The main purpose of this study was to develop an
analytical approach to utilize detailed dated criminal history
information in order to investigate whether, and to what extent,
incarceration is able to deter offenders from future offending. A
secondary purpose of the study was to demonstrate the utility of the
developed framework by applying it to a real-world dataset.
The study utilized a combination of
information-theory and event-history analysis to estimate
individual-specific offending micro-trajectories, project
counterfactual trajectories (the offending trajectory for each
individual had (s)he not been incarcerated), and to assess the actual
post-release offending patterns against the backdrop of these
counterfactuals. The study used data collected by the Bureau of
Justice Statistics (BJS), available as RECIDIVISM OF PRISONERS
RELEASED IN 1994: [UNITED STATES] (ICPSR 3355), which tracked a sample
of 38,624 prisoners released from prison in 15 states in 1994 over a
period of three years. The data consist of information on each
releasee's entire officially recorded criminal history. This includes
all recorded adult arrests through the end of the follow-up
period. These data were obtained from state and Federal Bureau of
Investigation (FBI) automated RAP sheets and include arrest,
adjudication, and sentencing information. Each arrest event includes
information on adjudication and sentencing related to that event, if
such action was taken. The data also include some demographic
information, type of release from prison, and type of admission into
prison. However, this information is only available for the 1994
release and not for all prior (or future) arrest events. This data
collection consists of the syntax for a SAS macro used to estimate
individual offending trajectories. Individuals that had some problem
in their arrest history records or were not included in the BJS report
were not included in the analysis sample. Arrest records were
restructured into a hierarchical person-event-level file. Arrest
events of each person were clustered in chronological order. The
arrest histories were then truncated after the first post-release
re-arrest event. For those persons that were not arrested after
release, the arrest age was set to the age at censoring.
After removing persons who either had some problem in
their arrest histories or were not included in the Bureau of Justice
Statistics (BJS) report, the remaining sample consisted of 32,628
persons across 15 states. Problem arrest histories were defined as
records that were not in proper chronological order, those in which
ages were incorrect or impossible, and those that were missing
information on all ages or that had gaps in their age variable. Since
the sample for California releasees was very large (nearly 60,000
person-events before prison release), a random subset of 2,500
individuals (21,838 person-events) from the California sample was used
for estimating the pre-prison criminal history accumulation process.
However, all individuals from the California sample were used in the
analysis of the post-release data. Therefore, the final pre-release
sample consisted of 21,226 individuals across the 15 states while the
post-release sample consisted of 32,628 individuals.
individual
RECIDIVISM OF PRISONERS RELEASED IN 1994: [UNITED
STATES] (ICPSR 3355)
program source code
The key independent variables used in estimating
the pre-release criminal history accumulation process included the
arrest number, the age at first arrest, whether or not the individual
was confined as a result of the previous arrest event, and a measure
of the number of years taken to reach each arrest event cumulated
through the last arrest event. The same set of basic variables were
used to model first re-arrest after release (recidivism).
Not applicable
None