National Archive of Criminal Justice Data
Techniques for Assessing the Accuracy of Recidivism Prediction Scales, 1960-1980: [Miami, Albuquerque, New York City, Alameda and Los Angeles Counties, and the State of California] (ICPSR 9988)
Principal Investigator(s): Cohen, Jacqueline; Zimmerman, Sherwood; King, Stephen
The purpose of this data collection was to measure the validity or accuracy of four recidivism prediction instruments: the INSLAW, RAND, SFS81, and CGR scales. These scales estimate the probability that criminals will commit subsequent crimes quickly, that individuals will commit crime frequently, that inmates who are eligible for release on parole will commit subsequent crimes, and that defendants awaiting trial will commit crimes while on pretrial arrest or detention. The investigators used longitudinal data from five existing independent studies to assess the validity of the four predictive measures in question. The first data file was originally collected by the Vera Institute of Justice in New York City and was derived from an experimental evaluation of a jobs training program called the Alternative Youth Employment Strategies Project implemented in Albuquerque, New Mexico, Miami, Florida, and New York City, New York. The second file contains data from a RAND Corporation study, EFFECTS OF PRISON VERSUS PROBATION IN CALIFORNIA, 1980-1982 (ICPSR 8700), from offenders in Alameda and Los Angeles counties, California. Parts 3 through 5 pertain to serious juvenile offenders who were incarcerated during the 1960s and 1970s in three institutions of the California Youth Authority. A portion of the original data for these parts was taken from EARLY IDENTIFICATION OF THE CHRONIC OFFENDER, [1978-1980: CALIFORNIA] (ICPSR 8226). All files present demographic and socioeconomic variables such as birth information, race and ethnicity, education background, work and military experience, and criminal history, including involvement in criminal activities, drug addiction, and incarceration episodes. From the variables in each data file, standard variables across all data files were constructed. Constructed variables included those on background (such as drug use, arrest, conviction, employment, and education history), which were used to construct the four predictive scales, and follow-up variables concerning arrest and incarceration history. Scores on the four predictive scales were estimated.
One or more data files in this study are set up in a non-standard format, such as card image format. Users may need help converting these files before they can be used for analysis.
These data are freely available.
Cohen, Jacqueline, Sherwood Zimmerman, and Stephen King. TECHNIQUES FOR ASSESSING THE ACCURACY OF RECIDIVISM PREDICTION SCALES, 1960-1980: [MIAMI, ALBUQUERQUE, NEW YORK CITY, ALAMEDA AND LOS ANGELES COUNTIES, AND THE STATE OF CALIFORNIA]. ICPSR version. Los Altos, CA: Sociometrics Corporation [producer], 1992. Ann Arbor, MI: Inter-university Consortium for Political and Social Research [distributor], 1999. http://doi.org/10.3886/ICPSR09988.v1
Persistent URL: http://doi.org/10.3886/ICPSR09988.v1
This study was funded by:
- United States Department of Justice. Office of Justice Programs. National Institute of Justice (86-IJ-CX-0039)
Scope of Study
Universe: Part 1: All youths in the Department of Labor job training programs in Albuquerque, Miami, and New York City. Part 2: All convicted felons in Alameda and Los Angeles counties. Parts 3-5: All young males arrested in the State of California who had spent time as juveniles in one of the three California Youth Authority facilities.
Data Types: event/transaction data
Data Collection Notes:
(1) ICPSR assigned values labeled as "missing" or "missing/NA" as designated missing values in the data definition statements for each part. For Part 2, values labeled as "unknown" were designated as missing values for cards 19-22. These declared missing values may be different than the missing values used by Sociometrics, Inc. to produce the tables in the user guide. The variable count provided by ICPSR for each part is the total number of variables as defined in the data definition statements and not a count of analyzable variables only. (2) The codebook and user guide are provided as a Portable Document Format file. The PDF file format was developed by Adobe Systems Incorporated and can be accessed using PDF reader software, such as the Adobe Acrobat Reader. Information on how to obtain a copy of the Acrobat Reader is provided through the ICPSR Website on the Internet.
Sample: For Part 1, the investigators selected individuals from the Department of Labor job training program identified as "high risk youths," aged 16 to 21, in Albuquerque, Miami, and New York City, who had arrests prior to their participation in the jobs training program. The arrest preceding participation in the jobs training program was marked as the target event for the application of the prediction scales. For Part 2, the sample consisted of matched samples of convicted felons who were sentenced either to prison or felony probation in Alameda and Los Angeles counties. The arrest associated with a 1980 conviction was used as the target for applying the prediction scales. The California Youth Authority samples, Parts 3-5, were male juveniles from an earlier study who were subsequently arrested after their 18th birthdays. The first arrest as an adult was the target event for the prediction scales.
- Standardized missing values.
- Checked for undocumented or out-of-range codes.
Original ICPSR Release: 1993-10-02
- 2006-01-18 File CB9988.ALL.PDF was removed from any previous datasets and flagged as a study-level file, so that it will accompany all downloads.
- 1999-02-17 A PDF codebook and SAS and SPSS data definition statements are now available for this collection. ICPSR assigned values labeled as "missing" or "missing/NA" as designated missing values in the data definition statements for each part. For Part 2, values labeled as "unknown" were designated as missing values for cards 19-22. These declared missing values may be different than the missing values used by Sociometrics, Inc. to produce the tables in the user guide. The variable count provided by ICPSR for each part is the total number of variables as defined in the data definition statements and not a count of analyzable variables only.
- Citations exports are provided above.
Export Study-level metadata (does not include variable-level metadata)