A diverting quarter century? Evidence-based police-led diversion 25 years on

Introduction

It is more than 25 years since Lawrence Sherman delivered his ‘Ideas in American Policing’ lecture on ‘evidence-based policing’ for the US Police Foundation. The original paper (Sherman, 1998), the equally seminal study on ‘preventing crime’ (Sherman et al., 1997) and the subsequent development of evidence-based policing in the ‘Triple T’ (Sherman, 2013) have been immensely influential as well as, for some, controversial. In 1998, Sherman challenged scholars, practitioners and policymakers to use the ‘best available evidence’ to guide policy and practice. He then posed four questions: ‘What is it? What is new about it? How does it apply to a specific example of police practice? How can it be institutionalized?’ (Sherman, 1998: 3).

Taking the lead from Sherman's questions, this contribution to the 25-year anniversary collection focuses on one example – police-led diversion – to illustrate the development and challenges of institutionalizing evidence-based policing since 1998. This is a very personal narrative: the author was the national police lead for out-of-court disposals and chaired the group set up to create conditional cautioning in England and Wales in 2002; as Chief Constable of Thames Valley Police, the author was one of the three police force sponsors of Sherman and Strang's programme of research into restorative justice, including its deployment as a part of diversion (Strang et al., 2013); then, as co-author with Sherman of Offender-Desistance Policing (Sherman and Neyroud, 2012), the author set up and managed a randomized controlled trial (RCT) – Operation Turning Point – in Birmingham that tested the effectiveness of deferred prosecution (Sherman et al., forthcoming); Turning Point was a central part of the author's own PhD, supervised by Sherman (Neyroud, 2017); the author was then commissioned by the National Police Chief's Council to assess the likelihood, based on the ‘best available evidence’ that a proposed ‘two-tier framework’ for police-led diversion would work (Neyroud, 2018). The article, therefore, charts a 25-year personal journey from practitioner to pracademic to academic as the author's role has changed from senior police leader to researcher and programme director for the Cambridge ‘Police Executive Programme’.

From simple cautions to conditions

Cautioning by the police in the UK has had a long history (Steer, 1970) but, until the past 20 years, there was very limited UK-based research on its effectiveness compared with prosecution (Neyroud, 2017). However, from a number of important studies that reviewed specific aspects of diversion we know that: celerity or the speed with which the process is concluded following an arrest matters (Giller in Jones, 1982 – the author assisted in this study of cautioning in Hampshire while still a probationer constable); cautioning may be discriminatory and one key explanation for this is the requirement to admit guilt (Lammy, 2017; Landau and Nathan, 1983); cautions have not always been applied consistently (Evans and Wilkinson, 1990; Giller and Tutt, 1987; Laycock and Tarling, 1985; Mott, 1983; Sandars, 1988); cautions using restorative justice appear to be effective (Strang et al., 2013); and diversion for drug offenders appears to be promising (Harvey et al., 2007). We also know from Petrosino et al. (2010) that formal processing through the criminal justice system does not have a crime control effect but rather, across the measures in their systematic review, appears to produce a backfire effect.

As for effectiveness, there is a collection of experimental studies dating back to the 1960s that were almost exclusively US-based and explored varied diversion approaches with variable quality. These were systematically reviewed by Wilson et al. (2018) and found to demonstrate a modest, but significant, effect in preventing reoffending. The ‘best available evidence’ (Sherman, 1998) suggested that police-led diversion, for young offenders at least, was a promising intervention.

One issue that Wilson et al. (2018) left open was the question as to whether diversion worked better when linked to conditions and treatments. The absence of clear evidence on this issue should perhaps have been given more attention when a major change to cautioning was introduced in England and Wales in 2002. Conditional cautioning was intended to divert adult cases that would otherwise have been prosecuted (Blakeborough and Pierpont, 2007). The legislation introduced the first statutory regime for diversion in England and Wales and the first to mandate the use of treatment conditions. The approach was piloted, and a process evaluation carried out, which identified some very significant implementation challenges (Blakeborough and Pierpont, 2007).

One of the key issues that Blakeborough and Pierpont (2007) highlighted was the skills gap in policing when it came to setting conditions. Despite this early warning, there was no substantive attempt to track or follow up the early research. The author reviewed the first 10 years of conditional cautioning when starting the work to develop Operation Turning Point. The data showed that more than 80% of the cautions had been given for low-level criminal damage and the overwhelming preponderance of conditions centred on an apology and the payment of small sums in reparation. In short, particularly given the very small numbers of conditional cautions given, the main policy aim, to divert significant numbers of cases, had clearly failed.

Offender-desistance policing and Operation Turning Point

With the start of a prolonged period of austerity in the UK in policing and criminal justice in 2010, the problem that conditional cautioning was supposed to address – the resource-thirsty overuse of prosecution for low-level offences and low-harm offenders – became more acute. Yet the new government's policy towards diversion in 2011 was driven not by the best available evidence, but by the combination of a think tank report from the Policy Exchange (Sosa, 2012) and an inspection by the Criminal Justice Joint Inspectorates triggered by that report (CJJI, 2011). The policy that eventually emerged (Ministry of Justice, 2014) was framed as ‘holding all offenders to account’ and to do so it was proposed to adopt a ‘two-tier’ framework in which both tiers – community resolutions and cautions – required conditions. Yet, the evidence gap on the effectiveness of conditions and the problems of implementing them effectively had not been addressed.

By contrast, the development of Operation Turning Point focused on using the best available evidence and testing it in the field with the first large-scale RCT to compare the effectiveness of prosecution with a diversion with conditions. Starting in 2011, and funded by the Monument Trust with the express aim of finding ways to reduce the use of prison for low-harm offenders (Sherman and Neyroud, 2012), the programme focused on offenders that the police had made the decision to prosecute and randomly assigned them to a control – charge and prosecution – or treatment – a deferred prosecution with conditions. If they consented, the treatment group met a police offender manager within 48 h and worked to agree a short set of conditions designed to support desistance.

Implementation of the experiment was far from straightforward (Neyroud, 2017). The first two stages demonstrated clearly that police needed training and support to set and track conditions (Slothower et al., 2015). There were lessons for setting eligibility criteria, for matching conditions to offenders and for tracking compliance. More than 400 offenders, 40% of whom had committed violent offences, 40% property offences and 20% drugs offences, were included in the trial. The offenders were block randomized into four groups: young offenders with and without personal victims and adult offenders with and without personal victims. The evaluation, using two years’ post-trial arrest and offending data, showed that the treatment was effective in reducing reoffending and harm committed by adult and juvenile offenders who had committed crimes with personal victims (Sherman et al., forthcoming).

Even before the Turning Point outcomes were available, a replication of key elements was already under way in Durham. Checkpoint used the same deferred prosecution treatment but incorporated two key changes: the use of trained ‘navigators’ rather than police office offender managers to set and manage conditions; and an algorithmic triage to determine the eligibility of low- and moderate-risk offenders. The Checkpoint experiment also had some challenges in implementation, but the outcomes reported were at least as effective as court prosecution and at a significantly better cost–benefit ratio (Weir et al., 2022).

Replication in policing and criminal justice is crucial but regrettably uncommon and, yet, Turning Point has now been replicated twice. The second – Turning Point (London) – has been led by Harber, who documented the challenges of replication and the process of implementation (Harber, 2019). Although the final evaluation has yet to report, the interim findings have followed a similar pattern to those of Turning Point (Birmingham) and Checkpoint (Harber and Neyroud, 2023).

From best available evidence to evidence-based policy?

Meanwhile policy was continuing to develop. The Ministry of Justice proposals for a two-tier model were developed into a three-force pilot, with an evaluation designed as a quasi-experiment (Ames et al., 2018). The intention of the pilot and evaluation – to test the two-tier model before wider implementation – was promising, as was the commitment, as part of the research, to carry out a full cost–benefit exercise. However, a number of the key lessons from the literature and the development of Turning Point were not incorporated into the pilots: for instance, a staged implementation incorporating learning and feedback, investment in training on condition setting and an active tracking regime. As a result, there were significant issues with quality and consistency in the trial. One of the main recommendations of the evaluation was that a six-month pre-implementation phase was required to train, test and secure providers for conditions. The problems that the absence of this run in phase caused may well have contributed to the findings that there was no significant difference between the control – the existing system of cautioning – and the two-tier treatment model in terms of reoffending or harm. This was compounded by the cost–benefit study, which found that the two-tier model compared unfavourably on cost (Ames et al., 2018).

Putting the cost–benefit findings to one side, the Ministry of Justice prepared to proceed with the two-tier model. However, the first results of the Turning Point and Checkpoint evaluations influenced a parallel strand of further research to explore the deferred prosecution model and the introduction of a new disposal – ‘Outcome 22’ – to the Home Office Counting Rules, which permitted forces to record deferred prosecutions as an official outcome.

Despite this door being left open to deferred prosecution, the momentum towards a new statutory diversion framework produced a section of the Police, Crime, Sentencing and Courts Act 2023 that, once enacted, would replace the existing cautioning framework with two new conditional cautions – a diversionary and community caution (Police, Crime, Sentencing and Courts Act, 2022).

Although it is possible to argue that the Turning Point and Checkpoint research has begun to answer some of the questions about diversion with conditions, there remain considerable questions about the most effective conditions for different offenders and offences. Wilson and Lipsey (2024) have recently completed a crucial systematic review of the current evidence on treatments. Findings on effective treatments are but one component of an evidence-based approach to a new diversion regime. It is one thing knowing that treatments might work, quite another to implement an evidence-based approach to diversion: that requires more attention to triage, eligibility, assessment and tracking.

Evidence-based policing and diversion 25 years on: Some concluding thoughts

There has been an exponential increase in the availability of systematic and experimental evidence in policing, including on police-led diversion, over the past 25 years. The author's own PhD research on RCTs in policing up to 2016 found only 122 completed in the 50 years since the first policing RCT in the 1960s (Neyroud, 2017). Yet by 2022, Mazerolle et al. (2023), using the Global Policing Database, found that that number had increased to nearly 500 in less than a decade.

In several areas of policing, the combination of systematic reviews and experiments has been highly influential on operational practice and government policy. The most obvious example of this is hotspot policing of violence. Braga et al.'s (2021) systematic review and two RCTs in Essex and Bedfordshire (Basford et al., 2021; Bland et al., 2021) have led to national roll out of Operation GRIP across 43 police forces in England and Wales in 2024.

Police-led diversion has been a more complicated area to apply evidence-based approaches and is a challenging area in which to achieve high-integrity field-testing. Despite having been involved as a practitioner and then as the national lead for policing on diversion, the author learned more about the key elements of success from conducting the Turning Point experiment than from operational practice alone. Shepherd (2002) recommended that police develop practice in field universities, akin to teaching hospitals. The lessons of the development and testing of the deferred prosecution model of police-led diversion appear to support Shepherd. Those lessons, set out above, also appear to suggest that such a model needs also to embrace policymakers and civil servants advising on the development of policy. Without this broader conception, field-testing in partnership between academic researchers and practitioners may produce important findings but fall short of framing the policy choices that lead to legislation.