8 Percent Problem Study Methodology

Note: This is a composite of the background section and original appendix from The "8% Problem": Chronic Juvenile Offender Recidivism (Kurz and Moore, 1994) which describes the research methodology used to examine juvenile offender recidivism in Orange County.

Background

Over the decade of the 1980s, the County of Orange became increasingly urban and more ethnically and culturally diverse. Statistics published by the County Administrative Office indicate that from 1980 to 1990:

  • The total population of Orange County increased by 26%.
  • The number of cities in Orange County grew to 31, with 10 over 100,000 population and five new since 1988.
  • The Hispanic population of Orange County doubled while its Asian population tripled. By 1990, one in four Orange County residents was Hispanic, while one in 10 was Asian.
  • 7,400 undocumented aliens were added each year for a total of 166,000 by 1990.

Over the same time period, there was dramatic growth in the number of criminally active gangs, violent crimes, and crimes involving the use of weapons. Many of these serious crimes have involved juvenile offenders. Information provided by the California Center for Law Enforcement Assistance indicated that, from 1980 to 1990, juvenile arrests for violent crimes increased by 45% - murder +130%, forcible rape +70%, and robbery +65% - while juvenile felony arrests involving the use of weapons increased by 48%.

The resources available to county and city governments to deal with the problems and challenges of increased urbanization could not keep pace. The trend was, therefore, to focus a greater proportion of available resources on the most difficult urban problems. Within the juvenile justice system of Orange County, as the number of referrals for serious crimes increased, the demand for programs and services that ensure a higher level of community protection also increased. According to the records of the Orange County Probation Department, from 1980 to 1990 Juvenile Court dispositions involving custody increased by 43%, commitments to local institutions were up 40%, and commitments to the California Youth Authority rose by 169%.

These higher control dispositional options are very costly. In 1990, local commitments in Orange County averaged 74 days and cost approximately $5,500 each, whereas commitments to the California Youth Authority averaged just over two years and cost about $55,000 per commitment. If a minor receives multiple commitments at any level, the custody costs alone become staggering!

Noting these disturbing trends, between 1985 and 1987 the Orange County Probation Department implemented a model for the management of probation resources recommended by the National Institute of Corrections (NIC). This model is predicated on the use of valid procedures for assessing the relative degree of risk various groups of minors present for committing multiple offenses. Therefore, as part of model implementation, the department developed and validated a Juvenile Offender Risk Assessment instrument, with technical assistance from the federal Office of Juvenile Justice and Delinquency Prevention and NIC staff.

Developed for use with a probation field supervision population, the instrument was initially constructed utilizing those factors that juvenile field supervision officers felt best predicted subsequent offense behavior. Data was collected on both new and active cases to test the proposed Initial Assessment and Reassessment instruments. This sample of over 500 cases was followed for a six-month period and any subsequent petition filings recorded. These included both alleged technical and new law violations.

Based on these data analyses, the Orange County classification instruments were fine-tuned to achieve maximum discrimination2 based on the dependent variables, i.e., the number of new law and technical violations over an initial six-month tracking period. The following factor sets were found to differentiate High-, Medium-, and Low-Risk juvenile probationers in Orange County, in order of their relative contribution to a higher risk of reoffending.

Prior Offense History: Multiple measures, ranging from prior arrests up to and including prior incarceration or out-of-home placement. (The more "priors," the higher the risk score.)

Multiple Problems: Includes lack of parental supervision and control; school attendance, behavior, and/or performance problems; substance abuse; chronic runaway behavior; and gang affiliations. (The more problems, the higher the Risk score.)

Age at Initial Assessment (15 or younger vs. 16 and older): This factor proved to be important in relation to the other two factor sets, particularly, the problem variables referenced above. (Younger minors with multiple problems had significantly higher recidivism rates than their older counterparts in both the Medium- and High-Risk groups.)

When these Juvenile Offender Risk Assessment instruments were formally adopted by the Probation Department and placed into department wide use, a Management Information System component was also established to record the profile of each minor at the Initial and most recent (or, final) assessment and document all technical and new law violations occurring during each minor's period of supervision. Thus, the capability was created to monitor and test the validity of these classification instruments, both over time and for longer periods of offense tracking than those utilized in the original validation effort.

Kara in Juvenile Hall

8% Problem Identification

In late 1988, a department-wide strategic planning effort was launched aimed at preparing for the anticipated changes and challenges in Orange County during the decade of the 1990s. Among the first questions raised by the participating managers were:

  • How are "we" - the Orange County Probation Department - doing with the youthful offenders of today?
  • Does the historical effectiveness information provide any clues as to how the department needs to change its service delivery systems to be more effective with the youthful offender populations of the next decade?

A significant amount of information was available on the outcomes achieved by individual program components administered by the department, such as the County's Juvenile Diversion program, probation field supervision and institutional treatment programs. However, there was very little information available by which to gauge the effectiveness of the department's overall juvenile operation or the relative costs of serving minors who re-offend with different frequencies.

In this latter regard, one data set offered some promise -- the Juvenile Court and Probation Statistical System (JCPSS) maintained by the California Bureau of Criminal Statistics (BCS). While containing a somewhat limited set of information,3 the Orange County data were both complete and known to be of consistently high quality, thanks to the efforts of both BCS and departmental clerical staff over the years of the JCPSS' existence.

In 1989 and again in 1990, two recidivism analyses were completed with assistance from the California Bureau of Criminal Statistics. Both involved the identification of an entire cohort of minors with their "first ever" application for petition during the first six months of either 1985 or 1987. In both study efforts, the number of minors identified equated to more than 3,000 individuals.

The 1985 cohort was tracked through the end of 1987 (and the 1987 cohort was tracked through the end of 1989) to identify:

  1. Any minors having subsequent applications for petition within each three-year study period; and
  2. The number and proportion of minors with 1, 2, 3 etc. subsequent applications for petition within each three-year tracking period.

The cohort was divided into three study groups stratified according to the total number of times minors were referred to the Probation Department during the tracking period.

The first study group included those minors who had only one application for petition during the entire tracking period. This group accounted for 71% of the cohort (2,234 of 3,164) and is referred to as the Non-recidivist study group.

The second study group (677 of 3,164) consisted of minors who had a total of two or three applications for petition during the follow-up period. This group comprised 21% of the study cohort and accounted for 45% of all re-referrals during the follow-up period. This group of minors is referred to as the Low-Rate recidivist group.

The third and final group included minors with a total of four or more applications for petition during the tracking period and is referred to as the Chronic recidivist (or "8% problem") group. These minors comprised only 8% of the study cohort, but accounted for 55% of all re-referrals during the tracking period.

Problem Definition (Relationship to Juvenile Risk Factors)

Having identified this "8% problem" group, the next phase of data exploration was aimed at broader definition of the problem, using the authors' knowledge of juvenile offender risk prediction factors to guide the analysis plan.

Noting that the department does not appear to be doing well with its "8%-ers" now, the specific study questions were:

  1. Are the "8%" minors different from the other two groups of non- and low-rate juvenile recidivists in ways that can be accurately and consistently identified early in their careers (at the first or second system referral)?
  2. Is there anything in the data regarding the profile or current handling of these "8%" minors that would provide us with clues as to problem solutions (i.e., more cost-effective intervention strategies and/or ways of allocating limited juvenile justice system resources)?

There were five steps involved in acquiring the study sample used for this analysis phase, once the 1987 study cohort had been partitioned into the three study groups described in Section II.

Step 1 required reducing the original 1987 cohort by 49 cases, to exclude those cases that were missing any of the original JCPSS data elements. (In this process, the proportional relationships between the original study groups were maintained.)

In Step 2, a sample of 220 minors was initially selected, using age distributions as the criteria to determine the appropriate sample size for each of the three study groups. Three subsamples were selected, one from each of the total cohort study groups, including 78 Non-recidivists, 70 Low-Rate Recidivists, and 72 Chronic Recidivists.

Steps 3 and 4 of the sample development process required a thorough review of the juvenile case files for each of the cases selected in Step 2. Step 3 confirmed that only minors with initial referrals during the first six months of 1987 were assigned to the study sample. Step 4 verified the number of system contacts during the initial tracking period. (This step included tracking minors into the adult system if they turned 18 during the 36-month follow-up period.) As a result of these two sample development steps, 13 cases were deleted from the study sample and 42 cases were reassigned to a different study group, as follows.

  • 22% of the non-recidivist cases were found to be Low-Rate recidivists and 5% were found to be Chronic recidivists. (73% remained in the Non-recidivist subgroup.)
  • 4% of the original Low-Rate recidivists became Non-recidivists while 22% were found to be Chronic recidivists. (74% remained in the Low-Rate recidivist group.)
  • None of the Chronic recidivists were moved to the Non-recidivist group while 6% were found to be Low-Rate recidivists. (94% remained in the Chronic recidivist group.)

Without this case review process, the size of the Non-recidivist group would have been overstated while the size of the Low-Rate and Chronic recidivist groups would have been understated.

During Step 5, the residence status of each minor was verified. As a result, a total of 36 cases were removed from the study sample due to being out-of-county residents.

Thirty-six percent of the Non-recidivist cases were out-of-county residents vs. 16% of the Low-Rate recidivists and a mere 6% of the Chronic recidivist subgroups.

The following chart provides a comparison of the original study sample with the one ultimately used for the follow-up study analyses:

Table 1: 8% Study Sample Development

Step 2: Original Sample (N = 220) Step 3: Removal of Cases Not in Study Time Frame (N = 207) Step 4: Adjusted Group (N = 207) Step 5: Removal of Out-of-County Cases (N = 171)
Non-Recidivist Subsample 78 (35%) 73 (35%) 56 (27%) 36 (21%)
Low-Rate Recidivist Subsample 70 (32%) 67 (32%) 69 (33%) 58 (34%)
Chronic Recidivist Subsample 72 (33%) 67 (33%) 82 (40%) 77 (45%)

From this sample selection process involving detailed case reviews, the authors concluded that the original "8% problem" designation might understate the chronic recidivism rate for juveniles initially referred to the juvenile justice system in Orange County. The actual rate may be somewhat higher, but would not be expected to exceed 15%.

First-time Ward Data Set (N = 905)

The 1987 Referral Cohort Study subsample analyses provided valuable information regarding the relationship between the identified problem factors, the minor's age and subsequent wardship status, and chronic juvenile offender recidivism rates. One limitation, particularly as related to the "multiproblem" factor testing, was the size of the study subsamples. Therefore, before moving ahead to develop intervention strategies for first-time wards with the identified high-risk profiles, the authors felt the need to "test" the predictive accuracy of the recommended target population "eligibility criteria" with a larger sample of first-time wards.

The Orange County Probation Department maintains a Management Information System component called the Juvenile Profile/Outcome System, which consists of the profile of each minor placed under field supervision at the time of his or her initial and most recent (or final) assessment for case classification purposes. This database was used to identify a one-year cohort of "first-time" juvenile wards, i.e., those with no prior arrests or referrals to the Probation Department in Orange County during CY-1987. A total of 1,039 cases were initially identified who met these selection criteria.

Cases from this group of first-time wards were then matched to their corresponding records in the JCPSS data file. 905 of the 1,039 first-time wards selected from the Juvenile Profile/Outcome system were found to also have complete records in the 1987 JCPSS data file. This cohort of 905 minors (referred to as the First-time Ward data set) was further partitioned into three subsamples, based on the total number of times the minor was referred on an application for petition during the three-year follow-up period.

Within the First-time Ward data set, 50% of the cases (450/905) were found to be Non-recidivists, 35% (315/905) were found to be Low-Rate recidivists (having one or two subsequent petitions during the three-year follow-up period) and 15% (140/905) were identified as Chronic recidivists with a total of four or more petitions during the three-year follow-up period.

In order to investigate the relationship between the previously identified problem factors and chronic juvenile offender recidivism rates, variables from the Orange County Juvenile Offender Risk and Needs Assessment scales were selected which most closely approximated the 8% risk factor definitions. These data were then subjected to the statistical analyses described in the text of this report.

Study Limitations/8% Exploratory Areas for Additional Study

As with any study dealing with risk prediction, the extent to which the 8% Problem study findings and conclusions may be generalized to other jurisdictions of comparable size and demographic make-up (much less those which are non-comparable) is an issue that should be approached with caution.

A number of other jurisdictions in California have used the Orange County Probation Department's recidivism analysis methodology to see what they come up with in terms of non-, low-rate, and chronic recidivist groups. This was, in fact, part of the technical assistance grant jointly awarded by the National Institute of Corrections to the Orange and Los Angeles County Probation Departments to assist in the design of potential "8% problem" solution(s) in these two counties.

The specific risk factors that were used in the development of the Orange County pilot intervention project, i.e. field test, and case identification procedures are comparable to those which other researchers have found to be significantly correlated with a higher risk of reoffending. They have nonetheless been specifically developed to accurately identify the youth in Orange County's juvenile justice system referral and/or ward population that are in need of a high level of resource coordination and focus.

Additionally, most of the proposed risk assessment variables have been validated and revalidated with various juvenile offender populations in Orange County over the past 10 years. Thus, the goal of the 1987 Initial Referral sample data analyses was not to create a new risk scale, but to confirm that the study subsamples conformed to the anticipated variable distribution patterns and to determine which factors best predict chronic juvenile offending.

The focus of the initial pilot project (or "field test") aimed at improving Orange County's "system batting average" with chronic juvenile recidivists was, as previously stated, younger, first-time wards in Orange County who present a multiproblem profile at the time of their first (or second) system referral. This is primarily because, based on the Orange County data analyses, this group had the highest rate of chronic recidivism. However, the authors wish to emphasize that we do not advocate, nor do any of our study findings support, any type of punitive action or the exercise of a higher level of control over the minor's actions based on the recommended target population criteria, than those already mandated or authorized by the Court in each individual case.

Further testing of the recommended "8%" case identification procedures has occurred and is still occurring to refine the estimates of the number of new "8%" minors that the system can be expected to receive in coming years. Major improvements have also been made in assessing the risk and needs associated with all first-time wards on a more timely and accurate basis and expediting the initiation of appropriate intervention programming.

System cost estimates need to be refined, using measures other than custody time savings. In the current experimental program evaluation, consultation with the County Executive Office and other juvenile justice system professionals is taking place to more fully address this important measure of project impact.

Under a separate state grant, the authors are also testing an early intervention program for older, first-time wards who become chronic recidivists. The analyses conducted to date indicate that substance abuse may be the most significant factor in the "late" onset of a chronic pattern of re-offending for this 8% problem group.

From the case study portion of this study effort, the authors were impressed with what appears to be a link between younger minors whose profile at initial referral includes a pattern of runaway behavior and child abuse/neglect. Unfortunately, this latter issue is often not recognized (nor does it come to the system's attention) until years after the initial system referral; however, minors subsequently report it as a long-standing issue, related to their earlier runaway episodes. This area is receiving closer attention in the pilot project case assessment, service delivery, and follow-on study plans.

Notes
  1. Local custody cost data provided by the Orange County Probation Department. State custody cost data provided by the California Youth Authority.
  2. Baird, C. (1984, March) Review of Orange County Juvenile Probation Classification System. Isthmus Associates. Madison, Wisconsin.
  3. The Juvenile Court and Probation Statistical System (JCPSS) database consisted of all minors referred for alleged criminal code violations for which the disposition of that offense occurred within a given calendar year. The database includes the minor's name, date of birth, gender, ethnicity, the highest referral charge - both the level (felony vs. misdemeanor) and type (criminal code section) - and all system disposition data and dates, e.g., Probation, District Attorney and Court, related to each discrete referral. The statewide system was terminated in October 1990, due to budget cutbacks at the state level. Efforts to bring this system back on line are currently under way. Orange County has maintained this database locally to the present time.