The Connection of Probation/Parole Officer Actions to Women Offenders’ Recidivism

Abstract

Because women offenders often have limited social networks and unique needs, the actions of probation/parole officers providing community supervision may be particularly relevant to outcomes. The present study examined the effects of probation/parole officer relationship style, attention to criminogenic needs, and intensity of supervision on women offenders’ arrests and convictions within a 24-month period. Contrary to findings from other studies, the measured elements of officer actions had no direct effects on recidivism for a sample of 226 women. However, the analysis revealed an indirect effect in which a non-supportive, punitive relationship was related to reactance and anxiety, which in turn were related to high recidivism. The discussion focuses on theoretical and methodological explanations for the null findings regarding direct effects. Moreover, it draws on the literature in psychology and communication to suggest approaches to reducing the reactance that can promote recidivism and to suggest related future research directions.

Keywords

recidivism women offenders relationship style criminogenic needs supervision effect reactance

The purpose of the study described in this article was to increase understanding of the effects of probation and parole officer (PO) interactions with women offenders. In particular, we sought to understand effects on recidivism. A social psychological theory of reactance, theories of corrections, communication theory, and feminist criminological theory influenced the focus on women and the choice of potential explanatory variables. The correctional theories included dual-role relationship (DRR) theory, the risk–needs–responsivity (RNR) model, and scholarship on the effect of intensive supervision (i.e., the supervision effect). Before we introduce the specific theoretical frameworks that informed the design of the research, we further explain the focus on community corrections and women offenders.

In 2012, one in every 50 U.S. residents was supervised in the community on probation or parole, and the smaller but sizable proportion of 1 in 108 adults was incarcerated in jail or prison (Glaze & Herberman, 2013). As the most common correctional intervention, community supervision (i.e., probation and parole) has substantial public cost both in terms of financial outlays to operate these programs and expenses resulting from recidivism. Recidivism costs extend beyond harm to victims, offenders, families, and communities to include expenditures on law enforcement, prosecution, courts, additional community supervision, and incarceration (Clear, 2007; Cohen & Bowles, 2010; National Association for the Advancement of Colored People, 2011). High rates of recidivism for individuals under community supervision and high costs of failure warrant increased research to understand effective practices (Guerino, Harrison, & Sabol, 2011; Kleiman, 2011; Meredith & Prevost, 2009; Ostermann, 2015; Ostermann, Miller, & Matejkowski, 2013; Schram, Koons-Witt, Williams, & McShane, 2006).

Women constitute a community supervision population of unique concern. Because a higher proportion of women than men are in the correctional system due to drug and alcohol misuse, it is important to focus research on women offenders with substance involvement (Belknap, 2014; Guerino et al., 2011; Langan & Pelissier, 2001; Mumola & Karberg, 2006). Other subgroups deserve study, but because groups follow different pathways into crime (Morash, 2010), each group must be large enough to allow for meaningful conclusions from the analysis. Women make up a growing proportion of individuals under correctional supervision, specifically 11% of those on parole and 24% of those on probation (Maruschak & Bonczar, 2013). Despite their representation, limited research has focused on the connection between women’s recidivism and POs’ actions and methods. This is problematic because although some offenders in the community attend programs other than supervision, for others the PO provides the only intervention or the impetus to participate in other programs (Gill, Hyatt, & Sherman, 2010).

Turning now to the theoretical framing of the research, DRR theory holds promise as an explanation of the effects of PO interactions on women. It considers professionals who have controlling authority over clients and who also are expected to help those clients. For these types of relationships, research finds that a PO’s supportive, non-punitive style of relationship is related to improved outcomes for offenders (Kennealy, Skeem, Manchak, & Louden, 2012; Skeem, Manchak, Vidal, & Hart, 2009). Although DRR study samples have included women, the small numbers have precluded women-specific analyses. As an exception, in a prior analysis of data considered in the present study, we examined short-term effects (e.g., reactance, anxiety), not recidivism (Morash, Kashy, Smith, & Cobbina, 2015). That analysis revealed that a more supportive, non-punitive style was related to lower anxiety and reactance, especially for low-risk women.

The present study contributes to the literature by focusing on the understudied group, women under community supervision. It extends research on DRRs and outcomes by examining the direct effect of PO relationship style on recidivism and the indirect effects through reactance and anxiety resulting from PO-offender interactions. To strengthen the test for significance of these effects, the analysis considers influences on recidivism identified by other correctional theories. RNR theory identifies two of these influences: the POs’ attention to crime causing needs (criminogenic needs) and the offenders’ initial risk for recidivism (Andrews & Bonta, 2006; Andrews et al., 1990; Cullen & Gendreau, 2000; Gendreau, Smith, & French, 2006; Singh et al., 2014; P. Smith, Gendreau, & Goggin, 2004). Research is inconsistent in its findings about a third influence, the intensity of supervision, which depending on the study increases, has no effect on, or decreases recidivism (Hawken & Kleiman, 2009; Kleiman, 2011; Petersilia & Turner, 1993). The literature review that follows provides additional information on DRR theory, shows a connection of DRR research to RNR theory, and explains the theoretical relevance of the concepts of reactance and anxiety to PO style and to recidivism. It also further explains the rationale for including POs’ attention to criminogenic needs, the offender’s risk for recidivism, and supervision intensity in the model that is tested.

Literature Review

DRR Theory

Skeem and colleagues advanced the development of DRR theory as it applies to corrections by creating and validating a measure of DRRs in which professionals are expected to both help and control clients and then using that measure in tests of the theory (Skeem, Encandela, & Eno Louden, 2003; Skeem, Eno Louden, Polaschek, & Camp, 2007). They found that for probationers with co-occurring substance abuse and other mental health disorders, relationships characterized by trust and caring/fairness (i.e., supportive, non-punitive relationships) led to positive outcomes. DRR researchers initially studied mentally ill offenders (Skeem, Eno Louden, Manchak, Vidal, & Haddad, 2009; Skeem et al., 2007), but they later replicated their findings with a general correctional population (Kennealy et al., 2012). Consistent with DRR theory, a study of case managers working with offenders showed that toughness, which is inconsistent with a supportive style, was positively related to recidivism (Angell & Mahoney, 2007). Studies of the working alliance also have provided support for DRR theory. In a working alliance, a helping professional and client agree on the most important problems to address and the desired outcomes. In correctional settings, the working alliance, a component incorporated into the measure of the DRR, promotes positive offender outcomes (Green et al., 2013; Taxman & Ainsworth, 2009).

Although DRR and related research have not emphasized women, Skeem and colleagues’ (Skeem, Eno Louden, et al., 2009; Skeem et al., 2007) studies of samples of offenders with co-occurring substance use and other mental health disorders makes their work particularly relevant to women offenders, many of whom have these co-occurring disorders (Fedock, Fries, & Kubiak, 2013; Maxwell & Freeman, 2007; Salina, Lesondak, Razzano, & Weilbaecher, 2007). The DRR research also has implications for research design; Skeem et al.’s (2007) DRR measure better explained probationers’ outcomes than more limited measures of therapeutic alliance. Thus, the revised instrument they developed (i.e., the Dual-Role Relationship Inventory–Revised [DRI-R]) is a preferred measure of PO style (Skeem et al., 2007).

RNR Theory

Apart from DRR theory, RNR scholarship specifies that concentrating interventions on high-risk offenders, addressing the crime causing (i.e., criminogenic) needs of those offenders, and responsivity promote low recidivism. RNR scholars have shown that interventions directed at offenders most at risk for recidivism decrease recidivism (Fielding, Tye, Ogawa, Imam, & Long, 2002; Lowenkamp, Holsinger, & Latessa, 2005; Lowenkamp, Latessa, & Holsinger, 2006). Research also confirms the RNR proposition that targeting services and supervision at criminogenic needs reduces recidivism (Schram et al., 2006; Singh et al., 2014; Vieira, Skilling, & Peterson-Badali, 2009).

Compared with studies of the concentration of correctional interventions on high-risk offenders, studies of responsivity have more varied findings. General responsivity refers to the styles and modes of service delivery that should be employed when providing correctional intervention services to most offenders. One examination of general responsivity found that correctional staff that discussed procriminal attitudes and cognitions and used cognitive techniques to evaluate and guide their focus on offenders’ procriminal attitudes had clients with lower recidivism (Bourgon & Gutierrez, 2012). There also is evidence that for optimal outcomes, offenders’ poor reasoning and communication skills, concrete-oriented thinking, and psychological functioning (e.g., depression) should be matched to both correctional programs and the staff in these programs, though there also are contradictory findings when the match to needs are statistically controlled (Bonta, 1995; Vieira et al., 2009).

More specific responsivity factors, such as an individual’s learning style, ability, motivation, mental health, and social skills are also relevant to whether correctional intervention has optimal effects (Andrews & Bonta, 2006). Specific responsivity is the RNR concept most directly relevant to DRRs. It refers to the fit of PO interactions with an offender’s characteristics, including gender (Andrews, Bonta, & Wormith, 2006). RNR theorists have not clarified specific modes of intervention with high responsivity to women. However feminist criminologists fill this gap with scholarship showing the efficacy of a positive and supportive style of relationship between the PO and female clients (Bloom, Owen, Covington, & Raeder, 2003; Bui & Morash, 2010; Covington, 1999, 2008; Maidment, 2006). They point out that women have relatively low levels of family support compared with male offenders (Mallik-Kane & Visher, 2008), which may heighten the effects of the PO’s supportive style. Also, feminist research shows that women place high value on relationships to other people, so the quality of these relationships is likely to have pronounced effects (Covington, 2008; Jordan, 2013; J. B. Miller, 1976). Our study did not examine differential effects of supportive style on women and men, but the feminist scholarship provided a rationale for focusing on relationship style for women offenders.

Intensity of Supervision

The second line of correctional research outside the DRR framework that is relevant to community supervision focuses on intensity of supervision, specifically the supervision effect. A seminal field experiment revealed that high levels of monitoring, testing, and contact with probationers and parolees in the absence of treatment and services led to increased violations, revocations, and incarcerations (Petersilia & Turner, 1993). Considerable research has confirmed this unintended negative connection of supervision intensity not only with revocations (Grattet, Lin, & Petersilia, 2011; also see Kubrin & Stewart, 2006; Sirakaya, 2006) but also with new offenses (Gill, 2010; Hanley, 2006; Lowenkamp & Latessa, 2004; Lowenkamp et al., 2006; Taxman, 2002). However, when combined with rehabilitation approaches, for some samples intensive supervision produced lower recidivism (Aos, Miller, & Drake, 2006; Jalbert & Rhodes, 2012; Jalbert, Rhodes, Flygare, & Kane, 2010; Latessa, Travis, Fulton, & Stichman, 1998; MacKenzie, 2006; Paparozzi & Gendreau, 2005; Warchol, 2000). Despite inconsistent evidence of the connection of supervision intensity to recidivism, a well-specified model to explain recidivism from supervision activities and relationships should consider the supervision effect.

Reactance and Anxiety

In DRRs, probation officers’ sole reliance on surveillance to restrict and control clients appears to promote negative arousal and a subsequent reactance to directives (Skeem et al., 2003). Social psychological research has firmly established that when people feel their personal freedoms are being limited or are about to be limited, this type of negative arousal occurs, which in turn instantiates a state of psychological reactance (Brehm, 1966; Brehm & Brehm, 1981). Reactance includes desires and actions to restore personal freedoms, including behaving contrary to directives (Dillard & Shen, 2005). In prior analysis of the data we consider in the present article, we sought to explain self-reports of substance use and probation/parole official records indicating substance-related violations of supervision conditions (e.g., failure to report for drug tests, use of drugs; S. W. Smith, Cornacchione, Morash, Kashy, & Cobbina, in press). We found that women who perceived their POs to threaten to limit their freedoms were most likely to report high levels of reactance, which were ultimately associated with self-reports of continued substance use and official violations involving drugs and drug testing (S. W. Smith et al., in press). The DRR items reflecting punitive PO style pertain to limiting personal freedoms (e.g., the PO’s rating on a scale ranging from strongly agree to strongly disagree, “I feel that it is sometimes necessary to punish this client,” “This client feels I make unreasonable demands on her”). If the negative arousal and reactance generated by punitive behavior on the part of the PO are related to recidivism, there may be an indirect effect in which DRR style is related to negative emotions (e.g., anxiety, anger) and psychological reactance, which in turn is related to official indicators of arrest and conviction.

Research Focus and Contribution

The literature suggested that a supportive, non-punitive PO style is related to low recidivism. In addition to this direct effect, prior research suggests the indirect effect that a more supportive, non-punitive PO style is related to offenders’ lower anxiety and reactance, which in turn are related to lower recidivism. Although reactance and anxiety have not been studied as accounting for indirect effects of either PO discussion of criminogenic needs or supervision intensity on recidivism, extending the logic for supportive relationship style, we examined the possibility that either the lack of needs-focused discussion or the provision of intense supervision might influence the arousal of negative emotional states and reactance, and thus be indirectly related to recidivism.

The research makes a contribution by considering a sample of women, a group that is understudied in correctional research because they make up a smaller proportion of offenders than do men. It also contributes by providing a strong test of the expected associations regarding DRRs, reactance, anxiety, and recidivism. It does this by including rival explanations (recidivism risk, match of supervision attention to criminogenic needs, and supervision intensity) in the models tested.

Method

Data

Data are from a longitudinal field study of women on probation and parole. Data were collected from multiple sources: three interviews with the offenders, two surveys of their POs, and official data from state agencies. When each PO was enrolled, she or he completed a survey that obtained PO demographic information, work-related information (e.g., training to work with women offenders), and a measure of general style of relating to women offenders. The offenders were interviewed 2 to 3 months after supervision began, and then twice more at 3-month intervals. We refer to the waves of data collection from the offenders as T1, T2, and T3. At T1, the Women’s Risk/Needs Assessment (WRNA; Van Voorhis, Salisbury, Wright, & Bauman, 2007; Van Voorhis, Wright, Salisbury, & Bauman, 2010) was administered to provide one measure of recidivism risk. The WRNA was developed to incorporate recidivism risk factors both unrelated to gender and those uniquely predictive for women (e.g., parental stress; Van Voorhis, 2012). At T2, women described their responses to interacting with their POs, and at the same time, each woman’s PO completed the DRI-R, which indicated her or his style of relating specifically to each woman. Data from official records included (a) prior arrests, convictions, and time in prison, (b) the frequency of PO-client contact and the length of supervision for an 18-month period from the start of supervision, and (c) arrests and convictions in the 24 months after the start of supervision. Thus, the data covered several time periods for each woman: Period 1—criminal history and incarceration prior to each woman’s start of supervision plus an assessment tool (i.e., the WRNA) measure of risk for recidivism at the start of supervision; Period 2—POs’ reports of their style of relating to each woman and women’s reports of their responses to interacting with their POs, both at approximately 6 months after the start of supervision; Period 3—PO case note information on intensity of supervision for 18 months; and Period 4—official recidivism data for 24 months after the start of supervision.

Sample

Michigan has a centralized statewide system of supervision for felony offenders and POs who specialize in supervising women. In densely populated counties, POs supervised only women on either probation or parole, but in less dense areas they had mixed gender or type of supervision caseloads. To guide POs’ interventions that address criminogenic needs, the department uses a needs assessment (Correctional Offender Management Profiling for Alternative Sanctions [COMPAS]) that incorporates subsections that consider women’s unique needs as identified by the WRNA (Brennan, Dieterich, & Ehret, 2009).

In the Michigan setting, an initial sample of 402 women felons was obtained by first recruiting 73 POs who supervised women. Corresponding to the small proportion of men supervising women statewide, just three POs were men. The proportion of POs recruited in each county corresponded to the proportions of women supervised in each of the 16 counties within a 1½-hr drive from the research office. These counties included 68.5% (6,759,961 of 9,876,187) of the 2011 state population, all major population centers (e.g., Detroit, Grand Rapids), and a mix of rural and suburban areas. To increase parolees to almost 25% of the total, POs were oversampled in relation to probation officers. A principal investigator reviewed the caseload list with each PO and assisted the PO in identifying eligible clients (i.e., women with substance involvement who had been supervised for approximately 3 months). POs facilitated recruiting women by (a) giving eligible women a project contact card or flyer so that (if interested) they could arrange a time to hear about the study, (b) introducing women to on-site project interviewers, or (c) seeking permission to share women’s contact information with interviewers. Interviewers who were hired and trained for the project did the direct recruiting of participants in private meetings with each woman.

Sample Attrition and Non-Response

The present analyses focus on 226 women, each of whom (a) was consistently supervised by the same PO across the study, (b) completed the T2 interview, and (c) had PO survey data concerning the PO’s interpersonal style with the specific woman. Implementation of a validated protocol for retaining participants in longitudinal research (Sullivan, Rumptz, Campbell, Eby, & Davidson, 1996) resulted in women offenders’ low attrition (94.3% retained for three interviews). A smaller proportion of POs (75.3%, or 55 of 73) both responded to the T2 survey and continued supervising a participating offender through the T3 interview. Workplace practices in the field setting (an operating statewide probation and parole agency) and PO non-response to the client-specific survey contributed to POs’ attrition and missing data, respectively. The most common reasons for discontinuation of supervision of a client was PO reassignment to different duties or a different caseload and medical and maternity leave. POs identified lack of time as the primary reason for survey non-response. To identify potential biases in the sample considered in analysis, we made two types of comparisons: (a) women offenders included and excluded in the analysis and (b) POs included and excluded.

The comparison of the 226 women with a continuing PO who responded to the client-specific survey and the 176 excluded women revealed no significant difference in offender probation versus parole status (for included women, 77.4%, n = 175 were on probation; for excluded women, 73.9%, n = 130 were on probation, χ²(1) = .69, p = .407). As shown in Table 1, included women had a slightly but significantly lower number of prior convictions than excluded women. However, the groups did not differ significantly on the two other criminal history measures, prior arrests and months in prison prior to the study. The groups also did not differ significantly on total risk, age, and 24-month recidivism. Overall, offender and PO attrition and non-response of POs to client-specific surveys did not appear to introduce bias into the sample of women with data available for analysis.

Table 1:

Comparison of Basic Descriptive Statistics for Included and Excluded Offenders and POs in the Sample

	Sample Included in Analyses		Sample Excluded From Analyses
	M	SD	M	SD	t
Offenders
No. of prior arrests	4.66	3.72	5.22	4.05	1.45
No. of prior convictions	5.51	4.93	6.66	5.66	2.18*
No. of prior months in prison	12.04	33.11	14.41	35.73	0.69
WRNA total risk score	22.16	7.98	23.57	8.71	1.69
Age	33.84	10.42	33.91	10.70	0.07
New arrests in 24 months	0.37	0.76	0.40	0.79	0.33
New convictions in 24 months	0.27	0.65	0.28	0.63	0.22
POs
Amount of PO training	3.80	1.21	3.78	1.40	0.06
DRR relationship style	5.91	0.54	5.89	0.57	0.14

Note. The df for t tests for offenders was 399 for months in prison and 400 otherwise. The df for t tests for POs was 71. PO = probation or parole officer; WRNA = Women’s Risk/Needs Assessment; DRR = dual-role relationship.

p < .05.

The comparison of the 55 POs who worked with women with complete data (i.e., included in the analysis) and the 18 who did not revealed few differences. Table 1 presents tests for differences in means. A measure of amount of PO training to work with women offenders (range: 1 to 5, with 5 indicating extensive training) did not significantly differ for the included and excluded POs. Also included and excluded POs did not differ significantly on their scores reflecting their general style of relating to women offenders (i.e., the DRR measure of supportive, non-punitive relationship style). Different proportions of included and excluded POs had a graduate as opposed to a bachelors degree, but the difference was not significant (for included POs, 32.7% [n = 18] held graduate degrees and for excluded POs, 16.7% [n = 3] did, χ²(1) = 1.71, p = .156). Finally, included and excluded POs did not differ significantly on whether their caseloads were all probation versus all or mixed parole (for included POs, 70.9% of 55 were only probation and for excluded 77.8% of 18 were only probation; χ²(2) = .725, p = .696.) Similar to findings from the comparison of women included and excluded from the analysis, the comparison of POs did not reveal bias introduced by attrition or non-response.

Participant Characteristics

The 226 women included in the analysis were supervised by 55 POs from 15 counties, with a concentration in Wayne County where Detroit is located (43.6%, n = 24). One PO reported high school as the highest degree, 65.5% (n = 36) held a bachelor’s degree, 30.9% (n = 17) a masters, and one a doctoral degree. The POs were about equally divided in their reports of being White (45.5%, n = 25) and Black (52.7%, n = 29), and one PO was Asian Indian.

Of the 226 women in the analysis, 22.6% (n = 51) were on parole and the remaining 77.4% (n = 175) on probation. Most women were White and not Hispanic (48.7%, n = 110), Black and not Hispanic (27.9%, n = 63), or were Hispanic or multiracial (22.6%, n = 51). Two women were Native American. Most (73.9%, n = 167) had not spent time in prison before the first interview, but 8.9% (n = 20) were incarcerated up to a year, 6.2% (n = 14) for 1 to 2 years, and 11.1% (n = 25) for more than 2 to nearly 19 years. Official records indicated that just 17.7% (n = 40) were first-time offenders, 31.9% (n = 72) had two or three prior arrests, and 50.4% (n = 112) had four to 23 prior arrests. In a similar pattern, 15.5% (n = 35) had no prior convictions, 27% (n = 61) had two or three, and 57.5% (n = 130) had four or more. Women ranged in age from 18 to 60 years.

Procedure

For each of the three offender interviews, trained interviewers conducted a face-to-face interview and entered data into a computerized program on a laptop. Interviews took place in locations identified as preferable by the offenders and included private offices in probation and parole centers and public venues that had private areas (e.g., a library). Participants received a $30 gift certificate for the first interview, a $50 gift certificate for the second, and a $75 gift certificate for the third.

POs responded to an on-line survey when first enrolled in the study and to a separate on-line survey for each client at the time of the client’s T2 interview. To ensure study participants understood the confidential nature of the research, POs and women offenders were informed of the university institutional review board (IRB)–approved protocol for protecting them and the National Institute of Health Confidentiality Certificate that further protected the data from subpoena.

Measures

In this study, we obtained multiple measures of three of the key predictor variables: recidivism risk, supervision intensity, and anxiety/reactance. Our intention was to develop a theoretically based latent variable model of how PO behavior predicts offender recidivism. Unfortunately, we encountered significant difficulties using the latent variable approach because our sample size was relatively small (i.e., less than 400) and the correlations among the indicators for each latent variable were only moderate to strong (i.e., r’s primarily in the .3 to .5 range). These problems resulted in unacceptable fit for the measurement model that included our five predictors, three of which were latent variables, χ²(46) = 150.02, p < .001; comparative fix index (CFI) = .825, root mean square error approximation (RMSEA) = .100. We therefore adopted the approach of creating factor scores, which allowed us to retain the benefits of having multiple measures for key constructs.

Outcome Variables

One of the two recidivism variables was total number of arrests at 24 months after the beginning of supervision. Of the 226 women in the analyzed sample, 169 had no arrests, 39 had one arrest, 11 had two arrests, six had three arrests, and one had five arrests. We also predicted the total number of convictions after the beginning of supervision (184 women had no convictions, 29 had one conviction, seven had two convictions, and six had three convictions).

Predictor Variables

PO supportive and therefore non-punitive style reported by the PO for each offender was computed as the mean for all items on the DRI-R (rated 1 = never, 7 = always) but with punitive items reverse scored (Skeem et al., 2007). Reliability of the scale with 30 items was .94, and the average interitem correlation was .39.¹

The variable, PO discussed criminogenic issues, is a count of the number of issues that the woman identified as problems at either T1 or T2 and for which she also reported PO discussion at either time (M = 1.84, SD = 1.96, range: 0-10). For each topic, at T1 interviewers asked women whether their PO discussed the topic with them as well as whether they felt it was a problem since supervision began. At T2, interviewers asked about discussion and the existence of a problem in the last 3 months (i.e., since T1). The topics were antisocial attitudes, educational needs, employment and financial problems, unsafe housing, antisocial friends, anger and hostility, mental illness, abuse as an adult, current substance abuse, and neighborhood crime.² We used T1 and T2 interview data so that the period covered corresponded to the period for which women were describing their reactions to supervision and to take advantage of the increased validity afforded by repeating the questions in two rounds of interviewing. To ensure that the needs discussed were the needs women perceived as problematic, our variable measuring PO discussion of criminogenic needs was a count of the number of issues that the woman identified as a problem which were specifically discussed by the PO. Our measurement approach is consistent with research on the therapeutic alliance; this research suggests that women’s perception that the needs they themselves identified were the needs discussed would be related to outcomes. It also is consistent with Trotter’s (2006) conclusions from a literature review that effective POs focus on problems identified by the client.

Supervision intensity was computed as a factor score based on data from agent case notes for 18 months since the start of supervision. Conclusions from a review of prior community supervision research supported the decision to consider several months of supervision, since intensity may vary over time, and to consider multiple indicators of intensity (Miofsky & Byrne, 2012). The supervision intensity factor includes length of supervision in months (M = 15.88, SD = 3.76, range: 3.68-23.13), number of home visits (M = 4.00, SD = 3.96, range: 0-33) and number of in-person contacts (M = 18.81, SD = 10.07, range: 4-56). Reliability of this aggregate of three variables was α = .69, and the average interitem correlation was .43.

Anxiety/reactance is a factor score aggregating women’s scores on three scales: anxiety, emotional reactance, and psychological reactance. The women were asked about their experience after a recent conversation (i.e., on the phone) or being with the PO.³ The anxiety scale was based on the Brief Symptom Inventory, which has been validated with offender samples; items were adapted to measure women’s experience of six anxiety-related states (e.g., nevousness, shakiness inside) after interacting with their POs (Boulet & Boss, 1991; Derogitas & Melisarotos, 1983). Each item was rated on a scale from 1 (not at all) to 5 (very much), M = 1.63, SD = 1.07, α = .95. Following a procedure suggested by Dillard and Shen (2005), emotional reactance after PO interactions was measured with five items (e.g., anger, irritation), and items were rated on a scale from 1 (not at all) to 5 (very much), M = 1.26, SD = .64, α = .88). Finally, psychological reactance after interactions with the PO was measured by adapting items from the Hong Psychological Reactance Scale (Hong & Faedda, 1996). Seven items (e.g., irritation when told what to do by the PO) were rated on a scale from 1 (very strongly disagree) to 7 (very strongly agree), M = 2.71, SD = 1.18, α = .92). The reliability based on the three standardized scale scores for anxiety, emotional reactance, and psychological reactance was .74 with an average interitem correlation of .50.

Recidivism risk was created as a factor score based on probation/parole status, number of all previous arrests before the study (M = 4.66, SD = 3.72, range: 1-23), number of all previous felony convictions before the study (M = 2.70, SD = 3.05, range: 0-21), the square-root of the number of months spent in prison prior to the study (raw number months, M = 12.04, SD = 33.11, range: 0-233), and the total needs score from the WRNA (M = 19.23, SD = 8.34).⁴ Reliability of this aggregate of variables was indicated by an alpha of .75 based on five standardized items; the average interitem correlation was .37.

Analytic Strategy

To determine the appropriate modeling technique, an initial set of multilevel models were computed to assess the degree to which women supervised by the same PO had similar outcomes. For the two outcomes, arrests and convictions at 24 months, we used a negative binomial intercept-only model, but the model could not be estimated (i.e., the Hessian matrix was not positive definite). When we treated these variables as if they were normally distributed, the intraclass correlation coefficient (ICC) for arrests was −.004, p = .911 and the ICC for convictions was .023, p = .256. These small and non-significant correlations provide little evidence that women who were supervised by the same PO have similar outcomes. In contrast, there was evidence of systematic PO variance for recidivism risk, likely because some POs supervised primarily paroled offenders who had considerably higher risk and other POs supervised offenders on probation. There was also evidence that POs varied by supervision intensity, their use of supportive style, and the degree of anxiety/reactance they elicited from the women. In sum, although the sample includes multiple offenders supervised by the same PO, because the outcome variables showed no evidence that women supervised by the same POs had similar outcomes, there was no need to model non-independence in analyses, and so standard analyses for independent data are appropriate.

We first examined the direct effects of each of our measures of PO behavior (i.e., supervision intensity, supportive style, and discussion of criminogenic needs) on arrests and convictions during 24 months. In each model, we controlled for recidivism risk. We used the program Mplus (Muthén & Muthén, 2015), which allowed us to treat the outcomes, number of arrests and number of convictions at 24 months, as negative binomial variables. Missing data were imputed using full information maximum likelihood, and we used bootstrapping with 1,000 bootstrapped samples to compute bias-corrected confidence intervals for the model parameters.

We next evaluated the indirect effect of each PO behavior on arrests and convictions during 24 months via women’s anxiety/reactance after meetings with the PO. As before, we controlled for recidivism risk in each analysis. The path model depicting this general approach in which the effects of PO behavior on women’s recidivism are indirect via anxiety/reactance is presented in Figure 1.

Figure 1:

General Model of Indirect Effects of PO Behavior on Women’s Recidivism via Anxiety/Reactance

Results

Descriptive and Bivariate Analyses

Table 2 presents the means, standard deviations, and correlations for the study variables. Importantly, only two variables show significant zero-order associations with number of arrests and number of convictions at 24 months: recidivism risk and anxiety/reactance responses after meeting with POs. Women with more evidence of recidivism risk had more arrests and convictions at 24 months, and women who reported responding more negatively to interactions with their PO had more arrests and convictions. Recidivism risk also correlated with the three measures of PO behavior such that women who had more evidence of risk for recidivism were supervised more intensely, their POs reported using less of a supportive interaction style, and the women reported that their PO discussed a greater number of the criminogenic needs that the women perceived to be a problem. Finally, women whose POs reported using a more supportive style experienced lower anxiety and reactance after interactions.

Table 2:

Means, Standard Deviations, and Correlations for the Study Variables

	1	2	3	4	5	6	7
1. No. of arrests at 24 months	—
2. No. of convictions at 24 months	.77**	—
3. Recidivism risk	.22**	.21**	—
4. PO supervision intensity	.03	.07	.29**	—
5. PO supportive style	−.02	.03	−.18**	−.02	—
6. No. of relevant issues PO discussed	.13	.08	.23**	.23**	−.07	—
7. Anxiety/reactance	.18**	.13*	.04	.10	−.27**	.08	—
ICCs for PO effects	.0^a	.0^a	.55**	.26**	.31**	.05	.14*
M	0.37	0.27	0.00	0.00	5.81	1.84	0.00
SD	0.76	0.65	1.00	1.00	0.68	1.96	1.00

Note. PO = probation or parole officer; ICC = intraclass correlation coefficient.

ICCs measuring similarity of arrests and convictions for women supervised by the same PO were not estimable using a negative binomial model due to a non positive definite Hessian matrix.

p < .05. **p < .01.

Analyses of Direct Effects of PO Actions on Recidivism

Considering number of arrests at 24 months as the outcome, the model in which PO supervision intensity was the predictor and recidivism risk was included as a control variable yielded a non-significant coefficient for supervision intensity, b = −.055, β = −.189, p = .685. Similarly, when PO supportive behavior was the key predictor of arrests, again controlling for recidivism risk the effect of PO behavior was not significant, b = .049, β = .114, p = .799. Finally, PO discussion of relevant issues also failed to predict arrests at 24 months, b = .076, β = .468, p = .189, controlling for recidivism risk. In each of these three models, the coefficient for recidivism risk was relatively large, bs > .25, βs > .78, and statistically significant, all ps < .05.

A similar set of results emerged for PO behavior predicting convictions at 24 months. Controlling for recidivism risk, supervision intensity did not significantly predict convictions, b = .026, β = .087, p = .874. Likewise, neither PO supportive behavior, b = .185, β = .390, p = .386, nor discussion of relevant issues, b = .017, β = .112, p = .846, predicted convictions at 24 months controlling for recidivism risk. As in the models predicting arrests, the effect of recidivism risk on convictions at 24 months was statistically significant in all three of these models, bs > .29, βs > .96, ps < .05. Finally, we conducted two multivariate analyses (one for arrests and one for convictions) in which all of the PO action and style variables were included in the model, and results were virtually unchanged.

Analyses of Indirect Effects of PO Actions and Style on Recidivism

Our next analyses examined the extent to which PO behavior may have an indirect effect on recidivism via the offender’s reaction to supervision interactions. Table 3 presents the results predicting number of arrests, and Table 4 presents the results predicting number of convictions. As can be seen in both tables, supervision intensity has no direct or indirect effect on recidivism. Likewise, there is no evidence that having the PO address specific problem areas identified by the woman directly or indirectly predicts recidivism. Indeed, the only indication that the PO’s behavior has any effect on arrests and convictions is an indirect effect of PO supportive, non-punitive style. This indirect effect suggests that POs who report that they adopt a more supportive style with a particular woman tend to elicit less anxiety/reactance from that woman, and ultimately the woman shows fewer arrests and convictions during the 2 years after the beginning of supervision. Finally, as with our analyses of the direct effects, we conducted two multivariate analyses (one for arrests and one for convictions) of the indirect effects of all of the PO variables simultaneously via anxiety/reactance, and as with the simpler indirect effect models, only PO supportive style showed a significant indirect effect on recidivism.

Table 3:

Total, Direct, and Indirect (via Anxiety/Reactance) Effects of PO Behavior on Number of Arrests at 24 Months, Controlling for Recidivism Risk

	Total Effects		Direct Effects		Indirect Effects
	b	95% CI	b	95% CI	b	95% CI
PO behavior variable
PO supervision intensity	−.057	[−0.33, 0.24]	−.083	[−0.37, 0.21]	.027	[−0.01, 0.10]
PO supportive style	.143	[−0.28, 0.63]	.278	[−0.13, 0.70]	−.135*	[−0.34, −0.01]
No. of relevant issues PO discussed	.089	[−0.03, 0.21]	.078	[−0.05, 0.19]	.011	[−0.00, 0.05]

Note. PO = probation or parole officer; CI = confidence interval.

p < .05

Table 4:

Total, Direct, and Indirect (via Anxiety/Reactance) Effects of PO Behavior on Number of Convictions at 24 Months, Controlling for Recidivism Risk

	Total Effects		Direct Effects		Indirect Effects
	b	95% CI	b	95% CI	b	95% CI
PO behavior variable
PO supervision intensity	.035	[−0.28, 0.38]	.009	[−0.33, 0.37]	.026	[−0.01, 0.10]
PO supportive style	.303	[−0.21, 0.89]	.459	[−0.08, 1.03]	−.156*	[−0.36, −0.01]
No. of relevant issues PO discussed	.020	[−0.13, 0.18]	.009	[−0.14, 0.16]	.011	[−0.00, 0.05]

Note. PO = probation or parole officer; CI = confidence interval.

p < .05.

Discussion

This article describes two important findings. The first is that the PO’s behavior has no direct effect on women’s official recidivism. More specifically, neither the intensity of supervision, the attention to women’s criminogenic needs, nor relationship style predicted arrests or convictions during 24 months. There are a number of possible explanations for these unexpected findings, the first of which is that supervision is simply ineffective in mitigating other central and strong influences on women’s recidivism. Research has shown that employment and related financial distress, poverty, unavailable and substandard housing, negative community conditions (i.e., low resource, high crime), and criminal associates are strong predictors of women’s (and for several of these variables, men’s) recidivism (Greiner, Law, & Brown, 2015; Holtfreter, Reisig, & Morash, 2004; Huebner, DeJong, & Cobbina, 2010; Makarios, Steiner, & Travis, 2010; Manchak, Skeem, Douglas, & Siranosian, 2009). Although individual choice and resources certainly can affect these predictors, many of them (e.g., availability of affordable housing, unemployment rates, neighborhood conditions) result from structural and contextual causes that neither women nor their POs can alter.

There also may be methodological reasons for what is essentially a null finding for the direct effects of PO behavior. For example, there may be insufficient variation in recidivism indicated by arrest and by convictions to demonstrate significance. Solutions are to extend the recidivism period or to use measures of other types of outcomes of supervision in future research (e.g., employment, paying fines, participating in drug treatment; Miofsky & Byrne, 2012). These solutions might be combined with examination of outcome measures after some period of supervision has passed, for example, after a year of supervision, because effects may be lagged (Scott, Grella, Dennis, & Funk, 2014).

There also may be problems with the measurement of the match of women’s needs to PO attention to those needs. Women’s perceptions of problems and of their POs’ discussion of them may not accurately reflect the existence of these needs and/or the PO’s response to them. We did explore alternative ways to measure this match. We used WRNA cutoff scores indicating a need for intervention and then constructed a measure based on whether women above each cutoff reported discussion of that topic. This alternative measure also was unrelated to both arrests and convictions. A simple count of the number of criminogenic needs discussed, regardless of whether the WRNA or the woman identified them as problematic, was similarly unrelated to recidivism. Finally, we substituted the number of identified needs ignored for the number of needs discussed; neither the number of WRNA-identified nor the number of offender-identified problem areas that were ignored by the PO was predictive of reactions to the PO or to recidivism. It may be that whether a PO discussed a need fails to provide the necessary depth of information on how fully and through what method (e.g., counseling, making referrals, establishing requirements) the problem was addressed. Case notes and both PO and offender descriptions of what the PO said and did (e.g., gave advice, made referrals, provided emotional support) may provide more valid indicators of PO responses to women’s criminogenic needs.

Finally, it also may be difficult to detect statistically significant effects of PO attention to needs because some women participate in multiple programs, for instance, mental health, substance abuse, and employment programs. These other programs may address women’s criminogenic needs, thereby making it difficult to isolate effects of supervision. Especially in light of the many women offenders who have substance abuse and mental health problems as well as difficulties finding and keeping steady jobs that pay adequate wages and benefits, future research should consider women offenders’ interactions with multiple professionals and programs that simultaneously provide assistance.

The second key finding is that the POs’ relationship style does have an indirect effect on official recidivism through its link to offenders’ negative responses to supervision interactions. The idea here is that women with whom POs use a less supportive, more punitive style tend to respond negatively to implied constraints on their freedom and in turn are more likely to recidivate. This is admittedly a correlational finding that cannot be interpreted within a causal framework. One possibility is that women who are higher in recidivism risk elicit more punitive behavior from their POs, respond to that punitiveness with anxiety and reactance, and ultimately recidivate. In other words, the whole process could be driven by risk. However, because recidivism risk was controlled in all of our analyses, this possible explanation seems unlikely. Therefore, our evidence highlights the need for additional exploration of the links between PO style, offender reactions, and recidivism.

Experimental research on the effect of training POs to avoid behavior that elicits reactance is one potentially fruitful approach to further study. Indeed, Bogue and Nandi (2012) included a chapter on dealing with offender resistance in their National Institute of Corrections sponsored manuscript on interviewing skills in correctional settings. They write that offender resistance is a function of agent skill and that the agent is responsible for developing a style that does not promote resistance.

Psychology and communication scholars have identified several ways to reduce reactance. Specific methods supported by empirical research that could be applied in the supervision setting include ending the conversation by stating that the offender still has the freedom to choose her actions (Bessarabova, Fink, & Turner, 2013; C. H. Miller, Lane, Deatrick, Young, & Potts, 2007) and identifying and stating some similarities between POs and offenders (e.g., both are single parents; Silvia, 2005). This second approach may be most feasible when there is gender matching of client and offender. It would be useful for future research to generate evidence to support or not support whether similarities are easier to find when offender and client are matched by gender, and thus there is a bona fide reason for gender matching. POs also could forewarn the offender that what they are about to say may seem to restrict the offender’s freedoms, but there are reasons not to ignore the message (Richards & Banas, 2015). Testimonials, narrative messages, and humor can also be used to mitigate reactance, as they occupy cognitive capacity, for example, with processing the narrative, and do not allow for as much reactance to form (Moyer-Gusé & Nabi, 2010; Skalski, Tamborini, Glazer, & Smith, 2009). It would be useful to consider which of these strategies are most applicable to community supervision settings for women and to design and empirically assess training to promote those most likely to be effective.

The present article drew on theory, measures, and suggestions for practical use of the findings across the disciplines of criminal justice, feminist criminology, psychology, and communication. This interdisciplinary project provided two key advantages. First, it led to a focus on criminogenic need areas known to be most relevant to women. Second, it led to the measurement of reactance and anxiety as hypothesized responses to PO relationship styles with their clients. Although we found no direct effect of type of DRR on either arrests or convictions, we were able to show the indirect relationship of PO behavior to recidivism through negative emotions and reactance. Third, the communication literature provided a rich and detailed body of research on very specific methods for reducing reactance in settings where people are likely to feel their freedoms limited. The research framework also considered rival explanations derived from alternative theories of corrections within the field of criminal justice, which allowed us to explore alternatives to the DRR explanation for influences on women’s recidivism.

Footnotes

This article is based on work supported by the National Science Foundation under Grant 1126162 and by a Strategic Partnership grant from the Michigan State University Foundation.

Notes

Merry Morash is a professor in the School of Criminal Justice, Michigan State University. Her books include Women on Probation and Parole: A Feminist Critique of Community Programs & Services (Northeastern University Press) and Understanding Gender, Crime, and Justice (SAGE). Her current focus is on women in correctional settings and violence against women.

Deborah A. Kashy received her MS in applied statistics (1990) and her PhD in social psychology (1991) from the University of Connecticut, and she is now a professor of psychology at Michigan State University. Her research interests include the analysis of non-independent data, models of small group and dyadic social interaction, interpersonal processes in close relationships, and relationship outcomes for cancer survivors and their partners.

Sandi W. Smith is a professor in the Department of Communication and director of the Health and Risk Communication Center at Michigan State University. She studies social influence and interpersonal relationships in the realm of communication.

Jennifer E. Cobbina is an associate professor in the School of Criminal Justice at Michigan State University. Her research interests focus on gender and prisoner reentry, desistance, and recidivism. Her work has appeared in Criminology, Justice Quarterly, Crime & Delinquency, Deviant Behavior, and Race and Justice.

References

Andrews

D. A.

Bonta

(2006). The psychology of criminal conduct. Cincinnati, OH: Anderson.

Andrews

D. A.

Bonta

Wormith

J. S.

(2006). The recent past and near future of risk and/or need assessment. Crime and Delinquency, 52, 7-27.

Andrews

D. A.

Zinger

Hoge

R. D.

Bonta

Gendreau

Cullen

F. T.

(1990). Does correctional treatment work? A clinically relevant and psychologically informed meta-analysis. Criminology, 28, 369-404.

Angell

Mahoney

C. A.

(2007). Re-conceptualizing the case management relationship in intensive treatment: A study of staff perceptions and experiences. Administration and Policy in Mental Health & Mental Health Services Research, 34, 172-188.

Aos

Miller

Drake

(2006). Evidence-based public policy options to reduce future prison construction, criminal justice costs, and crime rates. Olympia: Washington State Institute for Public Policy.

Belknap

(2014). The invisible woman. Belmont, MA: Wadsworth.

Bessarabova

Fink

E. L.

Turner

M. G.

(2013). Reactance, restoration, and cognitive structure: Comparative statics. Human Communication Research, 39, 339-364.

Bloom

Owen

Covington

S. S.

Raeder

M. S.

(2003). Gender-responsive strategies: Research, practice, and guiding principles for women offenders. Washington, DC: National Institute of Corrections.

Bogue

B. M.

Nandi

(2012). Motivational interviewing in corrections: A comprehensive guide to implementing MI in corrections. Washington, DC: National Institute of Corrections.

10.

Bonta

(1995). The responsivity principle and offender rehabilitation. Forum on Corrections Research, 7, 34-37.

11.

Boulet

Boss

M. W.

(1991). Reliability and validity of the Brief Symptom Inventory. Psychological Assessment: A Journal of Consulting and Clinical Psychology, 3, 433-437.

12.

Bourgon

Gutierrez

(2012). The general responsivity principle in community supervision: The importance of probation officers using cognitive intervention techniques and its influence on recidivism. Journal of Crime & Justice, 35, 149-166.

13.

Brehm

J. W.

(1966). A theory of psychological reactance. San Diego, CA: Academic Press.

14.

Brehm

J. W.

Brehm

S. S.

(1981). Psychological reactance: A theory of freedom and control. San Diego, CA: Academic Press.

15.

Brennan

Dieterich

Ehret

(2009). Evaluating the predictive validity of the COMPAS risk and needs assessment system. Criminal Justice and Behavior, 36, 21-40.

16.

Bui

Morash

(2010). The impact of network relationships, prison experiences, and internal transformation on women’s success after prison release. Journal of Offender Rehabilitation, 49, 1-22.

17.

Clear

T. R.

(2007). Imprisoning communities. How mass incarceration makes disadvantaged neighborhoods worse. New York, NY: Oxford University Press.

18.

Cohen

M. A.

Bowles

(2010). Estimating costs of crime. In Piquero

A. R.

Weisburd

(Eds.), Handbook of quantitative criminology (pp. 143-162). New York, NY: Springer.

19.

Covington

S. S.

(1999). Helping women recover: A program for treating addiction. San Francisco, CA: Jossey-Bass.

20.

Covington

S. S.

(2008). Women and addiction: A trauma-informed approach. Journal of Psychoactive Drugs, 40, 377-385.

21.

Cullen

F. T.

Gendreau

(2000). Assessing correctional rehabilitation: Policy, practice, and prospects. In Horney

(Ed.), Policies, processes, and decisions of the criminal justice system: Criminal justice 2000 (Vol. 3, pp. 109-175). Washington, DC: National Institute of Justice.

22.

Derogitas

L. R.

Melisarotos

(1983). The Brief Symptom Inventory: An introductory report. Psychological Medicine, 13, 595-605.

23.

Dillard

J. P.

Shen

(2005). On the nature of reactance and its role in persuasive health communication. Communication Monographs, 72, 114-168.

24.

Fedock

Fries

Kubiak

S. P.

(2013). Service needs for incarcerated adults: Exploring gender differences. Journal of Offender Rehabilitation, 52, 493-508.

25.

Fielding

J. E.

Tye

Ogawa

P. L.

Imam

I. J.

Long

A. M.

(2002). Los Angeles County drug court programs: Initial results. Journal of Substance Abuse Treatment, 23, 217-224.

26.

Gendreau

Smith

French

S. A.

(2006). The theory of effective intervention: Empirical status and future directions. In Cullen

F. T.

Wright

J. P.

Blevins

K. R.

(Eds.), Taking stock: The status of criminological theory (pp. 419-446). New Brunswick, NJ: Transaction Publishers.

27.

Gill

(2010). The effects of sanction severity on criminal conduct: A randomized low intensity probation experiment. Philadelphia: University of Pennsylvania. Retrieved from http://repository.upenn.edu/edissertations/121

28.

Gill

Hyatt

Sherman

L. W.

(2010). Probation intensity effects on probationers’ criminal conduct. The Campbell Collaboration. Retrieved from http://www.campbellcollaboration.org

29.

Glaze

L. E.

Herberman

E. J.

(2013). Correctional populations in the United States, 2012. Washington, DC: National Institute of Justice.

30.

Grattet

Lin

Petersilia

(2011). Supervision regimes, risk, and official reactions to parolee deviance. Criminology, 49, 371-399.

31.

Green

T. C.

Johnson

Harrington

Pouget

E. R.

Rhodes

A. G.

Taxman

F. S.

. . . Friedmann

P. D.

(2013). Parole officer-parolee relationships and HIV risk behaviors during community supervision. AIDS and Behavior, 17, 2667-2675.

32.

Greiner

L. E.

Law

M. A.

Brown

S. L.

(2015). Using dynamic factors to predict recidivism among women: A four-wave prospective study. Criminal Justice and Behavior, 42, 457-480.

33.

Guerino

Harrison

P. M.

Sabol

W. J.

(2011). Prisoners in 2010. Washington, DC: Bureau of Justice Statistics.

34.

Hanley

(2006). Appropriate services: Examining the case classification principle. Journal of Offender Rehabilitation, 42, 1-22.

35.

Hawken

Kleiman

(2009). Managing drug involved probationers with swift and certain sanctions: Evaluating Hawaii’s HOPE. Retrieved from https://www.ncjrs.gov/pdffiles1/nij/grants/229023.pdf

36.

Hipp

J. R.

Petersilia

Turner

(2010). Parolee recidivism in California: The effect of neighborhood context and social service agency characteristics. Criminology, 48, 947-979.

37.

Holtfreter

Reisig

M. D.

Morash

(2004). Poverty, state capital, and recidivism among women offenders. Criminology & Public Policy, 3, 185-208.

38.

Hong

Faedda

(1996). Refinement of the Hong Psychological Reactance Scale. Educational and Psychological Measurement, 56, 173-182.

39.

Huebner

B. M.

DeJong

Cobbina

J. E.

(2010). Women coming home: Long-term patterns of recidivism. Justice Quarterly, 27, 225-254.

40.

Jalbert

S. K.

Rhodes

(2012). Reduced caseloads improve probation outcomes. Journal of Crime & Justice, 35, 221-238.

41.

Jalbert

S. K.

Rhodes

Flygare

Kane

(2010). Testing probation outcomes in an evidence-based practice setting: Reduced caseload size and intensive supervision effectiveness. Journal of Offender Rehabilitation, 49, 233-253.

42.

Jordan

J. V.

(Ed.). (2013). The power of connection: Recent developments in relational-cultural theory. New York, NY: Routledge.

43.

Kennealy

P. J.

Skeem

J. L.

Manchak

S. M.

Louden

J. E.

(2012). Firm, fair, and caring officer-offender relationships protect against supervision failure. Law and Human Behavior, 36, 496-505.

44.

Kleiman

M. A. R.

(2011). Justice reinvestment in community supervision. Criminology & Public Policy, 10, 651-659.

45.

Kubrin

C. E.

Stewart

E. A.

(2006). Predicting who reoffends: The neglected role of neighborhood context in recidivism studies. Criminology, 44, 165-197.

46.

Langan

Pelissier

(2001). Gender differences among prisoners in drug treatment. Journal of Substance Abuse Treatment, 13, 291-301.

47.

Latessa

E. J.

Travis

Fulton

Stichman

(1998). Evaluating the prototypical ISP: Final report. Cincinnati, OH: University of Cincinnati & American Probation and Parole Association.

48.

Lowenkamp

C. T.

Holsinger

A. M.

Latessa

E. J.

(2005). Are drug courts effective? A meta-analytic review. Journal of Drug Issues, 31, 177-206.

49.

Lowenkamp

C. T.

Latessa

E. J.

(2004). Understanding the risk principle: How and why correctional interventions can harm low-risk offenders (Topics in community corrections). Washington, DC: National Institute of Corrections.

50.

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 & Delinquency, 52, 77-93.

51.

MacKenzie

D. L.

(2006). Reducing the criminal activities of known offenders and delinquents: Crime prevention in the courts and corrections. In Sherman

L. W.

Farrington

D. P.

Welsh

B. C.

MacKenzie

D. L.

(Eds.), Evidence-based crime prevention (pp. 330-404). New York, NY: Routledge.

52.

Maidment

M. R.

(2006). Doing time on the outside. Toronto, Ontario, Canada: University of Toronto Press.

53.

Makarios

Steiner

Travis

L. F.

(2010). Examining the predictors of recidivism among men and women released from prison in Ohio. Criminal Justice and Behavior, 37, 1377-1391.

54.

Mallik-Kane

Visher

C. A.

(2008). Health and prisoner reentry: How physical, mental, and substance abuse conditions shape the process of reintegration. Washington, DC: The Urban Institute.

55.

Manchak

S. M.

Skeem

J. L.

Douglas

K. S.

Siranosian

(2009). Does gender moderate the predictive utility of the Level of Services Inventory–Revised (LSI-R) for serious violent offenders? Criminal Justice and Behavior, 36, 425-442.

56.

Maruschak

L. M.

Bonczar

T. P.

(2013). Probation and parole in the United States, 2012. Washington, DC: Bureau of Justice Statistics.

57.

Maxwell

J. C.

Freeman

(2007). Gender differences in DUI offenders in treatment in Texas. Traffic Injury Prevention, 8, 353-360.

58.

Meredith

Prevost

(2009). Developing data driven supervision protocols for positive parole outcomes: Final project report 2005-IJ-CX-0029. Washington, DC: National Institute of Justice.

59.

Miller

C. H.

Lane

L. T.

Deatrick

L. M.

Young

A. M.

Potts

K. A.

(2007). Psychological reactance and promotional health messages: The effects of controlling language, lexical concreteness, and the restoration of freedom. Human Communication Research, 33, 219-240.

60.

Miller

J. B.

(1976). Toward a new psychology of women. Boston, MA: Beacon Press.

61.

Miofsky

K. T.

Byrne

J. M.

(2012). Evaluation research and probation: How to distinguish high performance from low performance programmes. In Gadd

Karstedt

Messner

S. F.

(Eds.), The SAGE handbook of criminological research methods (pp. 336-358). London, England: SAGE.

62.

Morash

(2010). Women on probation and parole: A feminist critique of community services and programs. Boston, MA: Northeastern University Press.

63.

Morash

Kashy

D. A.

Smith

S. W.

Cobbina

J. E.

(2015). The effects of probation or parole agent relationship style and women offenders’ criminogenic needs on offenders’ responses to supervision interactions. Criminal Justice and Behavior, 42, 412-434.

64.

Moyer-Gusé

Nabi

R. L.

(2010). Explaining the effects of narrative in an entertainment television program: Overcoming resistance to persuasion. Human Communication Research, 36, 26-52.

65.

Mumola

C. J.

Karberg

J. C.

(2006). Drug use and dependence, state and federal prisoners, 2004. Washington, DC: U.S. Department of Justice.

66.

Muthén

L. K.

Muthén

B. O.

(2015). Mplus users guide (7th ed.). Los Angeles, CA: Author.

67.

National Association for the Advancement of Colored People. (2011). Misplaced priorities: Over incarcerate, under educate. Baltimore, MD: Author.

68.

Ostermann

(2015). How do former inmates perform in the community? A survival analysis of rearrests, reconvictions, and technical parole violations. Crime & Delinquency, 61, 163-187.

69.

Ostermann

Miller

Matejkowski

(2013). Are “parole failures” parolees when they fail? Incorporating supervision terms and the timing of arrest events into parolee recidivism rates. Justice Research and Policy, 15, 1-20.

70.

Paparozzi

M. A.

Gendreau

(2005). An intensive supervision program that worked: Service delivery, professional orientation, and organizational supportiveness. The Prison Journal, 85, 445-466.

71.

Petersilia

Turner

(1993). Intensive probation and parole. Crime and Justice, 17, 281-335.

72.

Richards

A. S.

Banas

J. A.

(2015). Inoculating against reactance to persuasive health messages. Health Communication, 30, 451-460.

73.

Salina

D. D.

Lesondak

L. M.

Razzano

L. A.

Weilbaecher

(2007). Co-occurring mental disorders among incarcerated women: Preliminary findings from an integrated health treatment study. Journal of Offender Rehabilitation, 45, 207-225.

74.

Schram

P. J.

Koons-Witt

B. A.

Williams

F. P.

McShane

M. J.

(2006). Supervision strategies and approaches for female parolees: Examining the link between unmet needs and parolee outcome. Crime & Delinquency, 52, 450-471.

75.

Scott

C. K.

Grella

C. E.

Dennis

M. L.

Funk

R. R.

(2014). Predictors of recidivism over 3 years among substance-using women released from jail. Criminal Justice and Behavior, 41, 1257-1289.

76.

Silvia

P. J.

(2005). Deflecting reactance: The role of similarity in increasing compliance and reducing resistance. Basic and Applied Social Psychology, 27, 277-284.

77.

Singh

J. P.

Desmarais

S. L.

Sellers

B. G.

Hylton

Tirotti

Van Dorn

R. A.

(2014). From risk assessment to risk management: Matching interventions to adolescent offenders’ strengths and vulnerabilities. Children and Youth Services Review, 47, 1-9.

78.

Sirakaya

(2006). Recidivism and social interactions. Journal of the American Statistical Association, 101, 863-877.

79.

Skalski

Tamborini

Glazer

Smith

S. W.

(2009). Effects of humor on presence and recall of persuasive messages. Communication Quarterly, 57, 136-153.

80.

Skeem

J. L.

Encandela

Eno Louden

(2003). Perspectives on probation and mandated mental health treatment in specialized and traditional probation departments. Behavioral Sciences & the Law, 21, 429-458.

81.

Skeem

J. L.

Eno Louden

Manchak

S. M.

Vidal

Haddad

(2009). Social networks and social control of probationers with co-occurring mental and substance abuse problems. Law & Human Behavior, 33, 122-135.

82.

Skeem

J. L.

Eno Louden

Polaschek

D. L.

Camp

(2007). Assessing relationship quality in mandated community treatment: Blending care with control. Psychological Assessment, 19, 397-410.

83.

Skeem

J. L.

Manchak

S. M.

Vidal

Hart

(2009, March). Probationers with mental disorder: What (really) works? Paper presented at the American Psychology and Law Society, San Antonio, TX. Retrieved from https://webfiles.uci.edu:443/skeem/Downloads.html

84.

Smith

Gendreau

Goggin

(2004). Correctional treatment: Accomplishments and realities. In Van Voorhis

Braswell

Lester

(Eds.), Correctional counseling & rehabilitation (Vol. 5, pp. 315-324). Cincinnati, OH: Anderson.

85.

Smith

S. W.

Cornacchione

J. J.

Morash

Kashy

Cobbina

(in press). Communication style as an antecedent to reactance, self-efficacy, and restoration of freedom for drug- and alcohol-involved women on probation and parole. Journal of Health Communication.

86.

Sullivan

C. M.

Rumptz

M. H.

Campbell

Eby

K. K.

Davidson

W. S.

(1996). Retaining participants in longitudinal community research: A comprehensive protocol. Journal of Applied Behavioral Science, 32, 262-276.

87.

Taxman

F. S.

(2002). Supervision: Exploring the dimensions of effectiveness. Federal Probation, 66, 14-27.

88.

Taxman

F. S.

Ainsworth

(2009). Correctional milieu: The key to quality outcomes. Victims & Offenders, 4, 334-340.

89.

Trotter

(2006). Working with involuntary clients. Sydney, Australia: Allen & Unwin.

90.

Van Voorhis

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

91.

Van Voorhis

Salisbury

Wright

Bauman

. (2007). New classification instruments for female offenders. Corrections Today, 69, 96-97.

92.

Van Voorhis

Wright

E. M.

Salisbury

Bauman

. (2010). Women’s risk factors and their contributions to existing risk/needs assessment: The current status of a gender-responsive supplement. Criminal Justice and Behavior, 37, 261-288.

93.

Vieira

T. A.

Skilling

T. A.

Peterson-Badali

(2009). Matching court-ordered services with treatment needs: Predicting treatment success with young offenders. Criminal Justice and Behavior, 36, 385-401.

94.

Warchol

G. L.

(2000). Intensive supervision probation: An impact evaluation. The Justice Professional: A Critical Journal of Crime, Law and Society, 13, 219-232.

95.

Watson

Pennebaker

J. W.

Folger

(1987). Beyond negative affectivity: Measuring stress and satisfaction in the workplace. In Ivancevich

J. M.

Ganster

D. C.

(Eds.), Job stress: From theory to suggestion (pp. 141-157). New York, NY: Haworth Press.