Cynthia McDougall, Amanda E. Perry and David P. Farrington
This book had as its starting point a systematic review of criminal justice interventions aimed at reducing crime in the UK. The UK review was initiated to fill a gap in the current literature on the effectiveness of criminal justice interventions specifically relating to the UK. Its inspiration came from three sources: first, the UK Government's commitment to basing policy development on research evidence; second, the concurrent enthusiasm of practitioners for research knowledge; and third, from the influential publications by Sherman et al. (1997) on Preventing Crime: What Works, What Doesn't, What's Promising, and Sherman, Farrington, Welsh and MacKenzie (2002) on Evidence-Based Crime Prevention, reviewing research conducted in the US and internationally.
HM Home Office initially sponsored the systematic review and the findings are presented throughout this book. The UK studies are part of a programme that had three main aims: (i) to summarise the UK research evidence on reducing crime; (ii) to identify the gaps in research knowledge; and (iii) to stimulate research to fill these gaps. It was a similar process to that initiated by the US National Institute of Justice that led to the original Sherman et al. (1997) publication. This review of UK effectiveness studies is seen as complementary to the Sherman et al. (1997; 2002) publications, but also has some notable differences. The UK review has included only those studies that specifically reported on actual crime, self-reported crime or reconviction data. The latter constraint has resulted in the exclusion of a number of UK studies in the field of, for example, early childhood and school interventions, that have used as outcome measures, risk factors for crime, such as antisocial behaviour or truancy, rather than offence data. In addition, the UK review concentrates on the economic evaluations reported in the studies, focusing not only on what works with whom, but also at what cost. Therefore, the range of studies included in the review differs in some respects from that of Sherman et al. (1997; 2002). A similarity, however, is that recognised experts have been invited to update and comment on the systematic review findings, and to place them within a wider context of international research and policy development in their area of expertise.
Despite the slight difference in focus, we have recognised the value of the clear and practical framework adopted by Sherman et al. (1997; 2002). The UK review has followed their example by allocating studies to the categories of What Works, What Doesn't, What's Promising, What's Unknown, using their strict quality criteria for allocation to those categories. Where appropriate we have indicated the related supporting evidence in international research.
It will be noted that, in the original UK review, only a small number of UK studies found their way into the What Works category, despite the fact that inclusion in this category only required positive results from two well-conducted studies. This, in our view, makes the case for examining separately the research-evidence base of UK studies, but within the context of international research findings, so that gaps in our knowledge can be clearly identified.
There is little doubt that the exacting research quality standards, similar to those adopted by Sherman et al. (1997; 2002), have reduced the number of studies included in the review. From a starting point of 1,499 research papers identified from a range of databases, through the various stages of screening, the number of included studies was reduced to 62. There are competing views as to whether such a rigorous exclusion process is appropriate. Some researchers will argue that rigorous research in an operational setting is impossible to achieve, and that compromises are necessary and acceptable in order to learn from the large amount of research available. Indeed, some of the authors in this book support that view. There is, however, an alternative view that only by setting exacting standards will the overall quality of research, and hence the understanding of ways to reduce crime, improve. This is the view held by the Cochrane and Campbell collaborations (www.cochrane.org/; www.campbellcollaboration.org/): international organisations that strive to disseminate research findings on health and social science interventions, based on quality research and systematic review. Their commitment to quality research and systematic review has the objective of reducing bias in reporting research outcomes-through transparency of review methods, examining all available literature published in journals and in less formal publications, ensuring that results from both positive and negative outcome studies are reviewed, and reporting only on those studies that have been rigorously conducted so that bias attributable to methodology is kept to a minimum.
THE CASE FOR QUALITY RESEARCH
The case for quality research has long been established in the field of medicine (for example, the Cochrane Collaboration). It is now no longer questioned that a treatment should be thoroughly tested before being administered to patients. Concerns over potentially harming patients through the use of untested methods are rated above the ethical arguments regarding withholding treatments that might be beneficial. The Cochrane Collaboration has been the champion of this approach in the medical field, and is now followed by the Campbell Collaboration, which supports similar standards in the field of social sciences. The Campbell Collaboration Crime and Justice Group (Farrington & Petrosino, 2001) is an international organisation that is now seeking to raise standards of research in criminal justice, and to dispute the assumption that criminal justice interventions should be implemented without adequate safeguards.
It is widely assumed that, even if an intervention is not effective in reducing crime, it is unlikely to do any harm. Since the inception of the Campbell Collaboration Crime and Justice Group and its support for systematic reviews of criminal justice interventions, this view is being challenged. For example, Petrosino, Turpin-Petrosino and Buehler (2002) gave strong evidence to the contrary in their systematic review of 'Scared Straight' programmes. 'Scared Straight' programmes were originally proposed by prison inmates and based on an assumption that delinquency could be prevented by giving youngsters at risk of offending a taste of what it would be like to be imprisoned. The project started in Rahway Prison in New Jersey, and received a great deal of international attention when televised. 'Scared Straight' projects were eventually adopted in 38 states across the US, with a number of programmes introduced in the UK without being evaluated. When a randomised controlled trial was finally conducted to assess the effectiveness of the San Quentin Squires programme (Lewis, 1983), it was found that 81 per cent of the experimental group, compared to 67 per cent of the control group, had been arrested at a 12-month follow-up.
The review by Petrosino et al. (2002) describes this and other scientifically credible evaluations that have demonstrated that attempts to scare teenagers into better behaviour is not a successful enterprise, and can even be counterproductive.
A study which similarly showed that interventions can cause harm was cited in Sherman et al. (1997). Sherman described a 30-year follow-up of a rigorous mentoring study, the Cambridge-Somerville experiment, which began in 1937:
The results of this intensive mentoring showed no difference between treatment and control groups in criminal records, either in 1942 (Powers and Witmer, 1972) or in 1975-76 (McCord, 1978). The long-term follow-up, however, did show significantly higher levels of diagnosed alcoholism, serious mental illness, and stress-related physical health problems. A higher level of unfavourable life outcomes, although not specifically greater crime, among the treatment group seems clear. What is less clear is the meaning of the results for the value of mentoring programs today. (Sherman, et al., 1997).
This finding has not been satisfactorily explained, although possible reasons have been proposed by Sherman et al. (1997), for example: 'that mentoring is an artificial source of support which makes it harder for mentored boys to adjust as adults'; 'that the abrupt departure of these long-term counsellors from the boys' lives was as damaging emotionally to the boys as a divorce or other loss of parental involvement'; or that 'the treatment boys had no greater rate of personal problems, but when they had problems they were simply more likely to seek professional help of the kind the programme had taught them to seek'.
The conclusions drawn from the two interventions described above are that interventions that may have a superficial appeal as being likely to impact on reconvictions should be carefully evaluated before being widely implemented; there should be regular monitoring of interventions; and appropriate follow-up periods should be observed. In the UK review, follow-up periods ranged from 4 weeks to 10 years. Since reoffending rates vary across offences, the length of follow-up needs to take account of what is being assessed and the nature of the offending behaviour in question.
THE CASE FOR SYSTEMATIC REVIEWS
An important aspect of evaluation research is how the results are disseminated to practitioners. It is very easy in reviewing research literature to mislead as to what is effective through selectively choosing literature that supports the preferred review. It is true, however, that there is also implicit bias through negative findings being less likely to be published or cited in the literature of subsequent studies.
The Cochrane and Campbell collaborations support systematic reviews that examine all the available literature published in journals, presented at conferences, published in in-house journals, and any 'grey literature' available internationally, both positive and negative. The quality of the research is assessed, and only those studies that have been rigorously designed and conducted are included. Systematic reviews are published on the Internet and available to practitioners. This is established practice within the Cochrane Collaboration, and beginning to be implemented by the Campbell Collaboration.
THE CASE FOR RANDOMISED CONTROLLED TRIALS (RCTs)
Randomised controlled trials are generally recognised as the gold standard in evaluation design, and are widely used in medical evaluation. They minimise the methodological bias caused by selecting subjects whose selection criteria might influence results; for example, highly motivated people are more likely to do well than those less well motivated. Self-selected subjects may therefore be more likely to improve, and individual motivation rather than the type of intervention might be the most significant factor in behaviour change. Weisburd, Lum and Petrosino (2001) examined a range of studies that claimed effectiveness of interventions, and found that estimates of effect size were larger in studies where there were few precautions to minimise bias. The results were therefore not so dependable.
In the UK review, only two studies adopted a randomised control design: an in-prison alcohol education programme, and an experimental versus traditional probation supervision allocation. These demonstrate that it is possible to apply an RCT design in a criminal justice setting, both in prison and in the community. It is, however, recognised that it is much easier to conduct randomised controlled trials where the intervention is difficult to observe; for example, where some form of pharmaceutical medication is involved. One can quite easily randomly allocate patients to a pharmaceutical treatment or a placebo replacement without this being evident to those administering and those participating. This is more difficult where, for example, an offender is being allocated to one kind of judicial sentence rather than another. Furthermore, in this latter case, questions of just deserts, fairness of punishment, need for rehabilitation, deterrence and retribution are all introduced. Less frequently, questions of effectiveness are brought into the decision. When an offender is released from prison and reoffends, the conclusion is often not that prison didn't work, but that more prison is needed in order to be effective. Until we begin to seriously evaluate the effectiveness of interventions, this situation is unlikely to change.
Although it is recognised that it is difficult to bring about large-scale changes in sentencing and in the criminal justice system to allow for RCTs, this can be achieved. For example, the Home Office in the UK has recently funded a substantial evaluation project of restorative justice using random allocation of offenders, and this is now underway at three sites (Strang, 2005); and historically RCTs have been possible, exemplified by the Borstal Typology Study (HM Prison Service, 1971), which describes a study in which young offenders were randomly allocated to borstals with different kinds of regimes to evaluate effectiveness. There are examples in this book of diversions from court, and police-referred diversion to psychiatric treatment so, with determination, the wider use of randomised controlled trials is possible within the criminal justice system.
The development of RCT methodology in the UK Criminal Justice System (CJS) has revived interest in a growing number of researchers who strive to achieve the highest standards of quality and certainty on which to base their research results, and Home Office-funded RCTs are now being encouraged (Department of Health/Home Office, 2000, Part II, para. 6.53). Two RCT feasibility studies conducted in HMP Whitemoor in Cambridgeshire (Farrington & Jolliffe, 2002) and the Young Offender Institute (YOI) Thorn Cross in Warrington (Farrington et al., 2002) evaluated some of the practical issues associated with randomisation in a criminal justice setting. Threats to the feasibility of the RCT in HMP Whitemoor were linked to the small number of prisoners receiving treatment, the heterogeneity of the population, the likely length of the treatment, assessment case load, the occupation of treatment beds by previously assessed prisoners and the possibility of dropouts from the treatment (Farrington & Jolliffe, 2002).
Similar problems were encountered in the Thorn Cross study. The research aimed to identify 28 eligible participants every 5 weeks, of whom 14 were chosen at random for the intervention. Randomisation did not occur because of case flow problems and problems with identifying sufficient numbers of eligible participants. In the main, quasi-experimental studies or natural observational studies such as cohort studies are chosen to avoid such difficulties, and are used more commonly in the CJS to evaluate the effectiveness of different interventions and programmes. Despite these recognised problems, RCT methodology has been used with more frequency and success within the US CJS and should be attempted wherever feasible in the UK system.
With improvements in the quality of databases of offenders, an effective method of assessment is the analysis of data held on large databases, such as the Offenders Index database (Home Office, 1998) and the British Crime Survey (Home Office, 2000). There is the potential to use such approaches in order to examine effective interventions in offending behaviour (for example, Lloyd, Mair & Hough, 1995; Bowles et al., 2004). Methodologies for ensuring quality of cohort studies are described in Chapter 2.
Excerpted from Reducing Crime by Amanda Perry Copyright © 2006 by Amanda Perry. Excerpted by permission.
All rights reserved. No part of this excerpt may be reproduced or reprinted without permission in writing from the publisher.
Excerpts are provided by Dial-A-Book Inc. solely for the personal use of visitors to this web site.