Publication - Research and analysis

What Works to Reduce Reoffending: A Summary of the Evidence

Published: 8 May 2015

This is an updated version of the original review entitled ‘What Works to Reduce Reoffending: A Summary of the Evidence’, published in 2012.

What Works to Reduce Reoffending: A Summary of the Evidence


This chapter attempts to relate the evidence to the work of policy-makers and practitioners. It relates the findings of the evidence review to intermediate outcomes of offender interventions and non-criminogenic needs, then summarises the implications of the evidence for the way we work with offenders, and lastly outlines a recommended approach to evaluating projects in Scotland.

Intermediate outcomes - targets for interventions

The evidence reviewed in this paper suggests a number of areas in which work with offenders should be focused as intermediate outcomes with the ultimate aim of reducing reoffending (see Chapter 1, Figure 1). The following factors that have been identified as being associated with a reduced chance of an individual reoffending[506] are supported by the evidence reviewed, and reasonably consistently throughout the literature[507]:

  • Reduced or stabilised substance misuse;
  • The ability to access and sustain suitable accommodation;
  • Finding suitable employment;
  • Improvements in the attitudes or behaviour which lead to offending and greater acceptance of responsibility in managing their own behaviour and understanding of the impact of their offending on victims and on their own families;
  • Maintained or improved relationships with families, peers and community;
  • The ability to access and sustain community support[508].

A series of four rapid evidence assessments reports on intermediate outcomes and reoffending published by the Ministry of Justice in 2013[509] looked at a variety of intervention types (mentoring, family relationships, peer relationships and the arts) and the intermediate outcomes they sought to achieve. These addressed a similar range of outcomes, including improving educational outcomes, improved behaviour, improving or maintaining pre-existing relationships with partners and/or children, improving peer relationships, improved communication, improved employment outcomes, improved housing situations, and reductions in substance misuse.

Non-criminogenic needs

Non-criminogenic needs such as trauma and victimisation are highly prevalent in some offenders, but have not been found to have a direct association with reoffending behaviour. For example:

  • The evaluation of the Glasgow 218 Centre[510] reported that abuse and trauma were a significant feature of the lives of the women, and it cites direct interviews with arguably similar populations of women in HMP and YOI Cornton Vale in 1997, that also revealed high rates of abuse.
  • Monitoring data for Scotland's Women's Community Justice Services (WCJS)[151] show that, of the 107 women in the three WCJS recording domestic violence, abuse, or trauma, 70% entered the service with a history or symptoms of abuse. The range of mental health-related issues described by women and practitioners included (but were not limited to) confidence and self-esteem, anxiety, depression, isolation, stress, anger management, borderline personality disorder, and symptoms of complex trauma.
  • In a large-scale study of prisoners' mental health needs conducted on behalf of the Department of Health in England and Wales, over 66% of women in prison and 21% of female remand prisoners were found to have depression, anxiety and phobias (which compares with 20% of women in the community)[512]; however, these neurotic disorders (unlike personality disorders experienced by around 50% of female offenders) have not been found to be strongly related to reoffending[513].

Despite the lack of evidence linking these factors directly to reoffending - which is why they are not considered as primary outcomes for interventions aimed at reducing reoffending - some studies suggest that non-criminogenic needs such as poor mental health may have an indirect link with reoffending behaviour. For example, experiences of being victimised may contribute to the onset of mental health problems and other criminogenic risk behaviours such as drug abuse that may subsequently lead to reoffending[514]. This would suggest that that non-criminogenic needs are still important to address alongside criminogenic needs to help service users to sustain engagement with services (as discussed in Chapter 2 under the links between the GLM and RNR models), to address the underlying causes of behaviours such as substance misuse, and to benefit from interventions.

Furthermore, it should be borne in mind that support and interventions may be of value for other reasons than their perceived impact on recidivism. For example, such processes may be necessary in order to respect the human rights of former offenders and enact the duties of states to provide adequate care to their citizens.

Implications for approaches to working with offenders

One of the most consistent findings of this evidence review is that one-size-fits-all interventions are ill-suited to reducing reoffending, and that there are differences between individuals who offend. In practice this may mean that there are differences between how the criminal justice system supports individuals to stop offending based on the differences between individuals and variations in local context or service provision.

Research has also suggested that people who offend, and in particular women who offend, may face challenges in a number of the areas outlined above. Agency joint-working is likely to be important for this, but how joint working is experienced by service users is also important. In practice joint-working between agencies may in some cases be perceived as presenting additional bureaucratic hurdles to people who offend, and to reduce the quality of the relationship between supervisor and probationer[515]. Further work is required to evaluate partnership working efforts and how they are experienced by service users.

There is currently a lively debate in the field of desistance research, but the following contributors to desistance may be important to future decisions about how we work towards achieving outcomes with offenders:

  • Developing a non-offender identity is increasingly seen as important in reducing reoffending[516].
  • Motivation to change, and hope, may also be important to desistance[517].
  • Desistance research stresses the importance of individuals' self-efficacy and agency (that is, belief in one's own ability to complete tasks), and suggests that establishing a sense of agency is important in desisting from crime.

These factors typically represent the findings of a different type of research to those described above, reflecting offenders' experiences rather than statistical correlations with reoffending. Further work is required into exactly how these outcomes would be measured if they were to be investigated quantitatively[518]. It may also be the case that these findings represent the process by which offenders arrive at the outcome of reoffending[519], and so may relate more to ways of working with offenders rather than desirable outcomes for offenders. More research is required to understand the implications of this research for interventions in practice.

It is also worth noting that the literature seems to be divided as to the value of different approaches[520], which can complicate their interpretation for a policy or practitioner audience; it is not simply a case of basing recommendations on a neutral body of 'evidence' as the findings of different research projects also to some extent represent different researchers' theoretical perspectives. This theoretical disagreement may raise problems in trying to synthesise the findings of different researchers.

Evaluations of Scottish projects

A number of evaluation projects are currently investigating interventions and services to reduce reoffending that are being trialled in Scotland, and they are noted throughout this review. As per Scottish Government guidance, these evaluations are based on a logic model approach[521]. This approach seeks to measure the impact of interventions on intermediate outcomes derived from criminogenic needs (such as anti-social attitudes and peers, substance misuse, lack of employment, and homelessness) and based on the outcomes and evidence that are relevant to the intervention, rather than focusing directly on the long-term outcome of reduced reoffending. This provides an alternative to experimental designs and can be adopted where such trials are inappropriate due to, for example, a lack of participants, or technical or ethical constraints. Evaluations are then assessed on whether they worked in accordance with the logic model.

Whilst logic models cannot 'prove' the impact of an intervention in the same way as an experimental design, researchers have suggested that experimental designs by themselves are unsuitable to provide a full answer to the question of 'what works'[522] as they cannot investigate why an intervention may have worked. As such the logic model approach is intended to provide useful evaluation information about areas in which interventions should have worked, based on the available evidence.

Evaluations of projects in Scotland may be especially valuable given the distinctive nature of the Scottish justice system[523]. Sampson suggests that the aims of assessing whether a particular programme worked and whether a policy based on a study will work are not the same, due to the process of implementation which involves working across different contexts[524]. For example - although the reasons for this are debated - CBT evaluations show lower efficacy in the UK than they do internationally[525]. As such evaluating the implementation of projects in Scotland may be especially useful to inform policy-making in Scotland.


Email: Justice Analytical Unit