.7

Parole/probation violation

1

1.3

Other violent offense

2

2.6

Other property offense

4

5.1

Other neither offense

14

17.9

52

Table 6: Relationship Between Abuse History and Rearrest

Abuse History

df

χ2

p

Domestic Abuse

1

.09

.760

Sexual Abuse

1

.25

.614

Domestic & Sexual Abuse

1

.58

.456

53

TABLE 7: PAI Scores and Rearrest

PAI Subscore

N*

r

p

Anxiety

294

.092

.116

Borderline

294

.133

.023**

Depression

294

.041

.482

Non-support

294

.089

.127

Suicidal Ideation

294

.060

.303

*6 subjects were missing PAI data. **Significant at the p