Content uploaded by Andrew Bates
Author content
All content in this area was uploaded by Andrew Bates on Jun 16, 2016
Content may be subject to copyright.
Assessing Reconviction, Reoffending and Recidivism in a sample of UK
Sexual Offenders
Louise Falshaw*
Offending Behaviour Programmes Unit, HM Prison Service, UK
Andrew Bates
Thames Valley Project, Oxford, UK
Vaneeta Patel
Offending Behaviour Programmes Unit, HM Prison Service, UK
Carmen Corbett
Offending Behaviour Programmes Unit, HM Prison Service, UK
&
Caroline Friendship
Offending Behaviour Programmes Unit, HM Prison Service, UK
Address for correspondence:
Dr Louise Falshaw
Offending Behaviour Programmes Unit
HM Prison Service
Rm. 719 Abell House
John Islip Street
London
SW1P 4LH
Tel: 020 7217 5566 Fax: 020 7217 5824
Email: policy@prisons-obpu.demon.co.uk
Purpose: The base rate of sexual reconviction amongst sexual offenders is low.
Sexual reconviction, however, is currently the most commonly used outcome measure in
treatment evaluation studies. It is expected that offending behaviour treatment
programmes will reduce the likelihood of reconviction amongst participants. A low base
rate of sexual reconviction means that any reduction in reconviction (which could be
attributed to treatment) will be small and unlikely to be statistically significant. This
study aimed to assess other offence-related outcomes for sexual offenders, in addition to
reconviction.
Methods: The sample comprised 173 sexual offenders who had completed a
community offending behaviour treatment programme. Follow-up information was
collected for the sample from programme files containing multi-agency information.
Official reconviction rates were also obtained using both Home Office and police data.
Results: Collecting evidence of all sexually related offending behaviour for this
sample of sexual offenders raised the sample’s rate of sexual reconviction by a factor of
5.3.
Conclusions: The results show that broadening the outcome measure under observation
gives a better indication of the level of offence-related sexual behaviour perpetrated by
known sexual offenders. These results have implications for the outcomes measured in
treatment evaluation research for sexual offenders.
2
Offending behaviour treatment programmes are designed to reduce reoffending. To date,
reconviction rates have been the standard outcome measure used in England and Wales to
indicate the success or failure of offenders following their participation in a custodial or
community-based treatment programme. Reconviction does not represent reoffending
per se but has been accepted as a proxy outcome measure. The use of reconviction as an
accurate indicator of treatment success is dependent on the thorough recording of
conviction data. Friendship, Thornton, Erikson and Beech (2001), however, highlighted
the differences between the two main sources of conviction data in the UK, the Offenders
Index (OI), a Home Office database, and the National Identification Service (NIS), a
police database. They concluded that there were differences in the information recorded
by each source and that neither source appeared more reliable. The authors suggested a
combination of both datasets to increase the rate of reconviction in an offender sample.
The rate of sexual reconviction among sexual offenders i.e. the conviction of another
sexual offence during a defined period of time (Furby, Weinrott & Blackshaw, 1989), is
low and this is the case in England and Wales, Europe and North America (Grubin,
1998). In the UK, this rate has declined further over recent years (Friendship &
Thornton, 2001). Friendship and Thornton’s (2001) analysis of the sexual conviction
data indicated a decrease in the number of cautions/convictions for sexual offences in
England and Wales despite an increase in the number of reported sexual offences. In
fact, they found that the number of reported sexual crimes resulting in caution/conviction
in 1997 had halved compared to figures in 1981.
3
Friendship and Thornton (2001) also highlight how the low base rate for sexual
reconviction presents difficulties in the evaluation of offending behaviour treatment
programmes. It is anticipated that treatment will reduce the likelihood of reconviction.
When the base rate for sexual reconviction is so low to start with, it means the reduction
produced by treatment is likely to be small. A small reduction would be unlikely to
prove statistically significant and so the change could not be attributed, with confidence,
to treatment effect. This poses a problem for evaluators.
There are a number of ways in which this problem can be minimised, aside from the
combination of reconviction data from the OI and NIS. One such solution might be to
extend the follow-up period from which reconviction rates are derived from the standard
two years to five or ten years. This prolonged follow-up period, however, is impractical
in research terms and lacks the required immediacy for policy makers. An alternative
suggestion might be to access unofficial sources of data which provide a more real
assessment of reoffending. Marshall and Barbaree (1988) and Marques, Day, Nelson and
West (1994) have described the use of other outcome measures to supplement
reconviction data. Marshall and Barbaree (1988) assessed the unofficial level of sexual
reoffending among their sample by accessing the records of various local social agencies.
These results were then compared to officially recorded rates of sexual reoffending
provided by the Royal Canadian Mounted Police database. They found that the official
records identified the same sexual reoffenders as did the unofficial source, but the
unofficial source revealed many more. In fact, the unofficial records revealed 2.4 times
the number of reoffenders than the official source. Marques et al (1994) similarly
4
accessed both sexual and violent reoffence data through official records held by the
California Department of Justice and unofficial reports provided by the parole system.
The criterion for reoffence, in this case, was rearrest. Marques, Day, Nelson and West
(1993) discovered that the information provided by parole office records, combined with
official law enforcement records, increased the estimated number of serious crimes
committed by 33%.
In the criminological and psychological literature the terms reconviction, reoffending and
recidivism have been used interchangeably. In order to assess these different outcomes
some clarification of these terms is needed.
Reconviction can be defined as the conviction of an offence following a previous
conviction and is the narrowest outcome measure of the three. It refers only to offences
which are detected and successfully prosecuted. Reconviction is the most commonly used
outcome measure in UK evaluation studies due to its tangibility and accessibility through
official sources.
Reoffending refers to an illegal act committed by an individual who is already guilty of
previous criminal activity. Reoffending therefore incorporates offending behaviour
which has been officially recorded i.e. reconvictions but more than that also includes
offending undetected by the authorities.
5
Recidivism is a term mainly used in North American studies of offender rehabilitation.
Even in this literature, though, there is little consistency in the way this term is used and
the specific behaviour it refers to. In fact, Maltz (1984) identified the use of nine different
indicators of recidivism across a total of 90 US studies i.e. absconding, arrest,
incarceration, parole violation, parole suspension, parole revocation, reoffence and
reconviction. In the UK, recidivism is often used interchangeably with reconviction
although its definition encompasses much more. Recidivism means relapsing into
previous patterns of criminal behaviour (Maltz, 1984). By its definition, therefore,
recidivism does not only refer to behaviour that breaks the law, but can include any
conduct that is indicative of previous offence patterns. The authors would argue that
recidivism includes any offence-related behaviours. A specific example of this may be a
convicted child molester loitering outside a primary school. Recidivism is the broadest
definition and it incorporates both reconviction and reoffending.
The purpose of the current study was to assess the outcomes for sexual offenders,
specifically the rates of sexual reconviction, sexual reoffending and sexual recidivism in
a sample of sexual offenders. The study also aimed to calculate the differences between
these rates.
6
Method
Selecting a study area
To meet the aim of the study it was necessary to identify a regional area which was likely
to record a broad range of offending or offence-related behaviour for a group of sexual
offenders.
The Thames Valley Project (TVP) was chosen as the focus for this collaborative project
due to its multi-agency approach. TVP is a community-based assessment and treatment
provision working with male, adult sexual offenders across Oxfordshire,
Buckinghamshire and Berkshire. TVP is multi-funded and multi-agency and so has
connections with the probation service, social services, psychological services, health
authorities and Thames Valley police. Established in 1995, TVP has provided
therapeutic intervention to over 200 community-based sexual offenders. In March 2001
the treatment programme received accreditation1. Sexual offender referrals comprise not
only convicted sexual offenders, but individuals for whom there is evidence of sexually
abusive behaviour. These referrals are generally made by probation, often as part of a
community based penalty, police or psychiatrists. Offenders can even refer themselves
for intervention through their own GP. The main criterion for inclusion in the treatment
programme is acknowledgement by an offender of his own sexually abusive behaviour.
TVP also collects post-treatment information from various sources in relation to the
progress of its participants. These sources are the same as the referral agencies i.e.
1 In the UK an accreditation panel approves prison and probation run programmes to ensure they have a clear
theoretical framework, are supported by empirical research and are delivered to a prescribed set of standards.
7
police, health authorities, psychological services, probation services and social services
though probation appeared to be the most pro-active in feeding back information.
In this particular study, it was irrelevant that offenders had been through treatment as the
effect of treatment was not the focus of the study. This study aimed to assess access to the
wealth of follow-up information recorded by TVP from which we could calculate rates of
reconviction, reoffending and recidivism.
The Sample
Between July 1995 and May 1999, 173 sexual offenders had completed treatment with
TVP. In order to calculate reconviction rates of two years minimum for all participants
from the OI and NIS data, the end of May 1999 was chosen as the cut-off date. Follow-up
information was collected for each offender up to end May 2001.
The date of completion of the core treatment programme was taken, in this case, as the
beginning of the participants ‘at risk’ period in the community. As a result, the sexual
offenders in the sample had been at risk for varying periods of time (see Table 1). The
period at risk ranged from 2.0 years to 5.9 years with a mean of 3.9 years.
-----------------------------------
Insert Table 1 about here
-----------------------------------
8
At the time of treatment completion the average age of the sample was 40.6 years and the
age ranged from 17.8 years to 75.2 years.
Procedure
Follow-up information was gathered for all 173 offenders who had completed the TVP
treatment programme prior to May 1999. All information was held on file for each
participant. For those individuals on whom follow-up information was available, any
offence-related behaviour noted between the end of treatment and end of May 2001 was
recorded. Each incident was categorised by the researchers as sexual reconviction, sexual
reoffending or sexual recidivism using the definitions shown in Table 2. Conviction data
was obtained from both the OI and the NIS so that official reconviction rates could also
be calculated for the sample (see Friendship et al, 2001 for a more detailed description of
the data sources).
-----------------------------------
Insert Table 2 about here
-----------------------------------
The number of cases who had at any time been reconvicted of a sexual offence, sexually
reoffended or sexually recidivated during their at risk period, was identified by each
method of recording and these rates were then compared.
9
Results
Offenders Index
For 12% (n=21) of the sample two year reconviction rates could not be calculated as the
OI data-entry is six months in arrears. In addition, 50 offenders (29% of the sample)
were untraced by the OI. From the remaining 102 cases (59%), five (3%) sexual
reconvictions were identified.
National Information Service
Thirty-three cases (19%) were untraced by the NIS. Of the remaining 141 cases (82%)
for whom conviction information was available, 15 (9%) sexual reconvictions were
identified.
Thames Valley Project
From the total sample, TVP had a record of 10 offenders (6%) who had been reconvicted
of a sexual offence, 12 offenders (7%) who had sexually reoffended and 28 offenders
(16%) who had relapsed into previous offence-related sexual behaviour during their
follow-up period.
Figure 1 collectively displays these findings.
-----------------------------------
Insert Figure 1 about here
-----------------------------------
10
The results from the individual sources of data were combined (see Table 3). This
revealed the number of offenders who were recorded, either exclusively or collectively,
as having been involved in offence-related sexual behaviour during their follow-up
period.
-----------------------------------
Insert Table 3 about here
-----------------------------------
Combining the results for all methods of data collection produced the most complete
outcome information.
One offender was recorded by the OI as having been reconvicted of a sexual offence who
was not identified by any other source. The NIS identified six individuals as sexually
reconvicted who were not recognised by any other source. TVP had evidence of sexual
reconviction for two offenders who were not recorded by the NIS or OI, sexual
reoffending by four offenders not identified by either official source of data and sexual
recidivism for 20 offenders who were not identified by any other data source.
The sexual recidivism rate was 5.3 times greater than that provided by the OI and 1.8
times greater than that calculated from the NIS data.
11
Discussion
The sexual reconviction rate calculated from the OI data was 3% (n=5) for this sample.
This low figure is unsurprising and is supported by previous research (Friendship &
Thornton, 2001). Reconviction rates calculated from the OI are currently the most
common outcome measure used by treatment evaluators in the UK as the OI is the only
data source accessible to all researchers. The sexual reconviction rate for this sample
derived from the NIS was 9% (n=15). Again, this result was expected as the NIS has a
more inclusive remit which includes all offences as opposed to just the standard list2
offences recorded by the OI. Previous research has already highlighted that the OI and
NIS vary greatly in their identification of offenders who have been reconvicted of an
offence (Friendship et al, 2001). However, access to the NIS is for restricted personnel
only. Within the sample of 173 sexual offenders studied here, only nine were identified
by both data sources as receiving a sexual reconviction following their treatment
completion at the Thames Valley Project. The OI identified one additional other whilst
the NIS identified another six sexually reconvicted individuals. Which data source is the
most accurate is questionable, but even combining the results from the two sources (9%;
n=16) does not increase the low base rate for sexual reconviction substantially enough to
remedy the problems currently effecting treatment evaluation research.
Using the definitions suggested in this paper the Thames Valley Project provided
evidence of 10 (6%) sexual reconvictions among the sample. These results were an
improvement on those provided by the OI alone (n=5; 3%). They did not, however,
2 Standard list offences are all indictable or triable either way offences plus a few more serious summary
offences.
12
compare favourably with the 16 (9%) sexual reconvictions provided by the OI and NIS
combined. Measuring sexual reoffending (n=12; 7%) also improved on the rate
calculated from the OI, though again, it did not improve on the base rate provided by the
combined official sources of data. Adding the TVP reconviction rate and then the TVP
reoffending rate to the combined official reconviction rate, however, raises the base rate
to 10% and 12% respectively. Unofficial sources of reconviction and/or reoffending are
certainly beneficial when used instead of, but preferably in combination with, the OI
results.
The unofficial sexual recidivism evidence alone improved on all individual or
combinations of data discussed so far, producing a recidivism rate of 16% (n=28).
Combining this rate with those from the OI and NIS increased the combined official base
rate by 12%, to 21% (n=36) and substantially improved on the OI reconviction rate by
18%. Not only does using sexual recidivism as an additional outcome measure boost the
current low base rate for sexual reconviction, but it also provides a more sensitive
indicator of treatment success. In fact, Marques et al (1994) point out that reliance on
one outcome measure such as conviction or arrests, can produce unrepresentative
recidivism rates (Grossman, 1985) as well as an inaccurate indication of the effect of
treatment.
Unfortunately, measuring sexual recidivism is a resource intensive task and the accuracy
of the recording is only as good as the communication between sources. For instance, the
Thames Valley Project has good lines of communication with the referral bodies, in
13
particular, probation. However, they had evidence of 10 sexual reconvictions amongst
their treatment participants compared to the NIS’s 15 (only eight of which were identified
by both sources). This is why a combination of outcome measures is preferable to over-
reliance on one. In terms of the resource issue there is also a possible solution.
Whilst collecting recidivism rates from unofficial sources of data can initially be resource
intensive, obtaining official reconviction rates for the same sample allows the researcher
to calculate the relationship between the two rates. For example, Marshall and Barbaree
(1988) found that the unofficial rate was 2.4 times greater than that obtained through
official channels for their sample of sexual offenders. In this study, the recidivism rate is
5.3 times greater than the OI reconviction rate. If such a constant were calculated,
treatment evaluation studies need subsequently only collect official reconviction data to
obtain an indicator of the rate of recidivism.
This study has successfully indicated that collecting evidence of sexual recidivism both
raises the low base rate of sexual reconviction and provides a more accurate measure of
offence-related sexual behaviour perpetrated by known sexual offenders. In addition, this
piece of research has suggested how reconviction rates can be used to indicate the level
of recidivism in the same sample. The authors, however, propose that this project be
extended. The multiplication factor suggested here is unique to sexual offenders who
have completed the Thames Valley Project treatment programme and so is
geographically specific. The factor is also calculated from a sample which comprised
participants who had spent varying periods of time at risk in the community (from 2.0 to
14
5.9 years). If all offenders had spent an equal amount of time at risk and had completed
their treatment around the same time, the result would be more representative. In
addition, this study did not calculate a separate multiplication factor for different types of
sexual offenders. In order for the multiplication factor to be transferable, the recidivism
rates must be calculated for both prison and community based cohorts that are
geographically representative.
The results from this study show that broadening the outcome measure gives a better
indicator of the true level of offence-related sexual behaviour that is perpetrated by
known sexual offenders. Whilst this study was not concerned with evaluating the
effectiveness of a treatment programme, the results could have a notable effect on the
way that treatment evaluation research is conducted in England and Wales.
15
References
Friendship, C. & Thornton, D. (2001). Sexual reconviction for sexual offenders
discharged from prison in England and Wales. Implications for evaluating treatment.
British Journal of Criminology, 41, 285-292.
Friendship, C., Thornton, D., Erikson, M. & Beech, A. (2001). Reconviction: A critique
and comparison of two main data sources in England and Wales. Legal and
Criminological Psychology, 6, 121-129.
Furby, L., Weinrott, M.R. & Blackshaw, L. (1989). Sex offender recidivism: A review.
Psychological Bulletin, 105, 3-30.
Grossman, L.S. (1985). Research directions in the evaluation and treatment of sex
offenders: An analysis. Behavioral Sciences and the Law, 3, 421-440.
Grubin, D. (1998). Sex offending against children: Understanding the risk. Police
Research Series Paper 99. London: Policing and Reducing Crime Unit.
Maltz, M. (1984). Recidivism. London: Academic Press.
Marques, J.K., Day, D.M., Nelson, C. & West, M.A. (1993). Findings and
recommendations from California’s experimental treatment program. In G.C.N. Hall, R.
16
Hirschman, J.R. Graham & M.S. Zaragoza (Eds.), Sexual aggression: Issues in etiology,
assessment, and treatment (pp. 197-214). Washington, DC: Hemisphere.
Marques, J.K., Day, D.M., Nelson, C. & West, M.A. (1994). Effects of cognitive-
behavioral treatment on sex offender recidivism. Preliminary results of a longitudinal
study. Criminal Justice and Behavior, 21, 28-54.
Marshall, W.L. & Barbaree, H.E. (1988). The long-term evaluation of a behavioral
treatment program for child molesters. Behavior Research Therapy, 6, 499-511.
17
Table 1. Time at Risk (N = 173)
Time at Risk % (n)
2 to 3 years 30.6 (53)
3 to 4 years 21.4 (37)
4 to 5 years 22.6 (39)
5 to 6 years 25.4 (44)
18
Table 2. Definitions
Sexual Reconviction The conviction of another sexual offence during a specified
time period
Sexual Reoffending The perpetration of another illegal sexual act during a specified
time period (whether caught or not)
Sexual Recidivism The commission of an offence-related behaviour, legal or
illegal, during a specified time period with a clear sexual
motivation
19
Figure 1. Sexual reconviction, reoffending and recidivism identified using different
methods of measurement (N = 173)
20
0
2
4
6
8
10
12
14
16
Percentage
OI
reconviction
NIS
reconviction
TVP
reconviction
TVP
reoffending
TVP
recidivism
Table 3. Percentage of offenders recorded as having been involved in any offence-
related sexual behaviour when results from individual data sources merged
(N = 173)
Data Sources % (n)
Individual data sources
OI 3 (5)
NIS 9 (15)
TVPrecon 6 (10)
TVPreoff 7 (12)
TVPrecid 16 (28)
Combined data sources
OI & NIS 9 (16)
OI, NIS & TVPrecon 10 (18)
OI, NIS, TVPrecon & TVPreoff 12 (20)
OI, NIS, TVPrecon, TVPreoff & TVPrecid 21 (36)
21
