The Juvenile Court Statistics (JCS) series uses data provided to the National Juvenile Court Data Archive (the Archive) by state and county agencies responsible for collecting and/or disseminating information on the processing of youth in juvenile courts. These data are not the result of a uniform data collection effort. They are not derived from a complete census of juvenile courts or obtained from a probability sample of courts. The national estimates presented in this Report are developed by using compatible information from all courts that are able to provide data to the Archive.

Sources of Data

The Archive collects data in two forms: detailed case-level data and court-level aggregate statistics. Caselevel data are usually generated by automated client-tracking systems or case-reporting systems managed by juvenile courts or other juvenile justice agencies. These systems provide detailed data on the characteristics of each delinquency and status offense case handled by courts, generally including the age, gender, and race of the youth referred; the date and source of referral; the offenses charged; detention and petitioning decisions; and the date and type of disposition.

The structure of each case-level data set contributed to the Archive is unique, having been designed to meet the information needs of a particular jurisdiction. Archive staff study the structure and content of each data set in order to design an automated restructuring procedure that will transform each jurisdiction's data into a common case-level format. The aggregation of these standardized case-level data files constitutes the Archive's national case-level database.

Court-level aggregate statistics either are abstracted from the annual reports of state and local courts or are contributed directly to the Archive. Court-level statistics typically provide counts of the delinquency and status offense cases handled by courts in a defined time period (calendar or fiscal year). The compiled data from jurisdictions that contribute only court-level statistics constitute the national court-level database. Together, the two multi-jurisdictional databases (case-level and court-level) are used to generate the Archive's national estimates of delinquency and status offense cases.

Each year, many juvenile courts contribute either case-level data or courtlevel aggregate statistics to the Archive. However, not all of this information can be used to generate the national estimates contained in JCS.

To be used in the development of national estimates, the data must be in a compatible unit of count (i.e., case disposed), the data source must demonstrate a pattern of consistent reporting over time (at least 2 years), and the data file contributed to the Archive must represent a complete count of delinquency and/or status offense cases disposed in a jurisdiction during a given year.

In 2004, case-level data describing 1,083,986 delinquency cases handled by 1,785 jurisdictions in 36 states met the Archive's criteria for inclusion in the development of national estimates. Compatible data were available from Alabama, Alaska, Arizona, Arkansas, California, Connecticut, District of Columbia, Florida, Georgia, Hawaii, Illinois, Indiana, Kentucky, Maine, Maryland, Michigan, Minnesota, Missouri, Montana, Nebraska, Nevada, New Jersey, New Mexico, Ohio, Oklahoma, Pennsylvania, Rhode Island, South Carolina, South Dakota, Tennessee, Texas, Utah, Virginia, Washington, West Virginia, and Wisconsin. These courts had jurisdiction over 68% of the nation's juvenile population in 2004. Compatible courtlevel aggregate statistics on an additional 71,289 delinquency cases from 217 jurisdictions were reported from the states of California, Idaho, Illinois, Indiana, New York, and Vermont. In all, the Archive received compatible case-level data and court-level statistics on delinquency cases from 2,001 jurisdictions containing 77% of the Nation's juvenile population in 2004 (table A–1).

Table A-1: 2004 Stratum Profiles for Delinquency Data       Counties reporting compatible data       Number of counties Stratum County population ages 10-17 Counties in stratum Case-level Court-level Total* Percentage of juvenile population 1 Fewer than 13,030 2,614 1,497 175 1,672    66% 2 13,030–54,800    339     201   29     230 70 3 54,801–144,300    101      63     6       69 72 4 More than 144,300      31      24     7       30 98 Total 3,085 1,785 217 2,001 77 * Some counties reported both case-level and court-level data; therefore, the total number of counties reporting delinquency data is not equal to the number of counties reporting case-level data plus the number of counties reporting court-level data.

Case-level data describing 94,092 formally handled status offense cases from 1,863 jurisdictions in 35 states met the criteria for inclusion in the sample for 2004. The contributing states were Alabama, Alaska, Arizona, Arkansas, California, Connecticut, District of Columbia, Florida, Georgia, Hawaii, Illinois, Indiana, Kentucky, Maine, Maryland, Michigan, Minnesota, Missouri, Montana, Nebraska, Nevada, New Mexico, Ohio, Oklahoma, Pennsylvania, Rhode Island, South Carolina, South Dakota, Tennessee, Texas, Utah, Virginia, Washington, West Virginia, and Wisconsin. These courts had jurisdiction over 66% of the juvenile population. An additional 215 jurisdictions in 5 states (California, Idaho, Indiana, New York, and Vermont ) reported compatible court-level aggregate statistics on 15,317 petitioned status offense cases. Altogether, compatible case-level and court-level data on petitioned status offense cases were available from 2,078 jurisdictions containing 74% of the U.S. juvenile population in 2004 (table A–2).

A list of states contributing case-level data (either delinquency or petitioned status offense data), the variables each reports, and the percentage of cases containing each variable are presented in table A–3.

Juvenile Population

The volume and characteristics of juvenile court caseloads are partly a function of the size and demographic composition of a jurisdiction's population. Therefore, a critical element in the Archive's development of national estimates is the population of youth that generate the juvenile court referrals in each jurisdiction—i.e., the “juvenile” population of every U.S. county.

A survey of the Archive's case-level data shows that very few delinquency or status offense cases involve youth younger than 10. Therefore, the lower age limit of the juvenile population is set at 10 years for all jurisdictions. On the other hand, the upper age limit varies by state. Every state defines an upper age limit for youth who will come under the jurisdiction of the juvenile court if they commit an illegal act. (See “Upper age of jurisdiction” in the “Glossary of Terms” section.) Most states define this age to be 17 years; other states have set the age at 15 or 16. States often enact exceptions to this simple age criterion (e.g., youthful offender legislation and concurrent jurisdiction or extended jurisdiction provisions). In general, however, juvenile courts have responsibility for all law violations committed by youth at or below the upper age of original jurisdiction.

For the purposes of this Report, therefore, the juvenile population is defined as the number of youth living in a jurisdiction who are at least 10 years old but who are not older than the upper age of original juvenile court jurisdiction. For example, in New York, where the upper age of juvenile court jurisdiction is 15, the juvenile population is the number of youth residing in a county who are between the ages of 10 and 15.

The juvenile population estimates used in this Report were developed with data from the Census Bureau. 1 The estimates, separated into single-year age groups, reflect the number of whites, blacks, American Indians/Alaskan Natives, and Asians (including Native Hawaiians and Pacific Islanders) who reside in each county in the Nation and who are between the ages of 10 and the upper age of original juvenile court jurisdiction. 2

Estimation Procedure

National estimates are developed by using the national case-level database, the national court-level database, and the Archive's juvenile population estimates for every U.S. county. “County” was selected as the unit of aggregation because (1) most juvenile court jurisdictions in the United States are concurrent with county boundaries, (2) most data contributed by juvenile courts include the county in which the case was handled, and (3) youth population estimates can be developed at the county level. 3

The Archive's national estimates are generated using data obtained from its nonprobability sample of juvenile courts. There are two major components of the estimation procedure. First, missing values on individual records of the national case-level database are imputed using a hot deck procedure. Then the records of the national case-level database are weighted to represent the total number of cases handled by juvenile courts nationwide. Each stage of the estimation procedure will be described separately.

Record-level imputation. The first step in the estimation procedure is to place all U.S. counties into one of four strata based on their population of youth between the ages of 10 and 17. The lower and upper population limits of the four strata are defined each year so that each stratum contains one-quarter of the national population of youth between the ages of 10 and 17.

This information is added onto each record in the national case-level database. As a result, each record in the national case-level database contains 11 variables of interest to the JCS report: county strata, year of disposition, manner of handling, youth's age, youth's gender, youth's race, referral offense, source of referral, case detention, case adjudication, and case disposition.

By definition, the first three of these variables (i.e., county strata, year of disposition, and manner of handling) are known for every case in the database. Each of the others may contain a missing value code. The estimation procedure for the JCS report employs a multi-stage process to impute information for each missing value on each case record in the national caselevel database.

Within a county's set of records in the database there may be two types of missing information. For many counties a small proportion of their case-level records are missing valid codes in data elements that are not missing for the vast majority of the other records from that county. For example, the gender of a youth may be unknown on a few records while it is known for all the other youth in the county's database. This type of missing value is labeled “record-level missing.” There are also counties in which every record in the database has a missing value code for a specific variable. For example, some court data collection systems do not capture information on a youth's predisposition detention. Therefore, the variable “Was child detained?” in the national case-level data has a missing value code on each record from that county. This type of missing value is labeled “format-level missing.” (Table A–3 indicates the standardized data elements that were not available from each jurisdiction's 2004 data set.) The imputation process handles the two types of missing values separately.

The imputation of record-level missing values uses a hot deck procedure with a donor pool limited to records for the same county. First, all the records for a specific county are sorted by disposition date. Then the file is read again, a record at a time. When the imputation software identifies a record with a record-level missing value (i.e., the target record), it imputes a valid code in this data field. This is accomplished by locating the next record in the county file that matches the target record on all of its nonmissing values and has a nonmissing code in the field; this record is called the donor record. The imputation software copies the valid code from the donor record and overwrites the missing value code on the target record with this nonmissing value.

Once a donor record is used in the process, it is not used again unless no other matches can be found for another target record. Also there are a small number of instances in which no donor record can be found in the county file. When this occurs, the imputation software relaxes its record matching criteria. That is, instead of trying to find a donor record with identical codes on all known variables, the software ignores one nonmissing variable and attempts to find a match on all of the others. In the small number of cases where this does not lead to the identification of a donor record, a second variable is ignored and the file is reread looking for a donor. Although theoretically (and programmatically) this process can be repeated until all variables but county, year of disposition, and manner of handling are ignored to find a donor, this never occurred. The order in which variables are removed from the matching criteria are source of referral, detention, offense, adjudication, race, gender, and age.

Format-level imputation. After all the record-level missing values have been imputed, the process turns to formatmissing information, or information that is missing from a case record because that court's information system does not report this information on their cases. The process for imputing format-missing information is similar to that used in the record-missing imputation process with the needed difference that the donor pool is expanded. By definition, limiting the donor pool for format-missing records to records from their own county is fruitless. For format-missing records, the donor pool is defined as the records from all counties in the target record's strata with the same year of disposition and manner of handling.

Using this expanded donor pool, the imputation process follows the steps described above where a target record (i.e., one with missing data) is identified and the donor pool is scanned for a match. Once a match is found, the missing information on the target record is overwritten and the donor record is flagged as having been used. If a donor record cannot be found in the first pass through the donor pool, matching criteria are relaxed until a donor is found. There is one major exception to this process of imputing format-level missing information. This exception involves the process of imputing missing race for those counties that do not report this data element to the Archive. The racial composition of a court's caseload is strongly related to the racial composition of the resident juvenile population. Creating a donor pool that ignores this relationship would reduce the validity of the imputation process. So for those few formats that did not include race, donor pools were developed that limited the pool to counties with racial compositions similar to that of the target record county.

This was accomplished by dividing the counties in the U.S. into five groups defined by the percent of black juveniles in their 10–17 populations. This classification was then added to each case record and used as a matching criterion for finding a donor record within the set of potential donor records defined by strata, year of disposition, and manner of handling.

Using this process for the imputation of missing information in the national case-level database, the database then enters the next stage in the estimation process with no missing value codes in any of its data fields.

Weighting to produce national estimates. The Archive employs an elaborate multivariate procedure that assigns a weight to each record in the national case-level database that, when used in analysis, yields national estimates of juvenile court activity. The weights incorporate a number of factors related to the size and characteristics of juvenile court caseloads: the court's jurisdictional responsibilities (upper age); the size and demographic composition of the community; and the age, gender, and race profile of the youth involved in juvenile court cases.

The basic assumption underlying the weighting procedure is that similar legal and demographic factors shape the volume and characteristics of cases in reporting and nonreporting counties of comparable size and features. The weighting procedure develops independent estimates for the number of petitioned delinquency cases, nonpetitioned delinquency cases, and petitioned status offense cases handled by juvenile courts nationwide. Identical statistical procedures are used to develop all case estimates.

As noted earlier, all U.S. counties are placed into one of four strata based on the size of their youth population between the ages of 10 and 17. In the first step to develop the weights, the Archive determines for each stratum the number of juvenile residents who are under original juvenile court jurisdiction in three age groups: 10- through 15-year-olds, 16-year-olds, and 17-year-olds. The three age groups are further subdivided into four racial groups: white, black, American Indian (including Alaskan Native), and Asian/NHPI (including Native Hawaiian and Other Pacific Islander). Thus, juvenile resident population estimates are developed for 12 age/race categories in each stratum of counties.

The next step is to identify within each stratum the jurisdictions that contributed to the Archive case-level data consistent with JCS reporting requirements. The populations of these case-level reporting jurisdictions within each stratum are then developed for each of the 12 age/race categories. The national case-level database is summarized to determine within each stratum the number of court cases that involved youth in each of the 12 age/race population groups. Case rates (number of cases per 1,000 juveniles in the population) are then developed for the 12 age/race groups within each of the four strata.

For example, assume that a total of 3,039,000 white youth between the ages of 10 and 15 resided in the stratum 2 counties that reported caselevel data to the Archive. If the Archive's case-level database shows that the juvenile courts in these counties handled 52,605 petitioned delinquency cases involving white youth between the ages of 10 and 15, the number of cases per 1,000 white youth ages 10 to 15 for stratum 2 would be 17.3, or:

(52,605/3,039,000) x 1,000 = 17.3

Comparable analyses are then used to establish the stratum 2 case rates for black youth, American Indian youth, and Asian/NHPI youth in the same age group (54.6, 23.3, and 10.9, respectively).

Next, information contained in the national court-level database is introduced, and stratum-level case rates are adjusted accordingly. First, each court-level statistic is disaggregated into the 12 age/race groups. This separation is accomplished by assuming that, for each jurisdiction, the relationships among the stratum's 12 age/race case rates (developed from the case-level data) are paralleled in the aggregate statistic.

For example, assume that a jurisdiction in stratum 2 with an upper age of original juvenile court jurisdiction of 15 reported it processed 600 cases during the year. Also assume that this jurisdiction had a juvenile population of 12,000 white youth, 5,000 black youth, 500 American Indian youth, and 1,500 Asian/NHPI youth. The stratum 2 case rates for each racial group in the 10–15 age group would be multiplied by the corresponding population to develop estimates of the proportion of the court's caseload that came from each age/race group, as follows:

White: (17.3 x 12,000) / [(17.3 x 12,000) + (54.6 x 5,000) + (23.3 x 500) + (10.9 x 1,500)] = 41%

Black: (54.6 x 5,000) / [(17.3 x 12,000) + (54.6 x 5,000) + (23.3 x 500) + (10.9 x 1,500)] = 54%

American Indian: (23.3 x 500) / [(17.3 x 12,000) + (54.6 x 5,000) + (23.3 x 500) + (10.9 x 1,500)] = 2%

Asian/NHPI: (10.9 x 1,500) / [(17.3 x 12,000) + (54.6 x 5,000) + (23.3 x 500) + (10.9 x 1,500)] = 3%

The jurisdiction's total caseload of 600 would then be allocated based on these proportions. In this example, it would be estimated that 41% of all cases reported in the jurisdiction's aggregate statistics involved white youth, 54% involved black youth, 2% involved American Indian youth, and the remaining 3% involved Asian/NHPI youth. When these proportions are applied to a reported aggregate statistic of 600 cases, this jurisdiction is estimated to have handled 245 cases involving white youth, 322 cases involving black youth, 14 cases involving American Indian youth, and 19 cases involving Asian/NHPI youth age 15 or younger. The same method is used to disaggregate into the 12 age/race groups the aggregated case counts reported by those jurisdictions that could only report aggregate courtlevel statistics.

The disaggregated court-level counts are then added to the counts developed from case-level data to produce an estimate of the number of cases involving each of the 12 age/race groups handled by reporting courts (i.e., both case-level and aggregate reporters) in each of the four strata. The juvenile population figures for the entire reporting sample are also compiled. Together, these new stratum- specific case counts and juvenile population of reporting counties are used to generate a revised set of case rates for each of the 12 age/race groups within each of the four strata. Stratum estimates for the total number of cases involving each age/race group are then calculated by multiplying the revised case rate for each of the 12 age/race groups in a stratum by the corresponding juvenile population in all counties belonging to that stratum (both reporting and nonreporting).

After the stratum estimates for the total number of cases in each age/race group in each stratum has been calculated, the next step is to weight the records in the national case-level database. This weight is equal to the estimated number of cases in one of the stratum's 12 age/race groups divided by the actual number of such records in the national case-level database. For example, assume that the Archive generates a national estimate of 41,254 petitioned delinquency cases involving white 16-year-olds from stratum 2 counties. Assume also that the national case-level database for that year contained 25,758 petitioned delinquency cases involving white 16-year-olds from stratum 2 counties. In the Archive's national estimation database, each stratum 2 petitioned delinquency case that involved a white 16-year-old would be weighted by 1.60, because:

41,254/25,758 = 1.60

Finally, by incorporating the weights into all analyses of the national caselevel database (a database with no case-level information missing), national estimates of case volumes and case characteristics can be produced. More detailed information about the Archive's national estimation methodology is available on request from the National Center for Juvenile Justice.