ments, asking for computerized rap sheets for those prisoners. Separately, BJS also provided the set of identifiers for sampled prisoners to the FBI, asking it to query its databases—particularly useful to get information on offenses committed outside the state that released the prisoner. The state and FBI records were combined to form a master database—in 1994, one that included 6,427 variables on the 38,624 prisoners, 6,336 of which provided criminal histories for up to 99 “arrest cycles” per prisoner (Bureau of Justice Statistics, 2002:6–7).8

The four recidivism measures derived for each prisoner were rearrest, reconviction, resentence to prison, and return to prison with or without a new prison sentence. The BJS report on the 1994 study (Langan and Levin, 2002:2) follows the description of these four measures with a disclaimer that the measures are likely to be underestimates:

To an unknown extent, recidivism rates based on State and FBI criminal history repositories understate actual levels of recidivism. The police agency making the arrest or the court disposing of the case may fail to send the notifying document to the State or FBI repository. Even if the document is sent, the repository may be unable to match the person in the document to the correct person in the repository or may neglect to enter the new information. For these reasons, studies such as this one that rely on these repositories for complete criminal history information will understate recidivism rates.

BJS conducted the recidivism studies in-house, with assistance in data collection and processing from the Regional Justice Information Service. Funding for the 1994 study was received, in part, from the FBI and the Corrections Program Office within the Office of Justice Programs (Langan and Levin, 2002:16).

On a one-time basis, in 1986, BJS drew from criminal history records to track a sample of convicted felons for 3 years upon their entry into probation. This collection generated estimates of the percentage of probationers who were rearrested, reconvicted, or reimprisoned for new crimes during the study period.


In its final report, the President’s Commission on Law Enforcement and Administration of Justice (1967:10) was struck by the difficulty in studying “law enforcement” as a unified entity, where policy changes made on high directly affect the public experience at the street level:


Ten prisoners in the 1994 sample had more than 99 arrest cycles, and one had 176 different arrests on record. In these cases, the earliest arrests were dropped to fit the 99-maximum limit of the database (Bureau of Justice Statistics, 2002:7).

The National Academies of Sciences, Engineering, and Medicine
500 Fifth St. N.W. | Washington, D.C. 20001

Copyright © National Academy of Sciences. All rights reserved.
Terms of Use and Privacy Statement