Aggression and Violent Behavior 19 (2014) 686–691
Contents lists available at ScienceDirect
Aggression and Violent Behavior
Addressing gang-related violence in Glasgow: A preliminary pragmatic quasi-experimental evaluation of the Community Initiative to Reduce Violence (CIRV) D.J. Williams a,⁎, D. Currie a, W. Linden b, P.D. Donnelly a a b
School of Medicine, University of St Andrews, St Andrews KY16 0ES, UK Violence Reduction Unit, Glasgow G2 4LW, UK
a r t i c l e
i n f o
Article history: Received 18 August 2014 Accepted 27 September 2014 Available online 5 October 2014 Keywords: Violence Youth gang Weapons Intervention Prevention
a b s t r a c t Youth gang-related violence is a public health concern in Glasgow. The Community Initiative to Reduce Violence aims to address physical violence and weapon carriage among gang-related youths in a deprived area of Glasgow. It offers access to diversionary activity, personal development, and employment preparedness in exchange for adherence to a “no violence, no weapon” pledge. A preliminary post hoc before-and-after quasi-experimental design compared rates of criminal offending (including violent and non-violent offenses) for the 167 male youths (aged 16–29) who engaged with the initiative with data for one or two years follow-up for age-matched ganginvolved youths from an equally deprived area of the city. Violent offending reduced across all groups over the time of the study. In the cohort followed for 2-years the rate reduction was greater in the intervention group (52%) than the comparison group (29%). The reduction in the rate of physical violence was not significantly different between the intervention group and the comparison group; however, the rate of weapons carrying was reduced more in the intervention group than the comparison group (84% vs 40% respectively in the 2-year follow-up cohort). The study suggests that adopting a public health approach with gang-related youth was associated with reduced weapon carriage, which can prevent consequences for victims, offenders, and society. © 2014 Elsevier Ltd. All rights reserved.
Contents 1. 2.
Introduction . . . . . . . . . . . . . . . Methods . . . . . . . . . . . . . . . . . 2.1. Study design . . . . . . . . . . . . 2.2. Intervention . . . . . . . . . . . . 2.3. Study cohort . . . . . . . . . . . . 2.4. Dataset . . . . . . . . . . . . . . 2.5. Analysis . . . . . . . . . . . . . . 3. Results . . . . . . . . . . . . . . . . . 4. Discussion . . . . . . . . . . . . . . . . 4.1. Main finding of this study . . . . . . 4.2. What is already known on this topic? 4.3. What this study adds . . . . . . . . 4.4. Limitations of this study . . . . . . 5. Conclusions . . . . . . . . . . . . . . . 6. Competing interests . . . . . . . . . . . Acknowledgements . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . .
⁎ Corresponding author. Tel.: +44 1334 463481; fax: +44 1334 467470. E-mail address:
[email protected] (D.J. Williams).
http://dx.doi.org/10.1016/j.avb.2014.09.011 1359-1789/© 2014 Elsevier Ltd. All rights reserved.
. . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . .
686 687 687 687 687 687 687 688 688 688 688 689 689 690 690 690 690 690
D.J. Williams et al. / Aggression and Violent Behavior 19 (2014) 686–691
1. Introduction Glasgow has been blighted by high levels of violence (see Leyland, 2006) associated with a culture of knife-carrying, heavy drinking, deprivation, and youth gangs (Davies, 1998; McAra, 2008). Glasgow has a long history of youth gangs (see Davies, 1998, 2007, 2014; Patrick, 1973) that are closely associated with “territorial affiliations and rivalries manifest in gang fighting” (Bannister et al., 2010, p. 67). Indeed, involvement with such gangs not only increases the risk of involvement in “recreational violence” (Kintrea, Bannister, Pickering, Reid, & Suzuki, 2008) but also weapon carrying (Deuchar & Holligan, 2010) which is a key situational risk for violence and contributes significantly to homicide and injury (see Leyland, 2006). The roots of the youth gang issue have been attributed to a variety of structural factors, including: experiences of educational failure, worklessness and poverty (Centre for Criminal Justice, 2008), a lack of opportunities or facilities (Kintrea et al., 2008), and dysfunctional family life (Deuchar & Holligan, 2010). The public health approach focuses on addressing such social determinants in order to prevent violence (Dahlberg & Krug, 2002). The complex nature of gang violence has been acknowledged in initiatives in the US. For instance, the Cincinnati Initiative to Reduce Violence implemented a holistic, focused-deterrence approach to address gun-related gang violence through the provision of social support, with some success (Engel, Tillyer, & Corsaro, 2013). Based on the Cincinnati model, the Glasgow Community Initiative to Reduce Violence (CIRV) employed intelligence gathering, sheriff court selfreferral sessions (see Donnelly & Tombs, 2008), multi-agency individualized client support, and police enforcement in an attempt to reduce the long-standing involvement of young males in gang-related violence in Glasgow (VRU, 2009). Given the differences in the nature of the problem between Cincinnati and Glasgow including the type of weapon used (guns vs. knives and other sharp and blunt weapons), the age of the participants (older in Cincinnati compared to Glasgow), and the resources available (e.g. Scotland has a number of agencies already working with the target cohort), CIRV was adapted to function in this context. The initiative was led by Strathclyde Police with committed support from health, education, social services, housing, and community safety, and worked to a public health model which regarded violence as preventable (see Dahlberg & Krug, 2002). Previous qualitative work has illustrated the utility of CIRV in addressing issues pertinent to the lives of a small sample of the young males who engaged with the initiative (see Deuchar, 2013). This paper reports a preliminary post hoc quantitative evaluation of all the young men who engaged with CIRV through a comparison with a one-to-one matched cohort, using police data. 2. Methods 2.1. Study design A before-and-after quasi-experimental design with a comparison group was employed to evaluate the effectiveness of CIRV in reducing violent acts as measured through its two specific goals: reducing physical violence and weapon possession. The study utilizes routinely collected police data in the form of reports to the Procurator Fiscal. 2.2. Intervention CIRV ran from October 24th 2008 to 1st April 2011. Police intelligence systems were used to identify gang-related youths in BD and BA divisions of Strathclyde Police, corresponding to the area of Glasgow traditionally known as the East End. Fifty-five named groups were identified with a total estimated membership of approximately 700. Young men identified as being involved were invited to attend one (or more) of the 10 self-referral sessions held at Glasgow Sherriff Court (see Donnelly & Tombs, 2008). For those who attended, or became aware
687
of CIRV in other ways (e.g. word of mouth) calling a freephone number led to a meeting with a street worker who explained the engagement process and sought a written commitment to abstain from violence and refrain from carrying a weapon. No incentives were offered for signing-up to the initiative, other than accessing the services and opportunities made available through CIRV. The first engagements occurred on the first day of the intervention and the last on November 1st 2010. Access to services, opportunities and support was then triggered, based on a “needs analysis”, and provided through a combination of better focusing of existing statutory services (i.e. education, housing, social services, and health) and services provided by third-sector parties (i.e. diversionary activities, personal development, and job-readiness training) (VRU, 2009, 2010). Compliance with the “no violence, no weapon” pledge was robustly monitored through police systems, and services were temporarily withdrawn in the event of a breach. Within the limit that Scottish law allows, collective responsibility from gangs was expected, such that a breach by one gang member could result in the whole gang being temporarily excluded from CIRV. This was designed to encourage self-enforcement by the gang and sought to utilize its structure and social relationships as a positive force. 2.3. Study cohort Of the 214 individuals recorded as having engaged with CIRV, all male participants of any age who had completed one or two calendar years post-engagement (resulting in two cohorts: 1-year and 2-year) were included in the study and considered the “intervention group”. There were 47 exclusions, including: two females, one male with insufficient data, 21 duplicate individuals (i.e. their details appeared twice in the records provided by CIRV) and 23 whose details could not be verified by the research team (i.e. using name and variants of the name along with date of birth did not identify the individuals as having a criminal record or an entry into the police intelligence database as being gang-involved). The comparison group was selected from a neighboring nonintervention area (Strathclyde Police GA and GE divisions) in the south of Glasgow, with a youth gang problem (36 gangs listed on police intelligence systems with 921 suspected gang members). These divisions were chosen as they have similar socio-economic profiles to the BA and BD divisions served by CIRV. A simple random sample of 250 males was drawn from the 431 known gang members from the comparator divisions. Each member of the intervention group was then agematched on a one-to-one basis with a member from the comparison group which was necessary to control for the known effect of maturation on violent offending (see Junger-Tas, Terlouw, & Klein, 1994). This procedure was undertaken with no knowledge of the offending profile of the individuals involved, other than that they were involved with gangs. In total 167 individuals were identified in each of the intervention and comparison groups with a mean age of 17.8 (range 16–29 years). 2.4. Dataset Criminal records were accessed from Strathclyde Police for the intervention and comparison group members. In order to qualify as a countable incident police investigation had to reach the point where a formal submission was made to the Procurator Fiscal (PF) who, under Scots law, has discretion over whether a prosecution is pursued. The police in Scotland are obliged to make PF submission when they are clear that an offense has been committed, they have a prime suspect, and they have evidence to support a prosecution of that individual. The offenses fell into two broad categories: violent (i.e. assault, breach of the peace, serious violence, sexual, weapon possession: knife, and weapon possession: other) and non-violent (i.e. disorder/anti-social behavior, drugs, road-traffic, and theft). Before and after datasets were created as follows. For the intervention group, the point from which individuals were followed-up corresponded
688
D.J. Williams et al. / Aggression and Violent Behavior 19 (2014) 686–691
to the date on which they engaged with CIRV. Data was accessed for the period 28th October 2006 until 31st October 2011 to maximize the number of individuals accumulating at least one calendar year of follow-up data. Data was limited to one or two full calendar years as a way of avoiding the confounding effect caused by seasonal trends in violence (see Rock, Judd, & Hallmayer, 2008) and was dependent on when the individual engaged with CIRV. For example, an individual would be followed-up for one calendar year before their engagement date and one calendar year after. For the comparison group a notional date for follow-up was allocated based on the engagement date of their matched intervention group pair. As routinely collected data was utilized in the study it was possible to follow-up all members of the 1- and 2-year cohorts of the intervention and comparison groups thereby enabling an intention-to-treat analysis. 2.5. Analysis Descriptive statistics in the form of offense counts before and after engagement date were computed for each of the offense categories. Conditional fixed effects Poisson regression models were constructed to assess the significance of any difference in rate of offending before and after the engagement date. An interaction term between group (intervention or comparison) and time period (pre- or post-intervention) was included to assess whether any change in rate of offending between pre- and post-intervention periods differed between the comparison and intervention groups. Results are reported in tables as incident rate ratios and 95% confidence intervals, and in the text as rate reduction. 3. Results The numbers and mean ages of all those included in the analysis is shown in Table 1. Using independent samples t-tests there were no differences in age between intervention and comparison groups for the 1- and 2-year cohorts (p N .05). Table 2 shows a breakdown of offenses by category for the intervention and comparison cohorts with 1- and 2-years of data before and after the engagement date. There were reductions in most types of violent and non-violent offenses for both the intervention and comparison groups in the 1- and 2-year cohorts. Tables 3 focuses on violent and non-violent offending. The results from a series of conditional fixed effects Poisson regression models are presented. For non-violent offending there was a rate reduction of 14% in the 1-year cohorts and 34% in the 2-year cohorts. There was no evidence of a significant difference in incident rate reduction between the intervention and comparison groups. In the 1-year cohorts there was a significant reduction of 27% in the incident rate of violent offending, with no evidence that this reduction was different between intervention and comparison groups. In the 2-year data there was a rate reduction in both groups and the rate reduction in the intervention group (52%) was significantly greater than in the comparison group (29%). Table 4 presents conditional fixed effects Poisson regression models for types of offending specifically targeted by the intervention, namely physical violence and weapons possession. There was a 21% reduction in incident rate of physical violence offenses in the 1-year cohorts and 31% reduction in the 2-year cohorts. Rate of weapons carrying reduced Table 1 Characteristics of the 1-year and 2-year cohorts. 1-year
Sample size Mean age (SD) Total number of offenses in data
2-year
Intervention
Comparison
Intervention
Comparison
167 17.78 (2.06) 851
167 17.80 (2.04) 929
124 17.73 (2.09) 1182
124 17.74 (2.07) 1209
Table 2 Summary of offense counts for the 1-year and 2-year cohorts before and after engagement with CIRV for each offense category. Intervention Before After 1-year Violent offending Physical violence Assault 55 Breach of the peace (gangs) 31 Serious violence 31 Weapon possession Weapons: knife 19 Weapons: other 26 Non-violent offending Disorder/antisocial behavior 166 Drugs 29 Justice 47 Road traffic 17 Sexual 0 Theft 61 2-year Violent offending Physical violence Assault 60 Breach of the peace (gangs) 71 Serious violence 40 Weapon possession Weapons: knife 29 Weapons: other 58 Non-violent offending Disorder/antisocial behavior 284 Drugs 40 Justice 66 Road traffic 15 Sexual 0 Theft 80
Comparison Change
Before After Change
45 12 26
−18.2% −61.3% −16.1%
49 23 29
48 24 16
−2.0% 4.3% −44.8%
10 6
−47.4% −76.9%
14 39
20 20
42.9% −48.7%
136 36 27 8 0 63
−18.1% 24.1% −42.6% −52.9% 0.0% 3.3%
185 32 76 8 1 44
154 39 61 7 2 38
−16.8% 21.9% −19.7% −12.5% 100.0% −13.6%
54 19 33
−10.0% −73.2% −17.5%
59 43 31
58 26 18
−1.7% −39.5% −41.9%
12 4
−58.6% −93.1%
16 56
21 22
31.3% −60.7%
160 53 33 8 1 62
−43.7% 32.5% −50.0% −46.7% 100.0% −22.5%
288 38 105 13 2 64
185 44 72 5 2 41
−35.8% 15.8% −31.4% −61.5% 0.0% −35.9%
Note: two individuals in the intervention group each had one offense on their day of engagement. As we could not be certain if the offense occurred before or after the point of engagement the two offenses were removed.
more in the intervention than the comparison group. In the 1-year cohorts there was a rate reduction of 65% and 35% in intervention and comparison groups, respectively. In the 2-year cohorts rate of weapons possession reduced by 82% and 40% in the intervention and comparison group, respectively. 4. Discussion 4.1. Main finding of this study In the study area, young men who engaged with CIRV greatly reduced their carriage of weapons (principally knives). This rate of reduced knife carriage was statistically significantly greater than that of the age/sex matched comparison group in a non-intervention area of the city. The results also demonstrate that gang-related youth reduce their rate of physical violence after engagement with CIRV. This was most prominent among those who engaged with CIRV; however, youths in a comparison area also reduced their committal of physical violence, albeit to a lesser degree, despite the lack of a specific CIRV initiative in their area. Although it is not possible to infer causality from a quasi-experimental design, there is evidence of an effect that lasts beyond the initial year and is specific to one of the behaviors directly targeted by CIRV, weapon carriage. Such evidence of specificity and longevity go some way to indicating the strength of association between the intervention and reduced offending. Moreover, while CIRV took place at a time when violent crime throughout Scotland was decreasing the inclusion of a comparison group means that we can say that against a backdrop of a general decrease in levels of violent crime there is a greater reduction in weapon carriage among those individuals who engaged with CIRV.
D.J. Williams et al. / Aggression and Violent Behavior 19 (2014) 686–691 Table 3 Conditional Poisson regression models for violent and non-violent offenses. Incident rate ratios and 95% confidence intervals for 1-year and 2-year cohorts. Cohort Variable Violent offenses 1-year Group: interventiona Time period: afterb Ln pseudo-likelihood N (pairs) 2-year Group: interventiona Time period: afterb Group ∗ time period Ln pseudo-likelihood N (pairs) Non-violent offenses 1-year Group: interventiona Time period: afterb Ln pseudo-likelihood N (pairs) 2-year Group: interventiona Time period: afterb Group ∗ time period Ln pseudo-likelihood N (pairs) a b
pNz
IRR
Robust SE
z
.919
.116
−0.67 0.503 .720, 1.176
.726 −490.2 604 (302) 1.256
.074
−3.14 0.002 .594, .886
.710 .666 −520.7 488 (244)
.091 .111
.916
.132
−0.61 0.542 .691, 1.214
.856 −861.6 640 (320) .942
.062
−2.16 0.031 .743, .986
.137
−0.41 0.679 .708, 1.253
.665
.051
−5.32 0.000 .572, .773
.181
95% CI
689
A similar holistic approach to gang-violence prevention in Cincinnati provided a package of social support, and has been associated with reductions in gang-member involved homicides and violent firearm incidents at 24 and 42months post-intervention that were not evident in comparison areas (Engel et al., 2013). However, Engel et al. note that the provision of social services was found not to be responsible for the decline. Moreover, the nature of the violence and the composition of the gangs in this initiative differ from that in the UK, specifically those in Glasgow where the use of knives is central to much of the violence and homicides.
1.58 0.114 .946, 1.667 −2.66 0.008 .552, .914 −2.44 0.015 .481, .923
−958.470 492 (246)
Reference is comparison group. Reference is before engagement with CIRV.
An objective consideration of the impact of CIRV must also take into account other initiatives to reduce violence throughout the Strathclyde Police region. The main such initiative was the Gangs Task Force which started in 2008 and adopted a traditional policing and enforcement approach as opposed to the public health approach inherent in CIRV. It is likely that some of the effects observed in the current evaluation may be related to this initiative particularly as the task force has a City-wide remit. It can be assumed, therefore, that its impact would be consistent across the intervention (East End) and comparison (South) regions of Glasgow and that this may account for the lack of a statistically significant excess drop in violent offending in the study area. However, our findings do suggest that there is an additional and beneficial impact in relation to weapon carriage in adopting a public health approach to gang-related youth violence, over and above traditional police enforcement.
4.3. What this study adds The study offers the first preliminary evaluation of the effectiveness of a multi-agency, community-centered approach to addressing violence among gang-related youths in Scotland. It demonstrates that the model that originated in Cincinnati can be transferred to the UK. Crucially, it was shown to be associated with a reduction in levels of weapons possession, which is an important situational risk factor and key to lowering homicides and injuries among Glasgow's gang-related youth. The scale of violence reduction achieved appears to be comparable with the Cincinnati initiative. However, the American gang members were older, involved in organized crime, and subject to programs designed to disband their organizations. In contrast, CIRV focused on reducing violent behavior and weapon carriage rather than disrupting youth affiliations, and its clientele were younger and involved in what has been termed “recreational violence” (Kintrea et al., 2008) rather than organized crime. Additionally, they were invited to engage as opposed to being legally compelled. This study suggests that a process of gang rehabilitation originally implemented in the US can be adapted for the needs of gang-related youths in Glasgow. The direction, size and specificity of the effect in the intervention group when compared to the comparison group in relation to weapons carriage argues that CIRV is beneficial in reducing this key risk factor and is thus central to limiting the associated far-reaching adverse health consequences. The principle implication for violence prevention policy and practice is that a police-led, multidisciplinary intervention that provides health and social opportunities and incentives to change behavior, combined with enforcement can begin to help gang-related youths address their violent lifestyle, particularly through reducing weapons carriage.
Table 4 Conditional Poisson regression models for physical violence and weapons carrying offenses. Incident rate ratios and 95% confidence intervals for 1-year and 2-year cohorts. Cohort
4.2. What is already known on this topic? Violence poses a significant public health threat through its impact on the health and wellbeing of the victims and perpetrators, and the wider impact on family, friends and society (see Krug, Dahlberg, Mercy, Zwi, & Lozano, 2002). Scotland has a global reputation as a violent country, supported by epidemiological data and focused on its largest city, Glasgow (see Leyland, 2006). The variety of negative social determinants experienced by young males in Glasgow (i.e. experiences of educational failure, worklessness, and poverty, Centre for Social Justice, 2008; a lack of opportunities or facilities, Kintrea et al., 2008; dysfunctional family life, Deuchar & Holligan, 2010) put them at increased risk of joining a youth gang, and involvement in weapon carrying and gang violence, spurred on by a deep-rooted territorial affiliation. The public health approach emphasizes the importance of implementing interventions that address such social determinants in order to prevent violence (see Dahlberg & Krug, 2002).
Variable
Physical violence offenses 1-year Group: interventiona Time period: afterb Ln pseudo-likelihood N (pairs) 2-year Group: interventiona Time period: afterb Ln pseudo-likelihood N (pairs) Weapon carrying offenses 1-year Group: interventiona Time period: afterb Ln pseudo-likelihood N (pairs) 2-year Group: interventiona Time period: afterb Group ∗ time period Ln pseudo-likelihood N (pairs) a b
z
pNz
95% CI
1.047 .151 .795 .087 −375.2 544 (272) 1.163 .170 .690 .068 −422.8 468 (234)
0.32 −2.10
0.750 0.036
.790, 1.388 .641, .985
1.03 −3.74
0.302 0.000
.873, 1.550 .568, .838
.849 .171 .755 .171 −177.382 408 (204) 1.208 .234 .597 .121 .308 .107 −206.3 372 (286)
−0.81 −1.24
0.418 0.214
.572, 1.261 .484, 1.177
0.98 −2.53 −3.38
0.329 0.011 0.001
.826, 1.767 .401, .890 .156, .609
IRR
Robust SE
Reference is comparison group. Reference is before engagement with CIRV.
690
D.J. Williams et al. / Aggression and Violent Behavior 19 (2014) 686–691
4.4. Limitations of this study CIRV was designed as a public policy initiative and therefore had an operational focus as opposed to being an intervention specifically designed for research purposes. It is, nonetheless, important that the effectiveness of such interventions is evaluated in order to inform policy, and that evaluation is considered from the very earliest stages of intervention development. An RCT, while considered the gold-standard for evaluations of effectiveness (Gray, 2009) was not possible under the current circumstances. Consequently, a before-and-after quasiexperimental design with a comparison group was chosen as it offered a pragmatic yet robust approach (Robson, Shannon, Goldenhar, & Hale, 2001; Stoto & Cosler, 2008). It is, however, acknowledged that such a design is susceptible to a number of biases. First, the design is prone to selection bias (Stoto & Cosler, 2008) such that the intervention group differs in some fundamental way from the comparison group. In the current study the motivation to reduce violent offending could impact on the findings. For those individuals who voluntarily engaged with CIRV it can be assumed that they were motivated to change their behavior, whereas the motivational status of the comparison group is unknown. However, at the very least what this suggests is that there are gang-related youths in the East End of Glasgow who are motivated to change some aspects of their violent lifestyle (e.g. weapon carriage) and with appropriate support, can succeed. Secondly, the design is susceptible to maturation effects (Stoto & Cosler, 2008) particularly as violent behavior is known to peak around 17 or 18 and then decline with age (see Junger-Tas et al., 1994). In an effort to address this issue both intervention and comparison groups were matched for age, and were followed-up over the same period of time. A third threat to validity relates to reporting (Robson et al., 2001). The data comprised reliable records of criminal offending supplied by Strathclyde Police. There has been some debate regarding the use of criminal justice data to inform the evaluation of community violence prevention programs and some prefer to use emergency department data (Florence, Shepherd, Brennan, & Simon, 2011). While health data is appropriate in some circumstances, when evaluating violence prevention programs such as CIRV where the main outcomes included reductions in the perpetration of physical violence and the possession of weapons among a known cohort, such data can only be reliably collected through criminal justice sources. Moreover, the use of such data helps protect against dropout bias (Robson et al., 2001) as it is possible to follow-up the offending of all individuals who engaged with CIRV over the full study period even if they had withdrawn from the initiative or had been temporarily removed. Fourthly, artifact or observer bias is theoretically possible but seem unlikely as data collection practices were non-discretionary and consistent for the before and after periods, and between the intervention and comparison groups. One further concern relates to the possibility of study overinterpretation and thoughtless replication. In the wake of the English riots of 2011 CIRV received considerable political and policy attention (Carrell, 2011) with its widespread adoption being advocated at the highest level of Government as a means to tackle gangs and gang culture (Hansard, 2011). However, it is unclear whether all who advocated the adoption of CIRV really understood the balance of enforcement and enlightened rehabilitation inherent in its apparently successful application in Glasgow in reducing weapons carriage. Nor is it assumed that an identical program could be easily or appropriately replicated in very different English or even Scottish cities. Moreover, it is difficult to generalize quasi-experimental trials of this type to global settings, including other high-income countries as well as low- and middle-income countries. It is evident when comparing the Glasgow and Cincinnati initiatives that the intervention was tailored to the nature of the problem and the resources either already available or made available through specific funding. It seems, nonetheless, intuitive
that participation in diversionary activities and development opportunities may remain relevant in reducing violence even in the challenged economic environments that prevail in many low- and middle-income countries. What is, therefore, required is the thoughtful interpretation, development, and customization of the model in the way that was undertaken in Glasgow. Finally, however, we are cognizant of the challenge of sustainability that has faced such gang-related violence prevention programs in the UK and US, as reductions in violent crime are generally welcomed but all too often accompanied by a shift in political attention and funding. 5. Conclusions The preliminary evaluation of CIRV found a positive statistically significant effect on rates of weapon possession among those engaged with the initiative. The fact that no such outcome was found for physical violence, could mean that either CIRV did not fulfill its objective of preventing such violence, or may have been because the evaluation lacked sensitivity to assess the true impact on physical violence. In both scenarios, a more robust evaluation framework including both outcome and process evaluation would be necessary to better understand the effect of CIRV in terms of what was and was not effective, and why (Williams, Gavine, Ward, & Donnelly, in press). CIRV was designed as a public policy initiative with an operational focus, and the evaluation was implemented after the initiative had begun. This raises an important lesson concerning the necessity of developing an evaluation framework as early as possible prior to implementation to ensure that an appropriate evaluation is possible. 6. Competing interests WL is an employee of the VRU which is jointly funded by Police Scotland and the Scottish Government which was also the major funder of CIRV. DJW and PDD co-supervised a PhD student funded by the VRU who was based at the University of St Andrews (whose work does not form part of this paper) and have received competitive and non-competitive funding to undertake other projects. PDD sits on the VRU Steering Group and the VRU research and evaluation group. DC has no competing interest. Acknowledgements The authors wish to acknowledge the key role of the Violence Reduction Unit staff in particularly John Carnochan (now retired), Karyn McCluskey, and Robert Stevenson (now retired), and the important contribution of partner agencies and voluntary sector organizations in the delivery of CIRV. References Bannister, J., Pickering, J., Batchelor, S., Burman, M., Kintrea, K., & McVie, S. (2010). Troublesome youth groups, gangs and knife carrying in Scotland. Edinburgh, UK: Scottish Government. Carrell, S. (2011). Glasgow gangs chose route to peace in face of tough crackdown. http:// www.guardian.co.uk/uk/2011/aug/11/glasgow-gangs-peace-crackdown Centre for Social Justice (2008). Breakthrough Glasgow: Ending the costs of social breakdown. http://www.centreforsocialjustice.org.uk/UserStorage/pdf/Pdf%20reports/ BreakthroughGlasgow.pdf Dahlberg, L., & Krug, E.G. (2002). Violence: A global public health problem. In E.G. Krug, L. Dahlberg, J.A. Mercy, A.B. Zwi, & R. Lozano (Eds.), The world report on violence and health. Geneva: WHO. Davies, A. (1998). Street gangs, crime and policing in Glasgow during the 1930s: The case of the beehive boys. Social History, 23(3), 251–267. Davies, A. (2007). Glasgow's ‘reign of terror’: Street gangs, racketeering and intimidation in the 1920s and 1930s. Contemporary British History, 21(4), 405–427. Davies, A. (2014). City of gangs: Glasgow and the rise of the British gangster. London: Hodder & Stoughton. Deuchar, R. (2013). Policing youth violence: Transatlantic connections. London: Trentham Books. Deuchar, R., & Holligan, C. (2010). Gangs, sectarianism and social capital: A qualitative study of young people in Scotland. Sociology, 44(1), 13–30.
D.J. Williams et al. / Aggression and Violent Behavior 19 (2014) 686–691 Donnelly, P.D., & Tombs, J. (2008). An unusual day in court. British Medical Journal, 337, a2959. Engel, R.S., Tillyer, M.S., & Corsaro, N. (2013). Reducing gang violence using focused deterrence: Evaluating the Cincinnati Initiative to Reduce Violence (CIRV). Justice Quarterly, 30(3), 403–439. Florence, C., Shepherd, J., Brennan, I., & Simon, T. (2011). Effectiveness of anonymised information sharing and use in health service, police, and local government partnership for preventing violence related injury: Experimental study and time series analysis. British Medical Journal, 342(7812), d3313. Gray, M. (2009). Evidence based healthcare and public health: How to make decisions about health services and public health (3rd ed.). Edinburgh, UK: Churchill Livingston. Hansard (2011). House of Commons: Public disorder. http://www.publications.parliament. uk/pa/cm201011/cmhansrd/cm110811/debtext/110811-0001.htm Junger-Tas, J., Terlouw, G. -J., & Klein, M.W. (Eds.). (1994). Delinquent behaviour among young people in the Western world. Amsterdam: Kugler. Kintrea, K., Bannister, J., Pickering, J., Reid, M., & Suzuki, N. (2008). Young people and territoriality in British cities. http://herd.typepad.com/files/2278-young-people-territoriality. pdf Krug, E.G., Dahlberg, L.L., Mercy, J.A., Zwi, A.B., & Lozano, R. (Eds.). (2002). World report on violence and health. Geneva: WHO. Leyland, A.H. (2006). Homicides involving knives and other sharp objects in Scotland, 1981–2003. Journal of Public Health, 28(2), 145–147.
691
McAra, L. (2008). Crime, criminology and criminal justice in Scotland. European Journal of Criminology, 5(4), 481–504. Patrick, J. (1973). A Glasgow gang observed. London: Metheun. Robson, L.S., Shannon, H.S., Goldenhar, L.M., & Hale, A.R. (2001). Guide to evaluating the effectiveness of strategies for preventing work injuries: How to show whether a safety intervention really works. http://ssmon.chb.kth.se/safebk/safetybk.pdf Rock, D.J., Judd, K., & Hallmayer, J.F. (2008). The seasonal relationship between assault and homicide in England and Wales. Injury, 39(9), 1047–1053. Stoto, M.A., & Cosler, E. (2008). Evaluation of public health interventions. In L.F. Novick, C.B. Morrow, & G.P. Mays (Eds.), Public health administration: Principles for populationbased management (pp. 495–544) (2nd ed.). London: Jones and Bartlett Publishers. VRU (2009). Community Initiative to Reduce Violence: Year 1 report. http://www. actiononviolence.org.uk/sites/default/files/CIRV-year1-report.pdf VRU (2010). Community Initiative to Reduce Violence: Third quarter progress report. http://www.actiononviolence.org.uk/sites/default/files/documents/CIRV%20Third% 20Quarter%20Report%20May%202010.pdf Williams, D.J., Gavine, A.J., Ward, C.L., & Donnelly, P.D. (2014s). What is evidence in violence prevention? In P.D. Donnelly, & C.L. Ward (Eds.), Oxford textbook of violence prevention (pp. 125–131). Oxford, UK: Oxford University press (in press).