The Bureau of Justice Statistics (BJS) collects criminal history data from the FBI and state record repositories to study recidivism patterns of various offenders, including persons on probation or discharged from prison.
The latest study of state prisoners estimated the recidivism patterns of about 400,000 persons released from state prisons in 30 states in 2005. The findings from the 2005 study cannot be directly compared to those from BJS's previous prisoner recidivism studies due to changes in the demographic characteristics and criminal histories of the U.S. prison population, an increase in the number of states in the study, and improvements made to the quality and completeness of the nation’s criminal history records since the mid-1990s. BJS has also used criminal history records to examine the recidivism of persons placed on federal community supervision in 2005.
In addition to recidivism statistics based on criminal history records, BJS collects administrative data through the Annual Surveys of Probation and Parole to examine the rate at which offenders are at risk of being incarcerated for a new offense or for violating the conditions of their supervision.
Data collections vary in scope, burden, and frequency of collection - (see individual data collection descriptions for more information). Generally, BJS collects data both from administrative records and from interviews with prison and jail inmates. All data collections must be approved by the Office of Management and Budget (OMB) prior to fielding, which takes several months. Collections must be resubmitted for approval every 3 years (sooner if there are changes in the data collection). For data that are collected through inmate interviews, there must also be an Institutional Review Board (IRB) to protect human subjects (prior to OMB submission), and individual jurisdictions may require additional reviews prior to participation.
All data collection is voluntary. Without a specific mandate by Congress, no jurisdiction is compelled to participate in our data collections; individual surveys are conducted only with persons granting formal consent to participate. Most jurisdictions choose to participate because the information is helpful for policy and practice and may be used to allocate funding. It takes time to achieve a complete enumeration, particularly in times of staff shortages and budget cuts in many levels of government.
Administrative collections are sent out close to the reference date in the survey and are due to BJS 2 to 3 months later. Most respondents submit the data on time, but for various reasons, other jurisdictions take longer to submit the data. BJS staff or contractor staff work with jurisdictions to obtain the necessary information, which can take an additional 3 months.
After data are collected, they must then be cleaned, weighted (in the case of sample populations), and analyzed. BJS staff has several methods of release, including a formal report, statistical/electronic tables, or a summary brief. All data are fully verified prior to release. Keeping in mind that each data collection is different and the times may vary significantly depending on the collection of interest, provided below is an average data collection and processing timetable:
Collection, 5–6 months (from reference date) for administrative surveys; 8–12 for interview surveys
Cleaning/weighting, 1–2 months for administrative surveys; 3–6 for interview surveys
Analysis/verification, 2–12 months, depending on survey type and complexity of analysis
Preparation to disseminate, 2–3 months
Not all datasets are available for public use. We are actively working on mechanisms to make as many datasets as possible accessible to the public. Much of our data is archived at the University of Michigan’s National Archive of Criminal Justice Data. A lag period exists between the release of a report and the archiving of the data, as the dataset must be submitted to the archive with full documentation and may undergo disclosure review to protect the confidentiality of the respondents, when applicable.