Safety in the Suburbs: Social Disadvantage, Community Mobilisation, and the Prevention of Violence Rebecca Wickes, Ross Homel and Renee Zahnow Published as: Wickes, R., Homel, R. & Zahnow, R. (2015). Safety in the suburbs: Social disadvantage, community mobilization, and the prevention of violence. In J. Stubbs & S. Tomsen (Eds), Australian Violence. Sydney: Federation Press Introduction There has been a marked decline in violence in most advanced countries in recent decades. The United States has, for example, experienced its lowest levels of violent crime in a generation (Goldberger & Rosenfeld, 2008, p.xi). This development could be viewed as a resumption, after a post second world war disruption, of the downward trend in violence apparent in Western societies for hundreds of years (Pinker, 2011). Australia too has witnessed a marked decline in some types of violence (Australian Institute of Criminology, 2013a). For example, homicide victimisation rates have dropped by 29 per cent since the mid- to late-1990s (Australian Institute of Criminology, 2013a), though over the last century homicide exhibits essentially a stationary trend. Yet in the new millennium, the pattern for non-lethal violence is not following the same trajectory. Disturbingly, Australian rates of selfreported violent victimisation are consistently in the upper half of the range reported over recent decades by countries participating in the International Crime Victimisation Surveys (van Dijk, van Kesteren & Smit, 2007). Additional evidence suggests that rates of violent assault have risen by an average of two per cent each year since the early 2000’s, even though the rates of other crimes have declined (Australian Institute of Criminology, 2010). Violence is not randomly distributed throughout the community, but instead concentrates in particular types of places. In Australia, as is the case in many developed, western countries, higher rates of violence occur in communities characterised by socioeconomic disadvantage. Disadvantage directly impacts violence through increasing one’s exposure to violence and victimisation (Bingenheimer, Brennan & Earls 2005), but its effects are also indirect. Poverty sets in motion a cycle that undermines the capacity to come together to resolve local problems, which in turn sets in play a range of problems that deepen and reinforce poverty and low levels of social cohesion and trust (Sampson & Morenoff, 2006). 1 In Australia, several cross-sectional studies demonstrate the strong relationship between disadvantage, social cohesion and violence (Mazerolle, Wickes & McBroom, 2010; Vinson & Homel, 1975; Weatherburn & Lind, 2001). What is less well-understood is the extent to which these characteristics and community processes influence violence over time. We argue a longitudinal approach is necessary to, not only identify the key predictors of violence, but to better inform community crime prevention programs concerned with the reduction of crime and other social problems. In this chapter we therefore consider the community-level characteristics and mechanisms associated with changes in violence over time. Specifically, we explore recent social and economic changes in Brisbane state suburbs, and the impact of these changes on rates of officially recorded violence. We meld longitudinal data from the Australian Community Capacity Study (ACCS) with census and police incident data and identify two critical community-level predictors of reductions in levels of suburban violence over the past 15 years. These critical factors are: (i) decreases in levels of social disadvantage, and (ii) increases in levels of social cohesion and trust. Our analyses suggest that even relatively small changes in these factors over time can correspond to socially significant changes in levels of officially recorded violence. This is, on the one hand, very encouraging as evidence suggests that trust can be strengthened and rates of violence substantially reduced by funding community organisations to develop and implement local community improvement and prevention initiatives (Cooper, Bumbarger & Moore, 2013; Ramey & Shrider, 2014). On the other hand, our research indicates that any reductions in violence achieved through community building strategies might be undermined if economic events or government policies have the effect of increasing spatial social inequality, as has indeed been happening for several decades (Australian Social Inclusion Board, 2011). Social Disorganisation and Violent Crime A focus on violence in urban communities is important for several reasons. First, high levels of violence are associated with decreased social capital; lower property values; decreases in economic development; low levels of neighbourhood satisfaction; and increased fear among residents (Rosenberg et al, 2006). Second, in Australia it is mostly in the suburbs that family violence happens, child abuse is perpetrated, school children are bullied, youth crime is committed, and alcohol gets abused. A focus on place brings in a large slice of the violence problem, doubly so if the place is socially disadvantaged (Mazerolle et al, 2010; Vinson & Homel, 1975). 2 Social disorganisation theorists contend that the concentration of crime and other problems in socially disorganised areas occurs because community structural characteristics (particularly poverty, racial/ethnic concentration, and residential instability) break down the key regulatory processes (formal and informal) associated with maintaining order and that this, in turn, allows serious crime to flourish (Bursik & Grasmick, 1993; Sampson & Groves, 1989; Sampson, Raudenbush & Earls, 1997; Shaw & McKay, 1969). Traditionally it was argued that social order was only possible in neighbourhoods where close relationships between residents were present to foster the realisation of common values and the enforcement of informal social controls (Bursik & Grasmick, 1993; Kornhauser, 1978). In the mid-1990’s social disorganisation theory was transformed, shifting in emphasis from the role of strong ties in preventing crime to the neighbourhood’s capacity to exercise informal control. Sampson and his colleagues demonstrated that violence and violent victimisation resulted from the differential ability of neighbourhoods to realise the common values of residents and maintain effective social controls, which they referred to as collective efficacy. Even in communities characterised by weak ties, when residents trusted each other and were willing to work together to solve local problems, violence was lower (Sampson et al., 1997). In the Australian context the core tenets of social disorganisation theory also help to explain the spatial concentration of violence. For example, Vinson and Homel (1975) examined the concordance of social problems (including crime) in communities in Newcastle, with Braithwaite (1979) later examining the relationship between social status and crime across Australia. More recently, Weatherburn and Lind (2001) surveyed Sydney-area suburbs and proposed an epidemic model of growth in the offender population derived from measures of economic and social stress. In 2010, Mazerolle and her colleagues provided the most comprehensive Australian test of the contemporary social disorganisation thesis. Drawing on the first wave of the Australian Community Capacity Study, they examined the relationship between social structural characteristics (poverty, racial/ethnic concentration and residential mobility), collective efficacy, and rates of violence and violent victimisation. In Brisbane, as in Chicago, a significant proportion of variation in collective efficacy and violence was due to between-suburb differences in structural characteristics. Thus levels of social cohesion and trust and the willingness of residents to intervene in prosocial ways differed across the sample neighbourhoods. Although the absolute level of violence in Brisbane was nearly three times lower than in Chicago, violent crime clustered, as expected, in neighbourhoods characterised by disadvantage. 3 The main finding from this research was that high levels of collective efficacy were strongly associated with lower rates of violence and the likelihood of household victimisation (Mazerolle et al., 2010). Moreover, collective efficacy mediated the direct relationship between disadvantage and violent incidents and rates of victimisation for violence. This suggests that the spatial clustering of violence in Australia can be explained by the same socio-demographic characteristics and neighbourhood processes as found in studies in the United States, the Netherlands, Sweden and the United Kingdom (Markowitz, Bellair, Liska & Liu, 2001; Sampson et al., 1997; Sampson & Wikstrom, 2008; Steenbeek & Hipp, 2011). Longitudinal Studies of Social Disorganisation Theory Inherent in all forms of social disorganisation theory are assumptions about the temporal associations between social conditions, social processes and violence. To date, empirical studies concerned with exploring these relationships have relied almost exclusively on cross-sectional data or have only considered changes over two time points (Markowitz et al, 2001 and Steenbeek & Hipp, 2011 are exceptions). Because longitudinal survey data on neighbourhood social processes are so rare, few studies can thoroughly examine the temporal component of the social disorganisation thesis. Thus it is difficult to ascertain if changes in neighbourhood crime rates are indeed a function of changes to the neighbourhood sociodemographic structure and the social processes necessary for community regulation, as argued by many scholars (Hipp, 2011; Kubrin & Herting, 2003; Weisburd, Groff & Yang, 2012). So what do we know? There is evidence emerging from a handful of longitudinal studies of places that the structural characteristics associated with violence are stable over time. In Chicago, the percentage of families in poverty in 1960 is correlated with the percentage of families in poverty in 2000 at 0.78 (Sampson 2012). Across all census tracts in the United States the percentages of people receiving public assistance in 1990 and 2000 are correlated at 0.85. Racial segregation is also profoundly stable in the United States and this segregation, combined with high levels of poverty has devastating effects on the social processes necessary to combat violence (Sampson, 2012). Studies also show that the social processes necessary for crime control, like social cohesion, are stable over time (Markowitz et al., 2001; Steenbeek & Hipp, 2011). Drawing on six waves of panel and census data from 74 neighbourhoods in the Netherlands, Steenbeek and Hipp’s (2011) research established that social cohesion was remarkably stable. Although there was considerable variation between 4 neighbourhoods, over time areas low in social cohesion at one point displayed low social cohesion at the next time point. In the U.K. the story is same (Markowitz et al., 2001). How the stability (or change) in structural characteristics and neighbourhood processes, like social cohesion, influence crime longitudinally is less clear, though two studies provide some support for the causal associations depicted in social disorganisation theory. Using their community longitudinal data, Steenbeek and Hipp (2011) found that despite the stability in disorder over time, neighbourhood disorder increased residential instability, which in turn decreased informal social control. Markowitz and colleagues (2001, p.313) report similar results: “In areas of decline, levels of fear are higher, inhibiting the degree of neighbourhood cohesion and the capacity for the informal controls necessary to curb crime.” Thus while we know that social structure and community regulatory processes are strongly associated with violence, we have a limited understanding of how these processes influence stability or changes in violence over time. The Australian Community Capacity Study (ACCS) 1 is one of just four studies in the world designed to examine neighbourhood processes and crime and disorder prospectively over time. In this chapter, we draw on three waves of the ACCS, in combination with census and police incident data, to examine the temporal relationship between socio-structural characteristics, community cohesion, and violence. Three key questions drive our analyses: 1. How stable is violence over time and across suburbs? 2. How stable are community structures and processes over time? 3. What are the most important predictors of violence over time? Our final question, which we reflect on in our discussion, is shaped by the answers to the first three: 4. How can community prevention innovations be developed to address the factors that predict community-level changes in violence? The Australian Community Capacity Study The ACCS is a longitudinal panel study of urban communities in Australia (for further information please see http://www.uq.edu.au/accs/index.html). The primary goal of this research is to develop a longitudinal understanding of the socio-structural characteristics and social processes that influence violence over time. The ACCS can be compared to a 5 handful of international studies that prospectively study the relationship between neighbourhood processes and social problems, including the Project for Human Development in Chicago Neighborhoods (PHDCN), the Los Angeles Family and Neighborhoods Study (LA FANS) and Nieuw Utrechts Piel (NUP) in the Netherlands. The ACCS has been conducted in Brisbane and Melbourne. Of interest to this study is the Brisbane site. Brisbane is the state capital of Queensland and at the last census recorded a population of nearly 2 million people. The current research employs three waves of data collected in 2008, 2010 and 2012 representing the second, third and fourth waves of the ACCS conducted in the Brisbane Statistical Division (BSD).2 The ACCS sample comprises 148 randomly selected suburbs with a residential population ranging from 245 to 20,999 (total suburbs in the BSD = 429). Participants The survey was administered by the Institute for Social Science Research at the University of Queensland using computer-assisted telephone interviewing and random digit dialling. Interviews lasted approximately 20-24 minutes. The sample sizes for Waves 2, 3 and 4 of the ACCS were 4,244; 4,403; and 4,132 participants respectively. Each wave comprised respondents who had indicated their willingness to continue with the study at the previous wave, and an additional top up sample. The consent and completion rates for the ACCS were 69.3% for Wave 2; 68.5% for Wave 3 and 46.2% for Wave 4. These rates represent the number of interviews completed proportional to the number of in-scope contacts. The inscope survey population comprised all people aged 18 years or over with telephones who were usually resident in private dwellings in the selected communities. An analysis of attrition rates indicated that older respondents were more likely to take part in future research compared to younger participants, as were those with dependent children. Those who owned their home and those who had resided at their current address for longer were also more likely to participate in future waves. Our results are therefore perhaps somewhat skewed to the views of older, long-term residents and those with children. Variables Dependent variable. Our dependent variable was derived from the Queensland Police Service (QPS) incident data and is the suburb rate of violent crime per 100,000 residents for the period 2006 to 2011. Violent crime includes homicide; other homicide; assault; and robbery. We averaged the data for each ACCS year with the violence rates in the years 6 immediately before and after. For example, for Wave 2 (2008) we averaged the police data for 2007, 2008 and 2009 to yield the violence score for 2008.3 Independent variable. The measure of social cohesion and trust employed in the current study is a scale comprising survey responses to four items. Respondents were asked to state their agreement with the following statements on a scale of 1-5: People around here are willing to help their neighbours; This is a close-knit community; People in this community can be trusted; People in this community do not share the same values. At each wave, the internal reliabilities for this scale (alpha scores) exceeded 0.68. In our analytic models we estimated fixed-effects models that included indicator variables for all suburbs in Brisbane, as well as several individual characteristics that might systematically bias perceptions of social cohesion in the suburb. The estimated coefficients for each suburb from these analyses were then used as unbiased estimates of the amount of social cohesion in the suburb. Thus we followed standard procedures in constructing an ecometrically valid community-level measure of cohesion and trust (Raudenbush & Sampson, 1999), using part of the widely used 10-item measure of community collective efficacy (Sampson et al, 1997). Control Variables. Violent crime rates are influenced by several suburb characteristics, including socio-economic disadvantage; population density; family disruption; age structure; racial composition, and residential instability (Baumer, Lauritsen, Rosenfeld & Wright, 1998; Sampson & Groves, 1989; Sampson et al., 1997). We selected data from the 2001, 2006 and 2011 censuses to capture these suburb characteristics using the 2006 census boundaries to ensure geographical consistency across all three waves of data. These census years were selected since they precede the ACCS years (2008, 2010 and 2012) and allow us to investigate the stability of social structural factors over time.4 The control variables used in our analyses were: Disadvantage: Consistent with previous studies, we conducted a principal components analysis on variables that potentially measured an underlying dimension of ‘disadvantage’ (Land, McCall & Cohen, 1990; Sampson et al, 1997). A single factor comprising four variables emerged from the results (factor loadings are in parenthesis): percentage of neighbourhood households renting (0.872); percentage of neighbourhood households that had an average weekly income less than $799 (0.831); the percentage of unemployed residents (0.824); and the percentage of neighbourhood residents who identified as Indigenous (0.751). A general disadvantage factor was clearly identifiable, and a composite factor score of these four measures was used in all analyses. 7 Ethnic Diversity: Ethnic heterogeneity of the neighbourhood was measured by the Blau index of language diversity (Blau, 1977). 5 The index represents the likelihood that two randomly selected residents from the suburb speak a different language. A perfectly homogenous group would receive a score of 0 while a completely heterogeneous group would receive a score of 1. While we recognise that ethnic diversity encompasses more than just language, previous research shows that language diversity is more consequential for community social processes in the Australian context (Leigh, 2006; Wickes et al., 2013). Residential Mobility: Drawing on the social disorganisation perspective we included a measure of residential mobility in our analyses. This captures the percentage of residents at a different address five years prior to each census period (2001, 2006 and 2011). Population Density: Measures of population size or density are frequently included in studies of neighbourhood violence (Baumer et al., 1998; MacDonald, 2002). Research tends to find a positive association between population density and violent crime, but results vary depending on the sample and analytic strategy (Deane et al., 2008). The measure of population density included in the current study is calculated as the number of neighbourhood residents per square kilometre. Population density was converted to logarithms to reduce right skew. Results Our first research question was: How stable is violence over time and across suburbs? The short answer to the first part of this question is that, on average, rates of violence within the 148 suburbs did not change much between 2006 and 2011. Table 1 shows the (untransformed) violence rates for the three time points. Although there was a trend downwards over time, from 896.95 to 814.20 incidents per 100,000 population, this trend is not statistically significant. Indeed examination of the time series of monthly violence rates for the 148 suburbs for a longer time period, January 1996 to December 2011 (this graph is not shown), demonstrates a stationary pattern, reflecting the pattern for the whole of Queensland. TABLE 1 ABOUT HERE The answer to the second part of the first research question is, as expected, there was considerable variability between suburbs in their violence rates at any given time, and also considerable variation between suburbs in terms of their levels of within suburb change in 8 violence over time. To clarify this last point, although on average the 148 suburbs showed stability in violence rates, some suburbs became more violent between 2007 and 2012, some became less violent, and others stayed pretty much the same. Figure 1 shows that the suburbs that became more violent tended to cluster in the north (e.g., Caboolture, Mango Hill, Kallangur, Morayfield) or to run in a ragged ribbon across the south (Forest Lakes, Redbank Plains, Browns Plains, Shailer Park). Suburbs that dropped most in recorded violence included some surprises, such as Inala, Woodridge, Underwood, and Slacks Creek in the south, all with reputations for above average rates of crime. Other improving areas included wealthy near-city suburbs such as Paddington and Ashgrove, as well as Bayside areas such as Alexandra Hills. FIGURE 1 ABOUT HERE Our second research question asked, How stable are neighbourhood structures and processes over time? As shown in Table 1, living standards improved markedly, particularly between 2001 and 2006, evidenced by the decline in the percentage of residents with low incomes or who were unemployed. These trends are reflected in the decline in scores on the index of disadvantage (from 18.38 to 13.56), a statistically significant drop. There appeared also to be a drift of Indigenous people to the city, a growth in language diversity, and increasing residential stability. Population density showed little change. Notwithstanding this trend to greater affluence and more stability, the index of social cohesion and trust showed a slight but statistically significant decline, especially between Wave 2 and Wave 3 (Times 1 and 2 in Table 1). Figure 2, which shows the geographical distribution of high and low cohesion scores in relation to the distribution of disadvantage, suggests that lower cohesion suburbs are mostly more disadvantaged, although there is a cluster of communities on the south side that score in the higher range on cohesion but have moderate to average levels of disadvantage (e.g., Corinda). The strong negative relationship between cohesion/trust and disadvantage is confirmed by the correlation of -0.46, shown in Table 2. Suburbs that are diverse in languages spoken or with high levels of residential mobility also tended to score lower on cohesion and trust. FIGURE 2 ABOUT HERE TABLE 2 ABOUT HERE Our third research question asked, What are the most important predictors of violence over time? To examine suburb level predictors of the violent crime rate, specifically the association between changes in social cohesion and trust and changes in the violent crime rate 9 over time, we adopted a hybrid fixed effects analytic approach. A hybrid fixed effects model separates the within-neighbourhood, or time variant, effect of social cohesion and trust from the contextual or between-neighbourhood effect of social cohesion and trust. This is the most appropriate model because we are interested in examining the influence on violent crime rates of both within neighbourhood change over time and between neighbourhood variation in key characteristics (Andreson, 2012).6 TABLE 3 ABOUT HERE Overall the results conformed to theoretical expectations and the findings of previous research. As indicated in Table 3, there was a significant association between changes in suburb social cohesion and trust over time and rates of violent crime, such that suburbs that reported increases in social cohesion and trust experienced lower rates of violence (β = 0.463, p = 0.05). Changes in residential mobility, population density, and language diversity did not lead to higher violent crime rates, but increases in disadvantage had a marginally significant effect (p = 0.06). Looking to the pooled effects of our variables, on average, across suburbs and across time, social cohesion and trust had a negative association with rate of violent crime (β= 0.693, p = 0.02) while disadvantage had a positive association with the violent crime rate (β= 0. 094, p<0.001). These were the only variables that influenced violent crime rates in our model. Unfortunately the complexity of the model, which is essential in order to incorporate both within suburb and between suburb variations in rates of violence over time, precludes simple interpretation of effect sizes. Moreover the suburb level measure of social cohesion and trust has been modelled as a latent scale score so has no straightforward metric. For these reasons it is safest to draw conclusions in general terms, without attempting to infer exactly what reductions in violence would follow improvements in cohesion and trust or increases in social disadvantage. What the model does allow us to predict is that, at least over a relatively short time span, only changes in aggregate levels of cohesion and trust and in levels of social disadvantage will have statistically detectable effects on rates of police-recorded violence. None of the other variables we included in the model, selected in the light of social disorganisation research, can be predicted to have such effects. Future directions for community violence prevention 10 Our research set out to examine the stability of disadvantage, social cohesion and violence. Working within a social disorganisation framework our findings demonstrate the stability of these community characteristics and processes. We also find that disadvantage is a particularly strong predictor of violence. Over time, as disadvantage increases, so too does the level of violence experienced in the community. Yet we also find that increases in social cohesion can have a significant influence on violence. So how can community prevention innovations be developed to address the factors that we have shown to predict communitylevel changes in violence? The promise of community prevention rests on research that demonstrates the link between social cohesion and the willingness of residents to work together to solve local problems and lower rates of crime and delinquency. Even in communities characterised by poverty, racial/ethnic segregation and residential instability (Mazerolle et al 2010; Morenoff et al., 2001; Sampson et al., 1997) cohesion can lead to lower crime. For this reason crime prevention initiatives influenced by social disorganisation theory have historically emphasised community building strategies to empower residents to participate in decisionmaking, strengthen informal networks, and enhance community organisation (Morgan, Boxall, Lindeman, & Anderson, 2012; Wickes, in press). Crime prevention initiatives using the community empowerment approach face multiple problems, especially when attempted in high crime, socially disadvantaged areas. One explanation for the consistent failure of community crime prevention initiatives is that when multiple players and stakeholders attempt to implement their own ideas of what works the result is “muddled, inconsistent, and untheorised” prevention efforts (Hope, 1995; Wickes, in press). In particular, many initiatives apparently fail to strengthen the capacity of community members to do what social disorganisation theory says is important, namely respond in a timely and preventative way to unwanted behaviour in the neighbourhood. Proponents of the now dominant research-to-practice model of prevention (Flaspohler et al., 2008) would argue that the failure to draw on and properly implement evidence-based interventions is a key reason why community empowerment initiatives do not address effectively the risk factors for crime. One response to our findings, and to the disappointing outcomes of eighty years of attempts to reduce crime through community prevention initiatives, would be to conclude that suburban violence cannot feasibly be prevented through community development or mobilisation strategies. It can be argued persuasively that the geographical distribution of social disadvantage is determined by economic policies, the housing market, and societal 11 change, all of which are largely beyond the power of local communities to influence. Universal services such as preschool, health care, and the welfare safety net, as well as community-based services such as family support, may do something to buffer the impacts of markets and large-scale social change on the most vulnerable individuals and communities. Yet this may have little impact on rates of suburban violence, which can only really be reduced, if at all, through political and economic actions at the national (or international) level. Countering this perspective is one that asserts that in Australia “it is feasible to substantially reduce violent and other offending by implementing evidence-based prevention and early intervention strategies” (Toumbourou et al , in press). One particularly promising such strategy is Communities That Care (CTC), a community mobilisation approach that arguably synthesises the best of the old tradition of collaborative community action research with modern prevention science (Greenberg, Feinberg, Gomez & Osgood, 2005; Hawkins et al., 2009). A second strategy with demonstrated effectiveness for long-term violence reduction is Washington State’s Neighborhood Matching Fund (NMF), which provides city funding for neighbourhood improvement activities organised locally (Ramey & Shrider, 2014). A key link between these contemporary approaches and the older tradition of community prevention is the creation of community coalitions consisting of engaged residents (including young people) and representatives from local agencies. These coalitions frequently grow from grassroots initiatives, but in the CTC model receive extensive technical assistance to measure risk and protective factors for youth in the area and to implement and evaluate evidence-based initiatives that have the best chance of effectively addressing the small number of factors selected by each local coalition as a target for action. The prevention science elements of the NMF model are less clear, but what is clearly evident from both CTC and NMF is the critical importance of external funding. Ramey and Shrider found that funding was particularly beneficial for poorer communities where reductions in violence were greatest, but so far CTC researchers have struggled to demonstrate effectiveness in disorganised high crime urban settings (Brown, Feinberg & Greenberg, 2010). In Australia the CTC model has been trialled in three communities including the Mornington Peninsula, a moderately disadvantaged region where a prevention-oriented coalition emerged in 2000 through the initiative of a group of local activists and was then boosted through state government funding (Toumbourou et al, in press). In 2002, before the CTC process had begun to be properly initiated, a comprehensive survey of youth showed 12 that 22 percent of 13 year-olds engaged in one or more of a range of violent and antisocial acts, such as attacking someone with the intent of seriously hurting them. By 2012 this had dropped to 13 percent, while in the same period in the rest of Victoria the figure was virtually static at 17 percent in 2002 and 16 percent in 2012. There is every likelihood that many CTC and related initiatives have increased levels of community cohesion and trust as a simple function of their effectiveness in building wellfunctioning and comprehensive local coalitions with demonstrable effectiveness, but the research to test this hypothesis remains to be done. Such research should include, in a longitudinal design, the measures of community structure and processes that we have identified in this chapter, alongside the well-established measures of risk and protective factors, coalition functioning, and impacts on community social cohesion and trust and rates of violence. We recognise that experience over many years with community empowerment and crime prevention initiatives in disadvantaged, socially disorganised communities is not at all encouraging (Homel & McGee, 2012; Wickes, in press). This long history of failure leads us to temper our cautious optimism about new community prevention approaches with realism about the corrosive effects of poverty and increasing social inequality on the capacity of communities to focus on local problems and to get organised to address them effectively, even with external funding (Greenberg et al, 2007). Nevertheless on the basis of our analyses and even more in the light of recent advances in prevention science (Spoth et al., 2013), we propose that strategies to improve safety must combine the best insights from both collaborative community action research and prevention science (Weissberg & Greenberg, 1998), using a community mobilisation approach. Specifically, we argue that aligning the fruits of the many years of social disorganisation research with the emerging science of prevention, using CTC or a similar model, holds the promise of taking community prevention research to a new level, especially if experiments are focused on socially disadvantaged communities. Such endeavours also hold the key to identifying both the critical mechanisms that allow for the development of social cohesion and the ways in which social cohesion manifests in activities that lead to lower rates of violence. Endnotes 1. This work was supported by the Australian Research Council (LP0453763; DP0771785; RO700002; DP1093960; DP1094589 and DE130100958). 13 2. The first wave of the ACCS conducted in 2005 was not employed here because the geographic boundaries of the sampling unit shifted significantly between Waves 1 and 2. 3. The averaged incident rates were then converted to logarithms to reduce right skew. 4. The census years do however precede the ACCS years by variable periods, with the 2001 and 2006 censuses both preceding the second wave of the ACCS study (which is the first wave used in this analysis). If there were a stronger theoretical and empirical basis than is currently available for selecting an appropriate time lag between census and survey year we could have used data interpolation methods to estimate census characteristics for the appropriate years, but in the absence of that information the census variables should be regarded as providing an approximate picture of how suburb characteristics changed some years before each of the ACCS waves. 5. The Blau index (Blau, 1977) was calculated using the following ABS census language categories: Northern European; Southern European; Eastern European; Southwest Central Asian; Southern Asian; South East Asian; Eastern Asian Languages; Australian Indigenous Languages; English Only. The Blau index is defined as: 1 - Σ𝑝𝑖2 where p is the proportion of group members in a given category and i is the number of different categories. 6. When using panel data one can use a random or fixed effects estimator. The fixed effects estimator is based on the time series component of the data while random effects estimation uses both the cross-section and time-series components (Andreson, 2012). While the Hausman test suggests the fixed-effects model is most appropriate in this situation our research question is most fully addressed by the hybrid model (Philips & Greenberg, 2008). The hybrid model augments the fixed effects model by borrowing elements of the random effects model to reduce bias associated with the random effects strategy. It should be noted that the effects of previous violence, often incorporated as an independent variable into regression models, is factored into the present model as a function of the time-varying dependent variable. The model was fitted by the maximum likelihood method and all analyses were conducted in STATA 13.0. References Andresen, M, 2012, “Unemployment and crime: A neighborhood level panel data approach” 41 Social Science Research 1615-1628. 14 Australian Institute of Criminology, 2010, Australian crime: Facts & Figures 2010 Canberra: Australian Institute of Criminology. Australian Institute of Criminology, 2013a, Homicide in Australia: 2008-09 to 2009-10 National Homicide Monitoring Program annual report, Australian Institute of Criminology, Canberra. Australian Institute of Criminology, 2013b, Australian crime: Facts and figures 2012, Australian Institute of Criminology, Canberra. Australian Social Inclusion Board, 2011, Breaking cycles of disadvantage, Commonwealth of Australia, Canberra. Baumer, E, Lauritsen, JL, Rosenfeld, R & Wright, R, 1998, “The influence of crack cocaine on robbery, burglary, and homicide rates: a cross-city, longitudinal analysis” 35 Journal of Research in Crime and Delinquency 316-340. Bingenheimer, JB, Brennan, RT & Earls, F, 2005, “Firearm Violence Exposure and Serious Violent Behavior” 308 Science 1323-1326. Blau, P, 1977, Inequality and Heterogeneity, Free Press, New York. Braithwaite, J,1979, Inequality, crime and public policy Routledge & Kegan Paul, London. Brown, LD, Feinberg, ME & Greenberg, MT, 2010, “Determinants of Community Coalition Ability to Support Evidence-Based Programs” 11 Prevention Science 287-297. Bursik, RJ, 1986, “Ecological stability and the dynamics of delinquency”, in A Reiss, Jr & M Tonry (eds) Communities and Crime University of Chicago Press, Chicago, pp. 35-66. Bursik, RJ & Grasmick, HG, 1993, Neighborhoods and crime: the dimensions of effective community control, Lexington Book, Maryland. Cooper, BR, Bumbarger, BK & Moore, JE, 2013, “Sustaining Evidence-Based Prevention Programs: Correlates in a Large-Scale Dissemination Initiative” Prevention Science, Epub ahead of print, doi:10.1007/s11121-013-0427-1. 15 Deane, G, Messner, S F, Stucky, T D, McGeever, K, & Kubrin, CE, 2008, “Not ‘islands, entire of themselves’: exploring the spatial context of city-level robbery rates” 24 Journal of Quantitative Criminology 363-380. Flaspohler, P, Duffy, J, Wandersman, A, Stillman, L & Maras, MA, 2008, “Unpacking Prevention Capacity: An Intersection of Research-to-practice Models and Communitycentered Models” 41 American Journal of Community Psychology 182-193. Goldberger, AS & Rosenfeld, R (eds), 2008, Understanding Crime Trends: Workshop Report National Academies Press, Washington. Greenberg, MT, Feinberg, ME, Gomez, BJ, & Osgood, DW, 2005, “Testing a community prevention focused model of coalition functioning and sustainability: A comprehensive study of communities that care in Pennsylvania”, in Stockwell, T, Gruenewald, P, Toumbourou, JW, Loxley, W (eds) Preventing harmful substance use: The evidence base for policy and practice Wiley, London, pp 129 – 142. Greenberg, MT, Feinberg, ME, Meyer-Chilenski, S, Spoth, RL & Redmond, C, 2007, “Community and Team Member Factors that Influence the Early Phase Functioning of Community Prevention Teams: The PROSPER Project” 28 Journal of Primary Prevention 485-504. Hawkins, JD, Oesterle, S, Brown, EC, Arthur, MW, Abbott, RD, Fagan, AA & Catalano, RF, 2009, “Results of a Type 2 Translational Research Trial to Prevent Adolescent Drug Use and Delinquency: A Test of Communities That Care” 163 Archives of Pediatric Adolescent Medicine 789-798. Hipp, JR, 2011, “Spreading the Wealth: The Effect of the Distribution of Income and Race/ ethnicity across Households and Neighborhoods on City Crime Neighborhood” 49 Criminology 631-665. Homel, R & McGee, T, 2012, “Community approaches to crime and violence prevention: Building prevention capacity,” in Loeber, R & Welsh, B (eds) The future of criminology Oxford University Press, New York, pp172-177. Hope, T, 1995, “Community crime prevention” 19 Crime and Justice 21. 16 Kornhauser, R, 1978, Social sources of delinquency University of Chicago Press, Chicago. Kubrin, C & Herting, J, 2003, “Neighborhood correlates of homicide trends: An analysis using growth-curve modeling” 44 Sociological Quarterly 329-350. Land, KC, McCall, PL & Cohen, LE, 1990, “Structural covariates of homicide rates: are there any invariances across tie and social space?” 95 American Journal of Sociology 922-963. Leigh, A, 2006, “Trust, inequality and ethnic heterogeneity” 82 The Economic Record 268. MacDonald, JM, 2002, “The effectiveness of community policing in reducing urban violence” 48 Journal of Crime and Delinquency 592-618. Markowitz, FE, Bellair, PE, Liska, AE & Liu, J, 2001, “Extending social disorganization theory: modeling the relationships between cohesion, disorder, and fear” 39 Criminology 293-319. Mazerolle, LM, Wickes, RL & McBroom, J, 2010, “Community variations in violence: the role of social ties and collective efficacy in comparative context” 47 Journal of Research in Crime and Delinquency 3-30. Morgan, A, Boxall, H, Lindeman, K & Anderson, J, 2012, Effective crime prevention strategies for implementation by local government, Research and Public Policy series no. 120 Australian Institute of Criminology, Canberra. Morenoff, JD, Sampson, RJ & Raudenbush, SW, 2001, “Neighbourhood inequality, collective efficacy, and the spatial dynamics of urban violence” 39 Criminology 517-558. Phillips, JA & Greenberg, D F, 2008, “A comparison of methods for analyzing criminological panel data” 24 Journal of Quantitative Criminology 51-72. Pinker, S, 2011, The better angels of our nature: The decline of violence in history and its causes, Penguin, UK. Ramey, DM & Shrider, EA, 2014, “New parochialism, sources of community investment, and the control of street crime” 13 Criminology & Public Policy 193-216. 17 Raudenbush, SW & Sampson, RJ, 1999, “Ecometrics: toward a science of assessing ecological settings, with application to the systematic social observation of neighborhoods” 29 Sociological methodology 1-41. Rosenberg, ML, Butchart, A, Mercy, J, Narasimhan, V, Waters, H & Marshall, MS, 2006, “Interpersonal Violence” in Jamison, DT, Breman, JG, Measham, AR, Claeson, M, Evans, DB, Jha, P, Mills, A & Musgrave, P (eds) Disease Control Priorities in Developing Countries 2nd edition World Bank, Washington DC, pp.755-770. Sampson, RJ, 2012, Great American city: Chicago and the enduring neighborhood effect University of Chicago Press, Chicago. Sampson, RJ & Groves, WB, 1989, “Community Structure and Crime: Testing Social Disorganization Theory” 94 American Journal of Sociology 774-802. Sampson, RJ & Morenoff, J, 2006, “Durable Inequality: Spatial Dynamics, Social Processes and the Persistence of Poverty in Chicago Neighborhoods” in Bowles, S, Durlauf, S & Hoff, K (eds) Poverty Traps Princeton University Press, Princeton, pp.176-203. Sampson, RJ, Morenoff, J & Earls, F, 1999, “Beyond social capital: spatial dynamics of collective efficacy for children” 64 American Sociological Review 633-660. Sampson, RJ, Raudenbush, SW & Earls, F, 1997, “Neighborhoods and violent crime: A multilevel study of collective efficacy” 277 Science 918-924. Sampson, RJ & Wikström, PO, 2008, “The social order of violence in Chicago and Stockholm neighborhoods: A Comparative Inquiry” in Shapiro, I, Kalyvas, SN & Masoud, T (eds) Order, Conflict, and Violence Cambridge University Press, Cambridge, UK, pp. 97119. Shaw, CR & McKay, HD, 1969, Juvenile delinquency and urban areas, rev edn University of Chicago Press, Chicago. Spoth, R, Rohrbach, LA, Greenberg, M, Leaf, P, Brown, CH, Fagan, A, Catalano, RF, Pentz, MA, Sloboda, Z, Hawkins, JD & Society for Prevention Research Type 2 Translational Task Force Members and Contributing Authors, 2013, “Addressing Core Challenges for the Next 18 Generation of Type 2 Translation Research and Systems: The Translation Science to Population Impact (TSci Impact) Framework” 14 Prevention Science 319-351. Steenbeek, W & Hipp, JR, 2011, “A Longitudinal Test of Social Disorganization Theory: Feedback Effects Among Cohesion, Social Control and Disorder” 49 Criminology 833-871. Toumbourou, JW, Leung, R, Homel, R, Freiberg, K, Satyen, L & Hemphill, SA, in press, “Violence prevention and early intervention: what works?” in Day, A & Fernandez, E (eds) Violence in Australia: Policy, Practice and Solutions Federation Press, Sydney. van Dijk, J, van Kesteren, J & Smit, P, 2007, Criminal Victimisation in International Perspective: Key findings from the 2004-2005 ICVS and EU ICS United Nations Office on Drugs and Crime, Tilburg, The Netherlands. Vinson, T & Homel, R, 1975, “Crime and disadvantage: the coincidence of medical and social problems in an Australian city” 15 The British Journal of Criminology 21-31. Weatherburn, D & Lind, B, 2001, Delinquent-prone communities Cambridge University Press, Cambridge. Weisburd, D, Groff, E & Yang, S, 2012, The Criminology of Place: Street Segments and Our Understanding of the Crime Problem Oxford University Press, Oxford. Weissberg, RP & Greenberg, MT,1998, “Prevention science and collaborative community action research: Combining the best from both perspectives” 7 Journal of Mental Health 479492. Wickes, R, in press, “Social Disorganization Theory: Its History and Relevance to Crime Prevention” in Teasdale, B & Bradley, M (eds) Preventing crime and violence: Volume 2 of Advancing Prevention Science Springer, New York. Wickes, R, Zahnow, R, White, G & Mazerolle, L, 2013, “Ethnic Diversity and its Impact on Community Social Cohesion and Neighbourly Exchange,” 36 Journal of Urban Affair 51-78. 19 Figure 1. Change in violence rates in ACCS suburbs between Times 1 and 3 (2008 and 2011) 20 Figure 2. Spatial distribution of disadvantage and social cohesion and trust at Time 3 (2011) 21 Table 1: Descriptive statistics for all variables Time 1 Time 2 Time 3 Mean SD Min Max Mean SD Min Max Mean SD Min Max 897 1557 0 16949 854 988 0 7062 814 1440 0 14407 Trust 3.76 0.28 3.08 4.31 3.69 0.27 2.98 4.28 3.71 0.28 3.02 4.28 18.38 6.26 6.54 33.29 13.68 5.84 3.93 28.24 13.56 5.77 3.88 27.42 % low income 39.70 11.77 17.06 66.57 27.30 11.10 8.15 55.24 22.07 9.37 6.96 46.41 % unemployed 4.92 1.82 1.83 10.09 2.80 1.05 0.53 6.04 3.80 1.30 1.34 8.15 % renting 27.39 12.64 4.05 57.76 23.01 12.56 1.75 51.14 26.49 13.70 2.36 57.43 % Indigenous 1.53 1.19 0 6.83 1.58 1.56 0 9.08 1.88 1.66 0 9.09 Language Diversity 0.23 0.12 0.05 0.65 0.26 0.15 0.04 0.78 0.28 0.17 0.06 0.72 Residential Mobility 45.92 9.79 20.72 74.16 42.32 8.97 23.45 79.32 38.19 10.34 14.86 74.09 Population Density 9.75 7.23 0.11 29.30 8.99 8.29 0.08 33.81 10.17 9.02 0.10 36.66 Violent Crime Rate (2008, 2010, 2011) Social Cohesion and (Wave 2-2008; 3-2010; 4-2012) Census Variables (2001; 2006; 2011) Disadvantage 22 Table 2. Correlations of violence, social cohesion and trust, and the independent variables, averaged across Times 1, 2 and 3 Variables 1- Social Cohesion and Trust Scale 1 2 3 4 5 6 1 2- Disadvantage -0.46* 1 3- Language Diversity -0.46* 0.35* 1 4- Residential Mobility -0.04 0.25* 0.10 1 5- Population Density -0.32* 0.41* 0.47* 0.23* 1 6- Violence -0.29* 0.39* -0.05 0.05 0.27* 1 * Correlation is significant at the 0.01 level (2-tailed) 23 Table 3: Results for hybrid regression analysis examining the effect of suburb social cohesion and trust on rates of violent crime over time β SE p Social Cohesion and Trust -0.463 0.181 0.01 Disadvantage 0.016 0.009 0.06 Residential Mobility 0.004 0.004 0.38 Language Diversity -0.176 0.450 0.73 Population Density -0.062 0.053 0.24 Social Cohesion and Trust -0.693 0.291 0.02 Disadvantage 0.094 0.012 0.00 Residential Mobility -0.009 0.006 0.14 Language Diversity 0.172 0.415 0.68 Population Density -0.043 0.047 0.35 Intercept 5.298 0.283 0.00 Variable Within Neighbourhood Variation Between Neighbourhood Variation 𝑅2 50.2% 24