This thesis presents a new neighbourhood classification, the Leeds Classification for Community Safety (LCCS). This is used to demonstrate the usefulness of area classifications for providing area context information to crime analysis, and for identifying neighbourhoods with atypical crime profiles - given their neighbourhood type. The work can be seen as a development of the classifications produced by the Home Office for comparative performance purposes, but at a smaller, neighbourhood scale. There has been a recent trend among practitioners to use commercial geodemographic products for this task, but these tools are primarily designed for consumer segmentation applications and little is revealed about the way in which these classifications are constructed, or their ability to discriminate geographies of crime and disorder. The research presented in this thesis discusses critically both these issues. The research draws upon academic and policy literature on the geography of crime, environmental criminology and community safety policy, and describes the types of task undertaken by community safety analysts. Existing knowledge about the causes and motivations for crime are used to select variables from new national and local sources. The final partition was created using the fuzzy c-means clustering technique, but alternative techniques were also employed and levels of agreement between the different results were measured. The design process also involved measuring the ability of different partitions to discriminate neighbourhood crime rates. Numeric comparisons were made between the LCCS and existing general purpose classifications, and these show that the task-specific approach was better overall at discriminating crime rates. Practical applications of the LCCS are also demonstrated using recorded crime data for criminal damage and domestic burglary. Furthermore, variations in response to burglary target hardening are analysed using the LCCS, and the cost benefit to neighbourhoods of different types is shown. These practical demonstrations of the LCCS go to reinforce the assertion that area classification can be useful, practical tool to aid in the analysis and understanding of spatial patterns of crime and disorder.

Most studies of 'cohesion' between urban residents focus separately on either social network structure or ideations with very little attention given to the intersection between spatial, structural and ideational dimensions of networks. This is problematic on two levels: firstly because attitudes and practices are formed in the context of personal social networks; and secondly because social interactions between residents are physically embodied and therefore spatially constrained. This thesis explores empirically the relationship between spatial cohesion (the extent to which a network is geographically localised or dispersed), structural cohesion (the extent to which a network is tight-knit or fragmented) and ideational cohesion (the extent to which residents have similar attitudes and practices). The social networks, attitudes and practices of white-British residents living in deprived urban localities of North Manchester are studied (survey, n=409; interview, n=53). Variances in forms of cohesion were found to have consequences for residents and localities. At the individual level, the spatial and structural shape of a resident's network was linked to their attitudes and behaviours. Attitudes and practices were 'framed' in the context of personal network structure exhibited through a set of resident 'roles'. This matters for urban cohesion because a person's social network structure influenced whether they liked their neighbourhood, trusted other residents, felt a sense of community or had found jobs through contacts. Previous studies have argued that contemporary urban networks have become fluid, dynamic and spatially dispersed. Yet this research found that although some people had networks that were geographically spread, most resident networks were made of localised, tight-knit, stable, long-term relations. Moreover, people with these cohesive, localised networks framed their experiences of urban cohesion differently to those with geographically spread and/or disconnected social networks. Particularly because the attitudes and practices of residents with localised, cohesive networks were very often habitual and socially reproduced. Social networks focus people's activities in such a way that not only constrains or enables current attitudes and practices but can also affect an individual's ability to change their future behaviour. At the locality level, the type of 'deprived' locality seemed to influence network structure. The structural, spatial and ideational distribution of cohesion at locality level provided neighbourhoods with different portfolios of social capital. Qualitative differences were observed between homogeneous-deprived (very low income, white areas) and socially mixed-deprived (white deprived areas with some class/ethnic mix) localities. People living in deprived-homogenous localities concentrated their networks within the local area and had few ties to residents of bordering areas, a sign of social distance. Conversely, residents of socially mixed-deprived localities had more potential to bridge ties to other neighbourhoods because their networks were not overly focused within the local area. Given that attitudes and practices are framed in the context of social network structure, it was argued that residents of deprived-homogeneous and socially mixed-deprived areas may experience and interpret urban cohesion differently and this has implications for universal policies of cohesion in deprived neighbourhoods. The thesis illustrates the interplay between spatial, structural and ideational forms of cohesion and highlights consequences for individual action and the generation of neighbourhood social capital. The originality of analysis and data synthesis are used to advance a relational and contextualised theory of urban cohesion and contribute to wider academic and policy debates on urban social networks and neighbourhood deprivation.