Defining rural areas

Introduction

Rurality is now increasingly recognised as a phenomenon worthy of its own policies and ‘an essential component of policy making in general’ (Noble and Wright, 2000:294). Performance indicators are informing the strategies being drawn up by Regional Development Agencies in England, whilst government consultation on the Rural White Paper (DETR/MAFF, 2000) actively sought ideas for appropriate indicators for rural areas. Meanwhile, in a European context, the development of broader rural development policies within the Common Agricultural Policy similarly requires information on the incidence of disadvantage within rural areas beyond the agricultural sector (Hodge et al., 2000).

However, the definition of rurality is far from straightforward (see for example Halfacree, 1993) and the delineation of rural areas is highly dependent not only on the definition employed but also on the size of the spatial unit selected. This has important consequences for the ways in which levels of service need, provision and utilisation appear to vary between rural and urban areas. Martin et al. (2000) have examined how the representation of deprivation and low health status changes markedly using alternative measures of rurality. For instance, the profound problems faced by many peripheral communities are not always reflected when rural areas are defined on the basis of population density. Similarly, the aggregation of small areas of rural deprivation into larger areas and the incorporation within them of low-density affluent rural settlements serve to obscure many aspects of rural deprivation.

Unfortunately, there is no definitive way of describing what constitutes a rural area, or one definitive or preferred measurement of rurality. Most commentators choose definitions and measurement systems that are best suited to their own application, taking into account issues of data availability, quality, consistency and ease of collection. This has led some to question whether evidence of rural-urban differentials in health service need and utilisation has been subject to methodological artefact (Higgs, 1999).

Whilst there remains little agreement on what is meant by the term 'rural', it is possible to identify six broad approaches to the measurement of rurality (Chapman et al., 1998; Higgs, 1999). These, which are discussed in this section, include:

The latter tend to comprise a combination of primary characteristics of rurality (e.g. sparsity and peripherality) and secondary characteristics such as a limited range of employment, a lack of services, low levels of provision of public transport and a high proportion of pensioners. The merits and limitations of multivariate classifications are discussed below. At this stage, however, it may be useful to emphasise the fact that secondary characteristics ‘are often (historically, but by no means logically) consequences of being a small scattered population, and not defining features per se’ (Noble and Wright, 2000:297, original emphasis).

Settlement size

Settlement size has been commonly used to define rural areas. However, Shucksmith (1990) notes the lack of international consensus on settlement size thresholds. In England, the Countryside Commission defines rural parishes as those comprising populations of less than 1,000, while Phillimore and Reading (1992) used wards with a population below 5,000 in their study of rural health inequalities. By contrast in Sweden, settlements are only defined as rural if they comprise less than 500 people. In the United States, towns with populations over 2,500 are classified as urban and all other settlements are defined as rural, while in Greece, rural settlements are those with less than 10,000 residents.

In addition to variation in thresholds between countries, the use of settlement size is complicated by the difficulty of defining settlements in the first place. The parish is one of the most stable spatial units available in England. However, key data are not available for all parishes throughout the country. Area units such as electoral wards are used nationally. However, these vary erratically in size (Coombes and Openshaw, 1991) and their interaction with underlying settlement patterns is inconsistent. For example, an urban ward may be but a small part of a major settlement. In a rural area, a single ward may contain one or more different villages, whilst in a peri-urban location, a ward may contain a mix of different villages and part of a major urban area.

Population density

Population density is probably the most widely used measure of rurality (Martin et al., 2000) which, in addition to distinguishing urban from rural areas, can be used to discriminate between rural areas. Population density has the advantage of being intuitive and easy to calculate. As a continuous variable, the results can be ranked, making comparison with other areas straightforward. However, simple population density takes no account of the distribution of population within areas. For example, an urban industrial ward that contains a large open space would, by virtue of its low population density, emerge as 'rural' using this measure. Equally, population density does not adequately capture the type of rurality where very sparsely populated areas are interspersed by small towns. Craig (1987) examined how Cornwall and Somerset compare using population density as a measure of rurality. At the Enumeration District (ED) level, the counties have a similar urban/rural distribution of 65% to 35% respectively. However, in Cornwall, urban areas are smaller and more numerous, so Somerset has more purely rural wards.

As with other measures of rurality, decisions also have to be made about appropriate threshold values that define urban and rural population densities. Martin et al. (2000) outline how over time, researchers have used differing levels of population density to define rural areas. In the Standard Spending Assessment (SSA), measures of sparsity and super-sparsity, based on the average number of people per hectare in each ward at the 1991 census, are used with respect to education and other services (e.g. refuse disposal, leisure services, libraries, planning and benefit administration). ¹ The process by which these thresholds were established is not clear.

Like standard population density measures, sparsity and super-sparsity also fail to take account of population clustering, a factor that will bear strongly upon the economies of scale that can be achieved in providing services to meet population needs. Indeed, work conducted for the Rural Development Commission in 1996 suggested that the SSA adjustment bears little relation to the higher costs actually faced. ‘The sparsity factor…is worth only £16 per head to the receiving authorities – and little or no recognition is given to measures of rural deprivation such as the effects of social isolation on the need for social services for the elderly.’ (R Hale and Associates and IPF 1996;vii.) Suffolk County Council, for example, calculated that the costs of sparsity were under funded by £10 million for education and that a case could be made for an additional £500,000 in their SSA to reflect the effect of sparsity on domiciliary social services for older people (Suffolk County Council, 2000).

There are various forms of weighting that can be applied to population density to take account of the way in which populations are clustered around specific geographical points. These include measures such as the weighted population density (WPD), geometric mean density (GMD) and population potential. These measures are relatively simple to calculate. However, they are not as easy to interpret as the conventional measure of population density. They are also limited in their ability to differentiate between types of rural area as they do not adjust for factors such as proximity to built up areas.

Accessibility

An interesting alternative measure of rurality is that of nearest neighbour distance. The mean distance between all residential locations, for example, can be calculated using a geographical information system (GIS). This provides a continuous variable that, unlike standard population density, takes some account of the degree of clustering within the spatial unit of interest. This is important as two administrative areas will have very different values if the population of one is widely dispersed and the other is concentrated in a single settlement. Nearest neighbour distance is also complementary to those measures of accessibility that seek to identify distance to higher-level service centres, as it describes geographical isolation at the local level.

As discussed in section 5 below, measures of accessibility can be used to distinguish between rural and urban areas. Accessibility concerns the physical availability of facilities and services and attempts to capture the ease with which a defined population can access services. Until recently, this tended to be reduced to crow-fly distance. However, travel times can now be readily computed using Geographical Information Systems (GIS) and are increasingly being incorporated into attempts to model access and utilisation of services in rural areas. As access standards are gradually being established in very diverse areas of service delivery, there are growing demands for standards that explicitly acknowledge the difficulties that rural providers can face in delivering services to dispersed populations. As a proxy for rurality, however, accessibility suffers from similar inconsistencies in terms of interaction with settlement type as does population density. There are, for example, small towns that are further away from centralised facilities than farming communities on the edge of cities.

Peripherality

Peripherality is an important dimension of accessibility and has been used to highlight service access problems associated with rurality in a number of academic studies (e.g. Bentham and Haynes, 1986). Philips and Williams (1984) describe peripherality as areas that experience disadvantage in the core/periphery relationship. While terrain, climate and remoteness are the physical aspects, social aspects such as low incomes and ageing populations predominate. In major reviews of health resource allocation systems in Scotland and Wales peripherality has been identified by the term 'remote rural' to differentiate it from purely rural. The reviews have focused on the difficulties faced by remote rural areas where distance to the facilities and markets of large urban centres gives a level of disadvantage to the resident population. Other aspects of peripherality include geographical barriers such as an extensive coastline, mountains or poor roads. Thompson (1996), for example, notes that the physical landscape of Cornwall (which has a long indented coastline) results in the need to duplicate fire service facilities in a short linear distance. The location of settlements on the coast effectively reduces the potential service catchment area because much of this is sea. The long, narrow nature of the peninsula exacerbates problems of peripherality as there is reduced scope for achieving economies of scale by sharing service provision with neighbouring providers. England contains a number of peripheral areas which makes this concept of relevance to rural health in the country. In addition to Cornwall, these include East Anglia, Cumbria and North Yorkshire.

Land use

It can be seen that no one form of measurement of rurality will capture all the aspects needed to differentiate different types of rural areas. One way to avoid the problems of partial definition and threshold setting which are common to the preceding measures of rurality is to use multivariate area classification schemes, and to attempt to identify those area types which together define rural areas.

The Office for National Statistics (ONS) has defined every ED in the country as urban or rural on an urban land use basis (OPCS, 1992). What is not urban is by default rural. As EDs are the building blocks of wards, this allows wards to be classified as urban or rural, on the basis of their constituent EDs. Rurality can also be described simply in terms of economic activity, such as a baseline percentage of those involved in agriculture or the extractive industries, although with the demise of the numbers employed in both these industries it becomes a less reliable marker.

Multivariate data classifications

A range of multivariate data classification approaches is also available. Many of the standard geodemographic classifiers (Birkin, 1995) include rural area types, although these do not constitute a standard definition. The first major use of principal components to describe rural areas was undertaken by Cloke in 1977 and then revised using 1981 census data (Cloke and Edwards 1986). They analysed a variety of factors related to rurality, including population density and occupation, and constructed an index to group non-urban local authority districts into rurality quartiles from extreme rural to extreme non-rural, excluding districts defined as metropolitan or urban. The variables contained in this analysis included distance from the nearest urban node with a population of over 50,000, population density and persons over 65. ONS have also used principal components and cluster analysis on 1991 census data to produce a national ward classification comprising 14 area groups, two of which are 'rural areas' and 'rural fringe' and a total of 40 clusters (Wallace et al., 1995).

The theoretical advantages of these classifications are that they provide a multi-dimensional view of social circumstances allowing areas with a similar set of socio-economic and environmental characteristics to be grouped. A corresponding disadvantage is that there is no rank order to the classification. Techniques such as principal components analysis may help in understanding covariance structures and in the identification of the major factors characterising rurality. However, they do not result in clearly labelled definitions of rurality.

Conclusion

The debate about which measures most effectively capture rurality is not merely academic. There has been growing concern amongst health authorities serving rural populations that resource allocation mechanisms fail to incorporate rural dimensions of need which take into account supply side factors and the fact that there are additional costs involved in providing rural health services. The difficulties faced by such authorities will vary, however, according to the type of rural area they serve.

The point can be illustrated by returning to the example of Cornwall and Somerset which, according to some population density measures, have comparable levels of rurality. Somerset has long borders with Devon, Dorset, Wiltshire and Avon. It is therefore able to share services with neighbouring counties in a way that increases accessibility for its resident population and achieves economies of scale for its service providers. Cornwall, by contrast, is surrounded by sea on three sides and bordered by a river with few crossing places on most of its fourth side. As a result, the difficulties that this county faces in providing similar levels of service accessibility are considerably more pronounced.

The way in which we conceptualise rurality will therefore have important implications for the way in which service needs and problems of access are interpreted. There is no smooth continuum from 'urban' to 'rural' as implied by continuous measures such as population density. Such measures can conceal important differences between urban and rural areas, and between and within rural areas. Measures of rurality also alter with the scale at which such analyses take place. Key aspects of rurality and, as discussed in the next section, rural deprivation, cannot always be gauged from standard indicators where the unit of analysis is too large to capture the heterogeneity that characterises rural areas.