The comparative context of collective efficacy: Understanding neighbourhood disorganisation and willingness to intervene in Seattle and Brisbane

Abstract

The collective efficacy literature provides a framework to understand how neighbourhood structure influences violence. Existing findings have been based largely on American cities where disadvantage and ethnic segregation are more concentrated. Thus, they are not always representative of other Western cities where structural disadvantage has a different history as well as less variation across neighbourhoods. This paper explores the comparative effect of collective efficacy in Seattle, USA, and Brisbane, Australia. Findings show that collective efficacy is a significant predictor of violent victimisation in both cities. However, in Brisbane, traditional measures of structural disorganisation are less of an influence on victimisation than in Seattle, and that collective efficacy as a neighbourhood process can exist and vary across neighbourhoods without extreme disorganisation.

Keywords

Collective efficacy comparative criminology neighbourhoods

Introduction

The study of neighbourhoods and neighbourhood effects continues to be important in the study of crime both theoretically and empirically. The first criminological linking of neighbourhood conditions to explanations of crime by Park and Burgess (1921) and Shaw and McKay (1942) has been revitalised by notions of collective efficacy and other systemic theories of crime (Bursik, 1988; Bursik & Grasmick, 1993; Sampson, Raudenbush, & Earls, 1997). Collective efficacy is supported in the literature to explain why criminologists consistently see the association between high neighbourhood disorganisation (poverty, residential instability, and ethnic heterogeneity) and high levels of crime. Collective efficacy is defined as the ‘willingness of local residents to intervene for the common good’ (Sampson et al., 1997, p. 919). This study uses a comparative framework to understand and explore two international cities: Seattle, USA and Brisbane, Australia; how collective efficacy is formed and how violent victimisation can be explained by levels of neighbourhood collective efficacy and neighbourhood structural conditions. These results support collective efficacy as a mechanism that influences violent victimisation across international and cultural borders, and suggest that comparative design in the study of neighbourhoods and crime is beneficial to increasing the understanding of neighbourhood crime and social control.

The global importance of ‘collective efficacy’ has become increasingly popular for exploring varying rates of neighbourhood crime in cities around the world. Stockholm, Sweden (Sampson & Wikstrom, 2007), Peterborough, UK (Wikstrom & Treiber, 2009), and Brisbane, Australia (Wickes, 2010) are just some of the sites that have extensively used the theoretical concept of collective efficacy to explore urban crime. These sites vary widely in their levels of crime and structural and cultural contexts. For example, different immigration policies have led to different experiences for new immigrants and ways that residents respond to immigrants in each of these locations. Interestingly, findings overwhelmingly suggest that collective efficacy plays a similar role in mediating or partially mediating the relationship between structural conditions and crime across these locations.

While collective efficacy theory has been largely based on Chicago neighbourhoods, Chicago is not representative of other large cities in the United States or across the globe. Chicago experiences extreme levels of poverty, ethnic segregation and diversity and has one of the highest urban crime rates in the country (Kubrin & Weitzer, 2003). By exploring the role of collective efficacy and neighbourhood crime without those extremes and in cities that are culturally, structurally and historically different, our understanding of how collective efficacy influences violent victimisation in other American and international cities would be enhanced and significantly contribute to the literature on collective efficacy and crime.

Neighbourhood crime and collective efficacy theory

Collective efficacy, conceptualised first by social psychologist Albert Bandura (1997)¹ and in criminology by Sampson et al. (1997) and Sampson, Morenoff, and Earls (1999), suggests that a neighbourhood’s ability to mobilise resources and combat neighbourhood problems rests not in the strength and number of social ties and relationships between residents, but in shared beliefs and norms in the ability to produce collective results in response to neighbourhood problems. Therefore, collective efficacy in the context of criminology is a shared expectation in the willingness of others to intervene when issues of crime and social disorder arise.

The development of collective efficacy theory stems from research by Sampson et al. in Chicago using the Project on Human Development in Chicago Neighbourhoods (PHDCN) (Earls & Visher, 1997). Their work extends the social disorganisation literature by shifting the focus of social ties and neighbourhood relationships as a mediating mechanism between neighbourhood structure and crime toward the process of converting those ties and relationships into actions or desired collective goods or outcomes. In this vein, Sampson et al. (1999) argue that social capital refers to the resource potential of social ties and neighbourhood relationships, while in contrast; collective efficacy refers to the task-specific process of converting those ties into desired outcomes.

Previous social disorganisation literature has focused heavily on the accumulation of social capital and social ties among residents themselves (Kornhauser, 1978). However, modern conceptualisations of urban and community life suggest a shift away from obligations toward the collective to one that is driven by more individualistic preferences within communities (Putnam, 1993). Therefore, defined by more systemic frameworks of disorganisation, the notion that social relationships in number and strength make social order and social control possible within neighbourhoods (Bursik & Grasmik, 1993) suggests that the strength and number of social ties themselves are not adequate in understanding how modern urban neighbourhoods are able to combat problems of crime and social disorder (Mazerolle et al., 2010).

In a study of 343 Chicago neighbourhoods, Sampson et al. (1997) found that collective efficacy was the casual mechanism that accounted for perceptions of neighbourhood violence, violent victimisation and homicide rates. In order to assess the impact of collective efficacy on perceived violence, survey measures addressing social cohesion and expectations for informal social control were used to measure neighbourhood collective efficacy. In doing so, Sampson found that the relationship between social-structural composition of a neighbourhood and neighbourhood levels of crime is mediated by collective efficacy. Thus, traditional conceptualisations of the effect of social disorganisation (poverty, residential instability, and ethnic heterogeneity) do not solely account for perceptions of violence, violent victimisation, and neighbourhood homicide rates on their own.

In a later study, Sampson et al. (1999) assessed the effect of neighbourhood characteristics on levels of collective efficacy with respect to the social control of children. They found that neighbourhoods characterised by residential instability and concentrated disadvantage have much lower expectations for the social control of children, and specifically lower expectations that others will intervene on behalf of children in the neighbourhood. In contrast, residents in areas of concentrated affluence are more likely to be able to achieve an environment geared toward the social control of children; these neighbourhoods have more resources and social networks that can be used to monitor and supervise children’s behaviour. Furthermore, Sampson et al. (1999) were able to demonstrate the spatial dependence of both collective efficacy and social capital and acknowledge the fact that neighbourhoods exist in a larger urban context where resources and neighbourhood processes are shared. Recent work in Seattle has demonstrated similar finding of spatial dependence in regards to labour market distribution and violent crime (Crutchfield, Matsueda, & Drakulich, 2006).

The emergence of Sampson et al.’s (1999) work on collective efficacy as a causal mechanism linking neighbourhood characteristics to violent crime has continued to be an important component in the criminological literature. In criminology, collective efficacy in other urban areas is thought to have a similar effect on crime and victimisation. Similar results have been found in neighbourhoods in Peterborough, UK (Wikstrom & Treiber, 2009), and Stockholm, Sweden (Sampson & Wikstrom, 2007). Sampson and Wikstrom (2007) find that despite distinctly different urban and national contexts between Chicago and Stockholm, neighbourhood structure, informal social control, disorder and violence appear to manifest consistently with the social disorganisation and collective efficacy literature. Similarly, evidence from Peterborough, UK highlights that social cohesion, a critical element of collective efficacy, not only helps explain victimisation but also the offending actions of individuals (Pauels & Svensson, 2010; Wikstrom & Treiber, 2009).

The comparative context of Seattle and Brisbane

This study explores the role of neighbourhood collective efficacy in Seattle, USA and Brisbane, Australia, as well as how collective efficacy is formed in both sites. These sites are interesting from a comparative perspective because of their structural differences. Though both are western urban centres, Seattle and Brisbane differ considerably in a number of respects including levels of welfare support, concentration of poverty, levels of crime and ethnic diversity. However, similar to Sampson and Wikstrom’s (2007) rational for comparing Stockholm and Chicago, one of the only comparative studies of collective efficacy to date, the aim is to uncover characteristics that ‘transcend cultural and national boundaries’ (Sampson & Wikstrom, 2007, p. 98) that help to explain violence and victimisation. The purpose of comparing Brisbane and Seattle is to determine what common processes exist in two cities with different characteristics and urban histories. Sampson and Wikstrom (2007) point to the value in cross national comparative work in this area to understand how social, structural and community settings can generate violent acts. This paper aims to contribute to the currently limited comparative work in collective efficacy and social disorganisation. Both Seattle and Brisbane have also been sites of large data collection, of which some items have been based on the PHDCN. Therefore, the comparison between these two cities is a practical one in addition to the benefits of theoretically and empirically understanding common processes that influence violence in cities that have different characteristics.

Despite limited comparative work, criminologists have separately indicated that the same process has been observed in other US cities, despite structural differences to Chicago. Crutchfield et al. (2006) find that social disorder helps explain the relationship between labour market participation and violent crime. Specifically, lower levels of social disorder (utilising the same social cohesion measures used by Sampson et al. (1997)) in Seattle neighbourhoods can in part be explained by lower labour market participation, particularly in minority neighbourhoods. Peterson and Krivo (2010) also find support for the role of collective efficacy as a mediating mechanism for neighbourhood violence. However, their results highlight the need for investigating neighbourhood processes in culturally different contexts.

Table 1 summarises the demographic differences between Seattle and Brisbane based on census data. Structural and cultural differences between Seattle and Brisbane are interesting because they suggest that neighbourhoods might function differently in order to combat neighbourhood problems. For example, Australia’s more extensive welfare policies might suggest that poverty and inequality are not as important factors in neighbourhood culture or crime as they have been found to be in the US context. Capitalising on differences such as this allows for exploration that will lead to a greater understanding of how neighbourhoods both experience and react to crime. For example, Seattle and Brisbane have comparative levels of foreign-born residents; approximately 19% of the population in Seattle, and 22% in Brisbane (Australian Bureau of Statistics, 2006; U.S. Census of Population and Housing, 2000). While the overall percentage of foreign-born residents may be similar in Brisbane and Seattle, the nature of the typical immigrant is remarkably different. Approximately 37% of Brisbane residents are non-native speakers of English, as opposed to approximately 9% of Seattle residents. These nuanced differences are important because language barriers in addition to differences in social and cultural norms hinder immigrants from forming social ties with other residents (Shaw & McKay, 1942).

Table 1.

Demographics for Seattle and Brisbane.

Demographics	Seattle	Brisbane
Population	563,374	1,763,131
Average population per census tract	4559.22	8099.38
% Below poverty line	11.80	15.80
% Affluent (income >$100K)	31.50	20.20
% Unemployed	3.80	4.40
% Single parent families	10.30	16.10
% Home ownership	51.90	65.90
% Renters	48.10	29.90
% Foreign-born	18.80	21.70
% Non-native speakers	9.30	37.10

Seattle demographics taken from U.S. Census of Population and Housing 2000. Brisbane demographics taken from Australian Bureau of Statistics Census 2006.

Other differences include Brisbane’s higher rate of homeownership in comparison to Seattle, as well as higher rates of overall poverty and unemployment. Levels of affluence (residents with household incomes over $100K) are comparatively higher in Seattle than in Brisbane. However, given Australia’s well-developed welfare system it is unknown how these structural differences influence community outcomes such as crime and victimisation.

The US and Australia have generally had differing paths of capitalist development. While factories in the inner sectors of American cities attracted low-skilled workers and new immigrants, Australia did not have any factories where rental properties or low-income neighbourhoods could congregate. Australia relied heavily on other nations for raw materials and as a result there were few factories as part of the urban city (Winter & Bryson, 1998). In fact, the vast majority of Australia’s low-income housing was constructed by State and Federal Governments due to a housing shortage after the Second World War (Davidson, Dingle, & O’Hanlon, 1995). These areas were stigmatised as the least desirable in which to live, and even today overwhelmingly contain working class Australians on meager incomes. These are akin to the areas of low-income housing seen in the US context, but are mediated by an important structural factor: the development of Australia as a welfare state (Winter & Bryson, 1998).

Extensive social provisions in Australia, such as income support, free education and universal health care have prevented the emergence of the levels of deprivation seen in the US. Norton (1994) has argued that the Australian welfare state is an important factor in reducing outright deprivation, especially for the poorest 20% of Australians. As such, the State has been influential in shaping urban poverty and development in Australian cities. However, Australian cities still see some of the same social problems that plague American cities. In areas of ‘urban poverty’ – where residents need to supplement income with government assistance, there is widespread unemployment and crime, and many residents have not completed high school education (Vinson & Homel, 1975; Winter & Bryson, 1998).

Recent population and economic trends in Brisbane have continued to put pressure on the city’s urban environment. Specifically, this is due to the internal migration of new residents from other Australian cities – often referred to as ‘sunbelt migration’. These residents have moved north for employment in the expanding tourist industry and the technological hub that is growing in Brisbane (Stimson & Taylor, 1999). In addition, the cost of living in the Brisbane area was much lower than other urban centres such as Sydney and Melbourne (Stimson & Taylor, 1999).

Seattle has also undergone urban changes, although not as rapidly as Brisbane. Most internal migration to the Seattle area has been due to the city’s importance as a technological hub – it is home to large companies such as Microsoft and Boeing. In addition, Seattle has also had an increasing population of Latino immigrants. Passel and Suro (2005) found that Seattle, among other cities, has experienced a rapid increase of immigrants from south of the border. This is an important change in immigration trends because traditionally, Latino immigrants have settled in the cities of Chicago, New York City and Miami, Florida (Passel & Suro, 2005).

Race and ethnic relations are also very different in the Australian context in comparison to the US. In Australia, the term ‘race’ is generally not used as a measure of ethnic dispersion. Instead, it is more useful to talk about ethnic heterogeneity in terms of the percentage of foreign-born residents, region of birth or ancestry. This is mostly due to Australia’s recent history of liberal immigration policies (Nguyen, 2005). In fact, census data containing information about ethnicity do not create categories of ‘race’, but instead contains detailed information on regions of birth, spoken languages and length of time spent as an Australian resident. The one exception to this is the categorisation of Australia’s native peoples as Aboriginal or Torres Strait Islander (Australian Bureau of Statistics, 2006).

In the US context, race is an important aspect of sociological research. For those using social disorganisation as a theoretical framework in particular, the impact of race with respect to neighbourhood segregation of immigrants and African-Americans in the urban context cannot be ignored. In fact, sociologists often use ‘percent African-American’ as an indicator for measures of disadvantage (Sampson et al., 1997). In addition, the migration of African-American slaves from the south during the Great Migration has had a lasting effect on the nature of receiving urban cities such as Chicago and New York (Daniels, 1990).

Early notions of ‘ethnic heterogeneity’ suggested that multiple ethnic groups in one neighbourhood did not allow for the emergence of common norms and values (Shaw & McKay, 1942). Cultural differences and language barriers were thought to be just one of the specific barriers that led ethnic groups to be isolated from each other. However, in Brisbane most foreign-born and recent immigrants are from English-speaking nations such as the UK and New Zealand (Australian Bureau of Statistics, 2006). This suggests that while cultural differences may still be a factor in creating boundaries between ethnic groups, problems due to language differences are not likely to contribute to common problems of ‘ethnic heterogeneity’ set out by social disorganisation theorists.

Currently, it is not clear in the collective efficacy literature whether neighbourhood collective efficacy can be produced in neighbourhoods with high levels of disadvantage, diversity and instability, or whether low levels of collective efficacy are a direct result of a structurally disorganised neighbourhood state. The differences between Seattle and Brisbane outlined above provide an opportunity to explore whether collective efficacy is associated with the same type of biographical and neighbourhood characteristics and whether it can reduce crime in settings with vastly different levels and types of structural disorganisation.

The utility of comparative research

Comparative research involves the systematic and theoretically informed comparison of outcomes in two or more countries, cultures or places (Bierne, 1997). Scholars as far back as Durkheim have understood that the benefits of comparative research, particularly in criminology, enhance both policy and theoretical development (Leavitt, 1990; McClintock & Wikstrom, 1992; Nelken, 1997). Though comparison is often difficult due to differences in context and measurement, comparative research helps identify the common mechanisms in offending behaviour, victimisation and neighbourhood processes and tests theoretical ideas. Nelken (1997) argues that comparative research is critical for understanding differences in culture and place given that there are common experiences of crime problems. In addition, countries have a tendency to adopt the common causes and implications of those crimes to be universal. Comparative research allows for an understanding of the limits and assumptions of those causes, processes and mechanisms as they relate to different cultural contexts.

The social disorganisation literature was largely based on how American cities evolved over time, with particular attention to the plight of immigrants and minority groups and driven by neighbourhood disadvantage. However, the social disorganisation and collective efficacy frameworks are used to explain crime at the neighbourhood level in urban cities and rural towns around the world such as Australia (McCrea, Shyy, Western, & Stimson, 2005), Hong Kong (Lee, 2006), the UK (Wikstrom & Treiber, 2009), and Stockholm, Sweden (Sampson & Wikstrom, 2007). The prevalence of this framework suggests that comparative research in this area is useful for exploring the aspects of collective efficacy and disorganisation that are not well understood. More specifically, though the relationship of collective efficacy to crime has been established in many international sites, a detailed look at the comparative effects of neighbourhood structural conditions, collective efficacy and crime would provide further understanding of how different urban environments foster collective efficacy at the neighbourhood level.

Comparative research in this area of criminology has the ability to inform the theoretical perspective on neighbourhood ecology and crime. Much of the current literature on collective efficacy has been largely based on research from the Chicago area where a large part of the relationship between structural disorganisation, collective efficacy and crime is based on high rates of urban segregation. The extent of such segregation is not seen in other US cities, nor is it common in cities outside of the US. Historically, neither Seattle nor Brisbane has experienced extreme levels of poverty or racial segregation. In addition, Brisbane may experience even less the effects of poverty than Seattle due to Australia’s well-established welfare provisions. If the relationship between collective efficacy, crime and neighbourhood structure can be preserved without these extremes, our understanding of collective efficacy as a neighbourhood process will be enhanced. Understanding whether collective efficacy reduces crime in areas that are not characterised by high levels of ethnic segregation and disadvantage would be a significant contribution to the collective efficacy literature.

The literature on cross cultural and comparative research clearly highlights the benefits of comparative work across international sites (Gasparini, 2010; McFarlane, 2010). However, comparative and cross cultural research does carry some inherent difficulties that should be acknowledged here. First, though the survey instruments used in Seattle and Brisbane are highly similar, differences among respondents are likely to occur. Measures that gauge perceptions, for example, are likely to be answered within the context of the culture of each respective city (Angel, Angel, & Hill, 2008; McFarlane, 2010). However, the purpose of this comparative analysis is to understand the applicability of social disorganisation and collective efficacy theory in two cross national urban cities, and therefore these differences provide the ability to understand where the theory is generalisable and where it is limited in understanding neighbourhood processes of crime. An additional difficulty for comparative research in neighbourhood research is in the definition of the neighbourhood boundary. This inherent difficulty is discussed in detail in this paper.

There are two ways in which comparative work is most effectively used. In the first instance, comparisons are similar on a number of factors, with one major difference (Westfelt & Estrada, 2005). In the other instance (see McFarlane, 2010), which is the approach used in this paper, Seattle and Brisbane are different in a large number of aspects of interest such as levels of poverty, immigrant concentration, ethnic diversity, urban history and neighbourhood development. However, the process of collective efficacy in the neighbourhood context on violent victimisation is expected to be similar in both sites given the work previously discussed where comparative analysis has been utilised. It is through examining how these differences come together to explain violent victimisation with collective efficacy as a mechanism for crime reduction that the comparison is drawn.

In sum, the hypothesised relationships between each site include (1) that biographical and neighbourhood characteristics that are associated with higher levels of collective efficacy will be the same in each site, (2) that neighbourhood structural conditions, disadvantage, residential instability and immigrant concentration is associated with violent victimisation in both Seattle and Brisbane as had been demonstrated in Chicago, (3) that the effect of neighbourhood structural conditions will be smaller in effect size in Brisbane due to differences in the geographical spread of disadvantage and (4) that neighbourhood collective efficacy will mediate the relationship between structural conditions on violent victimisation in both Seattle and Brisbane despite different conditions in each site.

Data and methods

The effect of individual and neighbourhood characteristics on individual collective efficacy scores as well as the effect of collective efficacy on violent victimisation is examined using separate multilevel models for each site. In this way, not only are the unique differences between Seattle and Brisbane retained in the analysis but the integrity of slight data differences between Seattle and Brisbane is also protected. Though some measures are not identical, each measure was chosen because it represents the best measure of a particular construct for each site.²

The data used for the following analyses are from the Seattle neighbourhoods and Crime Survey (Matsueda et al., 2010) (N = 5812), US Census 2000, and Seattle Police Department (SPD) Crime data for the site of Seattle.³ In Brisbane, the Australian Community Capacity Survey (ACCS) (Mazerolle, Wickes, & McBroom, 2005), Australian Bureau of Statistics Census (2006) and Queensland Police Service (QPS) crime data are used.⁴ See Table 2 for complete descriptive statistics for both Seattle and Brisbane. Similar to the PHDCN, the Seattle and Brisbane neighbourhood surveys use residents as informants on characteristics of their neighbourhood: perceptions of social control and neighbourhood cohesion, victimisation and fear of crime. It utilises identical measures of collective efficacy to the PHDCN (Sampson et al., 1997), thus making this key variable comparable to measures used in other American sites. In Seattle, survey data were collected in 2002–2003, and stratified over all 123 census tracts in the city with a final sample size of 5757 respondents. In Brisbane, the survey was conducted in a sample of 82 Statistical Local Areas (SLAs). Eighteen core SLAs were chosen for analysis, and SLAs sharing borders with these core SLAs were included creating 18 clusters of SLAs drawn from Brisbane’s 224 SLAs. Survey respondents were selected for participation using a quota scheme to determine a sufficient number of respondents per SLA. Survey respondents were randomly selected via random digit dialing. The final sample size across 82 SLAs was 2859. Participants completed the survey over the phone with a trained interviewer. The response rate for survey participation was approximately 68% (Mazerolle et al., 2010).

Table 2.

Descriptive statistics for Seattle and Brisbane.

Level	Seattle, USA				Brisbane, Australia
Level	N	M, Mdn, Mo,%	SD	Range	N	M, Mdn, Mo, %	SD	Range
Individual
Age		48.22 (M)	17	17–103		35–39 (Mo)	3.1	18.5–70+
Annual household income
Less than $25K		14.30%	–	–
More than $75K		29.30%	–	–
Less than $39K						32.00%	–	–
More than $80K						23.40%	–	–
Home owners		68.10%	–	–		74.50%	–	–
Years at current address		11.65 (M)	12.9	<1–87		4.63 (M)	1.69	<1–7
Female		50.60%	–	–		59.40%	–	–
Married		44.90%	–	–		54.10%	–	–
Black/African-American		5.70%	–	–
Asian		8.27%	–	–
Latino		4.57%	–	–
Aboriginal or TSI						1.20%	–	–
Violent victimisation		11.30%	–	–		7.90%	–	–
	5757				2859
Neighbourhood
% Low-income households		25.7 (M)	14.2	7.2–77.5		31.37 (M)	11.3	0–62.6
% Same home 5 years or more		44.3 (M)	13.3	10.1–69.7		54.8 (M)	8.8	34.9–80.9
% Black/African-American		8.1 (M)	10.3	0–50.2
% Aboriginal or TSI						1.36 (M)	1.15	0–7.23
% Foreign-born		16.3 (M)	10.8	5.4–61.7		27.7 (M)	8.5	17.7–59.4
% Non-English speaking		21.4 (M)	12.9	8.6–79.9		10.9 (M)	8.6	2.2–46.9
4-year average violent crime rate		35.3 (M)	54.5	1–324.5		481.3 (M)	378.8	0–2204.3
Collective efficacy		2.97%	0.22	1.00–4.00		2.94 (M)	2.49	1.00–4.00
	123				82

TSI: Torres strait islander.

The neighbourhood as a unit of analysis

Typically, the measurement of ‘neighbourhood’ is an imprecise science. There are multiple definitions of what constitutes a ‘neighbourhood’ (Hillery, 1955), and there are likely to be individual differences among residents in what each consider to be the boundary of their community. In this study, the unit of analysis operationalised as a neighbourhood or community is an approximation. Hipp (2007) argues that social scientists should dedicate special attention to choosing the level of aggregation appropriate to both the individual urban structure of the research site as well as the outcome being studied. He argues that using administrative units that are too big can create too much heterogeneity within the neighbourhood and therefore obscure empirical relationships. Instead, Hipp (2007) argues that where possible, census blocks should be used to approximate neighbourhoods, or that multiple units of aggregation be used depending on the neighbourhood measure. However, Hipp (2007) also acknowledges that there is no perfect way to measure a neighbourhood and that social scientists should consider both the methodological as well as theoretical considerations. For example, Sampson et al. (1997) used neighbourhood clusters that are larger than census tracts to approximate neighbourhoods by combining census tracts to create geographically meaningful boundaries. Others, such as Guest, Kubrin, and Cover (2008) and Miethe and Meier (1994) and Coulton, Cook, and Irwin (2004) have utilised smaller units such as block groups. Still, Wooldredge (2002) argues that similar results are found regardless of the scale of aggregation particularly for measures of social class and residential stability.

In Seattle, census tracts are used to approximate neighbourhoods. I follow the work of Matsueda et al. (2006), Drakulich (2010), and Guest et al. (2008) who also utilise census tracts in their work on Seattle neighbourhoods. In Seattle, census tracts have an average population of 4559 persons (SD = 1714.99). The smallest census tract has 1091 persons, the largest contains 9047. In measuring collective efficacy, respondents were asked to think of their neighbourhood as ‘the three city blocks on each side of their home’, roughly approximating a census tract (Matsueda, Crutchfield, Guest, & Kubrin, 2010). Study investigators had reason to believe that census tracts were a good indicator for a neighbourhood given that tract boundaries also reflected natural breaks in Seattle’s geography, and generally map onto areas that are socially considered part of a neighbourhood. As such, census tracts have been used to approximate neighbourhoods in Seattle.

In Brisbane, the SLA is used to approximate the concept of a ‘neighbourhood’. These SLAs are often larger than US census tracts and typically map on to Brisbane suburbs.⁵ SLAs have an average population size of 7266 persons (SD = 6605.81). The smallest SLA has 250 persons, the largest contains 69,694. ACCS survey investigators produced a pilot study designed to investigate how Brisbane residents defined ‘community’ or ‘neighbourhood’. SLAs were found to be good approximations as residents thought of their community boundary as roughly within the parameters of their ‘suburb’ (see Mazerolle et al., 2010). In measuring collective efficacy, survey respondents were asked to think of their neighbourhood as approximately the suburb boundary (Mazerolle et al., 2010).

Though the size in population within each tract or SLA varies widely between Seattle and Brisbane, previous neighbourhood and community research in each respective site provides confidence that the use of census tracts in Seattle and SLAs in Brisbane are meaningful units of a neighbourhood for the residents and respondents within them even though the variation and average geographic and population size between the two are different (Guest et al., 2008; Mazerolle et al., 2010; Sampson & Wikstrom, 2007). From a comparative utility, the use of census tracts in Seattle and SLAs in Brisbane was deemed most appropriate for comparison. Smaller units in Brisbane would not contain enough survey respondents for fruitful analysis, therefore, limiting the use of smaller administrative boundaries. In Seattle, where smaller units of analysis can be used, they would be too different in size and nature to yield a helpful comparison between the two sites.⁶ Table 2 contains the descriptive statistics for all variables in the models.

Neighbourhood surveys

Following from Drakulich (2010), Mazerolle et al. (2010), and Sampson et al. (1997), a number of individual level variables were used in the analyses from each research site. Each of these variables was constructed so that they would be directly comparable in each site, with the exception of annual income levels and ethnic identity. Respondents’ gender (male = 0, female = 1), whether the respondent was a home owner (own = 1, rent = 0), years at current address, age of the respondent, and whether the respondent was married (yes = 1, no = 0) are included in both sites. In addition, two indicators were used at each site to control for annual income. In Brisbane, indicators included those who earned under $39K per year,⁷ and those who earned over $80K per year. In Seattle, indicators were used for respondents earning $25K and under, and $75K and over.⁸ Though income cutoff is not identical, they reflect the context of income and poverty in each respective site. For example, using the identical cut off of $25K and under in Brisbane would inaccurately reflect the relative quality cost of living in respect to that level for the Brisbane/Australian context; i.e. this would underestimate the number of people living below the poverty line in Brisbane.

Special attention has been paid to the measure of race and ethnicity in each site. In Seattle, the work of Crutchfield et al. (2006) and Matsueda et al. (2006) is followed and an individual measure of Latino identity is included from the survey. It is a-typical for Australians to talk about racial identity given the different context of diversity; however, a measure of Aboriginal/Torres Strait Islander (ATSI) is included for the Brisbane site. Though the percentage of this population is small in urban areas and are not highly represented in survey respondents (see Table 2), they do represent a minority group with a history of concentrated disadvantage.

Following the work of Sampson et al. (1997), collective efficacy was constructed as a scale of six items from the social control and cohesion component of the surveys of each city. These measures are identical and therefore directly comparable. For the social control component of the measure, respondents were asked to rate on a four-point Likert scale whether it was very likely, likely, unlikely or very unlikely that people in their neighbourhood would do something (a) if children were skipping school and hanging out on the street, (b) if children were spray painting graffiti on a local building, (c) if there was a fight or someone was being threatened in front of their house, and (d) if a child was showing disrespect to an adult. For the social cohesion component, respondents were asked to rate on a four-point Likert scale whether they strongly agree, agree, neither agree nor disagree, disagree, or strongly disagree that (a) people in the community are willing to help their neighbours, and (b) people in the community can be trusted. Responses to each of these were averaged and then aggregated to the census tract/SLA level (Sampson et al., 1997). Alpha coefficients for the collective efficacy scale in Seattle and Brisbane were 0.77 and 0.67, respectively. The reliability of the collective efficacy scale as a neighbourhood measure is discussed in greater detail as part of the results.

Violent victimisation in Brisbane and Seattle is measured as a binary response variable that indicates whether the respondent has been a victim of a violent crime (1) or has not (0). In Brisbane, respondents were asked whether they had been the victim of a violent crime in the last 12 months (yes/no). In Seattle, respondents were asked how many times in the last 12 months they experienced violent victimisation. In Seattle, this measure was recoded to match the Brisbane data where a count of 0 indicated no victimisation and any number above 0 indicated experience of victimisation (1). In both cities, violent victimisation was a rare event; 11.3% of respondents in Seattle and 7.9% of respondents in Brisbane indicated that they had been victimised in the last 12 months.

Census data

Following the work of Mazerolle et al. (2010), census tract/SLA level predictors included percentage of low-income households and homes with the same residents for five or more years. In Seattle, we followed the research from American scholars (Blau & Blau, 1982; Sampson et al., 1997, 1999) and included the percentage of the population who are African-American. For the Brisbane sample, racial categories similar to those used in the US are not available. However, the per cent of neighbourhood residents who identify as ATSI are included. In order to capture the ethnic diversity in Brisbane as well as Seattle, the percentage of the neighbourhood population who are non-native speakers of English and the percentage of neighbourhood residents who are foreign-born are used. The indicators help account for the vastly different ethnic composition in each site. These measures are comparable, but retain the context of diversity in each site (Crutchfield & Ramirez, 2012).

Police reports

Following Sampson et al. (1997), the four-year average violent crime rate was included as a control at the tract/SLA for each site. Violent crime rates were provided by the SPD and QPS and represent crime known to the police in each tract/SLA. The four-year average violent crime included the year of survey data collection and three years prior. Violent crimes included at each site included homicide, robbery, aggravated assault and rape, and are thus comparable. Though much of the comparative work on violent crime restricts measures to homicide rates/counts because of reporting reliability (Westfelt & Estrada, 2005; Young, 2005), it is essential to cast a wider view of violent crime in this comparative context. Due to low homicide rates in Australia as a whole, there is not enough variation in the homicide rate or count in Brisbane to produce a meaningful analysis. Given that police records in both Seattle and Brisbane provided violent crime broken down into similar offences, four-year averages included the violent offences that were defined similarly in each site.

Analytic strategy

It is worth noting that the analytical strategy used here does not aim to directly replicate the models of the Sampson et al.’s (1997) original paper. Instead, it aims to utilise Sampson et al.’s (1997) theoretical approach and ideas to understand whether similar relationships exist in a comparative context. Regardless, the differences to the Sampson et al. approach are acknowledged. First, Sampson et al. (1997) look at the effect of neighbourhood structure (disadvantage, immigrant concentration and residential stability) on violent victimisation, perceptions of violence and homicide rates. Given the interest here is in the effect of the different neighbourhood and individual indicators on violent victimisation in two different cities, it is important to retain unloaded factors at the neighbourhood level in order to see which aspects of disadvantage, immigrant concentration and residential stability are mediated by collective efficacy in each site, and understand how they might differ. This paper only uses violent victimisation as an outcome. There is not enough variation in homicide rates in Brisbane to detect any statistical effect, and no comparable data on perceptions of violence were available in both cities.

Data for each site were explored for missing data and influential outliers. Extreme outliers at the tract/SLA level in both sets were flagged and the following analyses were run with and without the outliers to determine if their inclusion dramatically changed the results of the models. No evidence of dramatic changes to multilevel regression results were detected, thus they have been included in all models. Missing data on the dependent variable were detected at both sites. In Brisbane, violent victimisation was missing in 0.1% of cases (three cases in total). In Seattle, violent victimisation was missing in 1.7% of cases. In both sites, these missing cases were excluded from the analysis. Seattle data also experienced missing data in annual household income, approximately 11.7%. Following the work of Drakulich (2010), I included a dummy variable for missing income as a control in the model.⁹

The decomposition of variance for the collective efficacy scale was examined in order to establish whether at least some of the differences among the 82 Brisbane SLAs and 123 Seattle census tracts were due to the variance at the neighbourhood level (see Sampson et al., 1997). Following Raudenbush and Bryk (2002), within group and between group variance of the collective efficacy scale was estimated in each research site by using a random intercept one-way analysis of variance where Y_ij is the collective efficacy score for individuals i within tracts/SLAs j in the following equation

Y_{ij} = μ + U_{j} + R_{ij}

(1)

where μ is the grand mean for the population, U_j is the random effect of tracts/SLAs j, and R_ij is the residual effect for individuals i within tracts/SLAs. The interclass correlation coefficient (ICC) is defined by

\overset{\land}{ρ} = \frac{{\overset{\land}{τ}}_{ij}}{{\overset{\land}{τ}}_{ij} + {\overset{\land}{θ}}^{2}}

(2)

where τ_ij is the estimate of the between groups variance in collective efficacy score for individuals i within tracts/SLAs j, and θ² is the estimate of the within groups variance in collective efficacy score.

The correlates of perceived collective efficacy are first analysed in linear mixed effect multilevel models for Seattle and Brisbane separately. The results of these models guide the inclusion of significant predictors in the final models of violent victimisation below. Linear mixed effect models were fitted with the equation

y = X β + Zu + ɛ

(3)

where y is a vector of responses, X is a matrix of fixed effect covariates, β is a matrix of fixed effect parameters, Z is a matrix of random effect regressors, u is a vector of random effects and ɛ is a vector of errors. This model uses maximum likelihood estimation to model the effects at level 1 and level 2 on individual responses of collective efficacy.

Following Raudenbush and Bryk (2002), the multilevel models of violent victimisation were performed in three steps. First, Model 1 estimates the multilevel effect of individual level predictors on violent victimisation. These included the age of the respondent, whether or not the respondent was female, married, number of years at current address, whether or not the respondent was a homeowner, and two annual household income indicator variables: one for respondents with low yearly household earnings (<$39K in Brisbane, and <$25K in Seattle), the other for respondents with high yearly household earnings (>$80K in Brisbane and >$75K in Seattle). Second, Model 2 includes the neighbourhood level predictors, such as percentage of households with non-English-speaking (NES) backgrounds; percentage Aboriginal or Torres Strait Islander (ATSI) (Brisbane only), or percentage African-American (Seattle only); percentage foreign-born; percentage of residents who have been in their home for five years or more; percentage of low-income households; and the four-year average violent crime rate.¹⁰ Finally, Model 3 includes the neighbourhood collective efficacy scale.¹¹

To test the relationship between neighbourhood and individual level indicators and violent victimisation, (yes/no to experiencing violent victimisation), generalised linear mixed multilevel models were fitted with a conditional binomial and logistic link function with the equations

y_{ij} ~ Bin (p_{ij})

(4)

and

\log (\frac{p_{ij}}{1 - p_{ij}}) = X β + b_{i}

(5)

where

y_{ij}

has a binomial distribution of the response of individuals (i = 1, … . n) within SLA/tract (j = 1, … 82123), and p is the probability of individuals i experiencing outcome =1 within SLA j. X denotes the individual and SLA level predictor variables, β is the fixed effect of said predictor variables and b_i are the random SLA/tract effects. These models use maximum likelihood estimation to model the level 1 and level 2 effects on the binary outcome violent victimisation.

Results

The composition of variance, ICC and neighbourhood level reliability of the collective efficacy scale were assessed in order to ensure that the scale had acceptable levels of reliability in both Seattle and Brisbane neighbourhoods. Table 3 shows the within group variance and between group variance of the collective efficacy scale as well as the ICC and neighbourhood level reliability measures of collective efficacy. The results show that in Seattle, 17%, and in Brisbane, 9% of the variation in the collective efficacy scale lies between tracts and SLAs in each research site. In addition, the collective efficacy scale is reliable at both the tract level in Seattle (α = 0.77), and the SLA level in Brisbane (α = 0.67).

Table 3.

Variance decomposition and neighbourhood reliability for collective efficacy scale in Seattle and Brisbane.

Variance components	Collective efficacy scale
Variance components	Seattle	Brisbane
Between group variance	3.37	1.50
Within group variance	16.62	15.66
Interclass correlation coefficient	0.17***	0.09***
Reliability	0.77	0.67
N	5629/123	2857/82

***

p < 0.001.

Characteristics of individual collective efficacy

First, Table 4 shows the multilevel models predicting individual perceptions of collective efficacy in Seattle and Brisbane. In Seattle, a number of individual level effects were significant in the model; older respondents (β = 0.027, p < 0.001), female respondents in comparison to male respondents (β = 0.045, p < 0.001), those who are married (β = 0.093, p < 0.01), those who earn more than $75K per year in comparison to those who earn less (β = 0.045, p < 0.05), and those who identify as Latino in comparison to other ethnic identities (β = 0.078, p < 0.05), report higher perceptions of collective efficacy. African-American and Asian ethnicity were not related to collective efficacy. At the neighbourhood level, controlling for individual level factors, the percentage of low-income homes (β = −0.004, p < 0.01), the percentage of African-American residents (β = −0.004, p < 0.01), and percentage of foreign-born residents (β = −0.009, p < 0.01), are related to lower levels of perceived collective efficacy. Housing tenure of five or more years (β = 0.005, p < 0.01) is positively related to perceived collective efficacy. Per cent youth and per cent non-English speaking residents were not associated with collective efficacy in Seattle.

Table 4.

Generalised linear mixed models for perceived collective efficacy in Seattle and Brisbane.

Level	Seattle			Brisbane
Level	Coef.	SE	Conf. in. 95%	Coef.	SE	Conf. in. 95%
Constant	2.662***	0.077	2.511–2.812	3.327***	0.192	2.951–3.703
Individual
Age	0.027***	0.003	0.021–0.034	0.009*	0.004	0.000–0.017
Gender	0.045**	0.014	0.017–0.073	0.086***	0.023	0.043–0.129
Homeowner	0.093***	0.02	0.054–0.132	0.133***	0.03	0.075–0.191
Married	0.099***	0.016	0.068–0.131	0.038	0.024	−0.009–0.084
Years at current address	−0.001	0.001	−0.002–0.001	−0.009	0.008	−0.024–0.007
Aboriginal				0.023	0.103	−0.178–0.225
Black/African-American	−0.039	0.032	−0.102–0.025
Asian	−0.051	0.029	−0.107–0.006
Latino	0.078*	0.035	0.010–0.146
Annual household income
$24K or less	−0.039	0.023	−0.083–0.006
$75K or more	0.045*	0.017	0.011–0.079
$39K or less				−0.059*	0.028	−0.114– −0.004
$80K or more				0.032	0.028	−0.003–0.086
Neighbourhood
% Low-income households	−0.004**	0.001	−0.007 – −0.001	−0.003	0.002	−0.007–0.001
% Aboriginal				−0.069***	0.017	−0.103– −0.035
% Black/African-American	−0.004**	0.001	−0.007 – −0.002
% Foreign-born	−0.009**	0.003	−0.015 – −0.003	0.0001	0.005	−0.009–0.01
% Non-English speaking residents	0.003	0.003	−0.003 – 0.008	−0.002	0.005	−0.01–0.007
% in home 5 or more years	0.005**	0.002	0.001 – 0.008	0.004*	0.002	−0.008–0.000
% Youth	0.002	0.004	−0.005 – 0.01	−0.015	0.012	−0.039–0.008
Variance of the random intercept	0.224			0.220
Log likelihood	−3013.106			−1293.996
N	4437/123			1929/82

***

p < 0.001, **p < 0.01, *p < 0.05.

OR: odds ratio; SE: standard error; Coef.: coefficient.

In Brisbane, similar patterns of perceived collective efficacy emerge with some differences. As age increases, so too do perceptions of collective efficacy (β = 0.009, p < 0.05). Females in comparison to males (β = 0.086, p < 0.001) and home owners in comparison to non-homeowners (β = 0.133, p < 0.001) are both more likely to report higher levels of perceived collective efficacy. Those earning $39K or less per year (β = −0.059, p < 0.05) are more likely than those who earn more to report lower levels of collective efficacy. At the neighbourhood level, the percentage of Aboriginal residents (β = −0.069, p < 0.001) significantly decreases reports of collective efficacy while the per cent of residents with housing tenure of five years or more (β = 0.004, p < 0.05) was positively related to perceived collective efficacy.

These results generally support Sampson et al.’s (1997) findings on the correlates of collective efficacy. However, Sampson et al. found no individual level effects of gender and ethnicity on collective efficacy as is found in both Seattle and Brisbane in this instance. At the neighbourhood level, Sampson found that indicators of disadvantage, immigrant concentration and residential stability were important correlates of collective efficacy. Though the models here do not look at the effect of these factors as a whole, there is some indication that similar patters emerge, particularly in Seattle. In Brisbane, only the concentration of Aboriginal residents and housing tenure are significant correlates of collective efficacy, whereas in Seattle, the per cent of low-income homes, African-American, and foreign-born concentration is also important.

Below, multilevel regression models are presented for each city. Model 1 shows the effect of individual level predicators on violent victimisation, controlling for neighbourhood clustering. These models also control for individual collective efficacy scores. Model 2 also includes neighbourhood level structural variables of immigrant concentration (% NES, % African-American, % ATSI, % foreign-born), residential stability (been in home 5+years), and disadvantage (% low-income households). The four-year average neighbourhood violent crime rate is also controlled for in Models 2 and 3. Finally, Model 3 includes neighbourhood levels of collective efficacy.

Violent victimisation in Seattle

In Table 5, Model 1, controlling for neighbourhood clustering, several individual level predictors are significant in explaining violent victimisation among survey respondents in Seattle. Older respondents (p < 0.001), those who are married (p < 0.05), those who have higher collective efficacy scores (p < 0.001) are more likely to report lower levels of violent victimisation. Income levels, ethnicity, gender, home ownership and years in current residence were not associated with violence victimisation.

Table 5.

Generalised linear mixed effect logistic models for violent victimisation in Seattle.

Level	Model 1			Model 2			Model 3
Level	OR	SE	Conf. in. 95%	OR	SE	Conf. in. 95%	OR	SE	Conf. in. 95%
Constant	1.434	0.392	0.839–2.451	0.388*	0.178	0.158–0.954	4.573	5.838	0.374–55.824
Individual
Age	0.928***	0.020	0.889–0.968	0.927***	0.020	0.888–0.967	0.927***	0.019	0.888–0.967
Married	0.881*	0.086	0.657–1.000	0.804*	0.086	0.652–0.993	0.810*	0.086	0.656–0.999
Female	0.868	0.082	0.721–1.045	0.881	0.083	0.732–1.061	0.884	0.083	0.734–1.064
Home owner	0.910	0.109	0.718–1.153	0.873	0.109	0.683–1.117	0.873	0.109	0.683–1.117
Latino	1.088	0.224	0.725–1.631	1.070	0.221	0.713–1.605	1.056	0.217	0.705–1.582
Years at current address	0.997	0.005	0.986–1.008	0.997	0.005	0.986–1.009	0.998	0.005	0.986–1.009
Annual household income
$24K or less	1.145	0.155	0.877–1.494	1.086	0.149	0.829–1.423	1.092	0.149	0.834–1.428
$75K or more	1.079	0.124	0.862–1.352	1.119	0.129	0.892–1.404	1.134	0.131	0.904–1.422
Individual collective efficacy score	0.560***	0.051	0.467–0.671	0.590***	0.056	0.489–0.712	0.612***	0.059	0.505–0.740
Neighbourhood
% Black/African-American				1.000	0.004	0.991–1.009	0.997	0.004	0.998–1.007
% Low income				5.662**	3.834	1.501–21.351	4.321*	2.91	1.154–16.177
% Foreign-born				0.995	0.005	0.985–1.006	0.991	0.005	0.979–1.002
% in home 5 or more years				1.016**	0.006	1.003–1.029	1.023***	0.007	1.009–1.038
Average 4-year violent crime rate				1.000	0.001	0.997–1.004	0.999	0.001	0.996–1.003
Collective efficacy scale							0.407*	0.178	0.172–0.960
Variance of the random intercept	0.32			0.009			2.24 e-12
Log likelihood	−1582.46			−1573.91			−1571.86
AIC	3186.92			3179.83			3177.72
BIC	3257.309			3282.2			3286.48
N	4439/123			4437/123			4437/123

***

p < 0.001, **p < 0.01, *p < 0.05.

OR: odds ratio; SE: standard error.

Prior to looking at the effect of neighbourhood collective efficacy, Model 2 assesses the impact of neighbourhood disorganisation on violent victimisation. Important individual factors such as the age of the respondent (OR = 0.92, p < 0.001), married respondents (OR = 0.88, p < 0.05) and individual collective efficacy scores (OR = 0.59, p < 0.001) all remain negatively associated with violent victimisation. At the neighbourhood level, the percentage of low-income households is five times more likely to be associated with reports of violent victimisation (OR = 5.66, p < 0.01). The percentage of the neighbourhood population with housing tenure of five years or more is significantly related to a 1.6% increase in violent victimisation (OR = 1.016, p < 0.01). Neighbourhood violent crime rate, the per cent of foreign-born residents and non-English speaking residents and per cent of African-American residents, is not significant in this model.

Model 3 shows the final and best fitting model (LL = −1734.316, AIC = 3177.72) of violent victimisation and includes neighbourhood collective efficacy. Individual level indicators show that older respondents experience a 9.2% decrease in the odds of victimisation, and married respondents experience a 19% decrease in the odds of victimisation. Individual collective efficacy scores remain significant and are negatively associated with violent victimisation (OR = 0.61, p < 0.001). Gender, annual household income, ethnicity (Latino) and housing tenure are not significant at the individual level. At the neighbourhood level, neighbourhood collective efficacy is negatively and significantly associated with violent victimisation (OR = 0.407, p < 0.05) even while controlling for individual levels of collective efficacy. This suggests that, like Chicago, collective efficacy is important in explaining reports of violent victimisation in Seattle. The per cent of residents who have been in their homes for five years or more remains significant and positively related to violent victimisation (OR = 1.023, p < 0.001). The relationship between low-income homes and violent victimisation is still significant when the model includes collective efficacy (OR = 4.32, p < 0.05); however, the size of the coefficient has decreased suggesting that neighbourhood levels of collective efficacy partially mediate the relationship between the per cent of low-income households on violent victimisation. This finding lends support to Sampson’s original finding that collective efficacy is the mechanism linking neighbourhood structure and violent victimisation.

Violent victimisation in Brisbane

Table 6 shows the multilevel logistic regression results of violent victimisation for the 82 SLAs in Brisbane. Controlling for SLA-level clustering, several individual level predictors are significant in explaining violent victimisation. Older respondents (OR = 0.90, p < 0.01) and home owners (OR = 0.62, p < 0.001) are less likely to report victimisation. Respondents with longer home tenure (OR = 1.29, p < 0.001) were significantly more likely to report experiences with violence. Individual collective efficacy score was also associated with lower reported victimisation (OR = 0.41, p < 0.001).

Table 6.

Generalised linear mixed effect logistic models for violent victimisation in Brisbane.

Level	Model 1			Model 2			Model 3
Level	OR	SE	Conf. in. 95%	OR	SE	Conf. in. 95%	OR	SE	Conf. in. 95%
Constant	0.998	0.548	0.339–2.931	0.463	0.451	0.068–3.120	576.812**	1387.776	5.165–64413.22
Individual
Age	0.905**	0.030	0.847–0.966	0'0.907**	0.030	0.848–0.969	0.914**	0.031	0.855–0.977
Married	1.032	0.192	0.716–1.487	1.052	0.195	0.731–1.515	1.025	0.190	0.711–1.477
Female	1.031	0.175	0.739–1.439	1.037	0.176	0.742–1.447	1.044	0.178	0.746–1.460
Home owner	0.623*	0.136	0.405–0.959	0.672	0.148	0.436–1.036	0.654	0.144	0.425–0.1008
Years at current address	1.291***	0.082	0.139–1.462	1.278***	0.081	1.128–1.447	1.284***	0.081	1.124–1.454
Annual household income
$39K or less	0.889	0.187	0.587–1.345	0.830	0.175	0.548–1.255	0.791	0.168	0.521–1.202
$80K or more	0.805	0.179	0.520–1.244	0.880	0.196	0.568–1.364	0.933	0.209	0.600–1.450
Individual collective efficacy score	0.401***	0.066	0.289–0.556	0.433***	0.072	0.312–0.601	0.497***	0.085	0.354–0.696
Neighbourhood
% Non-English speaking residents				0.987	0.026	0.937–1.039	0.986	0.026	0.936–1.039
% Foreign-born				0.979	0.025	0.930–1.030	0.976	0.025	0.927–1.027
% Low-income homes				1.029**	0.010	1.008–1.051	1.014	0.011	0.992–1.037
% in home 5 or more years				1.003	0.012	0.979–1.027	1.005	0.012	0.981–1.030
Average 4-year violent crime rate				1.000	0.000	0.999–1.000	0.999	0.000	0.999–1.000
Collective efficacy scale							0.093***	0.068	0.022–0.394
Variance of the random intercept	0.841			1.38 e-12			4.15 e-17
Log likelihood	−531.041			−520.64			−515.22
AIC	1084.08			1073.29			1064.44
BIC	1145.29			1162.32			1159
N	1928/82			1928/82			1928/82

***

p < 0.001, **p < 0.01, *p < 0.05.

OR: odds ratio; SE: standard error.

Model 2 includes additional SLA level variables that explore the impact of structural disadvantage and disorganisation on violent victimisation. Several individual level predictors remain significant in explaining violent victimisation. Older respondents (OR = 0.90, p < 0.01) were negatively associated with violent victimisation, while respondents who had been in their home for five or more years (OR = 1.27, p < 0.001) were more likely to report violent victimisation. Individual collective efficacy scores remain significant associated with lower levels of violent victimisation (OR = 0.43, p < 0.001). After accounting for neighbourhood structure, respondent marital status is no longer significant in predicting violent victimisation. At the neighbourhood level, the per cent of low-income homes (OR = 1.022, p < 0.05) is significant and positively related to violent victimisation. Neighbourhood violent crime rate, per cent foreign-born residents, per cent non-English speaking residents, and percentage of residents with tenure of five or more years in their home were not significant in explaining violent victimisation.

Model 3 shows the final and best fitting model (LL = −659.622, AIC = 1064.44). This model includes collective efficacy at the SLA level. At the individual level, results show that older respondents have an 8.6% decrease in the odds of experiencing violent victimisation. Respondents who have been in their homes for five or more years are 28% more likely to experience violent victimisation. Individual collective efficacy scores remain significant in their negative association with violent victimisation. Neighbourhood collective efficacy is significantly associated with a decrease in reports of violent victimisation (OR = 0.093, p < 0.001). Furthermore, when neighbourhood collective efficacy is added to this model, the percentage of low-income houses in the SLA ceases to be significant. This suggests that for Brisbane SLAs, levels of collective efficacy mediate the relationship between low-income areas and violent victimisation.¹² No other SLA level variables are significant in the final model.¹³

Discussion

Upon comparing the relationships between neighbourhood structure, collective efficacy and violent victimisation in Seattle and Brisbane neighbourhoods, important similarities and differences are evident. First, similar individual and neighbourhood characteristics predict individual collective efficacy scores in Seattle and Brisbane. Older residents, females and homeowners are all positively associated with higher levels of collective efficacy. In Seattle, being Latino and being married also predict higher levels of collective efficacy. Income predictors show similar patterns but emphasise the potential differences in context of each city; in Seattle, affluent residents had higher levels of collective efficacy but no distinguishable differences were seen between low-income residents in comparison to the rest of the sample. In Brisbane, low-income residents show lower collective efficacy scores but no differences were seen in affluent residents and the rest of the sample. Neighbourhood level predictors show similar effects across the two sites for ethnicity and housing tenure. In Seattle, the per cent foreign-born significantly predicts lower levels of collective efficacy but not in Brisbane.

The differences between Seattle and Brisbane in the effect of foreign-born residents likely reflect the differences in the foreign-born population. In Seattle’s trend in immigration is unique in that newer immigrants such as Hispanic speaking immigrants typically come with skills and social networks that allow them to bring resources to their new neighbourhood that are often in the suburbs and away from the most disadvantages neighbourhoods, leaving African-American and other minority groups in disadvantaged states (Wilson & Traub, 2006). This pattern of migration may explain why results show a positive association between Latino residents but a negative one for neighbourhood with higher percentages of foreign-born and African-American residents. In understanding the role of neighbourhood collective efficacy on violent victimisation in Seattle and Brisbane, it is clear that neighbourhood levels of collective efficacy significantly reduce the odds of reporting violent victimisation in both cities. This is strong evidence that despite contextual and structural differences between Seattle and Brisbane, collective efficacy is a stable predictor of violent victimisation despite overall crime levels being significantly lower in Brisbane than in Seattle. Interestingly, the role of neighbourhood structure (poverty, immigrant concentration, residential stability) in explaining violent victimisation in Seattle and Brisbane is remarkably similar. In Seattle, there are clear and significant effects of poverty, but not immigrant/ethnic concentration on victimisation before controlling for neighbourhood collective efficacy. Surprisingly, residential stability (housing tenure of five or more years) is positively related to violent victimisation.

Structural factors among Brisbane SLAs, particularly low-income households, are positively related to violent victimisation but completely mediated by collective efficacy. Overall, though the collective efficacy thesis holds across both cities, its relationship to neighbourhood structure is not as multidimensional in Brisbane and Seattle as is found in cities such as Chicago. Thus, widespread social disorganisation beyond levels of poverty does not explain violent victimisation at the neighbourhood level, and yet neighbourhood collective efficacy is still a strong predictor of violent crime.

What we learn from this comparison is that collective efficacy may still reduce crime even where structural differences between neighbourhoods are not apparent or not as extreme. Or, more specifically, emerging collective efficacy may not be strongly linked to social disorganisation with the exception of some levels of disadvantage or low-income areas. There are two potential explanations for this difference and these explanations are likely different in each site: the overall lower levels of crime and disadvantage across neighbourhoods and lack of variation between neighbourhoods on the neighbourhood level covariates, and the contextual difference between Brisbane’s and Seattle’s urban history compared to that of other cities.

First, as seen in Table 2, Brisbane’s neighbourhood level predictors indicative of social disorganisation theory (low-income households, residential stability, percentage foreign-born) all exhibit less variation in comparison to the same predictors in Seattle. For example, average percentage of low-income households across Seattle neighbourhoods is approximately 26% with a standard deviation of 14.2, whereas Brisbane neighbourhoods have on average 31% of low-income homes but a standard deviation of 11.27. In comparison to Seattle, a lack of variability in the data may explain why neighbourhood indicators of disorganisation do not prominently feature in understanding victimisation.

It may also be possible that the SLA, or suburb boundary is an inappropriate unit of measurement for a ‘neighbourhood’ in the case of Brisbane. Vinson and Baldry (1999) found that the typical suburb boundary in Australian urban areas is not a good indicator of between neighbourhood patterns of disadvantage or crime. Their research showed that within SLAs there are clusters of disadvantage, with some areas of the SLA being more affluent than others. Work by Hipp (2007) has suggested that using block groups, smaller than a census tract, may be more appropriate for picking up structural patterns in the data. In addition, Coulton, Korbin, Chan, and Su (2001) found that residents conceptualise their ‘neighbourhood’ with varying boundaries that often do not map onto census tracts. Despite this, census tracts remain a widely accepted unit of analysis in neighbourhood studies in the US. However, in Brisbane, work by Mazerolle et al. (2010) has shown that suburbs/SLAs are analogous to how residents conceptualise their neighbourhood; therefore, the SLA provides a reliable unit of analysis for measuring a neighbourhood. With the exception of low-income households, there may be more structural variation within suburbs than across them.

The lack of variability is not necessarily a statistical feature of the sample of Brisbane neighbourhoods. Brisbane’s historical and contextual differences are likely to explain why social structure plays a less dominant role in explaining crime in a social disorganisation framework. For example, Brisbane, and Australia in general, is established in social welfare provisions that, via extensive benefits, dampen the overall levels of poverty and disadvantage (Norton, 1994). This may also dampen the effect that disadvantage and poverty has on crime and victimisation. For example, American urban centres, including Seattle, have areas with high concentrations of disadvantage, such as neighbourhoods that have high levels of public housing. In contrast, Australia’s public housing has historically been less segregated and more spread out through Australia’s urban areas (Norton, 1994). Public housing residents were therefore more spatially integrated with middle class residents. This also means that the areas of highest residential instability are not as concentrated in particular areas in the city.

In Seattle, the percentage of foreign-born residents remains positive and significant in relation to perceptions of collective efficacy. In contrast, the percentage of foreign-born residents in Brisbane neighbourhoods fails to reach significance in multilevel models of collective efficacy. There is a stark contrast in the type, duration and ethnic composition of foreign-born residents in Seattle and Brisbane. A large portion of the foreign-born population in Brisbane, and in Australia in general, is from English-speaking, Commonwealth nations such as Britain, New Zealand and South Africa. These foreign-born residents not only share a common language with Australian-born residents but also share a common religion and cultural norms, and physically look more like native white Australians. Due to these commonalities, these residents may find it easier to blend into mainstream Australian culture (Hazelhurst, 1987), form network ties outside their own ethnic group (Davis, Erez, & Avitabile, 2001), and may be protected from negative perceptions of ‘immigrants’ that plague other ethnic groups migrating to Australia (Collins, 2005). Seattle’s foreign-born population, while smaller than that of other US cities, typically consists of a more ‘visible minority’: Hispanic, Asian and West African immigrants not only differ in cultural norms and religions but also bring language diversity to the city; barriers than may hinder the creation of social capital and collective efficacy. Overall, the above results show support for the collective efficacy thesis in two different international cities. In Seattle census tracts as well as in Brisbane suburbs, neighbourhoods with higher levels of collective efficacy are also those with lower incidences of violent victimisation. These findings contribute to the collective efficacy literature, which has largely been based on American cities that typically have more extreme segregation and poverty and different types of diversity than cities in other Western nations, particularly Australia.

Previously, it has been difficult to ascertain from the collective efficacy literature whether collective efficacy was truly a mediating factor between structural disorganisation and crime, or whether it was simply a manifestation of disadvantage itself. Sociological literature suggesting that collective efficacy emerges independent of structural conditions has been largely ethnographic in nature. For example, Patillo-McCoy (1998, 1999) found in her work on a middle class African-American neighbourhood in the Chicago area that low levels of disorganisation are not necessarily associated with low levels of crime. This neighbourhood had high residential stability and many families had lived there for many years because residential property is handed down through generations. However, Patillo-McCoy (1998, 1999) found that stable, or organised, neighbourhoods with dense network ties do not prevent the occurrence of criminal behaviour. Within the neighbourhood, law-abiding and criminal residents want both a safe and quiet neighbourhood; however, they disagree on the strategies to achieve this goal, with criminal residents opting for a criminal strategy. Because of dense network ties and the relative stability of the neighbourhood, it is difficult to keep known criminals from the neighbourhood at bay. Ethnographic studies such as this suggest that the notion of collective efficacy emerged as a separate component from the social structure. The finding above provides quantitative support that, even without widespread disorganisation, collective efficacy can work at the neighbourhood level to decrease crime.

Much like the findings from Morenoff, Sampson, and Raudenbush (2001), the finding here also shows that there are some structural elements that remain directly related to victimisation. Morenoff et al. (2001) found that the structure of the neighbourhood may be more independently related to crime and victimisation than first thought given that the effect of disadvantage and income effects prior homicide rates and is unmediated by collective efficacy and social ties. This study shows that in the case of Seattle, high levels of disadvantage in terms of annual income persist in their effect on reports of victimisation even after collective efficacy is included. In Brisbane, the findings show complete mediation. Further studies should continue to understand under what conditions collective efficacy and social ties operate in conjunction and independently of the social structure.

Though collective efficacy plays a similar role in the reduction of reported victimisation and violent crime rates in Seattle and Brisbane, there are some interesting differences that raise additional questions about the collective efficacy model. Contrary to the expectations of modern social disorganisation theory, the percentage of foreign-born residents in Seattle is significant but negatively related to collective efficacy but not significant in explaining violent victimisation. This is indicative of the complexity of immigration and its relationship to violent crime. In Seattle, a large proportion of immigrants are from Asian countries such as China and Japan. These immigrants are typically characterised by high levels of education and gainful employment, and low levels of victimisation and criminal offending (Jayasuriya & Sang, 1991). As such, though they are very much a part of Seattle’s foreign-born population, they are not likely to experience violent crime in the same way as other immigrants with different characteristics (Daniels, 1990).These patterns would make future comparison of cities within the US fruitful for understanding the direct link between neighbourhood structure, collective efficacy and crime.

In Brisbane, the percentage of foreign-born residents is not significant in any model run in the analyses above. However, it is unlikely, given the broad variation in Australian immigrant groups and recent trends in immigration in Australia that all groups would be equal in their experience of violent crime. Asian immigrants with a long-standing immigrant history in Australia are typically more like Asian immigrants in Seattle in that they are often middle class, gainfully employed and well educated. In contrast, newly arrived Sudanese immigrants have had more difficulty integrating into Australia’s social structure and economy. In Seattle, it is likely that variation in experiences of violent victimisation exists among different ethnic groups as well as among those who are recently emigrated. Future research should address whether different immigrant groups form different perceptions of social control in response to their environment. This may be particularly important to Australian cities such as Brisbane where there may be greater variation among ethnic groups within neighbourhoods rather than between neighbourhoods.

With comparative study there are often methodological concessions that need to be made in order to preserve the components and concepts that can be compared. In this case, a concession was made not to include indicators for spatial dependency because the data in the Brisbane site are not altogether contiguous and therefore cannot be used in spatial analyses that are comparable to Seattle. However, from this limitation comes an opportunity to enhance the argument for the utility of comparative work, as well as highlight the need for continued data collection that allows some of the more recent advances in theory, particularly around spatial dependency, to be collected. Here, I argue that what is gained from the comparative nature of a simplified model without accounting for spatial dependency helps to identify and understand the limits and relevance of social disorganisation and collective efficacy theory. In addition, the cross-sectional nature of these data sets prevents exploration of the temporal order of the relationships between collective efficacy and victimisation.

These limitations give rise to interesting opportunities for compelling comparative research. Given the difficulties with neighbourhood boundaries (Hipp, 2007), Hart and Waller (2013) as well as Coulton et al. (2004) argue that perceptual measures of the neighbourhood structure could replace official statistics and be used to estimate official boundaries of neighbourhoods that provide better estimates of defined neighbourhoods for analysis.

The overall findings of this study show support for the role that neighbourhood collective efficacy has in predicting violent victimisation and therefore underscores the relevance of the collective efficacy perspective in understanding crime in urban areas outside the context of the US. However, what is apparent from these results is that collective efficacy remains an important predictor of victimisation absent all the characteristics of disorganisation. Thus, it is not disorganisation itself that solely influences collective efficacy in all contexts.

Footnotes

Acknowledgements

The author would like to thank Robert Crutchfield, Alexes Harris, Jerry Herting, Steve Herbert, Rebecca Wickes, Ross Matsueda, Emma Antrobus, Harley Williamson, Elise Seargent, Gentry White, Angela Higginson, Jason Ferris and Adele Somerville for helpful reviews and suggestions on earlier versions of this paper. The author would also like to thank the members of the University of Washington, Department of Sociology Deviance Seminar for their comments and suggestions.

Funding

This research was supported by The University of Queensland Institute for Social Science Research and the Australian Research Council Centre of Excellence in Policing and Security.

Notes

References

Angel

R. J.

Angel

J. L.

Hill

T. D.

(2008) The Comparison of the health of older Hispanics in the United States and Mexico: Methodological challenges. Journal of Aging and Health 20(1): 3–31.

Australian Bureau of Statistics. (2006). 2006 census data: View by location. Retrieved from http://www.censusdata.abs.gov.au/ABSNavigation/prenav/ProductSelect?newproducttype=QuickStats&btnSelectProduct=Select+Location+%3E&collection=Census&period=2006&areacode=&geography=&method=&productlabel=&producttype=&topic=&navmapdisplayed=true&javascript=true&breadcrumb=P&topholder=0&leftholder=0&currentaction=201&action=104&textversion=false.

Bandura

(1997) Self-efficacy: The exercise of control, New York, NY: W.H. Freeman & Company.

Bierne

(1997) Cultural relativism and comparative criminology. In: Bierne

Nelken

(eds) Issues in comparative criminology (pp. 3–24), England, UK: Dartmouth Publishing Company Limited.

Blau

(1982) The cost of inequality: Metropolitan structure and violent crime. American Sociological Review 47: 114–129.

Bursik

(1988) Social disorganization theories of crime and delinquency: Problems and prospects. Criminology 26: 519–551.

Bursik

R. J.

Grasmick

H. G.

(1993) Neighbourhoods and crime: The dimensions of effective community control, New York, NY: Lexington.

Collins, J. (2005). Ethnic minorities and crime in Australia: Moral panic or meaningful policy responses. Sydney, Australia: Office of Multicultural Interest.

Coulton, C., Cook, T., & Irwin, M. (2004). Aggregation issues in neighborhood research: Comparisons of several levels of census geography and resident defined neighbourhoods. Retrieved from http://digitalcase.case.edu:9000/fedora/get/ksl:2006052511/Cook-Agression-2004.pdf.

10.

Coulton

C. J.

Korbin

Chan

(2001) Mapping residents’ perceptions of neighborhood boundaries: A methodological note. American Journal of Community Psychology 29(2): 371–383.

11.

Crutchfield

R. D.

Matsueda

R. L.

Drakulich

(2006) Race, labour markets and neighborhood violence. In: Peterson

R. D.

Krivo

L. J.

Hagan

(eds) The many colors of crime (pp. 335–356), New York, NY: New York University Press.

12.

Crutchfield

R. D.

Ramirez

S. R.

(2012) Using census data and surveys to study labor markets and crime. In: Gadd

Karstedt

Messner

S. F.

(eds) The SAGE handbook of criminological research methods (pp. 143–158), London, UK: Sage.

13.

Daniels

(1990) Coming to America: A history of immigration and ethnicity in American life, New York, NY: Harper Collins.

14.

Davidson

Dingle

O’Hanlon

(1995) The cream brick frontier: Histories of Australian suburbia, Clayton, VIC: Monash University Press.

15.

Davis

R. C.

Erez

Avitabile

(2001) Access to justice for immigrants who are victimized: The perspectives of police and prosecutors. Criminal Justice Policy Review 12(3): 183–196.

16.

Drakulich

K. M.

(2010) Estimating risk and expressing fear: Reactions to crime in Seattle neighborhoods. Dissertation Abstracts International, A: The Humanities and Social Sciences 70(8): AAI0821923.

17.

Earls

F. J.

Visher

C. A.

(1997) Project on human development in Chicago neighborhoods: A research update (PDF) [NIJ Research in Brief], Washington, DC: U.S. Department of Justice, Office of Justice Programs, National Institute of JusticeNCJ 163603 (9).

18.

Gasparini

(2010) Community and territorial belonging. Comparative Sociology 9(4): 433–462.

19.

Guest

A. M.

Kubrin

C. E.

Cover

J. K.

(2008) Heterogeneity and harmony: Neighboring relationships among whites in ethnically diverse neighborhoods in Seattle. Urban Studies 45(3): 501–526.

20.

Hart

T. C.

Waller

(2013) Neighbourhood boundaries and structural determinants of social disorganisation: Examining the validity of commonly used measures. Western Criminology Review 14(3): 16–33.

21.

Hazelhurst

K. M.

(1987) Migration, ethnicity and crime in Australian society, Canberra, Australia: Australian Institute of Criminology.

22.

Hillery, G. A. (1955). Definitions of community: Areas of agreement. Rural Sociology, 20, 111–123.

23.

Hipp

J. R.

(2007) Block, tract, and levels of aggregation: Neighbourhood structure and crime and disorder as a case in point. American Sociological Review 72(5): 659–680.

24.

Jayasuriya

Sang

(1991) Asian immigration to Australia: Past and current trends. Population Review 35(1–2): 35–56.

25.

Kornhauser

R. R.

(1978) Social sources of delinquency: An appraisal of analytic models, Chicago: Chicago University Press.

26.

Kubrin

Weitzer

(2003) New directions in social disorganization theory. Journal of Research in Crime and Delinquency 40: 374–402.

27.

Leavitt

G. C.

(1990) Relativism and cross-cultural criminology: A critical analysis. Journal of Research in Crime and Delinquency 27(1): 5–29.

28.

Lee

F. L.

(2006) Collective efficacy, support for democratization, and political participation in Hong Kong. International Journal of Public Opinion Research 18(3): 297–317.

29.

McClintock

F. H.

Wikstrom

P. O.

(1992) The comparative study of urban violence: Criminal violence in Edinburgh and Stockholm. British Journal of Criminology 32: 505–520.

30.

McCrea

Shyy

Western

Stimson

(2005) Fear of crime in Brisbane: Individual, social and neighbourhood factors in perspective. Journal of Sociology 41(1): 7–27.

31.

McFarlane

(2010) The comparative city: Knowledge, learning, urbanism. International Journal of Urban and Regional Research 34(4): 725–742.

32.

Matsueda, R. L., Crutchfield, R., Guest, A., & Kubrin, C. (2010). Seattle neighborhoods and crime survey, 2002–2003. Retrieved from https://dx-doi-org.web.bisu.edu.cn/10.3886/ICPSR28701.v1.

33.

Mazerolle

Homel

Stimson

Sampson

Stewart

Spence

(2005) Community variations in crime: A spatial and ecometric analysis (Australian Research Council Linkage Grant: LP 0453762), Australia: Griffith University.

34.

Mazerolle

Wickes

McBroom

(2010) Community variations in violence: The role of social ties and collective efficacy in comparative context. Journal of Research in Crime and Delinquency 47(1): 3–30.

35.

Miethe

Meier

R. F.

(1994) Crime and it’s social context: Toward and integrated theory of offenders, victims, and situations, Albany, NY: State University of New York Press.

36.

Morenoff

J. D.

Sampson

R. J.

Raudenbush

S. W.

(2001) Neighborhood inequality, collective efficacy, and the spatial dynamics of urban violence. Criminology 39(3): 517–558.

37.

Nelken

(1997) Whom can you trust? The future of comparative criminology. In: Bierne

Nelken

(eds) Issues in comparative criminology (pp. 3–25), England, UK: Dartmouth Publishing Company Limited.

38.

Nguyen

(2005) Immigration and crime: Do Asian immigrants bring more crimes to Australia? Policy and Governance 16: 1–25.

39.

Norton

(1994) Where to for welfare? Sustaining the welfare state remains a major problem. Review of “Welfare and Inequality: National and International Perspectives on the Welfare State” by Peter Saunders. Policy 10: 39–42.

40.

Park

R. E.

Burgess

E. W.

(1921) An introduction to sociology, Chicago: University of Chicago Press.

41.

Passel

J. S.

Suro

(2005) Rise, peak and decline: Trends in US immigration 1992–2004, Pew Hispanic Center. Available at: http://pewhispanic.org/files/reports/53.pdf.

42.

Patillo-McCoy

(1998) Sweet mothers and gangbangers: Managing crime in a black middle-class Neighborhood. Social Forces 76: 747–774.

43.

Patillo-McCoy

(1999) Black picket fences: Privilege and peril among the black middle class, Chicago, IL: The University of Chicago Press.

44.

Pauels

Svensson

(2010) Informal controls and the explanation of propensity to offend: A test in two urban samples. European Journal on Criminal Policy and Research 16(1): 15–27.

45.

Peterson

R. D.

Krivo

L. J.

(2010) Divergent social worlds: Neighborhood crime and the racial-spatial divide, New York, NY: Russell Sage Foundation.

46.

Putnam

(1993) The prosperous community: Social capital and community life. The American Prospect 13: 35–42.

47.

Raudenbush

S. W.

Bryk

A. S.

(2002) Hierarchical linear models: Applications and data analysis method, Thousand Oaks, CA: Sage Publications.

48.

Sampson

Morenoff

Earls

(1999) Spatial dynamics of collective efficacy for children. American Sociological Review 64: 633–660.

49.

Sampson

Raudenbush

Earls

(1997) Neighborhoods and violent crime: A multi-level study of collective efficacy. Science 277: 918–924.

50.

Sampson

Wikstrom

P. O.

(2007) The social order of violence in Chicago and Stockholm neighborhoods: A comparative inquiry. In: Shapiro

Kalyvas

Masoud

(eds) Order, conflict and violence (pp. 97–119), New York, NY: Cambridge University Press.

51.

Shaw

McKay

(1942) Juvenile delinquency in urban areas, Chicago: University of Chicago Press.

52.

StataCorp (2013) Stata statistical software: Release 13, College Station, TX: StataCorp LP.

53.

Stimson

R. J.

Taylor

S. P.

(1999) City profile: Brisbane. Cities 16(4): 285–298.

54.

U.S. Census of Population and Housing (2000) Summary population and housing characteristics: Washington, Washington, DC: Government Printing Office.

55.

Vinson

Baldry

(1999) The spatial clustering of child maltreatment: Are micro-social environments involved? Trends and Issues in Crime and Criminal Justice 119: 1–6.

56.

Vinson

Homel

(1975) Crime and disadvantage: The coincidence of medical and social problems in an Australian city. British Journal of Criminology 15(1): 21–31.

57.

Westfelt

Estrada

(2005) International crime trends: Sources of comparative crime data and post-war trends in Western Europe. In: Sheptycki

Wardak

(eds) Transnational and comparative criminology (pp. 19–48), England, UK: GlassHouse Press.

58.

Wickes

R. L.

(2010) Generating action and responding to local issues: Collective efficacy in context. The Australian and New Zealand Journal of Criminology 43(3): 423–443.

59.

Wikstrom

P. O.

Treiber

K. H.

(2009) Violence as situational action. International Journal of Conflict and Violence 3(1): 75–96.

60.

Wilson

W. J.

Traub

R. P.

(2006) There goes the neighbourhood: Racial, ethnic, and class tensions in four Chicago neighbourhoods and their meaning for America, New York, NY: Random House Inc.

61.

Winter

Bryson

(1998) Economic restructuring and state intervention in holdenist suburbia: Understanding urban poverty in Australia. International Journal of Urban and Regional Research 22(1): 60–75.

62.

Wooldredge

(2002) Examining the (ir)relevance of aggregation bias for multilevel studies of neighbourhoods and crime with an example comparing census tracts to official neighbourhoods in Cincinnati. Criminology 40(3): 681–710.

63.

Young

(2005) The use of national crime statistics in comparative research: Ireland and Scotland compared. In: Sheptycki

Wardak

(eds) Transnational and comparative criminology (pp. 49–68), England, UK: GlassHouse Press.