Volume 73 Number 1
Federal Probation
Parole Violations and Revocations in California: Analysis and Suggestions for Action1

Ryken Grattet, Ph.D., University of California, Davis
Joan Petersilia, Ph.D., University of California, Irvine
Jeffrey Lin, Ph.D., University of Denver
Marlene Beckman, J.D., National Institute of Justice

Understanding California Corrections and the Importance of Parole Violations
Study Research Questions and Data
Data and Analytic Approach
What Predicts Parole Violations?
What Predicts Parole Revocations and Return to Prison?
Policy and Research Implications


THE DEBATE ABOUT the costs and benefits of imprisonment is taking place all across the United States, but the stakes are highest in California. California’s 173,000 prisoners constitute the largest prison population of any state. One in seven state prisoners in the United States is incarcerated in California, and between 1980 and 2007, California’s prison population increased over sevenfold, compared with a fourfold increase nationally. And, despite a 2003 vow by Governor Arnold Schwarzenegger to reduce the state’s prison population, California’s prison population continues to grow; recent projections predict a prison population of 191,000 in the next five years.2

California’s prison expenditures are also among the highest in the nation—per inmate, per staff, and as a share of the overall state budget. In 2006-2007, the average annual cost of housing a California prisoner in was $43,287, 1.6 times higher than the national average of about $26,000. At the beginning of the prison building boom in the early 1980s, adult and youth corrections accounted for four percent of California’s general fund expenditures at $1 billion per year. Today, California’s budget for state corrections is over $10 billion a year, and growing at a rate of seven percent annually—the fastest-growing segment of the state’s criminal justice expenditures. State correctional costs now account for approximately ten percent of total state spending—nearly the same amount the state spends on higher education. Even after adjusting for inflation, general fund expenditures to support California Department of Corrections and Rehabilitation (CDCR) operations increased 50 percent between 2001-2002 and 2008–2009.3

Central to California’s debate over its prison system are the topics of parole violations and revocations (returns to prison). In September 2005, the National Institute of Justice (NIJ) funded the authors to undertake a three-year comprehensive study of the causes and consequences of parole violations and revocations in California. The study was supported fully by CDCR, the agency that oversees all of California state corrections. Their cooperation was essential to access and understand the extensive data that our project required. The project represents the largest, most comprehensive, and most rigorous study of parole violations and revocations ever conducted.

back to top

Understanding California Corrections and the Importance of Parole Violations

California’s recidivism rate, as measured by the return-to-prison rate, is 66 percent, compared to a 40 percent national average. Sixty-six percent of all parolees in California returned to prison within three years, 27 percent for a new criminal conviction and 39 percent for a technical or administrative violation (which can result from new crimes or violations of the conditions of parole).4 On any given day, six out of ten admissions to California prisons are returning parolees.5

Part of the explanation for California’s anomalously high parole return rate is its unique sentencing and parole system. California, for the most part, has a mandatory parole release system. California moved from an indeterminate to a determinate sentencing system in the late 1970s, and as a result, most offenders are released after they have served their original court-imposed sentence, less any accumulated good time credit. California’s Determinate Sentencing Law (DSL) allows offenders to earn, with some exceptions, dayfor- day good time, which can result in a 50 percent reduction in the amount of time they must serve. For about 80 percent of prisoners in California, there is no appearance before a parole board to determine whether they are fit to return to the community; instead, they are automatically released. Once released, nearly all prisoners are placed on formal parole supervision, usually for three years. California is virtually alone in this practice of combining determinate sentencing and placing all released prisoners on parole. Most other states either have an indeterminate sentencing system, where a discretionary parole board determines release dates, or reserve parole for only their most serious or risky offenders.

The state’s growing prison population, combined with its universal parole practices and lengthy parole terms, has resulted in California supervising far more parolees than any other state. The Bureau of Justice Statistics reports that in 2007, California supervised about 120,000 parolees on any given day, accounting for 15 percent of all parolees in the country.6

California’s parole population is now so large and its parole agents so overburdened that parolees who represent serious threats to public safety often are not watched closely, and those who wish to go straight often cannot get the help they need. About 80 percent of all California parolees have fewer than two 15-minute face-to-face meetings with a parole agent each month, and nearly all of these meetings take place in the parole agent’s office. It is estimated that two-thirds or more of all California parolees have substance abuse problems and nearly all of them are required to be drug tested.7 Yet few of them will participate in appropriate treatment while in prison or on parole. California’s recent Expert Panel on Adult Offender Recidivism found that 50 percent of exiting state prisoners did not participate in any rehabilitation or work program, nor did they have a work assignment during their entire prison stay. Many did not get the help they needed on parole either: 56 percent of parolees did not participate in any formal program while under parole supervision.8

Clearly, this low level of supervision and service provision has not prevented crime. As noted above, two-thirds of all California parolees return at least once to a California prison within three years. Due to their high failure rate, parolees account for the bulk of prison admissions in California; in 2006, nearly two-thirds (64 percent) of all persons admitted to California prisons were parole violators. Parole revocations have been rising nationally over the last 20 years, but California’s have increased more. Over the last 20 years, the number of parole revocations has increased about six-fold nationally. In California, the number of parole revocations has increased 30-fold.9

California’s unique decision-making process partly explains its high parole revocation rate. The decision to send a parole violator back to prison is often not made by a judge, but by a politically appointed deputy commissioner at the Board of Parole Hearings. Criminologists have coined the term “back-end sentencing” to describe how the parole revocation process centers on parole board practices.10 Not only are back-end sentences determined by correctional officials instead of judges in California, but the standard of evidence used (preponderance of the evidence) is much lower than is required in a court of law (beyond a reasonable doubt). This more lenient standard is deemed appropriate because in California, prisoners remain in the legal custody of the CDCR while on parole. Parole in California is not a reward for good behavior, as it might be in an indeterminate sentencing state, but rather an extension of a felon’s sentence and a period of extended surveillance after prison. As such, if parolees do not abide by the imposed parole conditions, the state has the legal right to revoke their parole terms and return them to prison.

California’s parole revocation process is also unique in another way. The maximum term for a parole violation in California is 12 months in prison. If a parolee is sentenced to that maximum term, there is usually a day-forday credit for time served in prison or in jail awaiting case disposition, assuming no prison rule infractions. This means that a parole violator who is not convicted of a new crime by a criminal court—totaling nearly 70,000 prison commitments in 2006—will only spend, on average, slightly more than four months in custody. Naturally, not everyone receives the maximum 12-month sentence; California’s Rehabilitation Strike Team found that of all parolees returned to a prison in 2004, 20 per-cent (one in five parole violators) served less than one month in a California prison.11

This system of “catch and release” makes little sense in terms of deterrence, incapacitation, treatment, and cost. Parolees quickly learn that being revoked from parole does not carry serious consequences, undercutting the deterrent effect of serving prison time; the resources of the police, the parole board, and parole officers, who have to reprocess the same individuals over and over again, are also wasted. The constant churning of parolees also disrupts community-based treatment, since parolees who are enrolled in community treatment programs are constantly having that treatment disrupted for what, in the treatment providers’ views, are predictable and minor rule violations (e.g., testing positive for drug use). Churning also encourages the spread of prison gang culture into the communities where inmates are discharged. Finally, given California’s overcrowding crisis in prisons, there is the high opportunity cost of occupying a limited number of prison beds that, in some cases, could be used for offenders who pose a greater risk to public safety.

Policymakers and practitioners agree that an overhaul of California’s parole system is urgently needed. In fact, more than a dozen reports published since 1980 have recommended changes in California’s parole revocation procedures.12 Unfortunately, California’s parole whaviolation process is so complex and involves decisions by so many parties, including the police, prosecutors, judges, parole agents and parole board commissioners, that it is unclear exactly what needs to be done to fix the problem.

back to top

Study Research Questions and Data

To better understand the complexities of the parole violation process and the characteristics of parolees who are returned to prison, we needed to unpack the “black box” of the parole violation and revocation process. We needed to study not only characteristics of parolees, but also characteristics of the supervising agency, parole agents, and the communities to which parolees return. We needed to identify the key decision points that ultimately lead to parole revocation and prison returns, and also how characteristics of the parole agent, caseload type, and variations in community characteristics impact the processes of violation and revocation.

We also had to better understand the critical role of Board of Parole Hearings (BPH), which has the ultimate responsibility for deciding which parole violators are returned to prison and which are allowed to remain in the community. In the mid-1990s, California adopted a “zero-tolerance” policy for “serious” and “violent” parolees (as defined in the Penal Code), such that parole agents are required to report every offender originally convicted of these crimes who violates any condition of parole to BPH for disposition. BPH is a politically appointed body with a history, especially in recent years, of returning to prison most parolees who come before it. BPH may be the most important gatekeeper of using prison for the sanctioning of parole violations, and yet its role and impact have gone virtually unnoticed and unstudied.

We assembled an extraordinarily large and complex database that tracked every adult on parole in California at any point during the calendar years 2003 and 2004. The resulting study sample consisted of 254,468 individuals. These parolees were responsible for 151,750 parole violations that made it to the court or BPH hearing level (thousands more were terminated at the parole unit level) over the two study years. These parole violation and revocation incidents were a central focus of our study. In addition to recording the details of each parolee’s behavior on a weekly basis during the two-year study period, we also merged data about each parolee, reflecting their personal characteristics and criminal histories, the nature and types of supervision to which they were subjected, the characteristics of agents who supervised them, and (using their addresses) the communities to which they returned. Using other statewide and national databases, we then collected information on “host” communities (e.g., services available, demographic and political characteristics of residents). The combined database allowed us to analyze the way in which three clusters of factors—characteristics of the parolee, the agency, and the community—interact to produce variations in parole outcomes.

We also investigated the major aspects of California’s sentencing and parole system that we believe impact parole revocations and prison returns. Because California releases nearly all prisoners subject to the Determinate Sentencing Law (DSL), with no opportunity to retain even the most likely recidivists, and then places all of them on parole supervision, the state’s parole agents end up supervising some individuals who pose a far more serious threat to society than the typical parolee in a state with discretionary release policies.13 In states that use discretionary release, these high-risk prisoners can be denied parole and kept in prison. Parole officers in California often point out that the high revocation rates are caused by the behavior of parolees who were almost certain to reoffend and should not have been released from prison in the first place.

On the other hand, since California law allows minor technical parole violators to be returned to prison (whereas some states do not), and these prisoners are also eventually released to parole supervision, California parole caseloads also include many less serious offenders as well. This point is critical to understanding parole violations in California: California parole caseloads likely contain an unusually high proportion of offenders at both extremes of the seriousness continuum— offenders who probably would not be on parole in other states, either because they are too serious to have been released from prison in the first place by parole boards operating in indeterminate states, or because they are such low-risk offenders that they would not have been assigned to post-prison parole supervision at release.

California’s DSL not only changed the way in which prisoners automatically got released and required all prisoners to serve a postprison parole term, but it also simultaneously and significantly increased both the length of the initial parole supervision term imposed and the length of the prison term that could be subsequently imposed if the parolee violated parole conditions. Before the passage of the law in 1977, prisoners released to parole were subject to a one-year period of parole. But DSL tripled the length of time on parole for most prisoners. Equally important, DSL also doubled the length of prison time that can be imposed upon parole revocation from six months to one year.14 And under California law, when a person is returned to prison for a parole violation, the “clock stops” on the time owed for parole supervision. So, when a person leaves prison after serving time for a parole violation, he still faces the remaining supervision time he owed the state before he went back to prison for the violation. In this way, parole supervision can stretch out for years for some individuals. Offenders often call it “doing a life sentence on the installment plan,” since they go in and out, never able to formally discharge from parole supervision.

In addition to changes in sentencing policy and the structure of parole in the last decade, the discretion held by California parole agents in the handling of violations has substantially eroded. In 1994, BPH implemented new regulations, referred to as the “Robin Reagan rules,” that significantly added to the list of parole violations the Division of Adult Parole Operations (DAPO) is required to refer to the parole board, thereby exposing more parolees to BPH decisions to return them to prison. These regulations were adopted as a result of a heinous murder committed by a parolee.

Whereas parole agents and supervisors once wielded discretion about how to handle many violations, now much of that authority has shifted to BPH. DAPO estimated that 85 percent of parole violations, including technical violations, were subject to mandatory referral policies in 2005. This means that parole agents and their supervisors have very little discretion in the handling of these cases and these offenders. BPH makes a decision about whether to return the parole violator to prison, and the vast majority of cases that go before BPH result in a return to prison. In 1993, about 65 percent of parolees referred to BPH for parole violations were returned to prison and 35 percent were continued on parole. By 2007, however, about 90 percent of parolees were returned to prison by BPH and only 10 percent were continued on parole. Whether these mandatory referral rules are appropriate or not is a political determination, but one thing is clear: parole agents, parole supervisors and DAPO retain discretionary decision-making power over a declining percentage of violations.

Like the role of discretion in California’s sentencing system, discretion in parole has shifted from corrections professionals to legislative and regulatory bodies that are politically elected or appointed. This change has occurred with virtually no discussion or public input, but the consequences are critically important. For one, it means that the extent to which changes in parole agent recruitment, training, or culture can reduce the number of parolee returns to prison is often overestimated. The parole agent recommends the disposition for the violation (e.g., to prison or not), but ultimately, the parole board has the sole authority to return a parolee to custody. These and other legal and procedural constraints are important to understanding the very complicated processes of prison release, parole supervision, and all too often, return to prison.

The growth of California’s prison population—combined with the policy of placing all exiting prisoners on parole supervision for three years, simultaneously reducing the discretion of parole agents to handle minor violations for an increasing proportion of parolees, and increasing the prison time served for violations—provides the requisite conditions for the growing contribution of parole violators to the state prison population. No other state has created this hybrid system (shifting simultaneously to fixed-term prison release and universal parole supervision) while at the same time reducing parole agent discretion and lengthening parole terms and prison terms upon revocation.

Our hope is that the empirical data analyzed in this report will permit policymakers in California to devise sounder parole supervision and revocation policies that better balance public safety and public resources. Importantly, such research should help advise policymakers on the “seriousness” of parole violators being returned to prison, which in turn can greatly influence the prison capacity discussion.

back to top

Data and Analytic Approach

Our statistical models separately investigate the prediction of parole violations, which are largely behavioral events, and the prediction of parole revocations, which reflect system responses to that behavior. We relied solely on official records rather than offender self-reports, even though we recognized that not all (or even most) parole violations came to the attention of authorities. Our database consisted of detailed information about every adult on parole in California at any point during 2003 and 2004. The resulting sample comprises 254,468 separate individuals. Some individuals were already on parole at the start of our study (January 1, 2003), whereas others were either free or in prison at the start of our study, but were released to parole at some time during the two-year study period. Study subjects were observed for two years (January 1, 2003 to December 31, 2004). However, since many subjects were already on parole at the start of the study period, we were able, through various methods of statistical estimation, to analyze violation and revocation patterns over longer periods of time.

We assembled a detailed personal and parole supervision profile for each parolee in the sample, consisting of their demographic characteristics and criminal records, the type(s) of parole supervision to which they were assigned, and all new technical and criminal recidivism events that occurred during the study period. We also recorded information about each parolee’s supervising parole agent (e.g., age, race, gender, job tenure) and, using the parolee’s address, characteristics of the community to which the parolee returned upon release from prison. Data were merged from over a dozen different state and national databases to create as comprehensive a profile as possible for each subject.

With the databases assembled and merged, we were then able to conduct our statistical analyses. In terms of parole violations, we structured the data for survival analysis—a multivariate method that examines both the likelihood and timing of violations. Violations were tracked on a weekly basis for each parolee throughout 2003 and 2004. The data format allowed us to construct multivariate survival models predicting the likelihood and timing of different types of violation behavior.

For our analyses of parole revocations, we created a dataset documenting every parole violation case heard in criminal court and/ or by BPH. These data, reflecting 151,750 violation reports, were used to estimate logistic regression models predicting revocation outcomes of interest—specifically, whether criminal violation cases were successfully prosecuted in court as opposed to being referred to the parole board, and whether cases heard by BPH were returned to prison or continued on parole. Thus, we were able to assess the relative impact of individual, organizational and community-level measures on numerous parole outcomes. In all of our analyses, we investigated the likelihood (i.e., probability) as well as the severity of the outcome. In addition to the administrative data we compiled, we also collected extensive qualitative information from field observations, staff interviews, and reviews of agency directives and policy memos.

back to top

What Predicts Parole Violations?

Nearly half (49 percent) of the parolees in our sample had at least one formal parole violation report during our study period, and 24 percent had multiple parole violation reports. Each report could contain multiple violations of any type (e.g., criminal, technical). Together, these parolees were responsible for 296,958 violation reports. CDCR tracks 247 different types of prohibited parolee behavior, ranging from violations of the parole process, usually referred to as technical violations, to serious and violent criminal offenses like robbery, assault with deadly weapons, and homicide.

Over a third (35 percent) of all the recorded parole violations were for noncriminal, or “technical,” violations. Two-thirds of technical violations were for absconding supervision, meaning that the parolee missed an appointment and/or his or her whereabouts were unknown. Other technical violations include weapons access, psychological endangerment, and various violations of the parole process such as violations of special conditions of parole imposed by a parole agent or deputy commissioner. Interestingly, the parole violation reports that pertained to drug use or drug sales (over 110,000 of them) accounted for over a third of all parole violation reports (37 percent) during our study period.

Two-thirds (65 percent) of all parole violations were for criminal behavior. Thirty-nine percent of the criminal violations were classified, according to the CDCR’s internal coding system, as Type I (the least serious—mostly drug use and possession); 17 percent were classified as Type II (moderately serious—e.g., forgery, drug sales, burglaries, battery without serious injury, driving violations); and 10 percent— nearly 29,000 violation reports—were classified as Type III (the most serious—e.g., major assaults, major drug crimes, robberies, rapes, and homicides).

In terms of the timing of violations among parolees in the study, the risk of all types of violations was highest during the first 180 days following release from prison, and declined thereafter. A major factor behind this declining risk pattern was that the most risk-prone parolees tended to violate early and be returned to prison. We theorized that the remainder were probably more compliant, less likely to violate, and more likely to successfully complete their parole period. Indeed, after 360 days on parole, a “surviving” parolee’s risk of violation had dropped 70 percent from what it was during the first two months of parole. From 360 to 900 days, a parolee’s risk only dropped another 10 percent. In other words, after about 360 days, a parolee’s risk of violation, while not zero, had substantially leveled off.

In terms of demographic and other personal characteristics, the youngest parolees (ages 18 to 30) posed the greatest risk of all kinds of violations except Type I criminal violations (the least serious). Male parolees posed significantly higher risks for all types of violations except absconding. Black parolees posed the same risks as nonblack parolees for technical violations but much greater risks than parolees from other racial backgrounds for the most serious and violent criminal violations. Parolees with a record of mental health problems had higher risks for all types of violations, and they had particularly elevated risks for the most violent criminal violations.

The best predictor of a parolee’s violation risk was the number of prior adult prison incarcerations in California. For all violation types, an offender coming out on his or her second release from prison had a 20 percent higher risk of violation than an offender on his or her first release. After a third release, an offender had a 39 percent higher risk of violation than an offender on first release. By the ninth release, an offender had a 124 percent higher risk of violation than an offender on first release.

In general, the extent of prior criminal record had more predictive value than the seriousness of prior record, but certain “seriousness indicators” did exhibit relationships to violation risk. Age at first adult commitment to a California prison, for example, predicted Type III (the most serious) criminal violations. For every additional year older a parolee was at his or her first prison commitment, the risk of a Type III violation decreased by 2.5 percent. However, parolees who were older when first committed to California prisons tended to present higher risks for technical violations and Type I criminal violations. This latter group may have been largely composed of drug offenders who had substance dependence driving their offending, and as a result of drug use, were prone to generating technical and Type I criminal violations, but were less likely to be involved in more serious criminal behavior.

The seriousness of the current commitment offense, while exhibiting a relationship to violation risk, did not predict violations in the ways that policymakers often assume. Parolees committed for violent and sex offenses, overall, had lower risks for most violations than those offenders committed for property and drug crimes. However, those who had been committed for violent offenses did show elevated risk for violent criminal violations and serious sexual violations.

Sex offender registrants posed lower risk for violations than other types of offenders for several types of violations (e.g., having any violation, absconding, Type I criminal violations). Sex offender registrants were no more likely to commit the most violent violations than other offenders.

Policymakers are particularly interested in the threat that paroled sex offenders pose to their communities, so we investigated these outcomes separately. We found that sex offender parolees were significantly more likely to be violated for sex crimes, but it is critical to note that these sexual violations were very rare—during the study period, reoffending sex offenders accounted for 1.5 percent of all violations and in about twothirds of the cases, the offenses were victimless (i.e., the paroled sex offenders failed to register as required by California Penal Code section 290). The majority of sexual violations, including the most serious violations involving rape, sexual assault, and child molestation, were committed by parolees who were not registered sex offenders. Setting aside the violations involving failure to register, of the 1,528 sexual violations committed during 2003 and 2004, just 25 percent were committed by sex offender registrants. The vast majority of sexual violations, including 78 percent of the most serious Type III sexual violations, were committed by paroled offenders who had not been previously sentenced for sex-related crimes.

Intensity of Supervision, Parole Agent Characteristics and Parole Organization

California parolees are assigned to one of five levels of supervision, with the assigned level determining the frequency and degree of oversight provided by parole agents. Twentythree percent of parole supervision performed during 2003 and 2004 was classified as “minimum service,” with the requirement that parolees see their parole agents only twice a year. Most contact between agents and parolees under minimum service supervision occurs through the mail; that is, parolees periodically mail a postcard to their agents to check in. Another 43 percent of supervision during our study period was classified as “control service;” parolees supervised at this level see a parole officer once every six weeks. These two classifications—in which relatively little supervision or programming is actually applied to parolees—accounted for 65 percent of the total supervision applied to parolees in 2003 and 2004.

Given that these offenders are placed in low-risk categories because they are not expected to be likely recidivists, a question to be considered is whether the effort expended to provide cursory oversight to so many former inmates is an effective use of resources. This issue is particularly pressing because California loses track of so many of its parolees; perhaps greater intensity of supervision or services for higher-risk parolees could help prevent new crimes, or the resources expended on low-risk parolees could be better used to locate those whose whereabouts are unknown. On any given day, nearly 17 percent of all California parolees—more than 20,400 people—are “parolees-at-large,” meaning they have absconded supervision. This is the highest rate of absconding in the nation and is far above the national average of seven percent.15

We found that, consistent with prior research, supervision intensity affected the risk of reported violations.16 More intensive parole supervision increased the risk of all violations, holding constant the offender’s personal attributes, offense background, and community conditions. The biggest differences in the effects of supervision on violation risk were found between minimum service supervision and active supervision (i.e., supervision at all other levels). Parolees who were on minimum service caseloads, which involved infrequent face-to-face or collateral contact, monthly mail correspondence, and no narcotics testing, had significantly lower risks for all kinds of violations than those parolees who were more actively supervised.

The differences in violation risk between parolees on minimum service supervision and active supervision were most pronounced among the most discretionary violations—technical violations not involving absconding and Type I criminal violations (the least serious, mostly involving drug use and possession). Compared to minimumservice parolees, actively supervised parolees had between two and three times the risk of technical and Type I criminal violations. Active supervision parolees also had consistently higher risks of absconding, Type II, Type III, and violent criminal violations, although the differences were not as great as among the more discretionary violation types. What became clear from the contrast between parolees on active supervision and those on minimum supervision was that more closely supervised parolees did not seem to be deterred from engaging in behavior that could result in parole revocation.

We also detected differences in violation risks among active supervision categories, but these differences were not as pronounced as those between active supervision and minimum service supervision. In general, parolees in more intensive supervision categories posed higher risks for violations.

California is subdivided into four parole regions, each supervising roughly one-fourth of the California parole population. The regions are understood to have differences in their organizational cultures and in the types of parolees they supervise. Region 3, which comprises Los Angeles County, is perceived to be the most overstretched part of the parole system, overseeing the supervision of the most serious parolees in the state. As a result, some believe that there is a lower rate of reporting of less serious violations in Region 3, as they have more serious criminal violations to contend with.

We found little support for regional differences in parole outcomes. Once the characteristics of parolees and communities were statistically controlled, Region 3 reported violations in a manner similar to the other three regions. Region 3 did report fewer drug use and possession violations (Type I criminal violations), but its reporting patterns for technical violations—both absconding and violations of the parole process—were no different from those of other regions. Nor was the risk that a parolee in Region 3 would be cited for a Type II or Type III criminal violation different from that of other regions.

We also explored the extent to which differences in parole outcomes were traceable to parole agent characteristics, and were able to detect some relationships between agent characteristics and violation risk. We found that female agents (who performed 28 percent of parole supervision during 2003–2004) appeared to exercise discretion in ways more “forgiving” of low-level criminal violations (i.e., Type I, mostly drug use and possession). Male agents, on the other hand, appeared to adopt a more lenient approach toward absconding than female agents. No gender differences were found in the reporting of the more serious Type II and III criminal violations.

Some research on black judges and black police officers suggests that, as a group, black parole agents might have more tolerance for violations.17 During our study, 32 percent of all supervision was done by black agents, 25 percent by Hispanic agents, 35 percent by white agents, and the rest was performed by Asian agents and those from other racial categories. We might theorize that black agents, like blacks in the rest of American society, may be more likely to have friends or family members who have had contact with the criminal justice system.18 As a result, they might be more sensitive to the conditions that foster criminal behavior and more wary of the effectiveness of system responses. Therefore, black agents might show more tolerance for less serious violations. Whether or not the theorized explanation is correct, our results supported this result. Parolees supervised by black agents had lower risks of technical violations and Type I criminal violations. But parolees with black agents were no different from other parolees in terms of their risks for Type II and III criminal violations.

Nearly half (48 percent) of parole supervision during 2003–2004 was done by parole agents with less than three years of job experience as a parole agent. Thirty percent of supervising agents were under age 40 and 83 percent of agents had previously worked in a CDCR correctional institution. We were told that older agents and those who have not worked in the prison system as correctional officers are more likely to see “shades of gray,” and thus tolerate some parolee behavior that other agents would elect to violate. Contrary to expectations, parolees assigned to agents with prior employment experience in a prison actually had an eight percent lower risk of the least serious Type I criminal violations than parolees assigned to agents with no prior prison employment. Prior employment in a prison did not affect the risks for any other type of criminal or technical violation. Moreover, neither parole agent age nor tenure on the job as a parole agent was significantly related to any type of criminal or technical violation.

CDCR announced a number of significant parole policy changes during 2003 and 2004. One policy, referred to as the “New Parole Model,” was announced with much fanfare in February 2004, before being scaled back significantly in April 2005. The New Parole Model proposed the greater use of intermediate sanctions for parole violators and the adoption of a parole violation matrix to standardize the handling of violations. We found no evidence that this announced policy change had any observable impact on parole decision-making or case processing outcomes at the aggregate level.

Community Conditions and Reentry Environments

Research suggests that community characteristics can have criminogenic or reintegrative effects on parolee behavior. In other words, neighborhood factors can either promote or discourage illicit activities. The following are common hypotheses made about the relation between neighborhood factors and parolee behavior:

  • Communities with greater financial resources may be able to fund more rehabilitation and work programs, which can provide parolees with pathways out of criminal lifestyles.
  • Communities with more progressive political views may have more tolerance for minor rule violations.
  • Less socioeconomically disadvantaged communities may provide better informal social supports that suppress criminal activity (i.e., increased residential stability)
  • On the other hand, socially disorganized (i.e., disadvantaged) communities may not be able to fund many alternatives to prison, and may exhibit other conditions that are conducive to criminal behavior.
  • Politically conservative communities may have less toleraehance for illicit bvior, and may exhibit an increased propensity to violate parolees.

To explore these ideas, we used parolee address records to link individuals to data about their communities. We mapped parolee addresses to U.S. census tracts to compile measures of poverty, unemployment, and public assistance. As a measure of service availability in parolees’ reentry environments, we drew data from the United States Substance Abuse and Mental Health Services Administration, listing the addresses of all substance abuse and mental health treatment providers in California that accept clients from criminal justice agencies.

We found modest support for the above hypotheses as they relate to an understanding of parole violations in California. Parolees who lived in neighborhoods that scored highly on socioeconomic disadvantage were at greater risk to abscond than parolees who lived in less disadvantaged environments. However, parolees residing in disadvantaged neighborhoods did not pose a greater risk to commit other kinds of violations than those from less disadvantaged neighborhoods.

Importantly, we found some evidence of the correlation between substance abuse and mental health treatment services on the one hand and lower risk of Type I (the least serious) criminal violations—which mainly involved drug use, drug possession, and misdemeanor violations of the law. If there is a cause/result relationship, it may be attributable to the effectiveness of these programs, but it may also have been due to a “parole agent effect”; that is, parole agents may have been less likely to violate parolees for lowlevel violations when they perceived that there were program opportunities that presented alternatives to initiating the formal violation and revocation process. Given that there were few alternatives to prison during 2003- 2004, this observed effect is important and might be strengthened if more programs were in existence.

back to top

What Predicts Parole Revocations and Return to Prison?

There are two ways parolees experience revocation: through county criminal courts and through the parole board (BPH). Courts only handle criminal violations—those that result from an arrest by a police officer or parole agent. BPH handles technical violation cases, as well as criminal violation cases that county courts do not successfully prosecute. The process by which cases are sorted through one venue versus the other, as well as the reasons that some parole violators are returned to custody while others are allowed to remain in the community, are not well understood.

The parolees in our study sample generated 151,750 parole violation cases in 2003 and 2004 that were processed either through the criminal court system or through the BPH. Eighty-four percent (127,742) of these cases involved new criminal violations. These criminal violation cases were heard first in criminal court; if a conviction could not be obtained in court, they were referred for assessment by the parole board. Sixteen percent (24,008) of all cases involved only technical violations, and these cases were heard by the parole board. Importantly, the board operates under a more lenient standard of evidence than the courts, and may return a parole violator to prison for no more than 12 months.

Of the 127,742 criminal violation cases reported during 2003 and 2004, 25 percent (31,417 cases) resulted in a new prison term delivered in criminal court. The other 75 percent (96,325 cases) were referred to the parole board. Among these referred criminal violation cases, the board elected to return 73 percent to prison. Not surprisingly, more serious criminal charges were more likely to result in a prison return. Type III criminal violation cases—the most serious—resulted in prison return 88 percent of the time. Moderately serious criminal cases (Type II) resulted in return almost as frequently; these parole violators were returned 80 percent of the time. The least serious criminal cases (Type I) only resulted in return to prison 52 percent of the time. Thus, when moderately serious and very serious criminal parole violations are evaluated by the board, the certainty of return is extremely high. The board appears to exercise greater discretion over cases involving Type I crimes—most of which involve drug use and possession violations.

A small but significant number of violent crimes such as homicide, robbery, and rape were processed through the parole board.19 These crimes carry lengthy prison terms when they are prosecuted in courts of law. However, when handled through the parole board, the maximum return time is capped at 12 months. Even though the proportion of homicide, robbery, and rape cases constituted a very small share of the total number of criminal parole violations returned to custody through the board, the fact that such cases were pursued in this arena is significant. The board was clearly not a venue that exclusively dealt with “smalltime” criminal cases. Further, because the board operates under a more lenient standard of evidence, there is a greater possibility that factually innocent criminal parole violators might be returned to custody.

Adding together the criminal violation cases that resulted in a new term through criminal courts and those criminal violation cases that resulted in a return to prison through the parole board, we found that among the 127,742 criminal violation cases officially recorded in 2003 and 2004, over three-quarters (77 percent) resulted in some form of prison return, either through the courts or through the board.

In addition to criminal violation cases referred from courts, the parole board also heard 24,008 technical violation cases (16 percent of all cases)—many of which involved absconding. Like criminal violation cases, technical violation cases heard by the parole board exhibited a high rate of prison return. About 85 percent of these technical violation cases resulted in a return to custody. Those cases involving technical charges (without absconding) were returned 79 percent of the time. Cases involving absconding (without other technical charges) were returned 85 percent of the time. Cases involving both technical and absconding charges were returned 91 percent of the time. Overall, the board returned 75 percent of all violation cases it heard.

We next turned to understanding the patterns and logic of the parole revocation process through multivariate statistical analyses (logistic regression). Our analysis was designed to answer two interrelated questions: What factors affected the sorting of violation cases through the courts versus through the parole board, and, once in front of the parole board, what affected the chances that a parolee would be returned to custody, as opposed to being continued on parole? As with our multivariate analysis of parole violations, we examined how parolee characteristics, organizational factors, and community characteristics correlated with and may have impacted revocation decision-making. We also investigated the relationship between case characteristics— such as the number and severity of violation charges—and revocation outcomes.

Case and Individual Characteristics

Violation case characteristics were critical to determining whether or not criminal violation cases were processed through criminal courts or through the parole board. They also influenced whether a case processed by the parole board resulted in a return to prison. As expected, cases involving more charges, and more serious charges, were likely to receive harsher treatment. In court decisions, the number of criminal charges contained in a case was not related to decisions to reimprison, but the severity of those charges did predict court sanctioning decisions. Board decisions were, for the most part, driven by both the number and severity of charges involved in violation cases.

In terms of individual factors, parolees with longer, and more serious, histories of criminal behavior were likely to be considered public safety risks by court and board decision-makers, and their cases were treated accordingly. Irrespective of the seriousness of their current parole violations, parolees’ histories of imprisonment, for example, were significantly predictive of harsher treatment in both decision venues. Those who had served more adult prison spells (both for new court-ordered terms and returns to custody on parole violations) in California were more likely to be sent back to custody by both the court and the board. Parolees on their “second strike” were also significantly more likely to be returned through the court than parolees without such status, and when their cases were referred to the board, they were significantly more likely to be re-incarcerated in cases involving criminal violations.

Statutorily-defined serious and violent offenders were actually less likely than others to experience court return to prison, but when their criminal violation cases were referred to the parole board, they were more likely to be returned to custody. Similarly, registered sex offenders were less likely than others to be returned to prison through court, but they were treated more severely by the board. One explanation for these findings is that the criminal violation cases of serious and violent offenders, as well as sex offender registrants, may have been unappealing to court decision-makers because they tended to lack compelling evidence. However, court decision-makers may have also referred these cases because they felt that the board, using a lower standard of evidence, could act quickly and decisively to reincarcerate parolees who were perceived as particularly threatening to public safety. The board sanctioned these types of parolees especially severely in lowlevel (Type I) criminal violation cases—the type allowing for the most discretion. It appears that low-level criminal activity, much of which is detected through parolee drug testing, was a crucial mechanism by which the parole board reincarcerated “high-profile” parole violators. Note that the criminal courts could not legally impose very harsh sanctions for these low-level crimes, and so they seemed to opt, through case referral, for the greater certainty of punishment that the board was able to provide.

Demographic characteristics were also somewhat predictive of case outcomes. Parolee age affected criminal court decisions, but not board decisions. Courts were inclined to prosecute the criminal violation cases of the youngest parolees (ages 18–30). Black parolees were more likely to have their cases referred to BPH—the more discretionary venue—and when their cases were heard by BPH, they were more likely to be incarcerated for criminal violations. Asian and Hispanic parolees were the most likely to be successfully prosecuted in criminal court, and Hispanics were also among the most likely to be returned to custody in criminal violation cases. White parolees, who had the lowest likelihood of court conviction, also had the lowest likelihood of return through BPH for criminal violation cases (although they were among the most likely to be returned when they absconded). These findings suggest that there may be observable or unobservable traits associated with parolees of different demographic groups that affect their case outcomes.

Organizational Factors

Over and above case- and parolee-specific characteristics, organizational factors also affected revocation decisions by the court and the parole board. Los Angeles County (Region 3) appeared distinct in its treatment of parole violators. Criminal violations in Los Angeles were more likely to result in reincarceration through the court. Board decisions were also uniquely patterned in Los Angeles. Technical parole violators were more likely to be returned to custody by the board in this region, while absconders were less likely to be returned by the board. These findings could have been due to many factors: differences in organizational culture across parole regions, unmeasured variation in local parolee populations that affected sanctioning decisions (e.g., gang affiliation, addiction and employability), or the effectiveness of policing practices in different regions.

Practical constraints on decision-making also appeared to play a role in violation case outcomes. A key practical constraint was available custodial space. We found that when available space in prison reception centers decreased, for example, the parole board was more likely to continue cases on parole, as opposed to returning parolees to prison. Moreover, in courts, workload pressures (measured as the ratio of felony cases to district attorneys in each county) were linked to an increased likelihood of case referral to the parole board. As felony court caseloads increased, courts were inclined to refer more criminal parole violation cases.

Community Factors

To generate a county-level measure of the “punitiveness” of different communities, we collected information on the results of ballot proposition voting and party registration from the Secretary of State. We selected data reflecting voting patterns of ballot propositions that pertained directly to state correctional practices—for example, Proposition 36, which allows some nonviolent drug offenders to receive treatment instead of incarceration for parole violations, and Proposition 66, which proposed a scaling back of California’s “three strikes” law. Our hypotheses were that community conditions and attitudes, as well as the availability of treatment, would be related to parole practices. Our statistical models showed that, net of all other measured factors, some characteristics of parolees’ communities were related to the treatment of parole violations in court and before the parole board. For example, more “punitive” counties—as measured by political party affiliation and electoral ballot voting—were more likely to return criminal parole violators through the court, and in violation cases heard by the board, these counties were more likely to return parolees to prison, regardless of whether the case involved a criminal violation, absconding, or other technical violations.

Community characteristics can also serve as cues to decision-makers that reflect something about individual parolees themselves. The extent of “racial threat” in a community, which has been examined in prior sentencing research, is illustrative of this point.20 Racial threat refers to the hypothesis that sanctioning officials may be sensitive to the prevalence of threatening minority groups in communities, and therefore punish offenders from these communities more harshly. Census tracts with higher proportions of black residents, and those with higher black unemployment rates, may be perceived as particularly unstable or crime-ridden, and parolees that live in these communities may be penalized by decision-makers because they come from, and are therefore representative of, these disadvantaged environments. In our models, parolees who came from communities that had a higher proportion of black residents, and higher black unemployment rates, were more likely to be returned by the court with a new term, as opposed to being referred to the parole board. When their cases were heard by the parole board, these parolees were generally more likely to be returned to prison, especially for criminal violations.

However, while community characteristics can have a stigmatizing effect on case outcomes, they can also have the opposite effect. For example, census tracts with more mental health and substance abuse services in close proximity were associated with more lenient outcomes among criminal violation cases and technical violation cases (not involving absconding) decided by the parole board. This may have been due to the fact that decisionmakers had more treatment options in these communities, and therefore more opportunities to keep parole violators out of prison, or that parolees from service-rich communities somehow appeared less threatening than parolees from communities that lack services.

A central implication of our analyses of revocations is that the response of criminal justice institutions does not totally derive from, and is not necessarily proportionate to, the extent of parolees’ criminal behavior, as is often assumed by policymakers, government officials and the public. While case characteristics matter in terms of court and board outcomes, so too do the characteristics of the individual, the organizations handling that individual’s case, and the community that the person comes from.

back to top

Policy and Research Implications

Our findings suggest a number of policy and research implications, the most important of which are:

  1. Concentrate supervision and services on the first six months. Parole should frontload services and surveillance to focus on a parolee’s first six months after release, when the risk of recidivism is the highest.
  2. Expand use of early and earned discharge. Parolees are most at risk of all kinds of violations during the first six months on parole. Parolees that make it to the sixth month without violation pose significantly lower risks than parolees who do not. The duration of the imposed parole term should be closely linked to an offender’s risk level or accomplishment of individual benchmarks. Low-risk offenders might not be assigned parole supervision at all, or those who adjust well to parole could be released after six months of supervision. Moderate-risk offenders might be assigned a year or two of parole, whereas high-risk offenders might serve two years or more, and very high-risk offenders might be assigned lifetime parole.
  3. Align parolee risk and supervision levels. Parole services and surveillance should be primarily risk-based rather than offensebased. CDCR needs to assign parole caseloads and supervision levels so that offenders are matched to types of surveillance and services that are most appropriate for them. Resources should be more heavily focused on higher-risk parolees, and very intensive (and expensive) programs should be reserved for those whose risk and need profiles suggest they will likely benefit from program participation.
  4. Employ a parole violation matrix. The parole division and the parole board should adopt policy-driven approaches to parole violations using a decision-making matrix and graduated community-based sanctions. This tool would allow parole officials to respond consistently to parole violations, using a well-developed range of intermediate sanctions. The response should reflect the original risk level of the parolee coupled with a proportionate response to the seriousness of the violation. Every major study on California’s prison system published since the 1980s has recommended the use of such a tool, but it has never been implemented, even though such instruments are used in over 20 other states.21 California is currently developing such an instrument and plans to pilot test it in winter 2009.
  5. Expand intermediate sanctions options. CDCR should implement additional intermediate sanction programs, particularly for drug-involved parolees. Current program offerings are woefully inadequate for appropriately dealing with the wide range of parole violations. CDCR cannot do this alone, as the most effective reentry programs and intermediate sanctions require community engagement and collaboration. The expansion of evidence-based intermediate sanctions should both reduce recidivism and save expensive prison beds for the most violent criminals.
  6. Encourage criminal prosecution. Parolees who commit new crimes should be prosecuted in criminal courts whenever possible. California’s “back-end sentencing” system allows some very serious criminals to evade the more severe criminal penalties that would have been imposed had their cases been criminally prosecuted as opposed to handled by the parole board, where the maximum term imposed was only 12 months. Further, we found some evidence that stresses on the capacity of California’s justice system—as measured by jail and prison overcrowding and district attorney caseloads—resulted in greater likelihoods that BPH would handle criminal violation cases. While case and offender characteristics are appropriate criteria for board referral decisions, system capacity should not affect these decisions.
  7. Track extralegal factors affecting revocation. CDCR should develop better evaluation methods to reduce the influence of extralegal factors—particularly parolee race—on violation case outcomes. We found that black parole violators were more likely to experience referral to the parole board, and more likely to be returned by the board for certain types of violations. We also found effects related to age, gender, and mental health status. The state must explore the causes and consequences of the influences of demographic and personal characteristics on sanctioning decisions.
  8. Expand substance abuse and mental health programs. Substance abuse-related violations and the violations of parolees with mental health problems make up a large share of all violations. These populations are not well-served by short returns to prison, where the few available services and sanctions are of insufficient duration to improve their outcomes. CDCR should expand intermediate sanctions specifically for these populations, so as to allow for community-based and in-custody treatment in a non-prison environment for sufficient time periods to address these criminogenic needs.

back to top



This study is just the first step toward a better understanding of California’s parole violation and revocation process. The data we collected were primarily administrative; other types of data, such as systematic interviews with parolees about their parole experiences, would highlight issues of discretion and sanctioning that are difficult to capture through quantitative analyses of official records alone. Future research on parole outcomes could also benefit from improvements to data quality. Some of our variables were underspecified (e.g., the community variables and parole agent characteristics). Other factors that may be related to parole outcomes, such as addiction and employability of parolees, were beyond the scope of our data collection effort. Data on the extent and type of programs parolees participated in could also expand on what we have done here. Given that many parolees are violated for program noncompliance, and that others may benefit from work and educational programming, it would be useful to know the degree to which parolees are engaged in assigned programming. Future studies might also address parole policies more specifically. Our research has generated many insights that can inform certain policies, such as early discharge from parole and the timing of service delivery.

It is important to note that our data is from 2003–2004 and California’s parole system is currently undergoing the most significant changes in its procedures since the late 1970s. Currently, California is implementing a new, evidence-based, parole violation decisionmaking instrument (PVDMI) to help agents and BPH assess risk and needs in determining sanctions. PVDMI was specifically designed for California parolees using another new instrument, the California Static Risk Assessment (CSRA). CSRA uses the offender’s past criminal history and characteristics such as age and gender to predict the likelihood that they will reoffend.

CSRA, combined with the severity ranking of all parole violations, has been incorporated into PVDMI, which results in a score that designates the appropriate violation response level. The response levels range from least intensive (e.g., community programs) to most intensive responses (in-custody drug treatment or return to prison recommendations). PVDMI is designed to focus California’s prison resources on higher-risk parolees while targeting less serious parole violators for community- based alternatives that address the root sources of their problems. DAPO is acquiring or redirecting treatment resources to plan for the expanded use of community- based sanctions in responding to parole violations in California. PVDMI was developed with the full participation and support of BPH and it is anticipated that the instrument will impact BPH’s decision-making as well. Implementation of PVDMI will be evaluated by the University of California Irvine Center for Evidence-Based Corrections.22

As these and other parole reforms move forward and parole data systems and knowledge about parole outcomes improve, it should be easier to implement studies that focus specifically on the potential effectiveness of various policy choices. We hope that this research will provide guidance for future research efforts and for the important discussions that will be taking place over the next several years about parole in California and the United States.

back to top