Abstract
There is much debate about the effects of punitive or treatment responses to the many women who are on probation and parole. This article examines whether types of technical violations (drug or nondrug related) and responses to them (treatment or punishment oriented) as well as supervision intensity predict recidivism. Study participants are 385 women on probation or parole for a felony offense, and official records of violations and recidivism are the data source. Negative binomial regression analysis revealed that for high-risk women, treatment responses to nondrug violations are related to reductions in recidivism, whereas punitive responses to nondrug offenses are related to increased recidivism. For low-risk women, treatment responses to non-drug-related violations are related to increased recidivism and punitive responses to violations unrelated to drug use are related to decreased recidivism. Study findings suggest differential reactions to common supervision practices depending on a woman’s initial risk to recidivate.
Individuals sentenced to probation instead of incarceration and those released from prison to complete their sentences on parole are expected to conform to numerous requirements. Requirements vary between offenders and jurisdictions, and include avoiding drug use and submitting to drug testing and treatment; paying fines, fees, and restitution; securing approved housing and employment; avoiding contact with known felons; and attending various treatment, educational, and job preparation programs. When offenders fail to follow rules and meet requirements, agents issue technical violations that are distinct from arrests and charges for the commission of new crimes (Campbell, 2016). In many cases, agents not only issue violations but also respond to violations by levying sanctions or they add or increase requirements for treatment. Supervision of substance-involved offenders often combines objectives, because substance misuse is viewed as a combination of criminal, medical, and behavioral problems that can be addressed through both sanctions to shape behavior and administer punishment, as well as through substance abuse and mental health treatment (Tiger, 2011).
The purpose of the study described in this article is to examine the relationship between patterns of issuing and responding to technical violations during an 18-month period and subsequent recidivism. The focus is on women, first, because compared with men, women are disproportionately in conflict with the law because of their substance involvement (Belknap, 2014; Guerino, Harrison, & Sabo, 2011; Langan & Pelissier, 2001; Mumola & Karberg, 2006). Thus, responses to technical violations are likely to include both sanctions and treatment requirements. Second, women make up an increasing proportion of offenders on probation and parole. At the end of 2014, more than 1 million U.S. women were supervised in the community, and they constituted 12% of the national parole population and 25% of the probation population (Herberman & Bonczar, 2014). Extrapolation of these numbers over multiple years indicates that community supervision practices affect a large part of the U.S. population, a majority of convicted offenders, and a sizable number of women. The high number of women who are supervised in the community is a major justification for research attention to them.
Research has documented considerable variation in the reasons for issuing technical violations and in responses to violations. For example, at one point, the California correctional system specified 247 different types of parole violations (Grattet, Petersilia, Lin, & Beckman, 2009). The many possible responses to violations include revocations of community supervision and incarceration in prison, short jail terms, increased monitoring and drug testing, verbal warnings, and encouragement or requirements to attend treatment and educational programs. Given these variations, scholars have pointed out the need for better understanding of whether recidivism outcomes are affected by complex combinations of the number of specific types of technical violations and the nature of responses to them (Clear, Harris, & Baird, 1992; Hamilton & Campbell, 2013; Hawken & Kleiman, 2009; Kleiman, 2011; MacKenzie, Browning, Skroban, & Smith, 1999; Rydberg & Grommon, 2016). The present analysis addresses this key question about the effects of issuing violations and of alternative ways of responding to them.
The lack of research on offenders in general and on technical violations in particular is most apparent for women, who are often understudied in correctional research either through omission or because they are a small proportion of mixed-gender samples (Belknap, 2014; Campbell, 2016; Schulenberg, 2007). It is especially important to study technical violations in a sample of women, because research in several correctional settings indicates that compared with men, women are subjected to closer supervision and have more or unique requirements to meet (Bosworth, 2007; Caputo, 2014; Carlen & Tchaikovsky, 1996; McCorkel, 2003). Heightened monitoring of women offenders in the community results from beliefs that women are at unique risk for victimization, because they reoffend for gender-related reasons (e.g., relationships with abusive men) or due to assumptions that women offenders’ high needs (e.g., poverty, mental illness) indicate high risk (Hannah-Moffat, 2004; Opsal, 2009; Pollack, 2007). The prior findings that in some correctional settings female offenders were monitored at high levels was one reason for our interest in the connection between detection and responses to women’s technical violations, and their subsequent recidivism.
The next section frames the research within the long-standing debate about whether punitive or treatment-oriented correctional supervision and intervention practices promote decreased recidivism. Then, we summarize the few studies that provide evidence regarding the connection of recidivism to supervision intensity; supervision intensity is an alternative explanation of the effect of supervision on recidivism, and thus must be considered as a control variable. Next, we examine the challenges involved in conceptualizing and operationalizing the dynamics of technical violations over time. Finally, we assess whether risk for recidivism alters (i.e., moderates) the connections between patterns of detection/responses to technical violations and recidivism.
Literature Review
Treatment and Punitive Responses to Violations
Historical shifts back and forth between treatment and punitive approaches characterize the history of U.S. correctional practice (MacKenzie, 2012; Steiner, Wada, Hemmens, & Burton, 2005). Related to these shifts, there is a long-standing debate about the effect of punishment versus treatment on recidivism (e.g., Cullen & Jonson, 2014; Cullen, Pratt, Micelli, & Moon, 2002; Lipsey, 2009; MacKenzie, 2006). This debate pertains to the choice of rehabilitation-oriented rather than punitive responses to technical violations.
Period-specific state and local policies influence the relative emphasis on punishment versus rehabilitation and treatment. The evolution of California’s parole policies illustrates the resulting fluctuation in emphasis. In the early 2000s, state policy required referral of even very minor technical parole violators to the Board of Parole Hearings, which typically returned referred offenders to prison (Grattet et al., 2009). California later took several steps to reduce prison populations, in part by instituting guidelines to limit revocations (Grattet & Lin, 2016; Grattet et al., 2009). Additional policy changes led to placing parolees who had committed the least serious offenses on non-revocable parole. Similar changes in states other than California have reduced the use of incarceration and increased the use of alternative punitive sanctions in the community and/or promoted community-based treatment interventions instead of punishment (Hamilton & Campbell, 2013; Taxman, 2008). In fact, some scholars have described a shift from mass incarceration to mass supervision (DeMichele, 2014; Phelps, 2013; Taxman, 2015). Despite these general trends away from punitive sanctions, there are other jurisdictions that have moved in the opposite direction, by increasing punitiveness through emphasis on the certain and swift delivery of a jail time in response to technical violations (e.g., see Alm, 2013; Hawken & Kleiman, 2009). Given the inconsistent trends and uncertainty about effective approaches, research on the effects of alternative responses to technical violations is especially timely.
Supervision Intensity as an Alternative Explanation of Recidivism
Besides examining the number and types of technical violations issued and the number of treatment and punitive responses to the violations, the literature identifies supervision intensity as an influence on both the detection of technical violations and on recidivism. Supervision intensity must, therefore, be incorporated as a control variable in analyses to understand the relationship of issuing/responding to violations and subsequent recidivism (Olson & Lurigio, 2000).
An early, nine-state field experiment to examine the effects of intensive supervision programs that emphasized close monitoring of offenders coupled with incarceration in response to violations revealed “no support for the argument that violating offenders on technical conditions suppressed new criminal arrests” (Petersilia & Turner, 1993, p. 342). Showing even more negative effects of punitive responses to violations, research in the state of Washington revealed that incarceration for probation and parole violations actually increased recidivism (Drake & Aos, 2012). Contradicting these findings, MacKenzie and Brame (2001; also see Paparozzi & Gendreau, 2005) found that, after controlling for potential influences on both conventional ties (e.g., to school or work) and recidivism, increased intensity in the supervision of Virginia probationers led to their greater involvement in conventional and therapeutic activities, and this involvement reduced recidivism. Similarly, research in Oklahoma City and Polk County, Iowa, showed that when responses to violations rested on evidence-based practices that addressed criminogenic needs, the increased supervision that came with reduced caseloads substantially decreased recidivism; also, the number of technical violations increased minimally or not at all (Jalbert & Rhodes, 2012; Jalbert, Rhodes, Flygare, & Kane, 2010). Considered together, these studies indicate the need to examine supervision intensity, number of technical violations, and the treatment versus punitive nature of responses to those violations in relation to recidivism. A key factor in successful (i.e., recidivism reducing) supervision intensity may be coupling it with treatment-oriented rather than punitive responses.
Conceptual Challenges in Studying Technical Violations
The conceptualization and related measurement of an offender’s experience of technical violations and responses to them is problematic, because of the multiple dimensions of possible variation. First, agents within and between correctional organizations vary considerably in their relative emphasis on punishment versus rehabilitation programming in responses to technical violations (Jones & Kerbs, 2007; Kerbs, Jones, & Jolley, 2009). For instance, Eno Louden, Skeem, Camp, and Christensen (2008) found that in agencies with policies and practices tailored to mentally ill offenders, agents spent more time on rehabilitation and were less punitive in responses to violations. Further complicating conceptualization of responses to violations, punitive and treatment responses are often mixed, sometimes because responses become less treatment-oriented and more punitive for an offender with multiple violations (e.g., see Cullen, Manchak, & Duriez, 2014; Duriez, Cullen, & Manchak, 2014; Kleiman, Kilmer, & Fisher, 2014). In other cases and settings, contradictory but coexisting assumptions that use of illegal drugs is a disease, a moral failing, a criminal behavior, or some mix of these three explanations accounts for inconsistent responses to continued substance use (Murphy, 2015). The complex and inconsistent patterns of response to violations require the use of a measure that reflects patterns of different types of violations and responses to them.
Just a few studies have investigated variation in both types of and responses to violations. Clear et al. (1992) took a unique approach to the study of violations in a multistate study of more than 7,000 offenders in six probation agencies. They examined the interconnections of the severity of violations, the severity of responses to violations, and the seriousness of new crimes. Findings of no relationship between severity of sanctions in response to violations and subsequent seriousness of misbehavior led them to conclude that the minor nature of most violations and the infrequency of repeated violations did not justify the use of controlling strategies such as electronic monitoring and intensive supervision. Differing from these findings, research on Virginia offenders showed that the number of technical violations was positively related to recidivism, but the nature of the agent’s response to violations was unrelated (MacKenzie et al., 1999). Finally, a study of California parolees concluded that parolee characteristics, the intensity of supervision, and the parole agent’s tolerance for deviance had independent effects on the length of time before a violation was issued; serious, violent, and sexual offenders were at increased risk for violating requirements in a short time, in part because they were supervised more intensely (Grattet, Lin, & Petersilia, 2011). As Wodahl, Boman, and Garland (2015) observed, with the few exceptions just noted, most studies of alternative responses to violations have focused primarily on incarceration as a sanction and not the full gamut of possible responses. Moreover, none of these studies focused on women offenders as a distinct subgroup. Finally, the very limited and inconsistent findings about the effects of alternative responses show the need for additional research.
Recidivism Risk as a Potential Moderator
Another neglected dynamic in technical violations research is the possibility that issuing and responding to technical violations have different effects depending on offenders’ initial risk for recidivism. Prior research indicates that women offenders at high and low risk for recidivism have different outcomes despite similar community supervision experiences (e.g., Kennealy, Skeem, Manchak, & Eno Louden, 2012; Morash, Kashy, Smith, & Cobbina, 2015; Skeem, Manchak, Vidal, & Hart, 2009; Smith, Cornacchione, Morash, Kashy, & Cobbina, 2016). In research that included both males and females, a meta-analysis of studies of the effect of correctional interventions further demonstrated that high- and low-risk offenders are differently affected by interventions, and this difference is greatest for females (Andrews & Dowden, 2006). More specifically, low-risk offenders do at least as well or even better when supervision is least intense (Aos, Miller, & Drake, 2006; Barnes et al., 2010), perhaps because low-risk offenders who are intensely supervised are mandated to take part in programs where they comingle with high-risk offenders who negatively influence them (Lowenkamp & Latessa, 2004). Consideration of the prior research findings demonstrates the need to examine moderating effects of risk for recidivism on the connection of recidivism to both the intensity of supervision and the pattern of issuing technical violations and responding to them.
Research Questions
The current state of knowledge led us to address the following research questions for a sample of women supervised in the community:
Method
Overview
The present analysis considers women recruited to take part in a study of probation and parole in Michigan. It examines the relationship between both the numbers of different types of technical violations and the responses to them and official measures of recidivism, net of the effect of the intensity of supervision. The data were time ordered, such that violation detection and response for an 18-month period, controlling for supervision intensity in that period, was used to predict recidivism in the 2 years following that 18-month period.
Research Setting
Unique from other states, the data collection site, Michigan, has a centralized system of probation and parole supervision for felony offenders. In counties with large populations, supervising agents specialize in probation or parole, and within the reporting centers, they have woman-only caseloads. In counties with smaller populations, agents have mixed gender and/or mixed probation and parole caseloads, but the agents who supervise women still specialize in working with women offenders as part of their caseloads. Michigan is also unique by virtue of (a) using an assessment instrument tailored to the needs of women for all women on parole and many of those on probation, and (b) having availability of several female-specific programs, including substance abuse treatment programs, for women offenders in the community (Holtfreter & Wattanaporn, 2014).
Predating and distinct from ongoing prison population reduction efforts, the Michigan Department of Corrections developed and implemented policies to ensure consistent responses to technical violations across the state. The policies also were intended to establish proportionality of responses to the nature of the violation and to protect public safety. Specifically, the official statewide policy and a review process sets limits on the choice of response to each type of violation, provides monitoring of responses, and requires approval of deviations from standards. Despite these regularities, however, there is variation in response to violations within the state. Some of this variation results from drug and alcohol courts in some counties or judicial responses to probation violations, and some of it results from agent differences in practice.
Sample
Because subgroups of women offenders follow different pathways into illegal activity (Daly, 1992; Morash, 2010; Reisig, Holtfreter, & Morash, 2006; Salisbury & Van Voorhis, 2009), the subgroups might be differentially affected by the same supervision experiences. Therefore, just the most common group, the two thirds of women offenders with substance abuse involvement, were recruited into the study (Belknap, 2014; Guerino et al., 2011; Langan & Pelissier, 2001; Mumola & Karberg, 2006).
In 2011 and 2012, an initial sample of 402 women felons was obtained by first recruiting agents. Seventy-seven agents with caseloads that included women were identified in 16 counties within 1½-hour drive to the research office. The proportion of agents invited to participate in each county corresponds to the proportions of women supervised in each of the 16 counties. These counties included 68.45% (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, parole officers were oversampled in relation to probation officers. Seventy-three of the 77 identified agents (94.8%) took part in the study. Of the four who did not take part, one withdrew, one refused, one was reassigned to supervise men, and one took a medical leave.
One of the authors reviewed the caseload list with each agent and assisted in identifying 846 eligible clients (i.e., women with substance involvement, convicted of a felony, who had been supervised for approximately 3 months). The discussion not only focused on the crime resulting in the present conviction but also included discussion of drug use, manufacture, and distribution in the past and present. Agents facilitated trained interviewers in recruiting 402 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. A comparison of available data on participants and nonparticipants revealed no statistically significant differences in official records of substance use, violations, arrests, misdemeanor convictions, and felony convictions in a 12-month period. Nonparticipants were slightly but significantly more likely to be in jail or prison, suggesting a small bias toward including women who were not incarcerated at 12 months.
For the present analysis to predict recidivism from supervision intensity and the number of and the responses to technical violations, the sample was reduced by 17 women. Twelve women were not available for any part of the 2-year period considered to measure recidivism. The reasons were that six left the state before the period, and thus we did not have access to detailed recidivism data, three were incarcerated for the entire period, and three passed away before the period began. Four women were on both probation and parole, and one had missing data, so they were also dropped from the analysis. Thus, the present analysis focuses on 385 women.
Description of the Sample, Violations, and Responses to Violations
The mean age of the 385 women was 33.68 (SD = 10.38, range = 18-60). The sample was ethnically and racially diverse and included 181 (47.01%) White women, 123 (31.95%) Black women, and 74 (19.22%) women who were Hispanic or multiracial. Two women were Native American, and five did not report race or ethnicity. In terms of their criminal backgrounds, 298 (77.40%) of the women were on probation, and 87 (22.60%) were on parole. Including the current offense for which the woman was on probation or parole, at the onset of the study, the average number of arrests was 4.77 (SD = 3.76, range = 1-23), the average number of prior convictions for felonies was 2.88 (SD = 3.07, range = 0-21), and the average number of months in prison was 11.85 (SD = 31.16, range = 0-233). During the 18 months after the start of supervision, on average women experienced 0.28 arrests (SD = .58, range = 0-3) and 0.21 convictions (SD = .52, range = 0-3). For recidivism, in the 2 years following the 18th month after the start of supervision, 64 women had one arrest, 21 women had two, 3 women had three, and 5 women had between four and eight arrests. The remaining 292 women had no arrests in the recidivism period that we considered.
Measurement
Dependent variable
Recidivism is a count of the number of arrests after the 18th month of supervision through the 42nd month. These counts were taken from official state police data, and unlike official data in some other states, they do not include arrests by either police or supervising agents in response to technical violations. To take into account months at risk for recidivism, what is referred to as an offset was included in the prediction models (Hilbe, 2014). For the recidivism period, months at risk was determined for each woman by subtracting months when recidivism could not occur or be determined (i.e., months in prison, months after death, months out of state). For the 2-year recidivism period, 351 (91.17%) women were in the community, and thus available to be arrested for the entire period. Due to death, moving out of state, or time in prison, for the remaining 34 women, the average time at risk for arrest was 12.55 months (SD = 6.71, range = 1.84-23.75).
Predictor variables
The case notes included the total of 240 pairs of a type of violation with a treatment or a punitive-oriented response to that violation. These were the types of responses of theoretical interest. 1 Counts were taken for four pairings of technical violation type and related response: (a) a drug-related violation and a treatment-oriented response (n = 110, 45.83% of 240 violation type-response pairs), (b) a drug-related violation and a punitive response (n = 64, 26.67%), (c) a non-drug-related violation and a substance abuse treatment response (n = 16, 6.66%), and (d) a non-drug-related violation and a punitive response (n = 50, 20.83%). To be more specific, all 240 violation type and response pairs of interest were considered in the analysis. The majority of women (n = 276; 71.69%) did not have any of the types of parings of violations and responses. Sixty-one women (15.84%) had one pairing, 20 (5.19%) had two, and the remainder (n = 28, 7.28%) had between 3 and 12. Drug-related violations included drug use and failure to comply with treatment or drug testing requirements. Non-drug-related violations included failure to report, pay fines and fees and restitution, notify of change of address or employment, avoid contact with a known felon, and complete community service. Treatment responses encouraged, required, changed, or increased substance abuse treatment. Punitive responses included jail time, community service, increased intensity or length of supervision, increased drug testing, and revocation of supervision resulting in incarceration.
Moderating and control variables
To indicate risk for recidivism, we used a factor score that combined probation versus parole status (probation = 0, parole = 1), months in prison before the start of the study, number of arrests prior to the start of the study, felony arrests prior to the start of the study, number of arrests in the 18 months after the start of supervision, and total convictions in the 18 months after the start of supervision. Due to a highly skewed distribution, the square root of the count of months in prison was used to generate the factor score. Official department of corrections data and arrest record data were the source of this information. The Cronbach’s alpha for this combination of variables is .64. 2 As other researchers have discovered (see review and analysis by Via, Dezember, & Taxman, 2017), these static factors were the strongest predictors of recidivism in the data that we used, and therefore they were appropriate control variables.
Intensity of supervision was a count of the number of office visits, home visits, and telephone contacts of the supervising agent with the offender during the 18 months after the start of supervision. This information was taken from official case notes, and it did not include violations for new crimes as indicated by arrests or convictions.
As additional control variables, offenders self-reported their age and race during an interview in the first few months of supervision, and race was coded to reflect the primary racial minority group in comparison with other women (0 = Other, 1 = Black).
Analytic Strategy
Negative binomial regression was used to predict recidivism. Independent variables included counts of the four pairings of violation types with response types, age, race, risk for recidivism, and supervision intensity. The model also included interactions between risk of recidivism and each of the four counts of pairings of violation and response types. To ensure that an omitted interaction does not distort the findings, the interaction between supervision intensity in 18 months from the start of supervision and recidivism risk also is included in the model (Hilbe, 2014). Negative binomial regression is the appropriate statistical technique because the dependent variable is a count with a high proportion of zero values and a mean (M = .370) lower than its variance (s2 = .759). All continuous variables were grand mean centered to avoid problems with multicollinearity (Aiken & West, 1991). Identification of a significant interaction was followed by simple slopes analyses examining the effects of the moderated variable for high and low recidivism risk, defined as ±1 standard deviation from the mean. For the multivariate analysis, the variance inflation factors (VIFs) are all below 2, which is considerably below the recommended critical level of 4.0 (Fisher & Mason, 1981). Robust standard errors were used for all Wald tests of the significance of estimated coefficients.
Due to the sampling approach, women were clustered in counties and within caseloads. A likelihood ratio test to compare an intercept only model with and without the multilevel effects showed a chi-square value close to zero and no significant difference when the county and agent effects were included in the model. Thus, multilevel modeling was not used.
Results
Table 1 presents the range, mean, and standard deviations for the variables of primary interest in the analysis. The correlations between these variables and the demographic variables are presented in Table 2. The indicator of recidivism risk is moderately and positively related to supervision intensity, being Black, being older, and the count of drug-related violations that resulted in a treatment response. Supervision intensity is moderately related to the number of drug-related violations that resulted in both treatment and punishment responses. There are also modest but statistically significant positive correlations among the four pairs of types of violations with types of responses. Thus, although there is a tendency for women to experience some mixing in the types and responses for violations, the counts of different pairings of violation type with response were not highly correlated. For example, a high number of punitive responses to drug-related violations is not highly correlated with a high number of treatment responses to drug-related violations.
Range, Means, and Standard Deviations for the Study Variables (N = 385).
Correlations for Study Variables (N = 385).
p < .05. **p < .01.
A residual deviance goodness of fit test was conducted for the multivariate model predicting recidivism (i.e., arrests). This test evaluates the difference between the deviance of the model being estimated and the maximum deviance of an ideal model, so a nonsignificant difference indicates a well-fitting model (Hilbe, 2014). This test result indicated that the model is well fit to the data (residual deviance = 260.4, df = 371, p = .99).
Table 3 presents the estimated coefficients, robust standard errors, incident ratio rates (IRRs; also commonly referred to as the exponentiated coefficients), and the results of the partial Wald tests of the significance of each coefficient. As would be expected, risk for recidivism is related to recidivism in the period from 18 to 42 months after the start of supervision. Both being Black and being older are negatively related to recidivism. None of the main effects for the pairings of violation type and response (e.g., a drug-related violation followed with a treatment-oriented response) were significantly related to recidivism. However, interaction effects between risk for recidivism and three of the violation-type response pairs are significant. Also, the interaction of risk for recidivism and supervision intensity is statistically significant.
Results of Test of Negative Binomial Regression Model (N = 385).
Note. IRR = incident ratio rate.
p < .05. **p < .01. ***p < .001.
For significant interaction effects, Table 4 presents results from simple slopes analyses to compare the effects of independent variables on recidivism for women with high and low risk for recidivism. First, we consider women at high risk of recidivism (i.e., 1 SD above average). For high-risk women, a one-unit increase in the number of treatment responses to nondrug violations is related to a reduction in recidivism by a factor of .685 (68.5%). Increases in punitive responses to nondrug offenses are associated with recidivism increase by a factor of 2.22, or 222%. Overall for the high-risk women, treatment responses to nondrug violations are related to reductions in recidivism, whereas punitive responses to nondrug offenses are related to increased recidivism. Moreover, for high-risk women, the intensity of supervision and the number of drug-related violations met by a treatment response were not significantly related to recidivism.
Results of Simple Slopes Analyses to Compare Women at High and Low Risk of Recidivism (N = 385).
Note. IRR = incident ratio rate.
p < .05. **p < .01.
Turning now to low-risk women (1 SD below average), there is a marginally significant (p = .062) negative relationship between number of drug violations met with treatment responses and recidivism: for each additional increase in the number of drug-related violations with a treatment response, the expected number of new arrests is reduced by a factor of .234, or 23.4%. For this low-risk group, a one-unit increase in the number of nondrug violations met with treatment is related to recidivism increased by a factor of 5.867, or 587.6%. Showing the opposite effect on recidivism, a one-unit increase in the number of nondrug violations met with punishment is related to a reduction in recidivism by factor of .648 (64.8%). Finally, a one-unit increase in intensity of supervision is positively related to recidivism by a factor of .029, or 2.9%. Thus, adding 10 additional agent contacts with a client in an 18-month period predicts an increase in recidivism of 20.9%, net of the effects of other variables. For low-risk women, reductions in recidivism are related to treatment responses to drug-related violations and nondrug violations met with punishment. However, for this low-risk group, nondrug violations with treatment responses and supervision intensity are related to increased recidivism.
Discussion
Research Question 1 asked about the relationship between the nature of detection/response to violations in an 18-month period and recidivism in the subsequent 24 months. An examination of direct effects of four pairings of number of violations (drug-related and other) and response type (treatment requirements and other) showed no significant connections. Wodahl et al. (2015), similarly, found no difference between the effects of jail sanctions and several alternative responses to violations, including written assignments, requirements to increase treatment participation, and increased community service hours. Gil (2010) also concluded from her research that low-risk offenders did not respond differently to responses to violations that were dissimilar from each other. And, in a study of responses to drug use, De Wree, De Ruyver, and Pauwels (2009) documented that community-based responses were equally effective as punitive jail sanctions in promoting desired recidivism outcomes. Notably, samples for all of these studies included women, or in the case of De Wree et al. (2009) violations by women. However, the proportion of male offenders and male-perpetrated violations was markedly higher than the proportion of female offenders and female-perpetrated violations, and no separate analyses were conducted for women and men. Thus, for all of our research questions, because prior research focused on mixed-gender samples, we cannot make a strong inference that our findings about women have been replicated with men, though it is possible that the gender groups do not differ.
Research Question 2 asks whether supervision intensity explains the relationship between violation detection and response patterns and subsequent recidivism. A test of the full model with interaction terms revealed that supervision intensity did not fully explain the effects of violation/response combinations, because several interaction effects between recidivism risk and the pattern of violations and responses were statistically significant. The interpretation of these interaction effects takes us to answering the third research question.
Research Question 3 asks whether risk for recidivism moderates the connection of violation detection/response patterns to subsequent recidivism. Recall that women at high risk for recidivism tended to be on parole rather than on probation, had spent more months in prison than other women before the start of the study, and had been arrested more often before and also in the 18 months after their supervision started. Showing the moderating effect of recidivism risk, for these high-risk women, treatment responses to non-drug-related violations were related to decreased recidivism, and punitive responses to nondrug responses were related to increased recidivism.
These findings about high-risk women raise questions about whether technical violations that on the surface appear to be unrelated to substance use may actually result from drug and alcohol problems that supervising agents detect and address through treatment. Note, however, that very few study participants experienced a substance abuse treatment response in response to a violation that was unrelated to drug use (n = 16), so the finding about this variable should be interpreted with caution. Future research should more fully explore whether the finding is replicated in settings where a larger number of participants experience a treatment response after a non-drug-related violation. If so, this would suggest that, at least for high-risk women, supervising agents would be most effective when they determine and address the reasons for the myriad types of non-drug-related technical violations, including drug dependence and addiction. Expanding the sample to men would show whether this finding can be generalized beyond women.
Also for the high-risk women, the connection between punitive responses to non-drug-related violations and increased recidivism raises questions about the efficacy of punishment. In the present study, women at high risk for recidivism did not have a statistically significant relationship between punitive responses to non-drug-related violations and recidivism as we measured it. Although not considered in the present study, some research suggests that punitive responses may even create new burdens that increase women’s lawbreaking. For instance, for a woman with multiple problems (e.g., mental illness, stress due to single-parenting, unemployment), who is failing to pay restitution or court fees, jail days or community service may just add to stressors that lead to drug use and other criminal activity. Prior research has described “the pain of punishment” that results from extensive requirements, and monitoring and sanctioning in community corrections settings (Crouch, 1993; Klingele, 2014; Maidment, 2006; May, 1994; Payne & Gainey, 1998; Petersilia & Deschenes, 1994; Petersilia & Turner, 1993). Interview studies with women under supervision could profitably further explore this possibility, and would shed additional light on the wisdom of emphasizing high levels of monitoring and related punishment for women offenders who are not meeting supervision requirements.
Regardless of whether women were at high or low risk for recidivism, the number of treatment responses to drug-related technical violations was unrelated to lower recidivism. This was unexpected, as a high proportion of responses to drug-related violations encourage or require substance abuse treatment, which should in the long-term reduce recidivism resulting from drug use. Morash’s (2010) study of women on probation and parole in a jurisdiction that emphasized gender-responsive treatment supports this conclusion. Based on qualitative data, she showed that probation and parole agents often used intensive supervision and the related detection of technical violations to justify and promote women’s participation in substance abuse treatment, and these practices were ultimately related to somewhat reduced recidivism. However, because the Department of Corrections that we studied routinely assessed substance abuse and emphasized treatment, it may be that detection of violations and efforts to encourage treatment just continued a process that was already in place. Alternatively, no matter how they are addressed (i.e., either with treatment or punitively), drug-related violations may indicate continuing substance abuse problems that are the reason for recidivism. The number of drug-related violations may indicate a continuation or a relapse to substance dependence or addiction, which often leads women to seek illegal income and to associate with people who are breaking the law, which in turn result in criminal behavior (Harm & Phillips, 2001; Huebner, Dejong, & Cobbina, 2010). Substance use may promote agents’ efforts to detect it and drive up the number of violations at the same time that it contributes to new arrests. Future research could fruitfully explore the timing of drug-related violations, engagement in treatment, and levels of recidivism to more fully understand whether drug-related violations promote women’s substance abuse treatment and eventual desistance from both drug use and related illegal activity.
The findings for women at low risk for recidivism are thought-provoking. For this group—women who have limited criminal histories or continued illegal activities and who are often on probation—treatment responses to drug-related violations are related to lower recidivism. These women may have less serious drug problems or may be more amenable to engaging in treatment, with the result that the supervising agent’s push for them to get treatment has the desired effect. Similarly, the low-risk non-drug-related violators may be effectively deterred from crime when they experience punishment or increased monitoring for technical violations that are unrelated to substance use. These findings lead us to reconsider the popular and empirically supported Risk-Needs-Responsivity (RNR) model, which emphasizes the benefits of focusing correctional resources on offenders who are at high risk for recidivism (Andrews & Bonta, 2010). Our finding that low-risk women had increased recidivism when they were sent to treatment for nondrug violations and when they were supervised intensely supports the RNR model; but it does appear that treatment responses to drug-related violations have a preventive effect on their recidivism. The crucial question is whether the cost of this prevention for the low-risk group is affordable, and this could only be determined by long-term study to see how many women who start out at low risk develop serious substance use and related crime problems over time.
A unique feature of the present research was its operationalization of the experiences of violations issued and responses to them. We examined drug- and non-drug-related technical violations separately, and we considered treatment and punitive/monitoring-oriented responses separately. Supporting the focus on drug- and non-drug-related violations, research in the District of Columbia found that a majority of violations are drug-related (Luallen, Astion, & Flygare, 2013). As noted above, drug-related violations are especially relevant to women offenders, who more often than men are involved in the justice system due to problems with alcohol or illegal drugs (Belknap, 2014; Greenfeld & Snell, 1999; Mumola & Karberg, 2006; Owen & Bloom, 1995). In addition to reflecting greater detail in types of and responses to violations, the operationalization we used seemed to have face validity, that is, women experienced a violation and a response nearly simultaneously, not as separate events. Future research that shows that this operationalization has explanatory value would increase our confidence in this approach to conceptualization and measurement.
Limitations
The research site and features of the study design raise caution about the strength of inferences drawn from the findings and suggest topics and methods for future research. By focusing on substance-involved women offenders, we accomplished the study aim of shedding light on this large group. The sampling approach provided an adequate number of substance-involved women for analysis, but it limits generalizability to this group. Specifically, women who are sex offenders, economic offenders with no substance involvement, and other less common types of women offenders are excluded. Research suggests that different subgroups of women offenders follow different pathways into criminality (Daly, 1992; Morash, 2010; Reisig et al., 2006; Salisbury & Van Voorhis, 2009), and thus may be differently affected by treatment and punishment responses. Thus, the findings are specific to this group and to the relatively treatment-oriented state where the study was conducted. Other researchers have found that punitive responses are concentrated on types of offenders absent from our study, specifically officially designated sex offenders (Grattet et al., 2011). Thus, other groups of women offenders and other settings need to be studied to establish generalizability of findings relating violations and responses to them during supervision to subsequent recidivism.
Some research on punitive responses to violations emphasizes the certainty and the immediacy of the response (e.g., Grommon, Cox, Davidson, & Bynum, 2013; Steiner, Markarios, Travis, & Meade, 2012). Swift and certain sanctions are related to avoiding criminal behavior at least during the period of monitoring (Grommon et al., 2013). We have not considered these dimensions of responses, but they do deserve further attention.
We could not separate nondrug violations into more specific types, because there were not enough in any category. Nor did we consider the potential moderating effects of degree of substance involvement on the connection of violations and responses to recidivism. It would be beneficial to know whether non-drug-related violations ever lead to interventions that could potentially reduce criminogenic needs and, if so, whether needs-oriented responses (e.g., requirements to attend employment enhancing programs) have a different relationship to recidivism than punitive responses. Likewise, it would be useful to know if violations and responses to them are differently related to recidivism for people with relatively minor drug use versus those with serious addiction. These types of examination would require a larger sample or a focus on just some very specific types of violations that occur frequently in particular settings.
A final limitation of the study is that we do not have information on whether women continued to be under supervision for their initial offense or for a new offense during the recidivism period we studied. Thus, the question we can answer is whether or not violations and related responses during an initial 18-month period predict subsequent recidivism for a 2-year period, regardless of supervision during the recidivism period. The effect of violations/responses and supervision after 18 months of supervision remains to be explored in future research.
Conclusion
Despite shortcomings of the present research, it provides insights into the nature of women’s technical violations and the connection of the violation process to recidivism. Some criminologists have raised questions about whether violations should be a primary focus of supervision (Duriez et al., 2014; MacKenzie et al., 1999). In fact, there is evidence of the importance of informal social controls that occur when offenders have strong bonds to family, school, employment, and prosocial intimate partners on recidivism (Committee on Community Supervision and Desistance From Crime, 2008; MacKenzie & Brame, 2001). Thus, it would be useful to know how alternative patterns of issuing and responding to violations influence informal social controls. Such information would inform assessment of the utility of detection and response to violations, and might even allow some fine-tuning of which violations and which responses are likely to reduce recidivism, perhaps through influence on informal social controls.
Turning back to the formal social controls provided by correctional supervision, the present study points to the need to compare supervision effects for low- and high-risk offenders, and to distinguish between different types of technical violations and the responses to them. Furthermore, the present research is relevant to the many jurisdictions where probation and parole agents heavily supervise some offenders, but where they combine these efforts with case plans that address the causes of illegal behavior and behavioral management strategies (Jalbert & Rhodes, 2012; Jalbert et al., 2010; Taxman, 2008). Finally, the findings underscore the need for future research to separate the intensity of supervision from the numbers of specific types of violations, and alternative responses to them.
Footnotes
Declaration of Conflicting Interests
The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: The data collection for this article was supported by the National Science Foundation under Grant No. 1126162 and a Strategic Partnership Grant from the Michigan State University Foundation.
