Spatial dependence and heterogeneity in patterns of urban deprivation

Developments in the provision and quality of digital data are creating possibilities for finer resolution spatial and temporal measurement of the properties of socio-economic systems. We suggest that the "lifestyles" datasets collected by private sector organisations provide one such prospect for better inferring the structure, composition and heterogeneity of urban areas. Clearly, deprivation and hardship are inextricably linked to incomes from earnings and transfer payments. In many countries (e.g. the UK) no small area income measures are collected at all, and this forces reliance upon commercial sources. Yet, the use of such data in academic research is not without considerable problems. In the same spirit as Gordon and Pantazis (1995) we thus think it necessary to retain some linkage to population census data ? but in a way which is much more sensitive to spatial context. A critical issue is thus to understand the scales at which both income, and the variables that are used to predict it, vary (see also Rees, 1998; Harris and Longley, 2002). We address some of these issues in the context of the debate about the intra-urban geography of hardship and social exclusion. Low income fundamentally restricts the abilities of people to participate actively in society (Harris and Longley, 2002), yet reliable, up-to-date income measures at fine spatial scales are rarely available from conventional sources. As a consequence, many indicators of deprivation are reliant upon data sources that are out of date and/or entail use of crude surrogate measures. Some measures bear little clear correspondence with hardship at all. Other widely-used indicators are spatially variable in their operation. The broader issue concerns the scale and extent of ?pockets? of hardship and the scale ranges at which difference is deemed manifests. The problems are further compounded if each of the range of surrogate measures used to specify a concept operates at different scales. Taken together, it remains unclear whether meaningful indicators of social conditions can ever be adequately specified, or whether generalised representations can be sufficiently sensitive to place. Using a case study of Bristol, UK, we compare the patterns of spatial dependence and spatial heterogeneity observed for a small area ("lifestyles") income measure with those of the census indicators that are commonly used as surrogates for it. This leads to specification of spatial dependence using a spatially autoregressive model, and accommodation of local heterogeneity using geographically weighted regression (GWR). This analysis begins to extend our understanding of the determinants of hardship and poverty in urban areas: urban policy has hitherto used aggregate, outdated or proxy measures of income in a less critical manner; and techniques for measuring spatial dependence and heterogeneity have usually been applied at the regional, rather than intra urban, scales. The consequence is a limited understanding of the geography and dynamics of income variations within urban areas. The advantages and limitations of the data used here are explored in the light of the results of our statistical analysis, and we discuss our results as part of a research agenda for exploring dependence and heterogeneity in spatial distributions.