Does parole supervision reduce the risk of re-offending?
Author | Don Weatherburn,Suzanne Poynton,Wai-Yin Wan |
Published date | 01 December 2016 |
Date | 01 December 2016 |
DOI | http://doi.org/10.1177/0004865815585393 |
Subject Matter | Articles |
Article
Australian & New Zealand
Journal of Criminology
2016, Vol. 49(4) 497–511
Does parole supervision reduce
! The Author(s) 2015
Reprints and permissions:
the risk of re-offending?
sagepub.co.uk/journalsPermissions.nav
DOI: 10.1177/0004865815585393
Wai-Yin Wan, Suzanne Poynton and
anj.sagepub.com
Don Weatherburn
NSW Bureau of Crime Statistics and Research, Sydney, New South
Wales, Australia
Abstract
Although a large number of offenders are released to parole each year, little is known about
the effectiveness of parole supervision in reducing re-offending. The few studies that have
been conducted provide mixed results and, for the most part, have been unable to rule out
the possibility of selection bias. The present study is the first to evaluate the effectiveness of
parole supervision using propensity score matching techniques. It compares two groups of
offenders, carefully matched in terms of factors likely to affect re-offending but differing in
terms of whether they are supervised. The results suggest that parole supervision does
reduce the risk of re-offending.
Keywords
Offence seriousness, parole, propensity score matching, re-offending, supervision
Introduction
In the United States, about two-thirds of prisoners (67.8%) are arrested for a new crime
within three years of release, while three quarters are arrested within five years (Durose,
Cooper, & Snyder, 2014). The situation in Australia is very similar. Jones, Hua,
Donnelly, McHuchison, and Heggie (2006), for example, found that 64% of offenders
released from prison in New South Wales (NSW) were reconvicted of a new offence
within two years of release.
The principal means by which correctional authorities attempt to minimise the risk of
re-offending among released prisoners is placement on parole. The Productivity
Commission (SCRGSP, 2013) puts the annual net recurrent cost of supervising an
offender in the community at $22.54 per offender per day (or $26.23 per prisoner per
day in NSW). Considering that more than 5000 offenders are released to parole in NSW
each year (CSNSW, 2014, see Table 12 of the report), parole programs constitute a
significant investment. It is somewhat surprising, then, to discover that the effectiveness
Corresponding author:
Don Weatherburn, NSW Bureau of Crime Statistics and Research, Sydney, New South Wales, Australia.
Email: don_j_weatherburn@agd.nsw.gov.au
498
Australian & New Zealand Journal of Criminology 49(4)
of parole in reducing re-offending remains the subject of considerable debate, not just in
Australia but in other countries as well.
Reviewing the situation in Britain, Shute (2004, p. 321) remarked that after 35 years
of research it was still unclear whether parole supervision has a beneficial effect on
recidivism. In its 2005 review of the impact of post-prison supervision, the Urban
Institute noted that, although 774,000 men and women in 2003 were under parole super-
vision in the United States, ‘remarkably little is known about whether parole supervision
increases public safety or improves re-entry transitions’ (Solomon, Kachnowski, &
Bhati, 2005, p. 1). A year later, Listwan, Cullen, and Latessa (2006, p. 20) expressed
similar sentiments: ‘We still know relatively little about the overall effectiveness of
parole, and even less about the effectiveness of the ‘newer’ re-entry programs’. Little
has changed since these reviews. McNeill (2006, 2009), for example, has discussed pos-
sible explanations for the frequent failure to find evidence of parole’s effectiveness.
Trotter (2012) and Bonta et al. (2010) have found evidence that re-offending on
parole is less likely when parole supervision is carried out by properly trained parole
officers. Convincing evidence that offenders released to parole are less likely than offen-
ders released without supervision, however, remains hard to find.
Some of the best evidence we have on the effectiveness of supervision in reducing re-
offending risk is now quite old. Petersilia and Turner (1993) conducted a nine-state
randomized trial of intensive supervision in the United States. That trial found little
evidence that intensive supervision reduces the risk of re-offending. Jackson (1983)
analysed the re-offending rates of 314 young offenders on parole in California who
were randomly assigned to discharge from parole or retention on regular parole super-
vision. He found no difference between the two groups in rates of arrest or conviction, or
in time to arrest or conviction. Solomon et al. (2005) compared rates of re-arrest among
38,628 prisoners released in 1994 from 15 American states who were either released
unconditionally, given mandatory parole (not supervised) or given discretionary
parole release (released under supervision). After adjusting for differences between the
groups in a large range of factors, they found no overall difference in re-arrest risk or
frequency between those released unconditionally and those placed on mandatory
parole, and only a slightly lower risk of re-offending among those placed on discretion-
ary parole release. Drake and Barnoski (2006) exploited the conditions of a natural
experiment created by a legislative change in Washington State that eliminated parole
supervision and then reinstated it 12 months later. They also found no evidence that
parole supervision reduced the risk of re-offending.
Other studies have found quite different results. Studies of offenders placed on com-
munity supervision orders (rather than imprisoned) generally find that more intensive
community supervision produces lower rates of offending than less intensive supervision
if accompanied by treatment or measures to address the offender’s criminogenic needs
(Aos, Miller, & Drake, 2006; MacKenzie, 2002). More recently, Ostermann (2013) com-
pared three years of post-release data from nearly 30,000 offenders released from prison
either unconditionally or conditionally between 2005 and 2007. The control variables in
his analysis were extensive, and included: age, gender, minority status, marital status, a
deprivation index of the county of return, a pre-release Level of Service Inventory–
Revised (LSI-R) score, the number of offenses for which the individual was incarcerated,
the type of offense for which the individual was incarcerated and the number of arrest
Wan et al.
499
events the individuals had on their official record prior to their release. His findings
indicated that, after three years, the reoffending rate of parolees was 1% lower than
that for unconditionally released inmates. The reoffending rate of parolees who were
assigned supervision terms of at least three years was eight per cent lower than among
those released unconditionally.
It is possible the inconsistent results obtained in studies of parole effectiveness simply
reflect differences between jurisdictions in the level of supervision and support provided
to parolees. Given the non-random nature of conditional/unconditional release decision,
it is also possible the conflicting results are a result of selection bias. The ideal defence
against selection bias is a randomized trial. In most situations, however, it is impossible
to randomly allocate offenders leaving prison to parole (viz., supervised release) or
unsupervised release. This makes it difficult to determine whether differences in
re-offending between those released on parole (treatment) and those released without
parole supervision (control) are due to the effect of parole or to pre-existing differences
between those treatment and control groups (i.e. selection bias). Most researchers resort
to some form of regression analysis to deal with the problem. A common approach is to
estimate the effect of parole on risk of re-offending post-release using a logistic regression
model to adjust for other factors (e.g. age, prior criminal record) thought likely to
influence both selection into treatment (parole) and re-offending risk. If the coefficient
on the variable indicating release to parole is significant after adjusting for these con-
founding factors, the researcher concludes that parole has had an effect on re-offending.
There are a number of problems with this approach to controlling for the influence of
extraneous variables in correctional program evaluation. Nagin, Cullen, and Johnson
(2009) have argued that the minimum controls required in any evaluation of the deter-
rent effect of prison are age, race, gender, index offence and prior convictions. There is
no reason why evaluations of parole should be any different; however, some studies do
not include controls even for this limited group of factors. Broadhurst and Maller
(1990), for example, only controlled for race. Broadhurst and Loh (2003) controlled
for race, sex, age and whether or not the offender had a prior criminal record but not
the index offence. To be fair, neither of these studies set out to evaluate the effectiveness
of parole. Studies that have set out with this objective, however, have also sometimes
included only limited controls. Ellis and Marshall (2000) did not control for the types of
prior convictions although research has since revealed that the type of prior conviction is
a predictor of re-offending (Poynton and Weatherburn, 2013). Solomon et al. (2005)
controlled for criminal history, age, race, and offence type but not for other factors, such
as previous breaches of community-based orders, that might have been expected to
influence whether or not a prisoner was released into the...
To continue reading
Request your trial