Technical Report. An Analysis of Probation Violations and Revocations in Maine Probation Entrants in Maine Statistical Analysis Center

Transcription

1 Technical Report An Analysis of Probation Violations and Revocations in Maine Probation Entrants in Submitted to the Justice Research and Statistics Association by Mark Rubin, Research Associate Muskie School of Public Service Michael Rocque, Ph.D. Student Northeastern University, Massachusetts Will Ethridge, Research Assistant Muskie School of Public Service December 2010 Maine Statistical Analysis Center USM Muskie School of Public Service

2 ABOUT THE UNIVERSITY OF SOUTHERN MAINE (USM) MUSKIE SCHOOL OF PUBLIC SERVICE The USM Muskie School of Public Service educates leaders, informs public policy, and strengthens civic life through its graduate degree programs, research institutes and public outreach activities. By making the essential connection between research, practice, and informed public policy, the School is dedicated to improving the lives of people of all ages, in every county in Maine and every state in the nation. ABOUT THE MAINE STATISTICAL ANALYSIS CENTER (SAC) The Maine Statistical Analysis Center (SAC) operates as a collaborative service of the USM Muskie School of Public Service and the Maine Department of Corrections. The SAC is partially supported by the Bureau of Justice Statistics and is guided by an Advisory Group of policy makers from the Maine Administrative Office of the Courts, Maine Department of Public Safety, Maine Department of Corrections, and Maine Criminal Justice Commission. The SAC collects, analyzes, and disseminates justice data and research reports to criminal justice professionals, policy makers, researchers, students, advocates, and the public. The Maine SAC website is located at: FUNDER This agreement is for the performance of a portion of the work originally awarded to JRSA from the Bureau of Justice Statistics, USDOJ, Award #2007-BJ-CX-K042, CFDA #16.734, under the direction of Stan Orchowsky, JRSA Research Director. The opinions, findings, and conclusions expressed in this publication are those of the authors and do not necessarily reflect the view of JRSA, the Bureau of Justice Statistics, the Department of Justice or the Maine Department of Corrections.

3 Table of Contents Section I: Introduction... 1 Maine s Sentencing Law and Probation Requirements... 1 Split Sentences... 3 Revocations... 4 Supervision of Probationers... 5 Section II: Descriptive Analysis ( )... 8 Demographic and Criminal-Legal Profiles of Maine Probationers... 8 Commitment Offense and Prior Criminal Record... 9 Probation Caseload and Probation Agent Characteristics Violations: Patterns and Processes Section III: Survival Analysis: Examining the Factors that influence Probation Violations Recidivism: All Violation Types... 4 Recidivism: Misdemeanor or Felony Analysis... 5 Section IV: Logistic Regression Survival Analysis: Examining the Factors that influence Probation Section V: Discussion References... 25

4 Section I: Introduction In 2009, Maine was selected (with four other states 1 ) to conduct a study of probation revocations in Maine, and to provide data to the Justice Research and Statistics Association (JRSA) for a multi-state study of parole/probation revocations. This study analyzes a sample of 4,725 offenders who entered probation between January, 2005 and December, 2006 from either prison or jail, and analyzes probation violations and revocations occurring in the sample population over a 24 month period. Maine s correctional system is unlike many others in the U.S. in that parole was abolished by the state legislature in However, probation often acts as de-facto parole in Maine, as more than two thirds of offenders enter probation from jail or prison (split sentence). Probation is a state-wide function administered by the Maine Department of Corrections (MDOC). MDOC supervises all Maine probationers across four probation regions. Since this study s main objective is to examine variations in probation violations and revocation practices, we must first understand how Maine s unique correctional laws and administration constrain and influence probation decision-making. The next section of this report briefly describes those aspects of Maine s sentencing system and probation supervision regulations that may influence probation recidivism rates. Maine s Sentencing Law and Probation Requirements In Maine, probation is a court-ordered term of community supervision with specified conditions for a determinant period of time that cannot exceed the maximum sentence imposed by a court. Probation is imposed on an adjudicated offender who is placed under supervision in lieu of or subsequent to incarceration, with a requirement to comply with certain standards of conduct. The probationer is required to abide by all conditions ordered by the court. Violation of these conditions may result in revocation by the court and imposition of an underlying sentence which was imposed at the time the offender was sentenced to probation. The probationer is generally required to pay the cost of supervision to the State of Maine, and may have additional conditions requiring payment of restitution, court costs and fines, public service, and various types of treatment. The probationer is usually required to visit his/her supervising officer in the local field office at intervals related to their risk of re-offending as measured by a risk/need assessment tool. If the probationer s assessment places him/her in the higher risk of re-offending levels, the officer will also contact the offender at his/her home and place of employment and maintain contact with service providers and other community members. Maine continues to have the lowest state prison incarceration rate per capita in the nation. By the end of 2008, Maine s 151 inmates per 100,000 residents was the lowest rate in the country. 1 The other states are New Mexico, Oregon, Utah and Wyoming. An Analysis of Probation Violations and Revocations in Maine Probation Entrants in

5 Maine s incarceration rate was three times lower than the national average (504). Maine had fewer total inmates (2,195) than New Hampshire (2,904), and a comparable number to Vermont (2,116). The lower incarceration rate reflects the overall crime rate in Maine which was 31% lower than the national average (Rubin, 2009). Between 1998 and 2007, the decline in Index crime rates in Maine (16.7%) and New Hampshire (16.2%) was less than the decline in Vermont (22.0%) and the U.S. overall (19.2%). From 2007 to 2008, Maine s state prison population grew an estimated 2.2%, continuing the growth trend of recent years. This was the ninth fastest growth in the country, and surpassed the national average of 0.8%. Since 2007, nearly half of all admissions to prison have been the result of prisoners being sent back to prison for a probation revocation. Those who are returned to prison on a probation violation are said to have had their probation revoked, either partially, meaning they will be released back onto probation, or fully revoked, where they are to serve the remainder of their probation in prison. Figure 1 - Percentage of New Admissions by Year 56.0% 54.0% 54.5% 52.0% 50.0% 48.0% 50.4% 49.6% 48.4% 51.6% 49.5% 50.5% Revocations 46.0% 45.5% New Sentence 44.0% 42.0% 40.0% Currently, nearly two-thirds of inmates were sentenced to state prison for a Class B or C felony crime (61.4 %). Overall, nine % of inmates in the state prisons have been convicted of murder 2, while only 0.4 % are in prison for a misdemeanor offense (Class D & E). Maine s state prison inmates serve an average of 7.9 years. Other than the 45 inmates in prison with a life sentence, the remainder (97.8%) will return to the community. More than half of the prison population has, on average, a sentence of five years or less. 2 In this table, most of the murders are in the A class category or in the Life sentence category An Analysis of Probation Violations and Revocations in Maine Probation Entrants in

6 Table 1 Prison Inmates by Crime Class and Average Sentence Length (as of 9/28/2010) Crime Class Frequency Percent Average Sentence Length A years B years C years D year E NA Murder years Life Life Total years In Maine, the increase in the county jail population has been noted by policy makers as a critical area in need of ongoing policy adjustments and reform. County jails are populated by two distinctly different types of inmates, those awaiting pre-trial hearings and those already convicted and sentenced. Generally, pre-trial defendants are in jail for a short period of time, and are usually released from custody, pending arraignment or other court hearing. Sentenced inmates generally are in the jails for a longer period of time, and are serving a jail sentence for a criminal conviction imposed by the court. The average in-house population of adult inmates in Maine s county jails has nearly doubled over the last ten years. The number of pre-trial inmates has nearly doubled (86.3%), and now represents the majority of inmates in the county jails. The number of convicted and sentenced inmates has also increased, but at a slower rate (17.8%). Split Sentences In Maine, prison sentences can be fully served while incarcerated, can be wholly suspended with probation, or can be split, with an unsuspended portion of the sentence served in incarceration followed by a period of probation. 3 This latter form is referred to as a split sentence. Throughout the entire period of probation, the offender is subject to having the whole suspended portion of the sentence, or any portion thereof, ordered served in incarceration as a result of a violation of probation. In 2005, Maine s sentencing laws were revised to reduce the length of time an offender can be sentenced to probation. 4 Probation lengths are based on the conviction offense: 3 See 17-A M.R.S.A. section 1152(2). 4 See 17-A M.R.S.A. section 1206(7-A) An Analysis of Probation Violations and Revocations in Maine Probation Entrants in

7 Table 2 Maine Probation Length by Class of Offense Class A B C D E Years 4 years 3 years 2 years 1 year 1 year Revocations A revocation of probation occurs when a court orders an offender to serve the remainder of his/her sentence in prison or jail, due to a new crime or technical (non-crime) violation. A technical violation occurs when a probationer violates one or more conditions set by the probation officer and the judge. Revocations for technical violations declined since 2005, falling to 17.0% in The decline in revocations for technical violations can be attributed in part to the increased use of alternative sanctions over the last four years. Alternative sanction refers to a punishment other than a revocation, including a verbal warning, a graduated sanction 5, amended conditions, etc. Maine s probation officers are far more likely in 2007 to issue a graduated sanction, or to give a warning to the probationer when the first violation is technical in nature. Maine s regional probation supervisors note that there are few formal revocation guidelines. 6 The most significant guideline was issued four years ago and requires a probation officer to consult with his/her supervisor for approval of a recommendation that exceeds 90 days of incarceration. One region offers that it has been difficult to maintain compliance with this directive, and now makes it mandatory for all recommendations to be documented by the officers and filed directly with the court. This formal documentation records the risk level, provides a narrative of the criminal history of the offender, and makes a recommendation concerning bail. As a result, the supervisors interviewed for this study believe that probation officers do a better job of maintaining consistency with the guideline. Although there are few mandated revocation guidelines, probation supervisors indicate that there is an effort to limit the use of incarceration for lower risk offenders and to keep technical violators out of prison in order to reduce prison overcrowding. Officers have a great deal of discretion to determine the array of sanctions to apply to individual cases. Probation officers refer to a list of 17 graduated sanctions and implement these based upon their discretion. Despite the great degree of discretion available to officers, all recommendations are required to be documented in the Department s Corrections Information System (CORIS). 5 Graduated sanction means any of a wide range of non-prison offender accountability measures and programs, including, but not limited to, electronic supervision tools; drug and alcohol testing or monitoring; day or evening reporting centers; restitution centers; forfeiture of earned compliance credits; rehabilitative interventions such as substance abuse or mental health treatment; reporting requirements to supervision officers; community service or work crews; secure or unsecure residential treatment facilities or halfway houses; and short-term or intermittent incarceration. 6 Interviews with regional probation supervisors conducted in March, An Analysis of Probation Violations and Revocations in Maine Probation Entrants in

8 Probation supervisors across the regions agree that the most common reasons for revocations are technical in nature- most often possession or use of drugs or alcohol. Some supervisors note that there is a perception that too many probationers are returned to prison and jail for technical violations. Supervision of Probationers The Level of Service Inventory- Revised (LSI-R) is used to assess the level of risk of recidivism of an offender. The LSI-R score is comprised of 10 categories (domains): Criminal History, Education/Employment, Finances Family/Marital, Accommodations, Leisure/Recreation, Companions, Alcohol/Drug, Emotional/Personal, and Attitude/Orientation. The total LSI-R score can range from 0 to 54, with lower numbers indicating less likelihood of recidivating than higher numbers. Many LSI-R domains are dynamic (can be changed) and are important for case planning and case management, as probation officers and treatment providers work with a probationer to effect positive behavior changes. Others, such as Criminal History, are static, and cannot be changed. Level of Supervision Maximum Risk (LSI-R: 32+) High Risk (LSI-R: 26-31) Moderate Risk (LSI-R: 21-25) Selected Probation Contact and Testing Requirements Contacts by the Probation Officer with an offender classified as Maximum Risk shall consist of five (5) contacts during a one (1) month period with at least one (1) contact per week. One (1) contact shall be in the person's home, two (2) shall be face to face contacts with the person and the remaining two (2) may be collateral contacts, to include at least one (1) employment check every two (2) months. In addition to the monthly contact in the home, when it is appropriate, officers shall incorporate home visits into their case plans for the purpose of monitoring identified risks. Contacts by the Probation Officer with an offender classified as High Risk shall consist of three (3) contacts during a one (1) month period including (1) face to face contact- and two (2) collateral contacts. At least one (1) face to face contact shall be made in the home per quarter. In addition to the quarterly home contact, when it is appropriate, officers shall incorporate home visits in to their case plans for the purpose of monitoring identified risks. Contacts by the Probation Officer with an offender classified as Moderate Risk shall consist of at least two (2) monthly contacts by the Probation Officer, with one (1) face to face and one (1) collateral An Analysis of Probation Violations and Revocations in Maine Probation Entrants in

9 Low Risk (LSI-R: 14-20) Administrative (LSI-R: 0-13) Contacts by the Probation Officer with an offender classified as Low Risk shall be at least one (1) every three (3) months. The contact may be satisfied either by an office contact in person or by telephone. A person classified as administrative is not required to report, except for the initial assessment and, if required by the supervising Probation Officer, sixty (6) days prior to termination of their supervision. In cases where there is a new criminal conviction or there is a citizen or law enforcement complaint involving an administrative case, the supervising Probation Officer shall take appropriate action. The probation officer responds to every violation of conditions of supervision, evaluating the nature of the violation and the offender s violation history in order to respond in an appropriate manner. The responses are documented in the Corrections Information System (CORIS), a fully integrated, web based MIS system designed to manage all aspects of MDOC data. This program has been in production since 2003 with detailed records for over 60,000 offenders. Graduated sanctions may be applied when appropriate to address the offender s behavior. Revocation proceedings are initiated when graduated sanctions are deemed ineffective or considered inappropriate. A study conducted MDOC and the University of Southern Maine s Muskie School of Public Service in 2005 led to two primary changes to the management of higher risk offenders by the Department. These were: (1) lowering the LSI-R threshold scores for Moderate, High and Maximum risk offenders; and (2) requiring a case plan for each higher risk offender. The re-calibration of Maine s probationer risk levels was made to better identify the higher risk probationers for increased supervision and case management. DOC also created a new LSI-R category of Low risk to provide a greater ability to manage clients according to their risk; consequently, the probation officers were able to avoid excessive contact with offenders that were less likely to reoffend, and instead could concentrate on giving greater attention to higher risk cases. While probationers are supervised according to their risk levels (above), there is currently no risk assessment tool in Maine for the purpose of recommending pretrial decisions. Corrections System Capacity While Maine has the lowest number of state prison inmates per 100,000 residents in the nation (151), the state s incarceration rate has risen 31.4% over the last ten years, and is projected to increase. An Analysis of Probation Violations and Revocations in Maine Probation Entrants in

10 Despite spending a smaller percentage of its general fund dollars on correction 4.6% -- than all but seven other states, Maine allocated $138 million on state corrections funding in 2007, a significant and escalating cost for a rural state with low population. In 2008, Maine froze the counties portion of its property taxes dedicated to jail costs to $52.5 million, with the remainder to be covered by state government. In 2005, the Corrections Alternatives Advisory Committee (CAAC) was established by Governor Baldacci to improve the efficiency and effectiveness of state and county level corrections systems, and to better manage costs. The CAAC found that the average length of stay (65 days) for pretrial defendants in a majority of Maine jails was more than three times higher than in other states. The increasing average length of stay for pretrial offenders in Maine jails was identified as one of the major factors contributing to the increase in county jail population. The CAAC identified changes in the bail code and pre-trial processes as essential elements to reducing county jail totals. As of 9/28/2010, there are 2,076 state prison inmates in Maine. The youngest is 18 years old, and the oldest is 79. More than half (58.7%) are under the age of 35, and 3.6% are over the age of 55 years old. Of the 1,644 prisoners in adult facilities for whom education data is available, a majority (58.4%) have less than a high school education, and nearly one-eighth (11.3%) have less than a 9th grade education. Overall, 41.7% of the inmates in Maine s prison system have a 12 th grade education or higher, compared with 89.4% in the general population. In 2005, Maine released 2,183 offenders back into the community, an increase of 14% over the previous year. Nearly half return to Cumberland County, primarily to the city of Portland. Though Maine has one of the nations lowest per capita rates of incarceration, the state s facilities were operating at overcapacity as of National Governors Association for Best Practices An Analysis of Probation Violations and Revocations in Maine Probation Entrants in

11 Section II: Descriptive Analysis ( ) This section describes the characteristics of the offenders who entered probation from prison or jail in 2005 and The section begins with a look at the basic demographics of the population, and then goes on to examine the prior criminal record of the offenders, as well as their offending history. The section concludes with a look at the differences in probation officer caseloads and practices across probation regions. Demographic and Criminal-Legal Profiles of Maine Probationers Gender, Age, Race Maine s probationers are mostly male, white, and between the ages of 18 and 30. About 85% of probationers in the cohort are male, and about 92% were white. Nearly 50% of offenders are age 18 to 30 when they enter probation (See Table 3). Table 3: Demographic Characteristics of Maine Probationers, 2005 and 2006 Number Percent Gender Male 4, % Female % Race White 4, % Black % American Indian % Asian/P.I % Other % Age Age , % Age , % Age % Overall Total 4,725 Of the 3,354 probationers for whom education data is available, 41.5 % have less than a high school (HS) education, and nearly one-fifth (17.1 %) have less than a 10th grade education. Of the seven educational attainment categories, the largest share of the population completed 12th grade (39.1 %), with an additional 14.3 % completing a GED program. Overall, 59.6 % of the offenders entering probation from jail or prison in 2005 and 2006 have a 12 th grade education or a higher level of education, compared with 85.4 % across the state. The majority of offenders entering probation from jail or prison in 2005 and 2006 were not married, and more than a third were not employed. An Analysis of Probation Violations and Revocations in Maine Probation Entrants in

12 Table 4: Education, Marital Status, and Employment Characteristics of Maine Probationers, 2005 and 2006 Number Percent Education Elementary % Some High School 1, % High School Graduate/GED 1, % Associate's Degree % Bachelor's Degree % Trade/Technical School % Graduate Degree 5 0.1% Marital Status Not married 3, % Married/Domestic Partner % Employment Full-Time 1, % Working (not full-time) % Not working % Commitment Offense and Prior Criminal Record Criminal histories of probationers in the sample were quite varied. The average (mean) age at first arrest was More than half of the probationers entered with a felony conviction, with a violent or drug offense as the most common offense type. As noted earlier, over three-quarters of the sample entered probation from jail, meaning they served nine months or less of their sentence in a confined facility. Table 5: Commitment Offense and Prior Criminal Record Characteristics of Maine Probationers, 2005 and 2006 Number Percent Age at 1st arrest Average NA Offense Type Felony 2, % Misdemeanor 2, % Presenting Offense Violent 1, % Property 1, % Drugs 1, % Sex % Other % Don't Know % Sentenced to Jail or Prison Jail 3, % Prison % An Analysis of Probation Violations and Revocations in Maine Probation Entrants in

13 Probation Caseload and Probation Agent Characteristics While this report found no differences in the demographic make-up among probationers across the four regions, differences (among caseloads and revocation guidelines) exist among the various regions. Interviews conducted with probation supervisors and line-staff from all four regions indicate some of the major distinctions in their practices and protocols. In general, there are very few formal revocation guidelines, and as discussed previously, a central directive was issued four years ago requiring all probation officers to gain approval for any revocation recommendation resulting in over 90 days incarceration. Additional guidelines are far more informal. For instance, across the regions there is a general effort to limit incarcerations for lower risk offenders and to keep technical violators out of prison. One regional supervisor said: There has been an effort to look at who is being sent to the state prison system. The big picture is implementing evidence-based principles (EBP) and making sure low risk offenders are not needlessly put into the system. We have been educated, trained, and reminded on this. However, I have no idea what other regions are doing. We are not being given specific directions, more an approach to how we handle offenders, violations, and supervision. By paying attention to these concepts, it guides our decision-making in a certain direction. Violations: Patterns and Processes Violations are aggregated into the categories listed in Table 6. Technical violations include noncriminal administrative violations experienced of the probation process. As a whole, technical violations account for nearly half of all violations experienced on probation (48.8%). Misdemeanors constitute 35.5 % of all violations and the majority of new criminal violations. Table 6: Violation by Type, 2005 and 2006 Violation Category Frequency Percent Total Violations % Technical Violations % Criminal Violations % Felony % Misdemeanor % An Analysis of Probation Violations and Revocations in Maine Probation Entrants in

14 Section III: Survival Analysis: Examining the Factors that Influence Probation Violations To examine the factors that influence probation violations we use a multivariate regression technique called survival analysis. Survival Analysis is a multivariate method that will enable examination of the likelihood and timing of violations. There are several different types of survival models (Allison 1995). We use the Cox regression model, which is named for the English statistician Sir David Cox, and which combines a proportional hazards model with partial likelihood estimation (Cox 1972). Given the nature of our data, the fact that violations are repeatable events, that the risk of a violation is continuous rather than experienced only in discrete time periods, and given the left truncation in our data, Cox models are the most appropriate statistical technique. In fact, Cox models have become a standard approach in studies of recidivism and parolee behavior (e.g., Benda, Toombs, and Peacock 2002; DeJong 1997; Langton 2006; Hepburn and Albonetti 1994). Table7 displays descriptive information for survival time. As is shown, the average number of days to first violation is 208 days and the average number of days to violation if the first violation was a crime (misdemeanor or felony) is 220 days. These results are only for probationers who violated within the study period (730 days). Note that in the dataset, several probationers recorded violations beyond the study period. In the survival analysis that follows, the raw number of days is included but the status variable (violation) indicates no violation if the days to violation variable is greater than 730 days. Table 7 - Descriptive Statistics Days to recidivism Days to violation Valid N (listwise) 1142 The results below indicate that nearly half of the probationers experienced a violation of any type (e.g., technical, felony, and misdemeanor). Note that these results record only violations that were founded in court. That is, numerous probationers had violations in which a court did not find enough evidence for a conviction. These violations are not included in the analysis. The survival analysis is also conducted on a probationer s first founded violation. Many probationers experienced multiple (up to 12) violations during the study period. Analyses (below) take into consideration a probationer s first technical, misdemeanor or felony violation, regardless of prior violation types. However, the analysis reported in this section examines the probationer s first violation, regardless of type. For example, if a probationer s first violation was a felony, s/he is included in the recidivism and any violation analyses (below). If a probationer s first violation was a technical violation, s/he is only included in the any violation analysis and not the recidivism analysis. In the first set of analyses, the dependent variable is any violation during the study period. Each model sequentially adds more covariates to determine the effect of additional variables. An Analysis of Probation Violations and Revocations in Maine Probation Entrants in

15 In the demographic models, age, sex, marital status and employment all show predicted relationships with survival. With respect to age, three categories were created (following Grattet et al., 2008): Probationers aged 18-30, aged 31-44; and 45+. For the analyses below, the middle category (31-44) is the reference category. Thus, probationers aged and 45+ are compared to those aged The first model (table 8) with covariates includes the first and last age categories. Both are significantly associated with the hazard of violation in a theoretically expected manner. For example, probationers aged have a 35% greater risk of violation than those aged Probationers aged 45 and up have a 38% lower risk of violation than those aged Table 8- Variables in the Equation B SE Wald df Sig. Exp(B) Age18 to Age45 and up The second model includes age and sex as predictors. The age category variables change relatively little. The sex variable (male) indicates that males have a 29% greater risk of violation than females. Table 9 - Variables in the Equation B SE Wald df Sig. Exp(B) Age18 to Age45 and up Male The third model includes marital status and employment. As expected, both exert a negative effect on violation. For example, married probationers have an 18% less risk of violation than unmarried probationers and a one unit increase in employment (from unemployed to part-time to full-time) is associated with a 23% decrease in risk. However, there is a considerable number of probationers with employment data missing (1425) resulting in a lower overall sample size for analysis. An Analysis of Probation Violations and Revocations in Maine Probation Entrants in

16 Table 10 - Variables in the Equation B SE Wald df Sig. Exp(B) Age18 to Age45 and up Male Married Employed The next set of results incorporates risk variables. Specifically, age at first arrest and the number of prior arrests are included along with demographic variables. As expected both age at first arrest (2% decrease) and prior arrests (4% increase) are associated with violation risk. These are both theoretically expected. For example, the older the probationer was at first arrest, the lower the hazard ratio of violation. One would expect that those who begin their criminal careers at an earlier age would be more prone to violation (Blumstein et al., 1986; Gottfredson and Hirschi, 1990; Moffitt, 1993). Interestingly however, the incorporation of these risk/criminal history variables render sex non-statistically significant. Thus, arguably, the effect of males on violation operates through criminal history. This is a somewhat surprising finding. The hazard ratio for marital status also decreases, to.797 (20% decrease in risk). Table 11 - Variables in the Equation B SE Wald df Sig. Exp(B) Age18 to Age45 and up Male Married Employed Age at First Arrest Prior Arrests The next model incorporates the LSI total risk variable along with demographic and criminal history variables. Interestingly, the inclusion of the LSI risk variable renders several demographic and risk variables non-statistically significant. This is not completely unexpected, considering that the LSI includes many of these or similar risk factors. For example, when sex is included without risk factors (table 10), males are shown to have an increased risk of violation. An Analysis of Probation Violations and Revocations in Maine Probation Entrants in

17 However, when age at first arrest, prior arrests and sex are included in the model, sex is no longer predictive of violation. Similarly, age at first arrest and sex are not predictive of the risk of violation when LSI is in the model (last table in violation section). However, as expected, the LSI is predictive of violation, even including other risk factors. A one unit increase in the LSI corresponds to a 6% increase in risk of violation. Table 12 - Variables in the Equation B SE Wald df Sig. Exp(B) Age18 to Age45 and up Male Married Employed Age at First Arrest Prior Arrests Total Risk Recidivism: All Violation Types The analysis below includes age, sex, and LSI domains as predictors of violation. This analysis provides insight into which of the components of the LSI are related to risk of recidivism. As is shown, age, sex, education/employment, criminal history, accommodations, companions and drug/alcohol history are significantly related to the risk of violation. Of these, age (18-30) has the largest hazard ratio, with a 95% greater risk of violation than probationers over the age of 45. With respect to the LSI domains, the accommodations domain is associated with the largest hazard ratio (1.13). Table 13 - Variables in the Equation B SE Wald df Sig. Exp(B) Age18 to Age31 to Male Criminal History Education/Employment Financial An Analysis of Probation Violations and Revocations in Maine Probation Entrants in

18 Family/Marital Accommodations Leisure/Recreation Companions Alcohol/Drug Emotional/Personal Attitude/Orientation Recidivism: Misdemeanor or Felony Analysis In this section, we analyze survival time to first violation conditional on whether the first violation was a crime (e.g., misdemeanor or felony). The analysis above examined all violations regardless of type. In this analysis, if the first violation was a technical violation, the time variable is coded as missing. This is because it would be misleading to code the time variable as if the individual did not commit an infraction (e.g., days to the end of the study period) when in fact they did have a violation. The other approach would be to code days to first criminal violation, regardless of how many previous technical violations occurred. That tactic is not taken here because it essentially implies that the person survived from date of first supervision until a criminal violation, and ignores the violations that occurred previously. We believe this is approach would be somewhat misleading. The analysis of survival time for criminal violations mirrors that which was done for all violations. It should be noted, however, that there are considerably more missing data in these models (which is caused, in part, by the exclusion of technical violations coded as missing). The first model (below) shows a survival analysis with no covariates. Table 14 - Recidivism Survival Analysis: No Covariates Case Processing Summary N Percent Cases available in analysis Event % Censored % Total % Cases dropped Cases with missing values % Cases with negative time 0.0% Censored cases before the earliest event in a stratum 0.0% Total % Total % An Analysis of Probation Violations and Revocations in Maine Probation Entrants in

19 Then age, sex, marital and employment status are included. The last models include demographics and criminal history/risk factors. In the initial models, a similar pattern of results is obtained. For example, compared to probationers aged 31-44, those aged have a higher risk of violation and those aged 45 and up have a lower risk of violation. Sex is also a significant predictor of the risk of violation, with males showing a 52% greater risk than females of violating with a felony or misdemeanor. Table 15- Variables in the Equation B SE Wald df Sig. Exp(B) Age 18 to Age 45 and up Male Married Interestingly, marital status does not appear to be a robust predictor of recidivism. It is not significant in any of the models (even that which only includes age and sex). This finding is surprising, given the recent attention paid to the importance of marriage as a predictor of desistance from crime (see Bersani et al., 2009; Laub and Sampson, 2003; Sampson et al., 2006). Employment (table 16) is a protective factor, reducing the risk of recidivism. Finally, age at first arrest and prior arrests are, as expected, positively related to the hazard rate of recidivism. Table 16 - Variables in the Equation B SE Wald df Sig. Exp(B) Age18 to Age 45 and up Male Married Employed Age at First Arrest Prior Arrest When the LSI total score is included in the model, age at first arrest and employment are no longer significant. An Analysis of Probation Violations and Revocations in Maine Probation Entrants in

20 Table 17- Variables in the Equation B SE Wald df Sig. Exp(B) Age18 to Age 45 and up Male Married Employed Age at First Arrest Prior Arrest Total Risk The LSI domains are similarly predictive of violations when examining misdemeanors and felonies. Here, only four domains are non-significant: financial, marital, accommodations, and leisure/recreation. In comparison to violations, the only differences are that accommodations are not significant and emotional/personal and attitude/orientation are significant predictors of risk of violation. In general, the story presented in table 17 suggests that the LSI domains in isolation are more related to violations that are misdemeanor or felonies than when violations are analyzed as a whole (including technical violations). Table 18 - Variables in the Equation B SE Wald df Sig. Exp(B) Age18 to Age45 and up Male Criminal History Education/Employment Financial Family/Marital Accommodations Leisure/Recreation Companions Alcohol/Drug Emotional/Personal Attitude/Orientation An Analysis of Probation Violations and Revocations in Maine Probation Entrants in

21 Section IV: Logistic Regression Survival Analysis: Examining the Factors that Influence Probation Revocation This section describes the results of our multivariate survival models predicting various types of probation revocations. Arguably, the most serious outcome that a probationer can experience is revocation of their probation status and return to jail/prison. This suggests that the probationer s violation upon release to the community was severe enough to warrant a new sanction and an adjudication that the probation has failed. Probation revocation is also an important assessment criterion for correctional programs and/or systems. As Grattet et al. (2008) argue, if one of the purposes of probation is to reduce prison crowding, probation revocations are important to avoid. Finally, Petersilia (2006) argues that revocation and return to prison based on decisions by parole/probation agencies rather than a traditional court process may conflict with expectations of procedural justice. In short, for a variety of reasons, it is necessary to scrutinize the causes and correlates of probation revocation in order to gain insight into why this outcome occurs. This information can be used to help ensure that Maine s correctional system can identify those most at risk to experience a revocation and to equip probationers with the tools they need to succeed. In the analyses that follow, we present descriptive and multivariate models to shed light on the factors related to probation revocation. In the multivariate models, we use logistic regression analyses, in which the outcome is a binary variable, scored 1 if the probationer experienced a revocation at any time during the study period. The predictors used include many of the same demographic, lifestyle and offense history variables used in the survival analyses. This analysis is particularly informative because previous research has suggested that the number of variables related to revocation is more limited than those which predict violations (Grattet et al., 2008). With respect to the full sample, table 19 shows that nearly half of the probationers experienced a revocation during the study period (n=2,345). However, individuals could incur more than one revocation. Summing the number of revocations, the data show that there were 4,691 revocations total. Thus, on average, each probationer that experienced a revocation had on average two revocations. Table 19- Descriptive Statistics N Minimum Maximum Mean Std. Deviation Revocation Valid N (listwise) 4725 An Analysis of Probation Violations and Revocations in Maine Probation Entrants in

22 Table 20 - Number of Revocations Frequency Percent Valid Percent Cumulative Percent Valid Total The multivariate results predicting revocation follow the analyses predicting time to violation. That is, we focus here on the first revocation for simplicity sake. Alternative strategies could examine the revocation as the unit of analysis and employ a multilevel or random effects model; however here we simply aim to provide a profile of individual level characteristics related to whether or not a person experienced a revocation. Thus, we employ multivariate logistic regression analysis. Our predictor variables are theoretically driven (see Grattet et al., 2008), and include demographic and risk factors. The models include additional independent variables in a stepwise manner such that model 1 in table 20includes only one predictor (age). As can be seen, age is related to revocation in a theoretically expected manner, with those aged having increased odds of revocation and those over the age of 45 having lower odds of revocation as compared to those aged As expected, males, those employed and marital status significantly predict revocations. For example, males have 51% greater odds of experiencing a revocation than females (in the model that includes only sex and age as independent variables Table 22). An Analysis of Probation Violations and Revocations in Maine Probation Entrants in

23 Table 21 - Age Predicting Revocation Variables in the Equation B S.E. Wald df Sig. Exp(B) Age18 to Age 45 and up Constant Table 22 - Age and Sex Predicting Revocation Variables in the Equation B S.E. Wald df Sig. Exp(B) Age 18 to Age 45 and up Male Constant Table 23 - Age, Sex, and Marital Status Predicting Revocation Variables in the Equation B S.E. Wald df Sig. Exp(B) Age18 to Age 45 and up Male Married Constant An Analysis of Probation Violations and Revocations in Maine Probation Entrants in

24 Table 24 - Age, Sex, Marital Status and Employment Status Predicting Revocation Variables in the Equation B S.E. Wald df Sig. Exp(B) Age 18 to Age 45 and up Male Married Employed Constant The next set of results incorporates age at first arrest and the number of prior arrests along with demographic variables. Again, the older an offender is at their age of first arrest decreases the likelihood of a revocation (2% decrease). Also, for each prior arrest, the risk of being revoked increases the odds by 9%. Again, the incorporation of these risk/criminal history variables have an effect on other variables: being married is rendered non-statistically significant. Table 25 - Demographics and Offense History Predicting Revocation Variables in the Equation B S.E. Wald df Sig. Exp(B) Age18 to Age 45 and up Male Married Employed Age at First Arrest Prior Arrest Constant An Analysis of Probation Violations and Revocations in Maine Probation Entrants in

25 The next model incorporates the LSI total risk variable along with demographic and criminal history variables. Again, age at first arrest and being married are not predictive of the risk of revocation when LSI (total risk) is in the model. A one unit increase in the LSI corresponds to a 9% increase in risk of revocation. Table 26 - Demographics, Offense History and LSI Predicting Revocation Variables in the Equation B S.E. Wald df Sig. Exp(B) Age18to Age45andup Male Married Employed Age at First Arrest Prior Arrest Total Risk Constant The analysis below includes LSI domains as predictors of any revocation. This analysis provides insight into which of the components of the LSI are related to risk of revocation. As is shown, criminal history, education/employment, accommodations, companions and drug/alcohol history are significantly related to the risk of violation. Of these, the criminal history domain is associated with the largest odds ratio (1.22). Table 27 - LSI DOMAINS Predicting Revocation Variables in the Equation B S.E. Wald df Sig. Exp(B) Criminal History ducation/employment Financial Family/Marital Accommodations Leisure/Recreation Companions Alcohol/Drug Emotional/Personal Attitude/Orientation Constant An Analysis of Probation Violations and Revocations in Maine Probation Entrants in

26 Table 28- Technical Violation Revocation Logistic Regression of LSI Domains Variables in the Equation B S.E. Wald df Sig. Exp(B) Criminal History Education/Employment Financial Family/Marital Accommodations Leisure/Recreation Companions Alcohol/Drug Emotional/Personal Attitude/Orientation Constant Finally, the LSI domains are similarly predictive of revocations following a misdemeanor or felony. In comparison to all revocation, only three variables are significant predictors of risk of revocation for a new crime: criminal history, education/employment and companions. Once again, the criminal history domain is associated with the largest odds ratio (1.19) of being revoked. Table 29 - Criminal Violations Revocation Logistic Regression of LSI Domains Variables in the Equation B S.E. Wald df Sig. Exp(B) Criminal History Education/Employment Financial Family/Marital Accommodations Leisure/Recreation Companions Alcohol/Drug Emotional/Personal Attitude/Orientation Constant An Analysis of Probation Violations and Revocations in Maine Probation Entrants in