Elsevier

Health & Place

Volume 13, Issue 4, December 2007, Pages 812-825
Health & Place

Modifiable neighbourhood units, zone design and residents’ perceptions

https://doi.org/10.1016/j.healthplace.2007.01.002Get rights and content

Abstract

Neighbourhood effects on health are partly determined by the way the neighbourhoods are defined (the modifiable areal unit problem), but few studies of place effects have incorporated alternative sets of areal units. This study compared computer-generated zones with areal units identified subjectively by local government officers as communities in the city of Bristol, UK. Automated zone design came close to replicating the subjective communities when the balance of objectives and boundary constraints was adjusted. The set of subjective community areas was compared with automated zone designs, which maximized the homogeneity of a social factor (deprivation) and an environmental factor (housing type), at three different geographical scales, with average populations of 2500, 3700 and 7500. All sets of areas were then matched against the neighbourhood perceptions and social behaviour reported by residents, measured as part of the Avon Longitudinal Study of Parents and Children (ALSPAC). Neighbourhood perceptions and social behaviour varied mostly between individuals, but there were significant small differences between all sets of areas. The neighbourhood perceptions of residents were found to match the areas identified by automated zone design as well as they matched the subjectively defined communities, suggesting that the neighbourhoods identified by experts were not more real to residents than synthetic areas. Differences in perceptions could be explained by variations in social and housing conditions at the very local scale of enumeration districts, with populations of about 500. The neighbourhoods with meaning for residents therefore appeared to be much smaller areas than those typically investigated in geographical studies of health.

Introduction

Many research studies over recent years have concluded that health inequalities between populations are partly the result of neighbourhood effects (Pickett and Pearl, 2001; Diez-Roux, 2001; Macintyre et al., 2002). Although everybody seems to accept that they live in a neighbourhood, the immediate environment that has the potential to influence residents’ health is difficult to define precisely. There are several competing definitions of neighbourhood, none of which has achieved universal acceptability, but most concentrate either on small geographical areas with similar attributes or areas whose residents interact with each other (Galster, 2001). Boundaries between neighbourhoods might coincide with administrative divisions, changes in physical environment, differences in residents’ characteristics or peoples’ perceptions (Diez-Roux, 2001). Some researchers have avoided the term neighbourhood, preferring “community” or “place effects”, but the problem of boundary delineation remains. Any study area can be divided into alternative plausible sets of small geographical units.

Most studies of local place effects on health have used administrative or census areas as the geographical units for convenience, because population data are available for them. Whether or not these areas are appropriate depends on the research question (Diez-Roux, 2001; Pickett and Pearl, 2001), but only a small number of studies so far have defined custom-made neighbourhoods to suit a particular investigation. Examples include the use of socially homogeneous areas (Reading et al., 1999; Law et al., 2005) and areas based on the local knowledge of key professionals (Ross et al., 2004).

The size of neighbourhoods designed to detect local place effects on health has varied enormously. Studies using administrative areas, such as census tracts in the US and wards in the UK, have worked with geographical units with populations mostly in the range 4000–5500 (Pickett and Pearl, 2001). Larger units with populations between 8000 and 40,000 have also been popular (Ellaway et al., 2001; Martikainen et al., 2003; Ross et al., 2004; Law et al., 2005; Subramanian et al., 2003; Shenassa et al., 2004). Few studies have investigated smaller areas (Coulton et al., 2001 is an exception). Although reviewers have called for comparisons of alternative neighbourhood schemes to achieve a better understanding of the underlying processes (Diez-Roux, 2001; Martikainen et al., 2003), most of the published studies of the effects of neighbourhoods on health have been based on a single set of area units. Some attempts to compare place effects on health at different geographical scales have been made (Haynes et al., 2003; Ross et al., 2004; Pampalon, 2005), but only one systematic comparison of alternative sets of areal units incorporating both boundary and scale changes has been reported, to our knowledge (Cockings and Martin, 2005).

Many authors (e.g. Openshaw, 1984) have demonstrated the modifiable areal unit problem (MAUP), whereby different definitions of areas—either in terms of average population size (scale) or choice of boundary (zoning)—will lead to different results for analyses based on those areas, such as area-level correlations. Therefore, careful consideration must be given to the definition and choice of areal unit for the analysis. Modifiable areal units are a problem only if they are arbitrary. If there is a hypothesis about the mechanism of the link between neighbourhood and health, then the set of areal units should be defined accordingly. When mechanisms are unclear, as they often are at an early stage of research, then it is important to test the sensitivity of relationships to the definition of the underlying areas.

A range of alternative areal units can be created using automated zone design procedures (Cockings and Martin, 2005) which group a set of basic areal units into a smaller number of zones which are in some sense optimal. The criteria used in the grouping process might include combinations of the number of zones required, constraints on the population size of each zone, the compactness of zone shape and a requirement to maximize the homogeneity of specified variables within each zone. Cockings and Martin used the technique to define zones with approximately equal populations at different scales and were able to demonstrate that these synthetic zones produced stronger relationships between morbidity and deprivation than census units, and that larger areas produced stronger relationships. They made no attempt in this exploratory study to design zones that were internally homogeneous in terms of environmental or social characteristics. Others have suggested that areas based on homogeneous characteristics produce stronger relationships than heterogeneous areas (Carstairs, 1981; Morgenstern, 1982; Haynes et al., 1999), so this might be a promising line of inquiry. Such procedures were used in the 2001 England and Wales census to define homogeneous census output areas (Martin et al., 2001).

Another issue is whether zones identified by automated programmes have any meaning for residents. After an accumulation of substantial evidence that neighbourhoods affect the health of residents in a variety of ways, we still need to understand how, and why some people are particularly affected, in some settings more than in others (Macintyre et al., 2002). Much attention has been given to the theory that a breakdown of social cohesion might be responsible for the link between general levels of health and income inequality within communities (Wilkinson, 1996; Kawachi and Kennedy, 1997). Associations have been found between levels of health and residents’ perceptions of both environmental problems, such as noise, fumes, road traffic, lack of safe places for children to play, few local facilities, and so on, and social problems, such as vandalism, assaults, burglaries, drug use, disturbances from teenagers and trouble between neighbours (Ross, 2000; Ross et al., 2000; Stafford and Marmot, 2003; Parkes and Kearns, 2006). Measures of neighbourhood social cohesion made up of the perceived attractiveness of neighbourhoods, the psychological sense of community and behavioural measures of frequency of contacts with neighbours have also been demonstrated to explain significant proportions of health variations (Ellaway et al., 2001; Simons et al., 2004; Pampalon, 2005).

Several studies have investigated the extent to which residents’ senses of environmental quality, safety, social cohesion, trust and so on, are shared within neighbourhood boundaries. Individual perception data of this type has been entered into multilevel models, nesting individual residents into the objectively defined neighbourhoods (usually census or administrative divisions) in which they lived, and then decomposing the variations in perception scores between those occurring at area level and those at individual level using the intra-class correlation (ICC) measure. Typically the results have shown significant variation in residents’ neighbourhood perceptions between census or administrative divisions, although perceptions varied much more at the individual level (Kingston et al., 1999; Ellaway et al., 2001; Subramanian et al., 2003; Raudenbush, 2003; Pampalon, 2005). Investigations which asked residents about their social behaviour (contacts with neighbours, participation in local activities, etc.) have found only very small or non-existent variation at area level after adjusting for individual factors (Kingston et al., 1999; Lindstrom et al., 2002).

Only one study, to our knowledge, has used this technique to try to identify the set of area units that maximizes the variance in residents’ perceptions at area level. Using five composite neighbourhood perception measures from residents of 10 US cities, Coulton et al. (2004) compared ICCs across alternative geographical units: entire central cities, sub-areas, census tracts and census block groups. They found that the ICCs were highest at the smallest unit of geography.

This study aims to combine these various recent approaches, by identifying alternative sets of neighbourhood units using zone design and comparing their characteristics, including their correspondence with the perceptions of residents. It will address the following questions:

  • Can neighbourhoods subjectively recognized by local professionals be replicated from census data using automated zone design?

  • What similarities and differences are there between a set of neighbourhoods subjectively defined using local knowledge and automated zone sets that maximize the homogeneity of environmental and social variables within zones?

  • To what extent do residents’ perceptions of neighbourhood characteristics and patterns of social contact coincide with areas recognized by local professionals and automated zones maximizing the homogeneity of environmental and social variables?

  • At what geographical scale are residents’ perceptions and behaviour most similar?

Section snippets

Methods: subjective communities

The study area was the city of Bristol in South West England. Bristol had a population of 375,000 at the 1991 census. The smallest census units at the time in Bristol were 814 enumeration districts (EDs) with average populations of about 500. EDs were grouped for administrative purposes into 34 wards, which were regarded as being too large for some planning purposes. In 1994, local government officers in Bristol City Council Planning Department divided Bristol into 101 small areas which, after

Automated zone design

Synthetic areas that optimized the homogeneity of social characteristics and environmental features were identified by grouping EDs with an automated zone design programme “A2Z” (Daras, 2006) derived from the AZP algorithm developed by Openshaw and Rao (1995). In common with other zone design programmes, A2Z starts with an initial configuration of zones and improves it by swapping basic areal units at the borders of zones while optimizing an objective function. It was designed to reduce some of

Measures to compare zone sets

The internal homogeneity of a particular variable in each set of zones was measured by the ICC, also known as the intra-unit correlation (Goldstein, 2003). MLwiN software was used to fit a simple null model with no predictors that separated the total variance into variance between EDs within zones and variance between zones. The ICC was calculated by dividing the between-zone variance component by the total variance, hence giving a measure of the proportion of the between-zone variation of the

Neighbourhood perceptions and social behaviour: data

The final step in this study was to assess which of the nine alternative divisions of the city best fitted the pattern of neighbourhood perceptions and social behaviour reported by residents. Perceptions and self-reported contact information were extracted from surveys conducted by the Avon Longitudinal Study of Parents and Children (ALSPAC) in Bristol and the surrounding area (Golding et al., 2001). In 1991 the women in the ALSPAC sample were questioned about their circumstances during their

Results: replicating the subjective scheme

Characteristics of the 101 communities in Bristol recognized by the planners and the various systems of 101 zones generated automatically are compared in Table 1. All zone sets in Table 1 combined 814 EDs (level 1 units) into 101 zones (level 2 units). The proportion of the total variation in ED values that occurred between zones is the ICC measure in the first two columns. On average there were only eight EDs per zone, so the ICCs tended to be high. The subjectively defined communities

Comparison of alternative zone sets

Automated zone set 1 in Table 1 was chosen to contrast with the planners’ communities: more homogenous in terms of levels of poverty, but less plausible as a set of neighbourhoods because the boundaries were much more complex. Zone set 6, maximizing the homogeneity of housing types but also with a weak shape constraint, was chosen as a third example. These are illustrated in Fig. 2, Fig. 3, both with deprivation scores in the smaller EDs in the background. Visual comparison with Fig. 1 confirms

Validation of zone designs against residents’ perceptions

ICCs for neighbourhood perceptions and social behaviour scores of residents mapped into the various zone sets (and on to the smallest areal units available: the 814 EDs) are shown in Table 3. Here the level one units were the 6510 mothers and the level two units were either 814 EDs or sets of 150, 101 or 50 zones. The low coefficients indicate that scores varied much more substantially between individual respondents than between areas. The variations at area level were comparatively small, but

Discussion

Most researchers accept that the characteristics of neighbourhoods affect health, but have tested neighbourhoods based on a single set of administrative and census areas rather than on areas defined according to a postulated mechanism. We have used both local planners’ predefined communities and areas produced synthetically by an automated zone design to compare alternative neighbourhood sets. Automated zone design is useful in such investigations because this technique can create

Acknowledgements

We are extremely grateful to all the families who took part, the midwives for help in recruiting them and the whole ALSPAC team, which includes interviewers, computer and laboratory technicians, clerical workers, research scientists, volunteers, managers, receptionists and nurses. The UK Medical Research Council, the Wellcome Trust and the University of Bristol provide core support for ALSPAC. This publication is the work of the authors, and Robin Haynes will serve as guarantor for the contents

References (42)

  • S. Alvanides et al.

    Zone design for planning and policy analysis

  • Poverty in Bristol

    (1994)
  • Poverty in Bristol 1996—An Update

    (1996)
  • Bristol City Council, 2003. Personal correspondence from Jayne Mills, Strategic and Citywide Policy, Department of...
  • V. Carstairs

    Small area analysis and health research

    Community Medicine

    (1981)
  • Coulton, C., Cook, T., Irwin, M., 2004. Aggregation issues in neighborhood research: a comparison of several levels of...
  • C.J. Coulton et al.

    Mapping residents’ perceptions of neighbourhood boundaries: a methodological note

    American Journal of Community Psychology

    (2001)
  • Daras, K., 2006. An information statistics approach to zone design in the geography of health outcomes and provision....
  • A.V. Diez-Roux

    Investigating neighborhood and area effects on health

    American Journal of Public Health

    (2001)
  • A. Ellaway et al.

    Perceptions of place and health in socially contrasted neighbourhoods

    Urban Studies

    (2001)
  • G. Galster

    On the nature of neighbourhood

    Urban Studies

    (2001)
  • Cited by (109)

    View all citing articles on Scopus
    View full text