A Preliminary Evaluation of the Supervision With Immediate Enforcement Probation Program for Adult Gang

Abstract

As of 2012, it was estimated that there were more than 30,000 active gangs in the United States with at least 850,000 members. Despite significant challenges that criminal justice agencies and personnel face in treating and supervising gang members, few studies have examined adult gang member outcomes and the effects of community supervision on gang-affiliated offenders. Recent research demonstrates mixed evidence that high-risk offenders have better outcomes in smaller problem-solving courts and programs, which have dual emphasis on rehabilitation and deterrence-based approaches to corrections. This study evaluates the efficacy of the Supervision with Immediate Enforcement (SWIFT) Court Program for young adult gang–affiliated probationers compared with non-SWIFT gang members and high-risk non-gang offenders. Findings indicated SWIFT had a moderate deterrent impact on offending compared with alternative probation sanctions. Results and discussion related to problem-solving courts and policy-related issues surrounding gang-affiliated and youthful violent offenders are offered.

Keywords

gangs problem-solving courts probation SWIFT youthful offender evaluation

Introduction

Of the 6.8 million adults under correctional supervision in the U.S. criminal justice system (Glaze & Kaeble, 2014), nearly 4.7 million were on some form of community supervision, with more than 80% of them placed on probation (Herberman & Bonczar, 2014). Probation is a common alternative to incarceration and involves offenders remaining in the community, but under the supervision of the courts and enforcement of conditions deemed appropriate by the judge and/or other court actors (Herberman & Bonczar, 2014). Probation provides a cheaper, more rehabilitative approach to punishment than incarceration (Piquero, 2003), yet it fails to efficiently protect against environmental stimuli (e.g., criminal peers) that may mitigate involvement in criminal lifestyle (see Scott, 2004). Specialized caseloads and special conditions of probation seek to address shortcomings of traditional probation by instituting supplementary and individualized treatment.

One particularly challenging group under community supervision is adult gang members. Although increased interest and literature on youth gangs has been noted (Pyrooz & Mitchell, 2015), little research to date has examined the impact of probation on adult gang members (Matz, Stevens-Martin, & DeMichele, 2014). When looking across this sparse literature, studies from a few states focus on recidivism outcomes for juveniles released from prison (e.g., Huebner, Varano, & Bynum, 2007; Krienert & Fleisher, 2001; Spooner, Pyrooz, Webb, & Fox, 2017; Sweeten, Pyrooz, & Piquero, 2013; Trulson, Haerle, Caudill, & DeLisi, 2016). Research shows adult gang members are at greater risk of recidivism, both in frequency and for drug-related offenses, than their non-gang counterparts (Olson, Dooley, & Kane, 2004; Spooner et al., 2017). In addition, gang members were twice as likely as non-gang-affiliated offenders to be rearrested while on probation (Adams & Olson, 2002). Yet few studies have examined adult gang members under community supervision to determine what, if any, conditions or treatments are most effective for reducing recidivism risk (Matz et al., 2014). To address this gap in the literature, the present study evaluates a specialized program that focuses on young adult gang–affiliated probationers within an urban jurisdiction in Tarrant County, Texas.

Literature Review

The national incarceration rate for high-risk probationers for revocation due to a new crime or technical violation remains stable at 5.4%; this figure translates into more than 250,000 probationers being remanded into custody annually (Herberman & Bonczar, 2014). Probation is the leading sanction implemented in the adult correctional system, and incarceration due to reoffending and revocation remains a real possibility for many high-risk offenders. Within the past decade, new programs, such as problem-solving courts, have been initiated nationally that aggressively address offenders who violate conditions of supervision.

Problem-solving courts take an interdisciplinary approach characterized by judicial oversight, community education, resource allocation, and community resources, relative to the community’s needs (Winick, 2002). Participants in problem-solving courts generally receive individualized treatments relative to their specific needs and risk for recidivism. Individualized treatment such as counseling, substance use treatment, and various self-help groups are common provisions given to offenders served through problem-solving courts and can improve long-term outcomes (Rossman et al., 2011). Judicial oversight of cases in problem-solving courts demonstrates judicial interaction can facilitate reduction in recidivism risk (Gottfredson, Kearley, Najaka, & Rocha, 2007; Taylor, 2012). These findings highlight the necessity of individualized treatments and modes of delivery to best ensure participation, successful completion, and, ultimately, better long-term outcomes for offenders sanctioned by the criminal justice system.

Coupled with the theme of therapeutic jurisprudence in the problem-solving court model is the theory of deterrence, which postulates that certain, severe, and swift punishment deters criminal behavior (Nagin & Porgarsky, 2001). Any delay between the commission of an offense and its associated punishment is postulated to reduce the deterrent effect of the sanction. Furthermore, if the punishment for the crime is not severe enough to cause any discomfort or inconvenience to the actor, he will not be deterred from engaging in additional criminal acts.

Problem-Solving Courts and Intensive Probation

A prime example of the deterrence-based problem-solving court model is the Hawaii Opportunity Probation with Enforcement (HOPE) program, which was developed in 2004 to improve compliance with probation among drug-involved offenders by changing the probation-enforcement process. Founded on the basic tenets of deterrence theory, the core premise is that swift and certain punishment will deter crime, thereby resulting in better compliance and less technical revocations and reoffending for offenders under community supervision (Alm, 2016). The central tenets of the HOPE program rely heavily on the mandates of targeted, high-intensity, and specialized treatments for participants and immediate appearances before the judge when violations are detected to increase accountability and decrease deniability of a problem for the offender (Hawken & Kleiman, 2009). Under HOPE, any infraction to supervision results in immediate sanctions of short-term jail time (normally only a few days) for the violated offender.

Initial evaluations of the HOPE pilot program were positive, with Hawken and Kleiman (2009) reporting significantly fewer positive drug tests, failures to report, arrests for new charges, or technical revocations for HOPE offenders versus the comparison group. However, a number of rigorous evaluations recently published in a 2016 issue of Criminology & Public Policy offer a more dire assessment of the efficacy of the HOPE program and its deterrent-based model (see, for example, Cook, 2016; Hamilton, Campbell, van Wormer, Kigerl, & Posey, 2016; Kleiman, 2016; Lattimore, MacKenzie, Zajac, Dawes, Arsenault, & Tueller, 2016). A notable exception discussed within this host of commentaries and evaluations of HOPE-based programs across the country which have shown positive outcome efficacy is the Supervision with Immediate Enforcement (SWIFT) Court in Tarrant County (see Alm, 2016, and Kleiman, 2016), which identifies high-risk gang offenders for inclusion in the program. Limited, preliminary evaluation of SWIFT was largely positive in outcomes (Martin, 2013).

To date, few studies of specialized programs targeting high-risk offenders have focused on adult gang members. As of 2012, it was estimated that there were more than 30,000 active gangs in the United States and at least 850,000 members; 5-year trends indicate an 8% increase in the number of gangs and an 11% increase in the amount of gang members (Egley, Howell, & Harris, 2014). It is speculated these increases are due to growing concentrations of gangs in large cities. Street gangs have a proclivity for engaging in violent crimes and often report lengthy criminal records (Matz et al., 2014). Research estimates that gangs are responsible for approximately 16% of all national homicides (Egley et al., 2014).

Despite the noted increase in gangs and gang members, very little research has been devoted to understanding adult gang members (Watkins & Moule, 2014), let alone evaluating the effectiveness of special conditions of probation for this population. What is known about gang members suggests they are at a higher risk of recidivism than non-gang member offenders (Spooner et al., 2017); in fact, gang membership is often listed as a significant risk factor for recidivism for all types of criminal offenders (Braga, Piehl, & Hureau, 2009). One of a few studies available on adult gang member recidivism found that gang members were not only more likely to be rearrested but also to be rearrested more quickly, more frequently, and for more violent or drug-related offenses than non-gang members (Olson et al., 2004). Reentering criminals may return or turn to street gangs as a source of economic and social stability, which in turn increases their risk of reoffending or rearrest (Scott, 2004). Adult gang research emphasizes environmental risk factors, such as returning to criminogenic or disorganized communities, as pressing concerns and where research efforts should focus (Olson et al., 2004). Cognitive-behavioral therapy (CBT) treatment centered around risk, needs, and responsivity principles and delivered via institutional mental health staff has been found to reduce the likelihood and severity of reoffending for adult gang members (Di Placido, Simon, Witte, Gu, & Wong, 2006) and in other high-risk offender populations (Barnes, Hyatt, & Sherman, 2017).

However, the exact mechanisms leading to recidivism, or how to best ensure reintegration, are unknown (Matz et al., 2014; Spooner et al., 2017). Adolescent gang research such as those examining the Gang Resistance Education and Training (GREAT) project finds that youth need teaching or reinforcement of cognitive-behavioral skills to be dissuaded from gang involvement (Esbenson, Osgood, Taylor, Peterson, & Freng, 2001). Maxson, Matsuda, and Hannigan (2011) suggested that youth gang members may actually be less subject to deterrence than non-gang youth, with gang members reporting less certainty of arrest and lower severity of potential punishment perceptions. A number of respected studies have found that gang-affiliated offenders have unique needs and risk factors that put them at greater likelihood of committing more serious offenses, drug sales and usage, and increased prevalence of general offending behaviors compared with non-gang youthful offenders (e.g., Battin, Hill, Abbott, Catalano, & Hawkins, 1998; Dong & Krohn, 2016; Thornberry, Krohn, Lizotte, Smith, & Tobin, 2003). Furthermore, Trulson and colleagues (2016) examined more than 3,000 youth sentenced in the state of Texas and determined that 59% were living in poverty, 71% reported a chaotic home life, and nearly a quarter of the youth reported some form of abuse or neglect. As young adult gang members tend to reoffend at especially high rates, commit more crimes, and demonstrate crime involvement that peaks during active gang participation, they are an especially challenging population to manage on community supervision (see, for example, Pyrooz, Turanovic, Decker, & Wu, 2016). Accordingly, some scholars argue that problem-solving court programs, such as the subset of HOPE programs like the SWIFT Court, offer the best opportunity to successfully rehabilitate gang-affiliated young adults back (Stevens-Martin, 2014).

The Present Study

Toward this end, the present study contributes to the literature on problem-solving courts targeting high-risk offenders by evaluating the SWIFT Court program on young adult gang–affiliated probationers. As described in detail below, the SWIFT Court is a specialized court program based on the HOPE program that targets felony offenders who are assessed at high or medium risk of reoffending, including gang members. To be clear, the SWIFT Court was not developed as a specific response for gang members in the jurisdiction. SWIFT was developed for high-risk offenders placed on community supervision who would benefit from more intensive supervision and a judicially coordinated response to violations of community supervision. Gang members represent a subset of the probation offender population supervised in this court.

Launched in 2011, the SWIFT Court in Tarrant County, Texas, is a relatively new program requiring lengthy involvement in supervision, which prevents evaluation of long-term outcomes such as recidivism after release from community supervision; however, the program can be evaluated on other important criminal justice factors. To determine the relative effectiveness in reducing both revocations and reoffending while on supervision, a quasi-experimental design was used to compare high-risk gang-affiliated probationers in the SWIFT program with two other probation groups: (a) gang-affiliated, non-SWIFT young adult offenders, who were similar to SWIFT in their gang involvement but sanctioned to regular probation, and (b) high-risk, non-gang, non-SWIFT-affiliated offenders, who were not gang-affiliated but received similar conditions of probation as SWIFT due to their high-risk status. Although this study does not offer a test of theory, null hypothesis significance testing was used to examine operationalized differences (see Kluger & Tikochinsky, 2001) between SWIFT and the comparison groups, which can inform similar populations and serve criminal justice practice. Specifically, the following hypotheses were tested:

Hypothesis 1: SWIFT offenders will be more likely to remain on active probation status and less likely to have their probation revoked than offenders in the comparison groups.

Hypothesis 2: SWIFT offenders will have fewer technical violations while on probation than offenders in the comparison groups.

Hypothesis 3: SWIFT offenders will have less in-program recidivism while on supervision than offenders in the comparison groups.

Method

Participants and Procedures

The data pertain to caseload information for the population of offenders on probation (N = 232) supervised on the specialized Gang/High Risk Offender (G/HR) caseloads, including a comparison group of high-risk, non-gang offenders, under the jurisdiction of the Tarrant County Community Supervision and Corrections Department (CSCD) between January 1, 2011, and December 31, 2013. The data were provided directly from CSCD in de-identified electronic format. All study procedures were approved by the university institutional review board (IRB).

The primary focus of this study was on gang-affiliated participants in the SWIFT Court program. The SWIFT Court program incorporates many drug court model tenets, including collaboration, risk and needs assessment, judicial interaction, monitoring and supervision, graduated sanctions and incentives, and treatment. In addition, the program utilizes swift and certain sanctions for every violation in combination with other elements to reduce violations of probation, increase successful completions of probation, and reduce recidivism. These sanctions are clearly articulated and conveyed to offenders and are expanded beyond the sanction of short-term jail time used in the HOPE approach to include writing assignments, additional community service, increased fines, electronic monitoring, substance abuse and other residential programs, and presentation in court.

When offenders are violated for a condition of their probation, a court hearing is scheduled within 24 hr. While each offender faces an immediate sanction of some type, including a minimum of 2 days in jail, offenders commonly are simultaneously referred to a host of therapeutic treatments. These interventions include inpatient and outpatient referrals to substance abuse interventions, community corrections facilities, parenting courses, and theft intervention programming. As a result of these extended therapeutic treatment plans, SWIFT offenders may be in program for several years. The goal of SWIFT is to increase accountability while also supporting offenders’ unique needs to help them be successful in the program.

Eligibility for the SWIFT specialty court requires offenders to (a) live in Tarrant County; (b) commit a first, second, or third degree felony; (c) be assessed as medium or high risk; (d) have at least 12 months remaining on supervision and at least 30 days of jail time remaining for the court to use as sanctions for violations; and (e) not be sex offenders or mentally impaired. An offender is considered at the end of the road, with a history of noncompliance, and facing prison before SWIFT is recommended and accepted by the court. Sentencing to SWIFT occurs only after all other types of intermediate sanctions have been tried and failed.

Five probation officers supervise 60 offenders each for a total capacity of 300. SWIFT supervision consists of three monthly face-to-face contacts, one monthly collateral contact with treatment provider(s), and six random monthly drug tests for 2 months and one random drug test for one subsequent month. Failures on drug tests, stalls or failures to show for drug testing, or failure to call the daily drug test line typically results in increased monitoring to six random tests per month. At the time of data collection, 72 young adult gang offenders participated in SWIFT and were included in the present study.

The two probation comparison groups reflect young adult offenders who were either (a) gang-affiliated, but not assigned to SWIFT or (b) high-risk, non-gang-affiliated young adult offenders. For the gang-affiliated, non-SWIFT group (n = 108), offenders were supervised on G/HRO caseloads with CSCD. Offenders who met the following criteria were eligible for G/HRO assignment: (a) between 17 and 25 years old; (b) identified as an active gang member by self-report, police reports, official gang identification system, and/or tattoos; (c) placed on felony supervision (case-by-case for misdemeanors) for at least 24 months or had 24 months remaining on supervision; and (d) had other risk factors for traditional probation, such as unemployment, no high school education, gang associations, instability, and current violations of probation.

Both SWIFT gang members and non-SWIFT gang members were assigned “gang member” designations via an official designation through a law enforcement–related classification system accessible to Tarrant County probation officers called GangNet. GangNET is governed by state and federal regulations and has web-based capabilities that allow certified users to enter officer comments, field notes, and/or view information on gang suspects and members who have been validated using standardized criteria. Within the state of Texas, to classify an individual as a gang member, the individual’s record must meet at least two of the seven criteria (e.g., self-admission, identification by reliable source, evidence of association with gang members, evidence of using gang dress, signals, tattoos, and/or symbols) of Texas Code of Criminal Procedure Section 61.02(c)(2), Compilation of Information Pertaining to Criminal Combinations and Criminal Street Gangs (2015). Although beyond the purview of the present study, there is criticism that identification of gang affiliation from official data and law enforcement may be sensitive to local politics, which in turn may exaggerate or underreport gang problems within a jurisdiction (see Curry, 2015; National Youth Gang Center, 2006).

The second comparison group represents young adult offenders on probation with no known gang affiliation, yet who were also considered high risk for recidivism. This comparison group is important to explore differences that may be attributed to treatment simply due to gang affiliation. This third group contained special caseloads of probation not under G/HRO supervision or high-risk young adult offenders who were not in gangs or the SWIFT Court program (n = 52). As the offenders were high risk, their conditions of probation shared many similarities with the SWIFT and gang-affiliated, non-SWIFT groups. To be clear, both comparison group members were classified as high risk and received intensive supervision but did not meet the restrictive criteria required to be ordered into the SWIFT Court. The primary difference for the high-risk, non-gang group was that they did not receive gang-related conditions of probation; however, they are similar on other risk factors.

In contrast to SWIFT gang offenders, clients supervised in both the non-SWIFT gang and high-risk specialized caseload groups received two monthly face-to-face contacts with a probation officer; one monthly collateral contact with a family, friend, or employer member; and one monthly collateral contact with a treatment provider to ensure compliance and forward progress. The length of time an offender could be on this specialized caseload ranged from a minimum of 12 to 24 months. During Phase I of the specialized caseload, offenders observed curfew, were placed in a CBT class, called in daily to a drug testing line, submitted random drug tests, and took classes for high school degree certification. Unemployed offenders reported weekly to their supervising officer; they also completed 20 hr of community service restitution and submitted 10 employment applications per week. During Phase II, high-risk probationers had later curfew, reported twice a month to their probation officer, and continued CBT. In the final Phase III, conditions were a further reduced curfew, bimonthly random drug tests, consistent employment search, and continued court-ordered class participation.

Measures

Group Membership

The key independent variable of interest was group membership. Categorical (dummy) variables were created to measure membership in the groups.

Sociodemographic, Mental Health, and Substance Abuse Characteristics

Several sociodemographic measures were included in the study. Gender was a dichotomous variable (0 = female; 1 = male). Age at time of placement on supervision was measures in number of years. Race was a dichotomous variable, with 0 = White and 1 = non-White (primarily Black/African American). Ethnicity was also a dichotomous variable measure (0 = non-Hispanic, 1 = Hispanic/Latino). Marital status was a categorical variable, with 1 = married, 2 = divorced, and 3 = single. Highest education was a continuous indicator measuring the number of years of education. Employment at time of placement under supervision was an ordinal variable, where 1 = unemployed; 2 = student, disabled, retired, or homemaker; 3 = employed part-time; and 4 = employed full-time. A dichotomous variable for employment was created for subsequent analyses, where 1 = unemployed and 0 = all else. Mental health and substance abuse were dichotomous variables reflecting whether the offender had a known mental health issue or substance abuse issue at the time of placement under supervision.

Prior Criminal History

Data were collected on official adult (aged 17 or older) criminal records for the population. Prior criminal record was a dichotomous measure of whether there was an official record of adult criminal involvement (0 = no, 1 = yes) prior to the supervision offense. Age at first arrest was measured in years. Continuous indicators for the number of prior arrests (felony and misdemeanor combined), prior felony arrests, prior misdemeanors, prior periods of supervision (i.e., probation), prior supervision revocations, prior county jail sentences, and prior state jail/prison sentences were included. Four of these indicators, prior felony arrests, periods of supervision, supervision revocations, and county jail sentences, were transformed using the natural logarithm (plus 1 prior to log transformation) to adjust for skewness.

Supervision Variables

Measures on probation supervision characteristics were included in the study. Dichotomous indicators of offense category measured the most serious offense for which the offender was placed on probation, as (a) drug-defined offense; (b) theft, fraud, or property offense; (c) other offense; or (d) violent offense, with the reference category depending on the analysis. Length of supervision on probation was measured in number of years. Probation status was a dichotomous indicator reflecting whether the offender was actively (1) on probation or not (0) at the time of data collection. Termination reason was a categorical variable measuring the reason the offender was no longer actively on supervision, where 0 = active, 1 = term expired, 2 = early discharge, and 3 = probation revoked. Revocation reason was also a categorical variable that measured the reason for the probation revocation, where applicable, with 0 = still active on probation, 1 = new offense arrest or conviction, 2 = technical violation, and 3 = not revoked, completed probation. In subsequent analyses, the revocation reason was recoded as 0 = still active or completed program (expiration of supervision or early discharge) and 1 = revoked due to new arrest charge, conviction, or technical violation.

Technical Violations for Probation

Technical violations represent a form of antisocial behavior or recidivism among the offending population and are often used as justification for increased sanctions and increased severity in sanctions as a deterrent for future noncompliant, antisocial behavior (Arrigona & Bryan, 1999; McRee & Drapela, 2012; Texas Department of Criminal Justice, 2014). CSCD classifies technical violations into high-, medium-, and low-severity violations. The technical violations were captured with summary scores for each severity level, adjusted for time in years on probation. High technical violations included eight violations for: contact with the victim; carry or possess firearm or other weapon; failure to comply with order to avoid prohibited locations; tampering with electronic monitoring device; unsuccessful discharge from treatment; failure to report to jail; search violation; and other high-level violations (e.g., sex offender conditions). Medium technical violations included four violations for: failure to follow curfew; positive blood alcohol content; no shows to court; and other medium-level violations. Low technical violations included seven violation items for: failure to report as directed; failure to pay court-ordered fees; absence from treatment, counseling, or education; leaving county without permission; failure to maintain employment; failure to perform community service; and other low-level violations. As drug use and gang activity represent major concerns affecting decision making for probation revocation and sanctions, these violations were included as separate measures in subsequent analyses, adjusted for time in years on probation. Drug-related technical violations included positive drug test, failure to submit drug test, and diluted urine tests. Gang-related technical violations included failure to comply with order not to associate with known gang members and avoid places where such persons congregate and failure to comply with order not to exhibit gang signals, symbols, or attire. Finally, the rate of total violations was created by dividing the total number of all technical violations by the number of years the offender was on probation. The rate of technical violation is commonly used by probation staff in decision making regarding probation status. Due to high skewness, the technical violation variables were transformed using the natural logarithm.

In addition, two indicators of sanctions administered as a consequence of technical violations were included as control variables in subsequent analyses. First, a measure was created for the number of progressive sanctions of jail used to address technical violations. This progressive sanction meant the offender on probation was sent to serve a short time in jail, usually 3 to 4 days, as a consequence of his or her technical violation(s), but did not have probation revoked.¹ It is important to analyze the number of progressive sanctions with jail as a condition of probation because the SWIFT program and the Hawaii HOPE models were founded on the principle that short stints in jail serve to deter crime. Second, a measure of the number of months sent to prison for revocation was included in subsequent analyses. Both of these indicators were skewed and were, therefore, transformed using the natural logarithm.

New Arrests and Arrest Charges While on Supervision

Another form of recidivism is new official arrests and charges received while on supervision. Often, this form of recidivism is referred to as “in-program” recidivism. In-program recidivism has a direct and indirect impact on success. In many jurisdictions, a new arrest or charge while on supervision can result in immediate revocation of probation and failure from the program. As such, in-program recidivism may be especially important to practitioners who are concerned with program completion as an indication of success. It is also of particular concern to the community, as probationers who continue to commit crimes while on supervision threaten public safety. Moreover, offenders may consequently serve additional jail/prison time or suffer increased sanctions. Recidivism during the in-program phase may also predict future recidivism (de Beus, & Rodriguez, 2007).

In the present study, new arrests and charges while on supervision were examined.

For any single arrest, multiple charges may be filed on an offender. Moreover, offenders may be charged with offenses without being arrested. Using official data obtained from the Texas Department of Public Safety, indicators of any new arrest and any new arrest charge during the probation period were created. These indicators were dichotomized to better permit the analyses to focus on the effects of probation group membership on the odds of in-program recidivism.

Results

Descriptive Statistics and Comparison of Sociodemographic and Offense History

Table 1 reports the descriptive statistics for the measures included in subsequent analyses comparing SWIFT with the gang and high-risk, non-gang groups, respectively. Almost all the offenders were male, single, had no history of mental health issues, and had a history of substance abuse issues. As these variables had little variation, they were excluded from subsequent statistical analyses. For subsequent regression analyses, diagnostics did not indicate problems with multicollinearity (variance inflation factor maximum = 3.04; minimum tolerance = 0.33; Fox, 1991).

Table 1:

Descriptive Statistics

Variable	SWIFT (n = 72)	Gang (n = 108)	Non-gang (n = 52)	Comparisons
Variable	SWIFT (n = 72)	Gang (n = 108)	Non-gang (n = 52)	SWIFT vs. gang	SWIFT vs. non-gang
Gender				NA	NA
Female	4.17%	0.93%	3.85%
Male	95.83%	99.07%	96.15%
Marital status:				NA	NA
Married	0.00%	2.78%	1.92%
Divorced	0.00%	1.85%	0.00%
Single	100.00%	95.37%	98.08%
Age at placement	19.82 (1.89)	19.72 (1.92)	19.40 (1.84)	t(178) = 0.34	t(122) = 1.22
Education at placement	10.22 (1.79)	10.07 (1.63)	9.83 (2.74)	t(178) = 0.57	t(122) = 0.97
Race (1 = non-White)	0.53 (0.50)	0.48 (0.50)	0.44 (0.50)	φ = 0.05	φ = 0.08
Ethnicity (1 = Hispanic)	0.44 (0.50)	0.51 (0.50)	0.53 (0.50)	φ = −0.06	φ = −0.08
Employment at placement				χ²(3) = 4.08	χ²(3) = 0.64
Unemployed	70.83%	80.56%	65.38%
Student, disabled, retired, homemaker	9.72%	4.63%	9.62%
Employed part-time	6.95%	8.33%	7.69%
Employed full-time	12.50%	6.48%	17.31%
Mental health issue (1 = yes)	0.07 (0.26)	0.11 (0.32)	0.08 (0.27)	φ = −0.07	φ = −0.01
Substance abuse issue (1 = yes)	0.99 (0.12)	0.95 (0.21)	0.96 (0.19)	φ = 0.09	φ = 0.08
Prior criminal record (1 = yes)	0.90 (0.30)	0.92 (0.28)	0.83 (0.38)	φ = −0.02	φ = 0.11
Age at first arrest	16.28 (2.29)	15.75 (2.72)	15.96 (2.49)	t(178) = 1.36	t(122) = 0.73
Prior arrests	3.56 (3.13)	3.56 (2.86)	3.48 (3.15)	t(178) = −0.02	t(122) = 0.13
Prior felony arrests (ln)	0.44 (0.61)	0.46 (0.56)	0.54 (0.61)	t(178) = −0.25	t(122) = −0.94
Prior misdemeanor arrests	2.62 (2.38)	2.69 (2.20)	2.38 (2.43)	t(178) = −0.17	t(122) = 0.55
Prior periods of supervision (ln)	0.91 (0.66)	0.98 (0.64)	0.93 (0.67)	t(178) = −0.64	t(122) = −0.15
Prior revocations (ln)	0.12 (0.30)	0.10 (0.29)	0.19 (0.39)	t(178) = 0.53	t(122) = −1.17
Prior county jail (ln)	0.44 (0.54)	0.37 (0.57)	0.44 (0.60)	t(178) = 0.84	t(122) = 0.00
Prior state jail/prison	0.07 (0.26)	0.17 (0.52)	0.12 (0.38)	t(178) = −1.47	t(122) = −0.81
Violent offense (0/1)	0.25 (0.44)	0.19 (0.40)	0.19 (0.40)	φ = 0.07	φ = 0.07
Drug offense (0/1)	0.15 (0.36)	0.20 (0.41)	0.17 (0.38)	φ = −0.06	φ = −0.03
Property offense (0/1)	0.42 (0.50)	0.32 (0.47)	0.46 (0.50)	φ = 0.09	φ = −0.05
Other offense (0/1)	0.18 (0.39)	0.28 (0.45)	0.17 (0.38)	φ = −0.11	φ = 0.01
Length of supervision (years)	4.94 (2.16)	4.43 (2.69)	4.55 (2.44)	t(178) = 1.35	t(122) = 0.96
Probation status (1 = active)	0.57 (0.50)	0.24 (0.43)	0.29 (0.46)	φ = 0.33**	φ = 0.28**
Termination reason				NA	NA
Active	56.94%	24.07%	28.85%
Expiration of supervision	0.00%	8.33%	9.62%
Early discharge	0.00%	0.93%	0.00%
Revocation	43.06%	66.7%	61.54%
Revocation reason				χ²(3) = 23.91**	NA
Not revoked, still active	56.94%	24.07%	28.85%
New offense arrest/conviction	25.00%	33.33%	30.77%
Technical violation	18.06%	33.33%	30.76%
Successful completion	0.00%	9.27%	9.62%
Revoked new offense or technical violation (0/1)	0.43 (0.50)	0.67 (0.47)	0.62 (0.49)	φ = −0.23**	φ = −0.18*
Rate of total technical violations (ln)	3.04 (0.78)	3.28 (0.73)	3.10 (1.09)	t(178) = −2.11*	t(122) = −0.37
Progressive sanction of jail (ln)	3.22 (1.14)	1.15 (1.43)	0.83 (1.27)	t(178) = 10.28**	t(122) = 11.01**
Months to prison for revocation (ln)	1.44 (1.86)	1.92 (1.68)	1.84 (1.90)	t(178) = −1.77	t(122) = −1.18
High technical violations (ln)	0.72 (0.66)	0.80 (0.70)	0.75 (0.71)	t(178) = −0.82	t(122) = −0.30
Medium technical violations (ln)	0.96 (0.70)	0.34 (0.65)	0.23 (0.44)	t(178) = 6.08**	t(122) = 6.59**
Low technical violations (ln)	2.72 (0.74)	3.09 (0.71)	2.97 (1.07)	t(178) = −3.41**	t(122) = −1.54
Drug-related technical violations (ln)	1.06 (0.76)	0.72 (0.80)	0.59 (0.73)	t(178) = 2.89**	t(122) = 3.46**
Gang-related technical violations (ln)	0.48 (0.56)	0.45 (0.57)	0.19 (0.41)	t(178) = 0.32	t(122) = 3.15**
In-program arrest charges (ln)	0.28 (0.49)	0.38 (0.48)	0.49 (0.52)	t(178) = −1.40	t(122) = −2.40
In-program arrest (1 = yes)	0.32 (0.47)	0.43 (0.50)	0.52 (0.51)	φ = −0.11	φ = −0.20*
In-program arrest charge (1 = yes)	0.31 (0.46)	0.43 (0.50)	0.52 (0.51)	φ = −0.12	φ = −0.22*

Note. Standard deviations or degrees of freedom in parentheses. SWIFT = Supervision with Immediate Enforcement.

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

Comparisons of other sociodemographic and criminal history characteristics for SWIFT versus each group were nonsignificant. On average, probationers were 19.68 (SD = 1.89) years old and had a 10th grade (M = 10.06, SD = 1.97) education. Approximately half of the probationers were persons of color (non-White) and/or Hispanic. More than half of the probationers in each group were unemployed at time of placement on probation. Overall, the probationers were 15.96 (SD = 2.54) years old at the time of their first official arrest and had 3.54 (SD = 3.00) prior arrests on their record at time of placement on probation. On average, the probationers had been placed on supervision before 0.95 (SD = 0.65) times and been sentenced to jail 0.41 (SD = 0.57) times before the current probation period. Few of the probationers served a prison sentence or had been previously revoked on probation. These results suggest the groups are similar in their sociodemographic and criminal history characteristics.

Comparison of Supervision Variables

Bivariate group comparisons (see Table 1) for the supervision variables found no differences across groups in the most serious offense category for which the offender was placed on probation or the length of supervision. The average length of supervision at the time of data collection was 4.62 (SD = 2.48) years, which reflects the high-risk status and long-term nature of the supervision requirements of each probationer. There were, however, significant differences in probation status, revocation of probation status, and reason for revocation or termination of probation when comparing SWIFT with the other two groups. SWIFT had significantly more participants still in active probation status at the time of the study, compared with the gang (φ = 0.33, p < .001) and non-gang groups (φ = 0.28, p = .002). A significant difference existed in the probation revocation reason when comparing SWIFT with the gang group, χ²(3, 180) = 23.91, p < .001. Furthermore, there were significant differences in the use of jail time as a sanction while on supervision—SWIFT vs. gang: t(178) = 10.28, p < .001; SWIFT vs. non-gang: t(172) = 11.01, p < .001—with SWIFT participants having significantly higher rates of jail time during supervision than the non-SWIFT group. This finding is consistent with sanctioning initiatives of the SWIFT program.

Post hoc power analyses were conducted to examine the probability of making an error (1 – β) in detecting an effect for the research hypotheses. Power analyses were conducted using the G*Power software (Faul, Erdfelder, Lang, & Buchner, 2007). For all models, a significance level of α = .05 was specified so that there was a small chance of erroneously rejecting the null hypothesis (Type I error) of no group difference in the proportion or means of the variables of interest. Consequently, power (Type II error) was weakened unless the effect size or ability of the model to explain variance in the outcome variable was moderate.

For probation status, revocation status, and progressive sanction of jail, the power probabilities for detecting differences between SWIFT and the gang group were greater than .96. The power probabilities for detecting a difference between SWIFT and the high-risk, non-gang group was greater than .94 for probation status and progressive sanctions of jail time, but only .64 for revocation of probation. These findings suggest caution should be used when making inferences about the differences between SWIFT and the non-gang group members regarding revocation.

Logistic regression analyses examined the odds of not being on active probation status for SWIFT versus each comparison group. First, regressions were conducted for the combined population to examine the group effect overall. Then, group-based models were estimated to explore differences across groups. As power was a concern, forward stepwise regression was used to pare down the models, retaining only significant and key variables across the models. Regression coefficients were compared across group-based models using Paternoster, Brame, Mazerolle, and Piquero’s (1998) test of equality of regression coefficients, and significant coefficient differences (p < .05) are noted by shared superscripts in the tables of results.

As shown in Model 1 of Table 2, SWIFT offenders (reference) were more likely to remain on active probation status than gang and high-risk, non-gang members, controlling for all else.² The odds of having inactive probation were 3.78 times greater for gang members compared with SWIFT and 3.21 times greater for high-risk, non-gang members compared with SWIFT. The group-based models (Models 2-4) allow for the exploration of differences within and across groups. For SWIFT, a one-unit increase in the rate of high-level technical violations increased the odds of inactive probation status by 13.21 times. For the gang group members, high-level technical violations was the strongest predictor (odds ratio [OR] = 38.39) for inactive probation status, but education and low-level technical violations were also significantly and positively related to the odds of inactive probation status and length of supervision, drug-related technical violations, and gang-related technical violations were significantly and negatively related to inactive probation status. For the high-risk, non-gang group members, none of the variables in the model were related to the odds of inactive probation status. Moreover, none of the significant coefficients in the group-based models were significant for tests of coefficient equality. This suggests that the impact of high-level technical violations is relatively equal across the groups. Power analyses of the logistic regressions in Table 2 indicated high power probabilities of greater than .90, using the strongest predictors in each model to assess power.

Table 2:

Logistic Regression of Inactive Probation Status

Variable	Combined (N = 232)		SWIFT (n = 72)		Gang (n = 108)		Non-gang (n = 52)
	Model 1		Model 2		Model 3		Model 4
	b (SE)	OR	b (SE)	OR	b (SE)	OR	b (SE)	OR
Gang group	1.33 (0.42)**	3.78
Non-gang group	1.17 (0.52)*	3.21
Education	0.18 (0.09)*	1.19	0.12 (0.17)	1.14	0.42 (0.19)*	1.53	0.26 (0.16)	1.30
Length of supervision (years)	−0.13 (0.07)	0.88	0.12 (0.15)	1.13^a,b	−0.29 (0.12)*	0.75^a	−0.29 (0.17)	0.75^b
High technical violations (ln)	1.98 (0.48)**	7.24	2.28 (0.86)**	13.21	3.65 (1.11)**	38.39	−0.54 (0.96)	0.58
Low technical violations (ln)	0.72 (0.28)*	2.05	0.62 (0.72)	1.86	1.76 (0.60)**	5.79	0.46 (0.45)	1.59
Drug-related technical violations (ln)	−0.49 (0.28)	0.62	−0.18 (0.58)	0.84	−1.25 (0.49)*	0.29	0.41 (0.62)	1.50
Gang-related technical violations (ln)	−0.52 (0.48)	0.59	−0.30 (0.80)	0.74	−2.54 (1.10)*	0.08	3.20 (2.16)	24.60
In-program arrest	0.94 (0.43)*	2.57	1.42 (0.78)	4.15	1.47 (0.84)	4.36	1.96 (1.12)	7.08
Intercept	−4.36 (1.29)**		−5.84 (2.68)*		−7.23 (2.66)**		−2.66 (2.49)
−2 Log likelihood	204.53		62.88		67.11		46.78
Model χ²/df	96.86/9**		35.54/7**		52.11/7**		15.70/7*
Pseudo-R²	.34		.39		.38		.26

Note. Reference category for gang and non-gang groups is SWIFT group. Pseudo-R² used is Cox and Snell. Superscripts indicate coefficient pairs with significant z scores (p < .05) for test of equality of regression coefficients. SWIFT = Supervision with Immediate Enforcement; OR = odds ratio.

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

Although examination of inactive probation status is informative, it provides a limited understanding of probation status. As noted in Table 1, a sizable portion (56.94%) of SWIFT was actively involved in probation, but only roughly one-quarter of the gang (24.07%) and high-risk, non-gang (28.85%) groups were actively involved in probation. Approximately 10% of the gang and non-gang groups were not actively involved in probation because they were released for either early discharge (i.e., successful completion) or expiration of supervision (i.e., successful or timed out), yet none of the SWIFT members had been released for early discharge or expiration of supervision. Therefore, comparisons of successful completion of probation could not be examined for SWIFT versus the comparison groups. However, the data did permit exploration of revocation status across the groups. As shown in Table 1, over half of the gang and non-gang group members had their probation revoked due to a new offense/conviction or technical violation, whereas 43.06% of SWIFT members had been revoked for these same reasons.

Logistic regression analyses were conducted to examine the odds of having probation revoked for SWIFT versus the comparison groups, controlling for sociodemographics, prior criminal involvement, and type of technical violations.³ As shown in Model 1 of Table 3, SWIFT offenders were less likely to be revoked from probation than gang and high-risk, non-gang members, controlling for all else. The odds of revocation were 2.65 times greater for gang members compared with SWIFT and 2.90 times greater for high-risk, non-gang members compared with SWIFT. Those who received more high-level technical violations and low-level technical violations had higher odds of being revoked, whereas those who received more medium-level technical violations had lower odds of being revoked.

Table 3:

Logistic Regression of Revoked Probation

Variable	Combined (N = 232)		SWIFT (n = 72)		Gang (n = 108)		Non-gang (n = 52)
	Model 1		Model 2		Model 3		Model 4
	b (SE)	OR	b (SE)	OR	b (SE)	OR	b (SE)	OR
Gang group	0.98 (0.46)*	2.65
Non-gang group	1.06 (0.54)*	2.90
Age	−0.02 (0.10)	0.98	−0.13 (0.19)	0.88^a	0.01 (0.16)	1.01^a,c	−0.04 (0.20)	0.96^c
Education	0.12 (0.08)	1.13	0.08 (0.17)	1.09	0.36 (0.18)*	1.43^c	0.03 (0.13)	1.03^c
Unemployed	0.24 (0.41)	1.28	1.27 (0.83)	3.55	0.07 (0.73)	1.07	0.34 (0.80)	1.41
Prior adult arrests	−0.05 (0.06)	0.95	−0.10 (0.14)	0.91^b	−0.16 (0.11)	0.85^c	0.11 (0.13)	1.12^b,c
High technical violations (ln)	2.24 (0.39)**	9.40	3.37 (0.91)**	29.17	2.74 (0.70)**	15.54	1.45 (0.71)*	4.27
Medium technical violations (ln)	−0.17 (0.32)**	0.84	0.12 (0.63)	1.12	−0.43 (0.48)	0.65	−0.05 (0.78)	0.95
Low related technical violations (ln)	0.93 (0.28)**	2.53	−0.16 (0.62)	0.95	1.93 (0.52)**	6.89	0.40 (0.46)	1.49
Intercept	−5.16 (2.35)*		−1.35 (3.99)		−9.72 (4.59)*		−1.64 (4.70)
−2 Log likelihood	204.44		61.74		75.44		50.71
Model χ²/df	111.57/9**		36.07/7**		62.05/7**		18.59/7*
Pseudo-R²	.38		.39		.44		.30

Note. Reference category for gang and non-gang groups is SWIFT group. Reference category for unemployed is student, disabled, homemaker, retired, part-time employed, or full-time employed. Pseudo-R² used is Cox and Snell. Superscripts indicate coefficient pairs with significant z scores (p < .05) for test of equality of regression coefficients. SWIFT = Supervision with Immediate Enforcement; OR = odds ratio.

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

The group-based models (Models 2-4 in Table 3) illustrated some differences within and across groups. For SWIFT, a one-unit increase in the rate of high-level technical violations increased the odds of revoked probation by 29.17 times. Neither medium- nor low-level technical violations affected revocation status for SWIFT members, controlling for other variables in the model. For the gang group members, high-level technical violations was the strongest predictor (OR = 15.54) for revoked probation status, and education and low-level technical violations were also significantly and positively related to the odds of revocation. For the high-risk, non-gang group members, high-level technical violations (OR = 4.27) significantly increased the odds of revocation. None of the significant coefficients in the group-based models were significant for tests of coefficient equality. This suggests the impact of high-level technical violations on revocation status, similar to probation status, was relatively equal across the groups. Power analyses of the logistic regressions in Table 3 indicated high power probabilities of greater than .95 for Models 1 through 3 but lower power for the non-gang group analyses in Model 4 (power = .84), using the strongest predictors in each model to assess power.⁴

Comparison of Technical Violations

As shown in Table 1, on average, SWIFT probationers had fewer total, high-severity, and low-severity technical violations than probationers in gang and non-gang groups, but more medium-severity, drug-related, and gang-related technical violations. There were significant bivariate differences in rates of technical violations across groups. Specifically, the means for total rate of technical violations were significantly lower for SWIFT compared with the mean for gang members, but the means were not significantly different comparing SWIFT with high-risk, non-gang members. The means were not significantly different for high-level technical violations across the two pairs of groups (SWIFT vs. gang and SWIFT vs. non-gang). For medium-level, drug-related, and gang-related technical violations, the means for SWIFT were significantly higher than the means of one or both of the comparison groups. For low-level technical violations, the mean for SWIFT was significantly lower than that for gang group members. Power analyses of the bivariate effects revealed strong power (>.90) for medium-level technical violations for both group comparison pairs, low-level technical violations for SWIFT compared with the gang group, and drug-related and gang-related technical violations for SWIFT compared with the non-gang group. The power for drug-related technical violations (power = .89) and total technical violations (power = .67) for SWIFT compared with the gang group was moderate to weak, and the power for the remaining comparisons for the technical violations variables was weak (<.60).

Ordinary least squares (OLS) regression analyses were conducted on the rate of total technical violations, controlling for sociodemographics, probation characteristics, prior offending, length of supervision, and group membership.⁵ As reported in Model 1 of Table 4, SWIFT members had significantly fewer total technical violations while on probation than gang group members. There was no significant difference between high-risk, non-gang offenders and SWIFT offenders in predicting the rate of technical violations, controlling for all else.

Table 4:

OLS Regression of Rate of Total Technical Violations

Variable	Combined (N = 232)		SWIFT (n = 72)		Gang (n = 108)		Non-gang (n = 52)
	Model 1		Model 2		Model 3		Model 4
	b (SE)	β	b (SE)	β	b (SE)	β	b (SE)	β
Gang group	0.27 (0.12)*	0.16
Non-gang group	0.01 (0.14)	0.01
Age	−0.09 (0.03)**	−0.21	−0.12 (0.06)*	−0.29^a	−0.05 (0.04)	−0.12^a,c	−0.20 (0.08)*	−0.34^c
Race (non-White)	0.26 (0.11)*	0.15	0.32 (0.19)	0.20	0.14 (0.15)	0.09	0.36 (0.29)	0.16
Education	−0.06 (0.03)*	−0.13	0.01 (0.06)	0.01^a,b	−0.05 (0.04)	−0.12^a	−0.09 (0.05)	−0.24^b
Probation offense
Drug	0.27 (0.17)	0.12
Property	0.34 (0.14)*	0.20
Other nonviolent	0.04 (0.16)	0.02
Property offense			0.20 (0.20)	0.13	0.20 (0.16)	0.13	0.18 (0.30)	0.09
Length of supervision (years)	−0.01 (0.02)	−0.01	0.04 (0.04)	0.12^a,b	−0.01 (0.03)	−0.01^a,c	−0.07 (0.06)	−0.16^b,c
Age of first adult arrest	−0.01 (0.03)	−0.02
Prior adult arrests	0.02 (0.02)	0.06	0.02 (0.03)	0.06^b	0.02 (0.03)	0.06^c	0.08 (0.06)	0.23^b,c
Prior jail sentence (ln)	0.31 (0.11)**	0.21	0.40 (0.21)	0.28^b	0.28 (0.15)	0.22^c	−0.01 (0.30)	−0.01^b,c
Intercept	5.04 (0.65)**		4.75 (1.09)**		4.40 (0.90)**		7.67 (1.69)**
Model F	4.24**		1.77		1.58		3.43**
R ²	.19		.16		.10		.35

Note. Reference category for gang and non-gang groups is SWIFT group. Reference category for race is White. Reference category for probation offense is violent offense. Reference category for property offense only at the time of intake is violent, drug, or other offense. Superscripts indicate coefficient pairs with significant z scores (p < .05) for test of equality of regression coefficients. OLS = ordinary least squares; SWIFT = Supervision with Immediate Enforcement.

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

Group-based models (Models 2-4 in Table 4) illustrated few differences within and across groups for total technical violations. For SWIFT, age at time of placement was the only significant predictor for total technical violations; younger SWIFT offenders had significantly more total technical violations. Similar results were seen in the non-gang group-based results (Model 4). For the gang group, none of the variables were significantly related to total technical violations. Power probabilities for the overall OLS regressions were strong for the combined Model 1 (power = >.99) and the non-gang group-based Model 4 (power = .97), but weaker for the SWIFT group-based Model 2 (power = .72) and gang group-based Model 3 (power = .67).

Comparison of New Offending While on Supervision

Table 1 reports the descriptive statistics and bivariate group comparisons for in-program arrests and charges. For the in-program arrests, the number of arrests ranged from 0 to 3 within each group, with 68.06% of SWIFT cases, 57.41% of non-SWIFT, gang cases, and 48.08% of non-gang cases possessing no arrests while on supervision. For in-program charges, the number of charges ranged from 0 to 9 (SWIFT: 0-3; non-gang: 0-4; gang: 0-9), with 69.44% of SWIFT cases, 57.41% of gang cases, and 48.08% of non-gang cases possessing no charges while on supervision. As the range of arrests/charges was small and in-program charges and arrests were highly correlated (r_s = .97, p < .001), subsequent analyses used the dummy variable for in-program arrest charges. As reported in Table 1, there were significant differences in the dummy variables for in-program arrest and in-program arrest charges when comparing SWIFT with the non-gang group only. A smaller proportion of SWIFT members received official arrests or charges while on supervision, compared with non-gang group members. The power for these bivariate effects was weak, ranging from .73 to .80 for the SWIFT to non-gang group effects and from .38 to .49 for the SWIFT to gang group effects. The power probabilities were highest for the in-program arrest charges dummy variable.

Logistic regressions were estimated for new arrest charge during supervision for SWIFT versus comparison groups, controlling for sociodemographics, length of supervision, progressive sanctions for jail, months sent to prison for revocation, and type of technical violations.⁶ As seen in Model 1 of Table 5, there were no significant differences between SWIFT and the other two groups in predicting the odds of in-program recidivism when controlling for all else. That is, SWIFT offenders were just as likely as gang member offenders and high-risk, non-gang offenders to recidivate while on supervision. Those who were under supervision longer and those who received more jail time as a progressive sanction were significantly less likely to have a new arrest charge while on supervision, regardless of group. Offenders with more high-severity technical violations were at significantly greater odds of having a new arrest charge while on supervision. Conversely, offenders with more low-severity technical violations were at significantly lower odds of having a new arrest charge while on supervision.

Table 5:

Logistic Regression of New Arrest Charges While on Supervision

Variable	Combined (N = 232)		SWIFT (n = 72)		Gang (n = 108)		Non-gang (n = 52)
	Model 1		Model 2		Model 2		Model 2
	b (SE)	OR	b (SE)	OR	b (SE)	OR	b (SE)	OR
Gang group	−0.16 (0.48)	0.85
Non-gang group	0.27 (0.55)	1.31
Age	−0.02 (0.09)	0.98	−0.36 (0.20)	0.70^a,b	0.11 (0.13)	1.12^a	0.16 (0.26)	1.17^b
Race (non-White)	−0.23 (0.35)	0.80	−0.39 (0.69)	0.68	−0.40 (0.52)	0.67	−0.06 (1.11)	0.94
Education	−0.19 (0.08)*	0.83	−0.23 (0.19)	0.80	−0.07 (0.15)	0.93	−0.21 (0.19)	0.81
Length of supervision (years)	−0.17 (0.07)*	0.84	−0.24 (0.15)	0.79^b	−0.30 (0.12)*	0.74^c	0.30 (0.22)	1.35^b,c
High technical violations (ln)	1.73 (0.34)**	5.62	1.01 (0.70)	2.73	1.77 (0.46)**	5.89	4.40 (1.34)**	81.60
Low technical violations (ln)	−0.64 (0.25)**	0.53	−0.84 (0.62)	0.43	−0.75 (0.41)	0.47	−0.91 (0.58)	0.40
Progressive sanction of jail (ln)	−0.28 (0.13)*	0.76	−0.34 (0.33)	0.71	−0.42 (0.18)*	0.66	0.05 (0.41)	1.05
Months to prison for revocation (ln)	0.36 (0.11)**	1.44	0.31 (0.20)	1.36^b	0.54 (0.21)*	1.71^c	0.39 (0.31)	1.48^b,c
Intercept	3.24 (2.14)		12.04 (4.67)*		−0.03 (3.16)		−3.53 (6.28)
−2 Log likelihood	232.92		66.98		105.41		34.95
Model χ²/df	81.77/10**		23.23/8**		41.93/8**		37.06/8**
Pseudo-R²	.30		.28		.32		.51

Note. Reference category for gang and non-gang groups is SWIFT group. Reference category for race is White. Pseudo-R² used is Cox and Snell. Superscripts indicate coefficient pairs with significant z scores (p < .05) for test of equality of regression coefficients. SWIFT = Supervision with Immediate Enforcement; OR = odds ratio.

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

The group-based models (Models 2-4 in Table 5) illustrated few differences within and across groups for in-program arrest charges. For SWIFT members only (Model 2), none of the variables were related to the odds of in-program recidivism. For gang group members only (Model 3), length of supervision and progressive sanctions of jail reduced the odds of new arrest charges while on supervision, but high-level technical violations and months sent to prison for revocation were associated with increased odds of in-program recidivism. The effect for months sent to prison for revocation may be an artifact of the in-program arrest charge itself resulting in imprisonment. Unfortunately, additional cases and data would be necessary to examine the potential spuriousness of this effect. For non-gang group members only (Model 4), high-severity technical violations increased the odds of in-program recidivism. The tests of coefficient equality revealed a few interesting results for the gang group results. For the gang group, the effect of length of supervision on in-program recidivism was significantly lower than the effect for non-gang members, and the effect of months sent to prison for revocation was significantly higher than the effect for non-gang member, but the effects of neither of these regression coefficients were significantly different for gang members compared with SWIFT members. Power analyses of the logistic regressions in Table 5 indicated high power probabilities of greater than .99 for the combined groups, separate gang group, and separate non-gang group models, using the strongest predictors in each model to assess power. However, the power probability for the SWIFT group–only model was weak at .70, using the strongest predictor in the model.

Discussion

The present study evaluated the SWIFT Court program on adult gang–affiliated probationers in the jurisdiction of Tarrant County, Texas. Based largely on the HOPE program, the SWIFT Court focuses heavily on key components, including targeted, high-intensity, and specialized treatments for participants coupled with immediate accountability for violations with appearance before the judge (see Hawken & Kleiman, 2009). Overall, SWIFT had only a mixed deterrent impact on offenders, compared with alternative probation strategies within Tarrant County community corrections.

In support of the first hypothesis, SWIFT gang offenders were more likely to remain on probation and less likely to have probation revoked than the two comparison groups, even when controlling for technical violations. The combined population model suggested high-severity (e.g., contact with the victim, carrying weapon, failure to report to jail) and low-severity (e.g., failure to pay court fees, absence from court-ordered treatment or counseling) technical violations served as risk factors for probation revocation. The impact of high technical violations continued to matter for the groups when examined separately, but did not depend on group membership. For the gang comparison group, probationers who had more low-severity technical violations, which primarily involved paying fines and fees, were at greater risk of having their probation revoked, but the effect was not significantly different compared with the other groups.

Two considerations bear noting regarding revocation status. First, very few of the cases (0% of SWIFT and less than 10% of comparison groups) had successfully completed probation—hence the reliance on in-program recidivism. Therefore, this study could not effectively examine protective factors that may improve SWIFT members’ success on probation. Second, low-severity technical violations were related to revocation status for the gang group members only. It should be noted that all offenders in the SWIFT program were probated out of various courts within the jurisdiction, but those judges agreed to transfer jurisdiction to the SWIFT Court while they were in the SWIFT program. Non-SWIFT offenders remained under the jurisdiction of the original court that probated them, which could have been any one of the 20 criminal courts in Tarrant County. It is possible that these judicial differences reflect the effect of low-severity technical violations, though this is only speculation.

Similar to other types of problem-solving courts (Boots, Wareham, Bartula, & Canas, 2016; Gover, Brank, & MacDonald, 2007), the SWIFT Court judge uses motivational interviewing to attempt to determine the underlying reasons for violations, regardless of the severity, and orders participants to corresponding programming multiple times throughout their supervision term. Moreover, the judge who oversees SWIFT Court is a former prosecutor with a reputation for being very tough; however, she believes in providing offenders every opportunity to change. This premise of rehabilitation is central to the HOPE strategy, which in turn became the basis of the SWIFT Court program. Steven Alm, the founding judge of the HOPE program, recently noted evaluations of HOPE programs often ignore the importance of the judge, probation officers, and treatment providers in facilitating the success of the program (Alm, 2016). Future extensions and replications should gather data that provide better detail about the nature of technical violations and the overseeing judge. Such details may explain group differences in the predictor variables and help identify potential forms of bias affecting success.

It was also hypothesized SWIFT would have more of a deterrent effect than probation for the other two high-risk offender groups. Results only partially supported this assumption. In support of the second hypothesis, at a bivariate level, SWIFT probationers had fewer total rates of technical violations compared with non-SWIFT gang and non-gang offenders, but higher rates of medium-severity, drug-related, and gang-related technical violations. Multivariate results provided partial support for the second hypothesis as well. SWIFT probationers had significantly fewer rates of total technical violations than gang group probationers, controlling for other variables in the model, but not the non-gang group probationers. This finding might be elucidated by the fact that SWIFT gang offenders (as opposed to high-risk non-gang) had more conditions of supervision, specifically prohibiting certain behaviors related to gang activity. Extant research shows length and stability of gang membership are positively correlated to increased violent offending and recidivism risk as well (see Melde, Diem, & Drake, 2012; Sweeten et al., 2013, for example). An alternative explanation is that these findings might reflect differential supervision practices, whereby offenders in SWIFT received more intensive supervision due to risk and thus more violations were discovered compared with the high-risk, non-gang group. In addition, due to the process of bringing SWIFT offenders quickly to court (usually within a day) for their violation, officers were forced to accurately reflect every violation of probation in the chronological case notes, whereas officers supervising the high-risk, non-gang offenders may not be as diligent with their record-keeping.

There was little support for the third hypothesis that SWIFT would have lower in-program recidivism than the other groups. Bivariate comparisons of in-program recidivism across the groups indicated SWIFT had significantly fewer arrest and charges than the high-risk, non-gang group, but not the gang group. In multivariate analyses, SWIFT gang offenders performed similarly to those in the probation comparison groups in their odds of receiving a new arrest/charge while on supervision (i.e., in-program recidivism). This lack of intervention effect on in-program recidivism is inconsistent with Hawken and Kleiman’s (2009) evaluation of HOPE and more consistent with recent evaluations of HOPE (e.g., Cook, 2016; Hamilton et al., 2016; Kleiman, 2016). In short, desistence may be difficult because leaving the gang lifestyle can be a dangerous and complex process involving threats of being beaten or raped, extortion, or being required to commit a crime (Decker & Lauritsen, 2002; Decker & Van Winkle, 1996). Alternatively, some scholars have postulated it is much less difficult to exit a gang and that people simply walk away (Decker & Pyrooz, 2011). Despite the common misperception that gang membership is for life, research has shown that most affiliations last less than 4 years and many end their involvement after 1 year (Melde et al., 2012; Thornberry et al., 2003).

Interestingly, progressive sanctions of short-term jail—a key component of the SWIFT Court program—had the desired deterrent effect on offenders. Short periods of jail time may correct noncompliance on probation. While this type of ongoing, close supervision did not reduce the risk of in-program recidivism among SWIFT probationers, each progressive sanction of jail reduced the odds of recidivism for gang group members by 34%. As gang membership is a robust predictor of recidivism when controlling for other factors and is a salient factor in determining whether to keep serious and violent youthful offenders in the community versus prison (see Trulson, Caudill, Haerle, & DeLisi, 2012), future studies should explore progressive jail sanctions as a deterrent mechanism for gang-involved probationers.

This preliminary study of the effectiveness of SWIFT probation has several limitations worth noting. First, the study did not use an experimental design and had smaller population subgroup (n) sizes. Although the three groups examined here were comparable on key risk and offending history variables and do not appear to differ significantly in their backgrounds, there may be unperceived differences that have biased the findings. In addition, the small n size affected power in the analyses. While we could have increased the level of significance from α of .05 to a higher level to conversely increase power, this would have increased the chances of committing a Type I error and erroneously rejecting the null hypothesis, so we opted to use a more conservative significance level. In most instances, the power of the bivariate and multivariate models was sufficient (>.80), but there were a few exceptions. Plans to extend this study by including more recent SWIFT and comparison group cases should improve power while maintaining a conservative significance level.

Second, although this study examined two forms of recidivism of technical violations and new arrests/charges while on supervision, it did not examine recidivism after supervision. Unfortunately, due to the lengthy period of community supervision and the short tenure of the specialized court program for gang and felony offenders, these data were not yet available for SWIFT cases. Future studies of recidivism after all supervision and conditions of release have concluded are the next step in evaluations of the SWIFT program to determine long-term efficacy.

Third, SWIFT provides for progressive sanctions other than jail time, which were excluded from these analyses due to multicollinearity issues. Furthermore, the data regarding the non-jail progressive sanctions were aggregated across the supervision period and did not offer details about the nature of the progressive sanction or temporal sequencing of technical violations following progressive sanctions of jail. Future research should examine the impact of progressive sanctions other than jail time while on supervision.

Finally, this study relied exclusively on official data. Future research should include data from probationers themselves to supplement and enhance validity of the official data. There may be reasons beyond gang membership and gang associations that affect violations of probation and recidivism. Such information could be used to improve SWIFT and similar programs.

There are important implications from this study of the efficacy of SWIFT in improving probation compliance. There continues to be lively debate within the field regarding the efficacy of intensive supervision programs, as scholarship indicates that these programs frequently result in more technical revisions, higher costs, and more recidivism (see Hyatt & Barnes, 2017) that may result in “backdoor sentencing” (see Medina, 2017). With the rise of such problem-solving courts such as HOPE that target various subpopulations of offenders, rigorous evaluations that explore the efficacy of programming for challenging correctional populations are an important charge for scholars in helping practitioners direct scarce community resources (Boots et al., 2016; Mitchell, Wilson, Eggers, & MacKenzie, 2012).

In keeping with the spirit of problem-solving courts and the ideals of therapeutic jurisprudence which dominate such programs (Wexler, 1993), SWIFT offenders were more likely to be brought before their judge quickly to ensure accountability and remain on probation and less likely to have their probation revoked than gang members on non-SWIFT probation in this jurisdiction. Moreover, the key reactive sanction component of immediate short-term jail time reduced the odds of new arrests/charges during the supervision period. These findings largely mirror the positive findings reported for the HOPE program (Hawken & Kleiman, 2009) while contradicting a large number of studies of sites that have implemented HOPE-based programs and found null or negative effects (see Cullen, Pratt, & Turanovic, 2016, for commentary). Future studies that explore the long-term effects of programs such as SWIFT in community-based sanctions are the next step in bettering our understanding of these programs.

Moreover, future studies might pursue an analysis regarding the cost benefit of the SWIFT Court compared with other community-based sanctions for high-risk offenders. An advanced cost–efficacy analysis that considers the range and benefit of costs and benefits as well as the likelihood and average cost benefit of positive outcomes can be especially insightful in judging the value of these types of programs.

Footnotes

Authors’ Note:

The authors thank all the Tarrant County Community Supervision and Corrections Department staff and offenders who contributed to this research. They also thank the Editor and reviewers for their thoughtful and constructive feedback on this manuscript. Finally, they thank the Research Design and Analysis Unit at Wayne State University for its assistance with the power analyses conducted in this study.

Notes

Denise Paquette Boots, PhD, is an associate professor of Criminology and Criminal Justice at the University of Texas at Dallas. Her research interests focus on violence against women, sexual assault, child abuse, capital punishment, and evaluation of community-based programs.

Jennifer Wareham, PhD, is an associate professor in the Department of Criminal Justice at Wayne State University. Her research interests focus on juvenile justice and delinquency, intimate partner violence, mental health, and evaluation of justice programs.

Kelli Stevens-Martin, MA, is the research director for Tarrant County Community Supervision and Corrections Department and an adjunct instructor at Tarrant County College. Her research interests include community corrections, specialty courts, and special offender populations.

Nina Barbieri, PhD, is an assistant professor of Criminal Justice at the University of Houston–Downtown. Her research interests include delinquency and adolescence, focusing particularly on at-risk youth and school-related experiences.

References

Adams

Olson

(2002). An analysis of gang members and non-gang members discharged from probation. On Good Authority, 6(2), 1-4.

Alm

S. S.

(2016). HOPE probation: Fair sanctions, evidence-based principles, and therapeutic alliances. Criminology & Public Policy, 15, 1195-1214.

Arrigona

Bryan

(1999). Measuring the impact of progressive sanction guidelines. Austin, TX: Criminal Justice Policy Council.

Barnes

G. C.

Hyatt

J. M.

Sherman

L. W.

(2017). An implementation and experimental evaluation of cognitive-behavioral therapy for high-risk probationers. Criminal Justice and Behavior, 44, 611-630.

Battin

S. R.

Hill

K. G.

Abbott

R. D.

Catalano

R. F.

Hawkins

J. D.

(1998). The contribution of gang membership to delinquency beyond delinquent friends. Criminology, 36, 93-115.

Boots

D. P.

Wareham

Bartula

Canas

(2016). A comparison of the Batterer Intervention and Prevention Program to alternative court dispositions on 12-month recidivism. Violence Against Women, 22, 1134-1157.

Braga

A. A.

Piehl

Hureau

(2009). Controlling violent offenders released to the community: An evaluation of the Boston reentry initiative. Journal of Research in Crime and Delinquency, 46, 411-436.

Compilation of Information Pertaining to Criminal Combinations and Criminal Street Gangs, Code of Criminal Procedure, Tex. Art. 61.02 (2015).

Cook

P. J.

(2016). Behavioral science critique of HOPE. Criminology & Public Policy, 15, 1155-1161.

10.

Cullen

F. T.

Pratt

T. C.

Turanovic

J. J.

(2016). It’s hopeless: Beyond zero-tolerance supervision. Criminology & Public Policy, 15, 1215-1227.

11.

Curry

D. G.

(2015). The logic of defining gang revisited. In Decker

S. D.

Pyrooz

D. C.

(Eds.), The handbook of gangs (pp. 28-58). West Sussex, UK: Wiley-Blackwell.

12.

de Beus

Rodriguez

. (2007). Restorative justice practice: An examination of program completion and recidivism. Journal of Criminal Justice, 35, 337-347.

13.

Decker

S. H.

Lauritsen

J. L.

(2002). Leaving the gang. In Huff

C. R.

(Ed.), Gangs in America (3rd ed., pp. 51-67). Thousand Oaks, CA: SAGE.

14.

Decker

S. H.

Pyrooz

D. C.

(2011). Leaving the gang: Logging off and moving on. New York, NY: Council on Foreign Relations. Retrieved from http://www.cfr.org/counterradicalization/save-supporting-document-leaving-gang/p26590

15.

Decker

S. H.

Van Winkle

(1996). Life in the gang: Family, friends, and violence. New York, NY: Cambridge University Press.

16.

Di Placido

Simon

T. L.

Witte

T. D.

Wong

S. C. P

. (2006). Treatment of gang members can reduce recidivism and institutional misconduct. Law and Human Behavior, 301, 93-114.

17.

Dong

Krohn

M. D.

(2016). Dual trajectories of gang affiliation and delinquent peer association during adolescence: An examination of long-term offending outcomes. Journal of Youth and Adolescence, 45, 746-762.

18.

Egley

Howell

Harris

(2014). Highlights of the 2012 National Youth Gang Survey. Washington, DC: U.S. Department of Justice, Office of Justice Programs, Office of Juvenile Justice and Delinquency Prevention.

19.

Esbenson

F.-A.

Osgood

Taylor

T. J.

Peterson

Freng

(2001). How great is G.R.E.A.T? Results from a longitudinal quasi-experimental design. Criminology & Public Policy, 1, 87-118.

20.

Faul

Erdfelder

Lang

A.-G.

Buchner

(2007). G*Power 3: A flexible statistical power analysis program for the social, behavioral, and biomedical sciences. Behavior Research Methods, 39, 175-191.

21.

Fox

(1991). Regression diagnostics (Quantitative applications in the social sciences series [07-079]) Newbury Park, CA: SAGE.

22.

Glaze

Kaeble

(2014). Correctional populations in the United States, 2013. Washington, DC: Bureau of Justice Statistics.

23.

Gottfredson

Kearley

Najaka

Rocha

(2007). How drug treatment courts work: An analysis of mediators. Journal of Research in Crime & Delinquency, 44, 3-35.

24.

Gover

A. R.

Brank

E. M.

MacDonald

J. M.

(2007). A specialized domestic violence court in South Carolina: An example of procedural justice for victims and defendants. Violence Against Women, 13, 603-626.

25.

Hamilton

Campbell

C. M.

van Wormer

Kigerl

Posey

(2016). Impact of swift and certain sanctions. Criminology & Public Policy, 15, 1009-1071.

26.

Hawken

Kleiman

(2009). Managing drug-involved probationers with swift and certain sanctions: Evaluating Hawaii’s HOPE: Executive summary. Washington, DC: National Institute of Justice.

27.

Herberman

Bonczar

(2014). Probation and parole in the United States, 2013. Washington, DC: Bureau of Justice Statistics.

28.

Huebner

B. M.

Varano

S. P.

Bynum

T. S.

(2007). Gangs, guns, and drugs: Recidivism among serious, youth offenders. Criminology & Public Policy, 6, 187-222.

29.

Hyatt

J. M.

Barnes

G. C.

(2017). An experimental evaluation of the impact of intensive supervision on the recidivism of high-risk probationers. Crime & Delinquency, 63, 3-38.

30.

Kleiman

M. A. R.

(2016). Swift-certain-fair: What do we know now, and what do we need to know? Criminology & Public Policy, 15, 1185-1193.

31.

Kluger

A. N.

Tikochinsky

(2001). The error of accepting the “theoretical” null hypothesis: The rise, fall, and resurrection of commonsense hypotheses in psychology. Psychological Bulletin, 127, 408-423.

32.

Krienert

J. L.

Fleisher

M. S.

(2001). Gang membership as a proxy for social deficiencies: A study of Nebraska inmates. Corrections Management Quarterly, 5, 47-58.

33.

Lattimore

P. K.

MacKenzie

D. L.

Zajac

Dawes

Arsenault

Tueller

(2016). Outcome findings from the HOPE demonstration field experiment. Criminology & Public Policy, 15, 1103-1141.

34.

Martin

K. D.

(2013). The Supervision with Immediate Enforcement (SWIFT) court program: An overview of outcomes for the initial pilot program. Fort Worth, TX: Tarrant County Community Supervision and Corrections Department.

35.

Matz

Stevens-Martin

DeMichele

(2014). Barriers to effective gang-member reentry: An examination of street gang-affiliated probationer revocation in a southwestern state. Journal of Gang Research, 21, 32-51.

36.

Maxson

C. L.

Matsuda

K. N.

Hannigan

(2011). “Deterrability” among gang and nongang juvenile offenders: Are gang members more (or less) deterrable than other juvenile offenders? Crime & Delinquency, 57, 516-543.

37.

McRee

Drapela

(2012). The timing and accumulation of judicial sanctions among drug court clients. Crime & Delinquency, 58, 911-931.

38.

Medina

J. C.

(2017). Making the decision to extend probation supervision at a local agency. Crime & Delinquency, 63, 1712-1730.

39.

Melde

Diem

Drake

(2012). Identifying correlated of stable gang membership. Journal of Contemporary Criminal Justice, 28, 482-298.

40.

Mitchell

Wilson

Eggers

MacKenzie

(2012). Assessing the effectiveness of drug courts on recidivism: A meta-analytic review of traditional and non-traditional drug courts. Journal of Criminal Justice, 40, 60-71.

41.

Nagin

Porgarsky

(2001). Integrating celerity, impulsivity, and extralegal sanction threats into a model of general deterrence: Theory and evidence. Criminology, 39, 865-892.

42.

National Youth Gang Center. (2006). National Youth Gang Survey, 1999-2001 (NCJ 209392). Washington, DC: Office of Justice Programs, Office of Juvenile Justice and Delinquency Prevention.

43.

Olson

Dooley

Kane

(2004). The relationship between gang membership and inmate recidivism. Chicago: Illinois Criminal Justice Information Authority.

44.

Paternoster

Brame

Mazerolle

Piquero

(1998). Using the correct statistical test for the equality of regression coefficients. Criminology, 36, 859-866.

45.

Piquero

(2003). A recidivism analysis of Maryland’s community probation program. Journal of Criminal Justice, 31, 295-307.

46.

Pyrooz

D. C.

Mitchell

M. M.

(2015). Little gang research, big gang research. In Decker

S. D.

Pyrooz

D. C.

(Eds.), The handbook of gangs (pp. 28-58). West Sussex, UK: Wiley-Blackwell.

47.

Pyrooz

D. C.

Turanovic

J. J.

Decker

S. H.

(2016). Taking stock of the relationship between gang membership and offending: A meta-analysis. Criminal Justice and Behavior, 43(3), 365-397.

48.

Rossman

S. B.

Zweig

J. M.

Kralstein

Henry

Downey

P. M.

Lindquist

(2011). The multi-site adult drug court evaluation: The drug court experience (Vol. 3, Document No. 237111). Washington, DC: Urban Institute Justice Policy Center.

49.

Scott

(2004). “It’s a sucker’s outfit”: How urban gangs enable and impede the reintegration of ex-convicts. Ethnography, 51, 107-140.

50.

Spooner

Pyrooz

D. C.

Webb

V. J.

Fox

K. W.

(2017). Recidivism among juveniles in a multi-component gang reentry program: Findings from a program evaluation in Harris, County, Texas. Journal of Experimental Criminology, 13, 275-285.

51.

Stevens-Martin

K. D.

(2014, September/October). The Supervision with Immediate Enforcement (SWIFT) court: Tackling the issue of high-risk offenders and chronic probation violators. Corrections Today, 76, 74-77.

52.

Sweeten

Pyrooz

D. C.

Piquero

A. R.

(2013). Disengaging from gangs and desistance from crime. Justice Quarterly, 303, 469-500.

53.

Taylor

C. J.

(2012). Balancing act: The adaptation of traditional judicial roles in reentry court. Journal of Offender Rehabilitation, 51, 351-369.

54.

Texas Department of Criminal Justice. (2014). Statistical report, fiscal year 2014. Huntsville: Author.

55.

Thornberry

T. P.

Krohn

M. D.

Lizotte

A. J.

Smith

C. A.

Tobin

(2003). Gangs and delinquency in developmental perspective. New York, NY: Cambridge University Press.

56.

Trulson

C. R.

Caudill

J. W.

Haerle

D. R.

DeLisi

(2012). Cliqued up: The postincarceration recidivism of young gang-related homicide offenders. Criminal Justice Review, 37, 174-190.

57.

Trulson

C. R.

Haerle

D. R.

Caudill

J. W.

DeLisi

(2016). Lost causes: Blended sentencing, second chances, and the Texas Youth Commission. Austin: University of Texas Press.

58.

Watkins

A. M.

Moule

R. K.

(2014). Older, wiser, and a bit more badass? Exploring differences in juvenile and adult gang members’ gang-related attitudes and behaviors. Youth Violence and Juvenile Justice, 12, 121-136.

59.

Wexler

(1993). Therapeutic jurisprudence and the criminal courts. William & Mary Law Review, 35, 279-299.

60.

Winick

B. J.

(2002). Therapeutic jurisprudence and problem solving courts. Fordham Urban Law Journal, 30, 1054-1103.

A Preliminary Evaluation of the Supervision With Immediate Enforcement Probation Program for Adult Gang–Affiliated Offenders in Texas

Abstract

Keywords

Introduction

Literature Review

Problem-Solving Courts and Intensive Probation

The Present Study

Method

Participants and Procedures

Measures

Group Membership

Sociodemographic, Mental Health, and Substance Abuse Characteristics

Prior Criminal History

Supervision Variables

Technical Violations for Probation

New Arrests and Arrest Charges While on Supervision

Results

Descriptive Statistics and Comparison of Sociodemographic and Offense History

Comparison of Supervision Variables

Comparison of Technical Violations

Comparison of New Offending While on Supervision

Discussion

Footnotes

Authors’ Note:

Notes

References