Who Returns to Prison? A Survival Analysis of Recidivism among Adult Offenders Released in Oklahoma, 1985 – 2004
| Author | Andrew L. Spivak,Kelly R. Damphousse |
| Published date | 01 December 2006 |
| Date | 01 December 2006 |
| DOI | http://doi.org/10.3818/JRP.8.2.2006.57 |
| Subject Matter | Article |
Who Returns to Prison? • 57
* Who Returns to Prison? A Survival Analysis
of Recidivism Among Adult Offenders Released
in Oklahoma, 1985 – 2004
Andrew L. Spivak
Kelly R. Damphousse
University of Oklahoma
* Abstract
Identifying and evaluating the inuence of factors that predict offenders’ post-
release performance is central to the study of recidivism. In this project, 60,536
adult prison releases from the Oklahoma Department of Corrections between
1985 and 1999 were tracked until May 31, 2004. Recidivism was measured as a
return to incarceration and cases were analyzed with a Cox Proportional Hazards
Survival Regression, which allowed for the assessment of the relative hazards of
returning to prison over time. Predictor variables included offense type, release type
(probation, parole, discharge), number of prior incarcerations, sentence length, time
served in prison, security classication, education, age, sex, and race. A greater
hazard of recidivism was associated with being a property offender (as opposed
to a drug, violent, or sex offender), being released to probation (as opposed to
being discharged), having a violent offense history, having a greater number of past
incarcerations, and being young, male, and a racial minority. Sentence length and
time served in prison had weak (but signicant) associations with recidivism, while
security classication and proportion of sentence served (as proxies for institutional
performance) indicated modestly positive, signicant relationships.
An earlier version of this paper won the Graduate Student Paper Competition held in
conjunction with the Bureau of Justice Statistics/Justice Research and Statistics Association
conference in 2005. The authors gratefully acknowledge the comments of the anonymous
reviewers of the earlier paper, and the comments of the journal reviewers. The authors also
gratefully acknowledge William B. Chown at the Oklahoma Criminal Justice Resource
Center for compiling data extracts used in the study, and Paul D. Allison at the University of
Pennsylvania, who assisted with a better understanding of survival analysis methodology.
JUSTICE RESEARCH AND POLICY, Vol. 8, No. 2, 2006
© 2006 Justice Research and Statistics Association
58 • Justice Research and Policy
Who Returns to Prison? • 59
* Introduction
Recidivism can be dened as the reversion or relapse to criminal acts of an of-
fender who has been subject to punitive sanctions or rehabilitative treatment for
previous conduct (Maltz, 1984). The issue of prison recidivism is particularly
important to the modern correctional goal of rehabilitation, a relatively new ob-
jective in criminal justice when compared with punishment and incapacitation.
Coinciding with the early development of the social and behavioral sciences, the
idea of rehabilitation reected a transition from strictly punitive and isolation
approaches in past criminal sanctioning to a type of medical “treatment” model.
In the treatment approach, imprisonment ought to do something that makes of-
fenders less likely to offend after release than they otherwise would have been.
This is the same objective as the “specic deterrence” goal of a punitive model,
which in the tradition of classical criminological theory is meant to prevent
future offending through fear of sanctions (see Beccaria, 1789). The extent to
which the specic deterrence of punishment or the rehabilitation of treatment
are preferable to one another or even effective at all is a matter of some uncer-
tainty in criminology (Sechrest, White, & Brown, 1979).
Relevant to each of these correctional goals (the theoretical merits of which
are beyond the scope of this project) is the examination of who reoffends af-
ter release from custody, who does not, and how long they remain successful.
Studying recidivism in this way can be useful in two aspects of criminology.
First, factors that are found to affect the post-release performance of offenders
may elicit insights on the nature and causes of crime, while data that measure
recidivism can be used to evaluate programs intended to improve success after
prison release (Barton & Turnbull, 1981). Second, being aware of the predictors
of recidivism can lead to a better assessment of correctional goals, sentencing
and parole policies, and the possible effectiveness of such proposals as selective
incapacitation (Blumstein, Cohen, Roth, & Visher, 1986).
While recidivism is clearly an important measure of the effectiveness of
correctional sanctions, and its minimization a primary objective of justice sys-
tems in general, other measures of offender adjustment do exist. Personality
change, attitude improvement, employment success, and family relations can
also act as indicators of post-release performance (Waldo & Griswold, 1979).
Nonetheless, whether or not a former inmate reoffends or returns to prison is
perhaps the most important measure of success or failure, and thus this study
examined recidivism among Oklahoma prison inmates released between 1985
and 1999, tracking each inmate until 2004. Recidivism is operationally dened
as returning to prison in the custody of the Oklahoma Department of Correc-
tions. This denition is by no means the only way that recidivism can be stud-
ied. Other denitions of recidivism have been used in past studies, some of the
most common including rearrest, reconviction, and absconding (Maltz, 1984).
Reincarceration includes the rst two measures, as well as technical violations
58 • Justice Research and Policy
Who Returns to Prison? • 59
of parole. Also, if there is any degree of generality in the phenomena related to
a lapse to criminal behavior, factors predicting a greater likelihood of returning
to prison should also tend to predict a greater likelihood of being rearrested,
reconvicted, or absconding.
A medical analogy of recidivism suggests that only new offenses that are the
same as the previous offense should be considered. Just as a return to the hospi-
tal for an ingrown toenail is not a medical recidivism for past hepatitis, some re-
searchers have questioned whether, for example, an arrest for gambling should
be accepted as recidivism for a past burglary sentence (Waldo & Griswold,
1979). However, if we consider recidivism, and the factors associated with it,
to reect a similar generality observed more broadly in criminal behavior, then
a measure of post-release performance should consider any criminal relapse (or
other deviant relapse, in the case of technical probation and parole violations)
as a recidivated event (Gottfredson & Hirschi, 1990).
As used in this study, the operational denition of recidivism has the limita-
tion of not capturing persons released from Oklahoma prisons who are then in-
carcerated in other states. Reports from the Bureau of Justice Statistics, however
(Beck & Shipley, 1989; Langan & Levin, 2002), indicate that the proportion of
persons released from one state and then arrested in another within three years
is small (between 5.5% and 7.4%), and the proportion reincarcerated likely
even smaller. The long tracking period (between 5.5 and 19.5 years) allowed
for a powerful measure of various possible predictors of recidivism, including
sex, age, race, education, number of prior incarcerations, offense type, length
of stay, release type (to parole, to probation, to discharge), sentence length, and
security level.
* Literature Review
Previous studies on recidivism have found a number of variables to be consis-
tent predictors of criminal recidivism, just as they tend to predict criminality in
general, especially age, gender, and having past offenses; younger persons are
more likely to reoffend than older ones, men more than women, and those who
have offended often in the past more than those who have offended less often
in the past (Beck & Hoffman, 1976; Gendreau, Little, & Goggin, 1996; Jones
& Sims, 1997; McGuire, Flanagan, & Thornberry, 1988; Orsagh & Chen,
1988). Other static characteristics that may predict whether offenders return
to prison are less established, but mentioned in criminological literature and
governmental reports. Jones & Sims (1997, p. 336), for example, note recidi-
vism often associated with “being young, African American, and unmarried,”
while Stantz and Tewksbury (2000, p. 326) similarly characterize the highest
risk subjects in a Kentucky home incarceration program as being “younger,
male, and African American.” Schmidt & Witte (1989, p. 151), in
a study of
To continue reading
Request your trial