Urban–Rural Differences in Commuting in England: A Challenge to the Rural Sustainability Agenda?

Abstract

While housing affordability and homelessness worsen in rural England, sustainability concerns are being used to reinforce the 60-year-old goal of urban containment. Yet there are claims that the evidence for rural communities being intrinsically unsustainable in terms of their carbon emissions remains weak. This paper aims to improve the evidence base through a focus on commuting behaviour. Following a review of recent studies of commuting that have included a rural dimension, it presents the results of new census-based analyses of differences between places in the distance that residents travel to work. Both indicate that a more nuanced approach is needed towards new development than a simple urban–rural dichotomy.

Introduction

As urban–rural migration continues apace despite signs of an ‘urban renaissance’ in England (Champion, 2006; Parkinson et al., 2006), increasing concern is being voiced about its implications for the sustainability agenda. One major policy challenge is the way in which local workers and their families are increasingly being priced out of the countryside, threatening the economic viability of rural enterprises and the goal of a ‘living working countryside’ (Taylor Review, 2008). Across the regions, relative to local wages, housing is now more affordable in the urban areas than in the rural ones (Commission for Rural Communities [CRC], 2005), prompting Green (2007) to highlight the urgent need to address the rural housing shortfall. Similarly, Gallent (2008) argues for a more systematic planning response to the gentrification of England's villages, based on a new conviction to rebalancing their social composition. Most recently, the Campaign to Protect Rural England (2008) call for action on rural housing is underpinned by evidence that in the 4 years to 2007 waiting lists for affordable housing grew by an average of 14,500 a month to reach almost 700,000, while rural England's share of the nation's homeless rose from 16% to 37%. In terms of environmental sustainability, a particularly worrying outcome is the increase in the numbers of people commuting from urban homes to rural workplaces, up by 417,000 between 1981 and 2001 (Frost, 2006).

As far as commuting and its implications for the nation's carbon footprint are concerned, however, much more attention is being given to the challenge posed by the in-migrants to rural England, with their greater tendency to commute longer distances, often back to jobs closer to their previous urban homes (see Champion et al. [2008] for a review of the literature). This is perhaps not surprising in the context of 60 years of town and country planning that has sought to minimize the urban land take and to protect rural areas. Indeed, the current concerns over climate change (see, for example, Stern Review, 2006) are, according to some (for example, Lowe & Ward, 2007; Shucksmith, 2007), being harnessed as the latest tool in a long-running and well-orchestrated campaign that views development in the countryside as bad and development in cities as good—a campaign that initially drew on the need for ensuring national food self-sufficiency, then, in an era of food mountains, switched its case to the protection of landscape and aesthetic quality and has now embraced the sustainability agenda for its main justification. Indeed, according to Gallent et al. (2003, p. 12), within Western Europe England provides the archetypal example of a ‘divisive culture’ that is associated with ‘an unstable regime in which power moves … to reflect what are sometimes seen as incompatible “rural” and “urban” agendas’ (see also Hoggart, 2003).

On the other hand, as the body charged with ensuring that government policies reflect the real needs of people living and working in rural England, the CRC is keen to dispel claims that rural communities are intrinsically unsustainable. In its report Planning for Sustainable Rural Communities: A New Agenda?, the CRC (2007a) identifies and challenges four myths. Two of these—that the high projected levels of new housing will concrete over the countryside and that we should protect the countryside for its own sake alone—the CRC reckons it has little trouble refuting. A third—that new housing should be located only in settlements with a minimum level of services—it sees as a more compelling argument, but one that the CRC still feels confident in challenging on the grounds that ‘the key settlement approach is based on a mythical hierarchy of villages and towns … which fails to recognize the complexity and diversity of functions and networks that have developed in recent decades in rural England’ (CRC, 2007a, p. 6). In relation to the myth that rural communities are intrinsically unsustainable because they generate higher levels of carbon emissions, however, the verdict is more cautious, merely pointing out that people in all types of areas contribute to global warming and saying that ‘we await robust evidence on whether there is any overall systematic difference between the contributions of rural and urban areas’ (CRC, 2007a, p. 6).

The aims of this paper are to review and extend the evidence base on the differences between rural and urban areas in the personal travel that leads to carbon emissions and, in so doing, to provide the basis for a more balanced discussion of the relative degree of sustainability of different types and locations of settlement. The focus is the commuting element of personal travel, partly because of its importance but mainly because of data availability. Following an introduction to population trends and the sustainability issues they raise, the paper summarizes the findings of previous studies of urban–rural differences in commuting. Data from the 2001 Population Census is then used to compare types of places on the distances that their residents travel to work. Given the finding that some aspects of rurality appear to impose a greater carbon footprint than others, the ensuing discussion recommends that a more nuanced approach is needed towards rural development than a simple urban–rural dichotomy.

Population Trends and the Sustainability Agenda

If it is the case that rural settlements are intrinsically unsustainable, then the challenge for the sustainability agenda is indeed a colossal one according to the evidence of the population statistics. Between 1997 and 2005, rural England saw annual average growth of 110,600 people due to within-UK migration—an increase that is equivalent to just over 0.6% a year (CRC, 2007b). Moreover, the urban–rural population shift seems set to continue. Official population projections indicate a rise in rural England's population by 2.57 million in the 21 years to 2025, with the most rural type of district being tipped to see the strongest growth in both absolute and relative terms (CRC, 2007b; see also Champion & Shepherd, 2006; Taylor Review, 2008).

These patterns of population change sit rather uneasily alongside spatial planning strategies that have ‘urban containment’ as their central rationale. The latter is, of course, no recent development, being rooted in the 1944 Scott Report and the 1947 Town and Country Planning Act, but at that time the main thrust of the development restrictions was to minimize agricultural land loss through higher building densities and to reduce the length of the physical boundary between residential land and farmers' fields through more compact layouts, subsequently reinforced by green belts (see, for instance, Hall et al., 1973; Champion, 2002). Under these arrangements building could still take place in more rural areas, but normally it had to be attached to existing settlements and preferably be in the form of infilling rather than extending the ‘urban fence’.

In recent years, however, the tenor of planning guidance seems to have shifted towards a greater presumption against development in the countryside. There would seem to be two main drivers behind this. One is urban decline, which was first recognized as a policy challenge in the mid-1970s but became the central goal of spatial planning with the setting up of the Urban Task Force and the publication of Towards an Urban Renaissance (Rogers Report, 1999). A key element in the subsequent Urban White Paper was increasing the proportion of house-building taking place on recycled urban land (Department of the Environment, Transport and the Regions, 2000), which is much more abundant in the cities than in rural areas. Such ‘brownfield sites’ have recently been accounting for nearly 75% of this development, considerably higher than the 60% target (Communities and Local Government [CLG], 2007a, p. 16).

The other main driver is the quest for sustainable development. While this also had its origins three decades ago in the debate following the Club of Rome's Limits to Growth and the Brundtland Report, the concept was explicitly adopted as a basis for England's settlement planning in the Sustainable Communities Plan (Office of the Deputy Prime Minister, 2003). More recently still, this goal has been reinforced by the government's response to the perceived threats posed by climate change. Following the recommendations of the Stern Review (2006), guidance on measures to restrict carbon emissions by reducing the need to travel, among other things, has been set out as a supplement to the planning policy statement on delivering sustainable development (CLG, 2007b).

The current emphasis in planning thought is amply evident in the official documents. In answering the question ‘What is a sustainable community?’, the government stipulates—among other things—that ‘they must offer … good public transport, schools, hospitals, (and) shops’, ‘feature a wide range of jobs and training opportunities … [and] economically viable and attractive town centres’, and provide ‘transport facilities … that help … reduce dependence on cars’ (CLG, 2008a, pp. 1–2). A central principle in Planning Policy Statement 1 on delivering sustainable development is that ‘Planning should actively manage patterns of urban growth to make the fullest use of public transport and focus development in existing centres and near to major public transport interchanges’ (CLG, 2005, p. 12). Specifically in relation to rural areas, the guidance is for most new development to be focused on local service centres, as ‘This should help to ensure these facilities are served by public transport and provide improved opportunities for access by walking and cycling’ (Office of the Deputy Prime Minister, 2004, p. 8).

In relation to the guidance prompted by the climate change agenda, two of the 7 ‘key planning objectives’ listed in the relevant supplement to Planning Policy Statement 1 are spatial strategies, which:

in providing for the homes, jobs, services and infrastructure needed by communities and in renewing and shaping the places where they live and work, secure the highest viable resource and energy efficiency and reduction of emissions … [and] … deliver patterns of urban growth and sustainable rural developments that help secure the fullest possible use of sustainable transport … and … reduce the need to travel, especially by car. (CLG, 2007b, p. 10)

In their decision-making, all planning authorities are required to apply the principle that ‘The proposed provision for new development, its spatial distribution, location and design should be planned to limit carbon dioxide emissions’ (CLG, 2007b, p. 10). The goals of reducing travel demand and getting a greater proportion of journeys to be made by walking, cycling or public transport are also enshrined in the current planning guidance on housing provision (Planning Policy Statement 3), industry, commercial development and small firms (Planning Policy Guidance (PPG) 4), town centres (Planning Policy Statement 6) and transport (PPG13).

On the other hand, the planning guidance also seems to recognize the need for some flexibility when dealing with the rural context. For instance, when discussing the expected rural impact of implementing measures designed to curb carbon emissions, the consultation document on planning and climate change stated that it was not the intention to reduce the total amount of development taking place in the countryside. The aim was merely to direct it to the most sustainable locations, with the result that ‘this may lead to less development being brought forward in some rural locations’ (CLG, 2006, p. 72; original emphasis; see also CLG, 2008b, p. 37). Identifying the latter is presumably left to local discretion within the framework of the full set of planning guidance documents, but as voiced by the CRC (see the introductory comments above) the key concern arising from this is the dearth of ‘robust evidence’ about which locations are the most and least sustainable in economic, social and environmental terms. In particular, the answer should not be determined purely on the assumption that people will commute to the nearest jobs, but rather on the basis of evidence about people's actual choices of where to live, where to work and how to travel between them. It is this expressed behaviour with which the remainder of this paper is concerned, looking first at the findings of recent studies and then presenting the results of new analyses.

Review of Recent Studies

Over the past decade there have been a number of commuting-related studies that have examined urban–rural differences in commuting or have taken a specifically rural perspective. The main difficulty in using them to build up a robust evidence base, however, is that their aims have differed and so too have their approaches. In particular, while some are primarily concerned with differences between population aggregates using census data for wards and districts, others focus more on the individual people and their relative disadvantage in accessing work. They also vary by whether the commuting distances for area aggregates are calculated by reference to where people live or where they work and, most relevant for present purposes, by the way in which the geographical context is classified in terms of rural and urban areas.

In their analysis of commuting trends in England and Wales, Banister and Gallent (1998) used area-based data mainly from the 1991 census, looking at the distribution of commuting trip lengths by county types and then mapping average length of journeys to work at the district level. In relation to the former, they found that, in terms of average commuting distances measured from workers' places of residence, there was relatively little difference between their three English county aggregates: highest for the shire counties at 9.3 km, next highest for Greater London at 8.4 km and lowest for the metropolitan counties at 7.1 km. The picture, however, is significantly different when analysed by workplace, with people working in Greater London having much longer average commutes (12.1 km) than either the shires (8.2 km) or the metropolitan counties (7.7 km). With the mapping of commuting distances for residents revealing the outer suburban ring around London as the country's major concentration of longer commuting trips, this raises concern about the sustainability of such a large employment centre in terms of commuting-related sustainability, even though the environmental impact would be tempered by the greater use of the train to access the jobs there. By contrast, ‘Areas with the higher number of short trips are mainly rural, or the self-contained large towns and cities outside major conurbations’ (Banister & Gallent, 1998, p. 336). This study, however, did not attempt any analysis based on formal urban–rural definitions, while their follow-up study (Banister & Gallent, 1999) did not get into any more spatial detail than the census region.

Again using 1991 Census data, Coombes and Raybould (2001) also examined the differences in the lengths of commuting trips made by people in different parts of England and Wales, and went on to analyse some of the factors behind them. They, too, found that the proportion of residents commuting short distances was lowest in the provincial metropolitan areas and in the rest of the country outside the South East. Then, in modelling ward-level values of the proportions commuting less than 5 km, they initially found a positive association with both population density and a measure of accessibility. However, when the list of factors was extended to include several GIS-derived variables relating to the strength of the labour market, the effect of an ‘urbanization index’ turned out to be negative, indicating more short-distance commuting in less urbanized areas, all other things being equal.

Similarly, Boyle et al. (2001) used 1991 census data for a nationwide analysis, although in this case with the specific aim of identifying any distinctiveness in commuting behaviour by degree of rurality. Taking for this purpose the then current six-fold non-rural/rural classification of wards, they found that rural wards did not record as high a level of out-commuting as urban ones. In fact, there was a quite regular decline in the out-commuting share of working residents from 66% for the wholly urban wards to 63% for the predominantly urban ones, 61% for mixed urban, 59% for both mixed rural and predominantly rural wards, and 54% for the wholly rural ones. They attributed this partly to the higher levels of self-employment and home-working in the countryside, but likely to be at least as important is the larger geographical extent of the more rural wards, such that the average rural resident has to travel further than an urban resident to cross a ward boundary and become an out-commuter. Certainly, when average distances to work were calculated, commuters living in urban wards were found to travel shorter distances than those residing in more rural wards. Yet, after mapping the ward averages across England, they were at pains to point out that the pattern was not simply one of long commuting distances in the remote rural fringes and shorter commutes in other areas, for—confirming the results of Banister and Gallent's (1998) broader-scale analysis—they observed that long distances were also common in more accessible rural areas, particularly in a wide belt around London.

In the same report, Boyle et al. (2001) also presented the results of modelling individual people's journeys to work using the 1991 Census 1% Household Sample of Anonymised Records (SAR). The primary advantage of this microdata-set was that the effect of place of residence could be separated out from that of other factors known to be associated with differences in commuting behaviour, such as age, gender, occupational level and car ownership. In this case, place was treated as a binary variable, with rural areas being defined as the combination of the ‘rural’ and ‘rural fringe’ categories of the Office for National Statistics (ONS) 14-way ward group classification and with all other categories (making up 90% of the total population) being ‘non-rural’. On this basis, it was found that the relationships found in the aggregate analyses still held firm in the individual-level analyses after allowing for these other factors. Rural people are less likely to commute than the non-rural and, considering only commuters, the rural ones commute further than the non-rural. Additionally, the rural commuters are more likely to travel by car than the non-rural. This study, however, did not display the odds ratios for any results, so it is not possible to know high much higher or lower the odds were for rural residents compared with non-rural residents. Moreover, the dichotomy between these two types of residential area is a gross oversimplification of England's settlement pattern.

Frost (2006) has updated and extended this picture, using 2001 Census Special Workplace Statistics data for output areas that average about one-twentieth of the population size of wards and breaking down commuting flows by different types of urban and rural settlements. Partly echoing the results from the 1991 census above, Frost's analysis found that out-commuting is higher from settlements in the more densely populated rural areas that are located around towns and cities than it is from settlements in more sparsely populated areas. As regards the former, around one-half of the employed residents of these ‘less sparse rural areas’ travelled to work in some component of urban England (defined as all built-up areas with a population of at least 10,000). Moreover, the proportion was only marginally higher for the larger rural settlements (‘rural towns’ at 52%) than it was for the smallest settlement category (‘hamlets and isolated dwellings’ at 46%). By contrast, the dominant sources of employment for residents of the more sparsely populated areas were found to be more localized, with the proportion of residents working in an urban place ranging between one-sixth and one quarter depending on settlement size. But this study undertook no analyses of commuting distances.

Two further studies based on the latest census have included a rural dimension as well as looking at commuting distances. Like Boyle et al. (2001), Green and Owen (2006) used micro-level modelling that allows for non-geographical factors associated with the length of people's journeys to work. Their analysis of the 2001 Individual SAR examined the propensity to commute short distances, defined as less than 5 km. They found that the odds of a short-distance commute was significantly lower for people living in areas classified as ‘accessible rural’ than for those living in ‘remoter rural’ areas. This geographical contrast was particularly strong for people with the highest level of qualifications, but it also applied to the group at the other end of the skills ladder with no or low qualifications. In the context of the full set of localities across the country, however, the rural–urban contrast in commuting distances was found to be very clear. For both skill groups analysed, the four most rural types of areas were among the bottom five of the nation's 13 local authority area types in terms of the odds of being home to short-distance commuters, the other being Outer London. In other words, longer-distance commuting is more common in rural areas than in almost every type of urban area. Among the latter, the two highest odds of short-distance commuting were found for residents of large and small non-metropolitan cities. A separate regression analysis based on data aggregated to wards revealed lower commuting distances in more sparsely populated areas and longer commutes in the urban fringe wards of more densely populated areas.

Champion et al. (2008) also used the Individual SAR, but focused on long-distance commuting, taken as 20 km or more. Their analysis was restricted to the working residents of rural England and was undertaken with the specific aim of seeing whether being a recent arrival in a local area increased the odds of a long commute. They used several measures of the geographical context of people's residence including location in south-eastern England or not, in which of the three rural district types (in descending order of degree of rurality: Rural-80, Rural-50 and Significant Rural), and in which of four urban–rural settlement types (in descending order of rurality: hamlet and isolated dwelling, village, town/fringe, and urban areas with at least 10,000 people). Allowing for other factors associated with people's commute, it was found that living outside south-eastern England significantly reduced the odds of long-distance commuting, but that the odds rise with the rurality of local authority district. Perhaps surprisingly, the odds of a long commute were found to be lower for people living in the two smallest settlement types than for those in towns and larger urban areas, although the differences were not statistically significant at the 5% level.

Standing back from the detail of these studies, there appear to be some common features among their findings. Generally, labour markets in the most remote rural areas tend to be the most localized, while the incidence of out-commuting is highest in more densely populated areas around the larger cities. On the face of it at least, however, there also seem to be some contradictions, such as the longer average distance of commutes in non-metropolitan counties and more rural local authorities and the lack of a significant relationship between long-distance commuting and settlement size. This latter finding is something that needs to be probed further in relation to settlement planning issues, and there is a need for a more systematic examination of urban–rural differences in commuting distances across the country. The analyses below attempt to provide greater clarity on these sorts of issues.

Extending the Evidence Base

The purpose of the analytical part of this paper is to build on the work undertaken by Champion et al. (2008). The main development from that study is to extend the population analysed so that it covers the whole of England rather than just its rural districts. A second step forward is to probe more fully the role of the measures of geographical context, particularly the two urban–rural measures relating to type of local government district and size of settlement. Thirdly, although the main modelling exercise is again focused on long-distance commuting, the results of this are related to the wider picture of all working residents including those who work at home.

The data are from the 2001 Census Individual SAR or, more specifically, the Controlled Access Microdata Sample (CAMS) version of it that can be accessed only in the secure setting of the ONS's Virtual Microdata Laboratory, as used by Green and Owen (2006) and Champion et al. (2008). The CAMS version contains a larger set of variables and greater category detail for many of these. Particularly important for present purposes, it includes the settlement size variable, the identity of each local government district of residence (allowing district-level information to be attached to people's records), and both banded and unbanded distance to work, the latter recorded to the nearest 100 metres of straightline distance between home and workplace address. It should be noted, however, that the CAMS gives no further details about the workplace, not even the local government district in which it is located. That type of information is available only from the Special Workplace Statistics, as used by Frost (see above).

The key variables for this study are the two that represent urban–rural dimensions. The settlement size variable is based on the morphology of built-up areas and residential addresses, and involves the classification of the 2001 Census Output Areas (areas that contain an average of 125 households) into four basic categories. Areas where the majority of their residents live in a built-up area with at least 10,000 inhabitants are treated as ‘urban’, although in order to be clear here they are always referred to as ‘urban areas with at least 10,000 people’. The remaining ‘rural’ settlement is split three ways on the basis of a complex set of rules that take into account the population density profiles around each hectare-sized grid square (see Bibby & Shepherd, 2004, esp. Annex 1; see also Countryside Agency et al., 2004): ‘small town and fringe’ (referred to below as ‘town/fringe’), ‘village’ and ‘hamlet and isolated dwelling’. In the full definition these four settlement types are each divided two ways according to how sparsely populated is the wider regional context of each settlement area, but this further breakdown is ignored for present purposes because the ‘sparse’ part of England is restricted primarily to the more remote areas of north-east England and Cumbria, together with small areas in eastern Lincolnshire, north-west Norfolk, west Shropshire, west Herefordshire and the more northerly third of the South West peninsula. In all, sparse England including its towns of 10,000 or more was home to only 1.4% of England's population at the 2001 Census. On the four-way basis, urban areas with at least 10,000 residents made up 80.7% of England's total population in 2001, leaving rural England with 19.3% shared between the town/fringe category's 9.0%, villages 7.2% and hamlets and isolated dwellings 3.1% (for further details and a colour map, see CRC, 2007c, pp. 7–8).

The classification of local government districts (officially called ‘unitary and local authorities’) builds on this detailed settlement-based classification. Districts are allocated to one of six categories according to population size and the proportion of their population that lives in rural settlements (as defined above) and market towns of up to 30,000 people. Table 1 presents the detailed specifications. On this basis, rural England comprises 36.5% of England's 2001 population, based on the three most rural types of districts (for further details and map, see Department for Environment, Food and Rural Affairs [DEFRA], 2005; CRC, 2007c, pp. 15–16). This classification provides the most up-to-date way of distinguishing the degree of rurality and urbanness of the sub-regional context in which the individual settlements and their residents are located.

Table 1. DEFRA's rural–urban classification of the unitary and local authorities of England, 2005

District type (n; %)	Definition
Rural-80 (73; 11.7%)	Districts with at least 80% of their population living in rural settlements and market towns
Rural-50 (52; 11.7%)	Districts with at least 50% but less than 80% of their population living in rural settlements and market towns
Significant Rural (53; 13.1%)	Districts with more than 26% of their population or more than 37,000 living in rural settlements and market towns
Other Urban (55; 13.7%)	Districts (that are not Major Urban or Large Urban) with less than 26% of their population or fewer than 37,000 living in rural settlements and market towns
Large Urban (45; 14.8%)	Districts with 50% or more of their population or at least 50,000 living in one of 17 urban areas with a population of between 250,000 and 750,000
Major Urban (76; 35.0%)	Districts with 50% or more of their population or at least 100,000 living in one of 6 urban areas with a population of more than 750,000
England (354; 100.0%)
Note: n, number of districts; %, percentage share of England's 2001 population.
Source: DEFRA (2005). See also CRC (2007b, pp. 15–16).

Distance to Work Distributions for Two Urban–Rural Dimensions

This section examines how far commuting distance varies across the two urban–rural dimensions just described. This analysis is based on all of England's residents aged 16–74 who were working in the week leading up to the 2001 Census, which amounted to 684,585 people in the 3% CAMS sample, and uses the banded distance to work data there. Nationally, 9.2% indicated that their home was also their workplace and so do not have a length of commute assigned to them. Two-fifths (39.9%) commuted between 0.1 and 4.9 km, a further one-third (33.7%) travelled at least 5 km but under 20 km, and 1 in 8 (12.5%) travelled further than this to a regular workplace. Finally, just under 1 in 20 (4.7%) had no fixed workplace or were working outside the UK at that time.

Figure 1 shows how much variation from this national pattern there was across the four settlement types. Rather little difference is evident. The proportion of residents with a commuting journey of less than 5 km falls from 40.3% for the residents of urban areas with at least 10,000 people to 36.3% for those living in hamlets and isolated dwellings, but this pattern is almost exactly compensated for by the regular increase in the proportion working at home as settlement size falls. The latter is more than one-half as much again in hamlets and isolated dwellings as it is for urban England. Meanwhile, just 1.6% points separates the two extremes of the settlement hierarchy in terms of the proportion with no fixed workplace or working overseas, and there is a similarly small range in the proportion with regular journeys of 20 km or more. Finally, the proportion travelling 5–20 km to work is highest for urban areas, and falls progressively with settlement size.

Figure 1. Distance to work, 2001, England and size of settlement. Notes and source: see Table 2.

Moreover, these patterns differ little between the rural and urban parts of the country. Table 2 presents the same information for the four settlement types separated out into those located in the three urban district types and compares it with that for those in the three rural ones. The proportion of people working at home or in the other category rise progressively with falling settlement size in both urban and rural districts, while also in both the proportions travelling less than 20 km generally fall with settlement size. The only distinction in pattern is that the proportion commuting 20 km or more falls with settlement size in the rural districts, but the percentage point difference across the four settlement types is small in both types of districts.

Table 2. Distance to work, by settlement type, for urban and rural districts of England

Settlement type by urban/rural districts	At home	<5 km	5 to < 20 km	20 km+	Other	All in work
Urban districts
Urban area >10 km	7.8	42.8	35.5	9.6	4.3	100.0
Town/fringe	8.4	43.1	32.9	10.4	5.3	100.0
Village	9.6	41.8	31.8	11.2	5.6	100.0
Hamlet and isolated dwelling	10.6	40.5	32.2	10.6	6.1	100.0
Rural districts
Urban area >10 km	10.6	35.7	31.6	17.3	4.8	100.0
Town/fringe	11.3	35.0	31.8	16.7	5.1	100.0
Village	13.1	33.2	31.3	16.6	5.7	100.0
Hamlet and isolated dwelling	16.1	32.2	29.3	16.1	6.2	100.0
Note: Other refers to no fixed workplace (4.3% nationally) and working outside the UK (0.3%).
Source: 2001 Census Individual CAMS, © Crown.

What is far more striking from Table 2, however, is the broad contrast between the urban and rural districts, with the latter displaying substantially higher proportions of commutes of 20 km or more and also of people working at home, and substantially lower levels of journeys of under 5 km. These district-level contrasts are explored in more detail in Figure 2, which shows the distribution of commuting distances for the full breakdown of the six district types. The most important feature of this graph is the progressive rise in the proportion of workers travelling 20 km or more, with the 18.1% of the Rural-80 districts being more than twice the 7.8% level of the Major Urban category. On the other hand, the latter has by far the highest proportion of people travelling 5–20 km, as well as a lower proportion travelling under 5 km from home than the next three types. If the greatest environmental sustainability is viewed as a commute of less than 5 km (including people that work at home), then the Other Urban category appears to score highest (57.0%), with the proportion falling away both up the urban hierarchy to the Major Urban category (46.9%) and down it to the Rural-80 districts (44.4%). The latter, of course, is helped by its proportion of people working at home (12.8%) being at least one-half as much again as for the three urban district types.

Figure 2. Distance to work, 2001, England and type of local authority district. Notes and source: see Table 2.

Isolating the Effect of Urban–Rural Context

A potential explanation for these urban–rural differences in commuting distance is variation in population composition between places, such that some contain a higher proportion of the sorts of people who nationwide tend to make longer journeys to work while others comprise more of the sorts that have a greater tendency to work nearer to or at home. In this section the Individual CAMS data are used to separate out the effect of a wide range of personal characteristics that previous studies have shown to be associated with the distance that people travel to work, so as to try and isolate the intrinsic effect of where people live. This analysis focuses on the probability of a person commuting 20 km or more, as not only is this the most important aspect in sustainability terms but it also appears from the previous section to be the one that most strongly differentiates urban and rural districts.

The full results of the binary regression modelling of whether or not a person has a commute of at least 20 km are presented in Appendix 1. As regards the role of personal characteristics, the results conform to the general picture conveyed by the studies reviewed earlier. Higher odds of a long-distance commute compared with other people are found for those aged 30–44 years, males, the only or main breadwinner in the household, households with 2 or more cars, recent migrants, those with a degree, professionals and managers, full-time employees, and those working outside the primary sector. Some of these are particularly powerful discriminators, with the full-time self-employed having under one-third and part-time workers only two-fifths of a full-time employee's likelihood of a long commute, with low-skill workers having less than two-fifths the odds of higher professional or managerial workers, and with people in households with at least 2 cars at their disposal having 2.4 times the odds of those with no car.

Figure 3 shows the effect that three geographical context variables have on the odds of making a journey to work of at least 20 km while allowing for both these personal characteristics and each of the other context variables. Included because of previous studies' findings about the London effect, location of residence outside south-eastern England (defined as London, South East and East of England regions) reduces the odds by one-quarter compared with living in it. Secondly, allowing for these other factors produces little change for the settlement size variable from the picture conveyed by the proportions of longer commutes shown in Figure 1. The likelihood of a resident of a town/fringe, village or hamlet and isolated dwelling commuting at least 20 km is not significantly different (at the 5% level) from that of someone living in a larger town or city.

Figure 3. Likelihood of commuting 20 km or more, 2001, by place of residence, allowing for other personal and geographical factors. Notes: The first bar in each of the three panels is the reference case, set at 1.0. NS, not significantly different from the reference case (at 5% level). Source: Appendix 1.

Finally, district type remains almost as strong a discriminator now that the effect of all the other variables is allowed for. Figure 3 reveals progressively higher odds of a commute of 20 km or more as one moves from Major Urban districts (the reference category set at 1.0) to each less urban district type. Generally, the odds are now lower than the relativities that can be calculated from the data shown in Figure 1. For instance, there the difference in the proportion of workers commuting at least 20 km between Large Urban districts (10.3%) and Major Urban districts (7.8%) amounts to an excess of 32% of the former over the latter, whereas the modelling results in Figure 3 show the effect of living in a Large Urban district instead of a Major Urban one raises the odds of a long commute by less than 2%. The respective figures for Other Urban category are 78% and 39%, and for Significant Rural are 110% and 85%. By contrast, for the two most rural categories, the effect of allowing for all the other variables is much smaller, with figures of 117% and 108% for Rural-50 and 132% and 129% for Rural-80. The conclusion must be that, for the two most rural district types, any aspects of population composition or geographical context that might help to raise the likelihood of a longer commute would seem to be offset by aspects that reduce it.

Urban–Rural Differences in Average Distance to Work

The aim of this final analysis is to provide greater precision on the urban–rural differences in distances to work. It has just been shown that where people live in terms of district type still has a substantial effect on their odds of a long commute, even after allowing for other factors related to distance to work. On the other hand, it is the case that rural areas contain higher proportions of certain types of people who nationally display lower odds of a longer commute, as shown by previous studies and confirmed by the model results in Appendix 1, notably higher proportions of older workers, the self-employed and part-timers. Figure 2 has also shown that the proportion of people working at home is higher in rural areas. How do these various differences affect the amount of actual work-related travel undertaken in urban and rural areas? The availability of unbanded distance to work data (rounded to the nearest 100 metres) in the Individual CAMS allows the calculation of the average distance per worker.

The results of doing this for the 6 district types, plus the aggregates of urban and rural England, are presented in Table 3. For England as a whole, the total distance between home and workplace for our sample of 646,439 people in work in the week before the 2001 Census was just less than 7.6 million km, giving an average distance of 11.7 km per worker (including those working at home, deemed to have a commute of 0 km). When calculated separately for the 6 district types, the average commute rises progressively from 9.9 km for residents of Major Urban districts to 14.6 km for those of Rural-80 districts. The averages for the three urban types are all below the average, while those for the three rural types are above it. The average across the former come to 10.5 km compared with the overall average for rural England of 13.7 km.

Table 3. Average distance to work, 2001, England by type of local authority district and for urban and rural aggregates

Local authority district type	Total distance (×1,000 km)	Total in work (×1,000)	Average distance per worker (km)	Ratio to England average	Ratio to Major Urban	Ratio to next most urban type
England	7,578.0	646.4	11.72	1.00	1.18	n/a
Major Urban	2,157.5	218.0	9.90	0.84	1.00	n/a
Large Urban	1,039.2	94.1	11.05	0.94	1.12	1.12
Other Urban	1,038.4	90.3	11.50	0.98	1.16	1.04
Significant Rural	1,150.2	88.8	12.95	1.10	1.31	1.13
Rural-50	1,063.1	77.9	13.64	1.16	1.38	1.05
Rural-80	1,129.6	77.3	14.61	1.25	1.48	1.07
Urban England	4,235.1	402.4	10.53	0.90	1.06	n/a
Rural England	3,342.9	244.1	13.70	1.17	1.38	n/a
Notes: Sample comprises 646,439 people living in households in England at the 2001 Census who had a workplace address and were living at a known address in England 1 year ago.
Source: 2001 Census, Individual CAMS, © Crown.

The final three columns of Table 3 express the differences between these averages in relative terms. Firstly, in relation to the national average commuting distance of 11.7 km, the average for the Major Urban districts is 16% lower, while that for Rural-80 districts is 25% higher. The equivalent figures for the urban and rural aggregates are 10% lower for the former and 17% higher for the latter. Secondly, if the Major Urban districts are used as the reference point, England's average commute is found to be 18% further than for this, the whole of urban England's is 6% higher and rural England's is 38% higher. In terms of the other 5 district types, the Large Urban type has a 12% longer average commute than the Major Urban, the Other Urban a 16% longer one, the Significant Rural a 31% longer one, the Rural-50 a 38% longer one and the Rural-80 a 48% longer one.

The final column of Table 3 shows the relativities between adjacent district types, giving an indication of how much extra commuting would result from people living in the next less urban type of district. As just seen, people living in the Large Urban type have a 12% longer average commute than those in the Major Urban type. The difference between Other Urban and Large Urban is much smaller, at just a 4% lift. There is then a larger rise of 13% between Other Urban and Significant Rural, but then smaller ones of 5% and 7%, respectively, as one moves from Significant Rural to Rural-50 and then from the latter to Rural-80.

Alternatively, the relativities can be expressed the other way round, so as to indicate the ‘savings’ that might—under very hypothetical circumstances admittedly—be made in terms of total commuting travel if a particular number of people lived in a more urban type of district. For instance, if an element of national growth comprising an extra 1,000 people in work was to live in the Major Urban rather than Large Urban type, on the basis of the evidence of Table 3 their total distance to work would fall from 11,050 to 990 km, a reduction by 10.4%. Similarly, if the same number were to live in the Rural-50 rather than the Rural-80 type, their total distance to work would fall from 14,610 to 13,640 km, a reduction of 6.6%. Taking the most extreme scenario possible from this typology, if it could be arranged that 1,000 workers who would otherwise have been living in the Rural-80 type could instead be accommodated in a Major Urban type, then there would be a 32.3% reduction in total distance travelled.

Discussion

It should be stressed that the calculations of ‘savings’ presented above are put forward for illustrative purposes only. For one thing, in this hypothetical situation of shifting people between district types, it is being assumed that their characteristics and commuting behaviour will change from the averages of the one type to those of the other. For another, the calculations assume that the average behaviour of a particular type will apply to all districts of that type, whereas in reality there will be considerable diversity in average commuting distances between the individual districts of any particular type. Thirdly, the analysis is based on 2001 Census data which, judging by the substantial change in commuting patterns between 1981 and 2001 observed by Frost (2006), will not be a very reliable guide to the current situation.

Rather, these calculations are put forward to demonstrate the possibility of developing more sophisticated ways of thinking about the sustainability agenda in so far as it relates to commuting and rural areas than has been evident in recent debates. Most importantly, they help to reinforce the evidence of previous studies that there is much more to urban–rural differentiation across England than the polarization between the urban and the rural that, in the eyes of commentators like Hoggart (2003), Lowe and Ward (2007) and Shucksmith (2007), is not just long-established but becoming ever more entrenched in policy circles. What the above analyses of the Individual CAMS data have shown is that, even by taking the relatively crude geographical framework provided by DEFRA's district-level typology, there is a much more nuanced picture of relativities that goes well beyond an ‘urban good, rural bad’ approach to locating new development. Just as there are clear variations in commuting behaviour between the three separate city-size groups that form the urban half of that typology, so too there is considerable variation across the three categories that make up rural England.

At the same time, this ability to marshal statistical evidence on residents' commuting behaviour is also very helpful in allowing the importance of this aspect of the ‘carbon footprint’ to be weighed alongside the other environmental, economic and social aspects of spatial planning's sustainability agenda. As recognized by the planning guidance consultation document on sustainable economic development (CLG, 2007c), these include providing the wherewithal to allow the countryside to reach its full potential in contributing to the national economy. Set against this are the economies of scale that can be achieved by concentrating individual services and facilities in larger units, as well as the possibility that further cost saving may be gained by clustering these various functions together in larger settlements. Such strategies are likely to be most cost-effective for functions that are primarily oriented to the residential population, but on the other side of the balance sheet a policy of concentrating rural residents in larger towns can also impose other sustainability costs. Not the least of these is that more people will need to commute out to their rural workplaces. Indeed, according to Frost (2006), already by 2001 rural areas were relying quite heavily on in-commuters from urban homes. In the less sparse territory that accounts for the vast majority of the rural population (see above), the rural towns then drew on urban areas for almost one-third of their workers, with this proportion rising to 38% for the villages and to 46% for the areas of dispersed settlement.

Even so, a great deal of further work would be needed on the commuting aspects before there is likely to be enough information to plug into a more balanced assessment of rural sustainability. The key tasks arising from the above results would seem to be the following:

•	Considering other aspects of commuting behaviour besides home-to-workplace distance. Mode of transport is particularly important for the sustainability debate, especially the degree of reliance on the car and whether travelling as driver or passenger. Becoming increasingly important is the frequency of the commute, most notably how many days a week and weeks a year that the workplace is accessed. There is also the need for accurate data on actual trip length rather than crow-fly distance, as well as the matter of multi-purpose or chained trips where the commute is combined with other activities and thus reduces the separate travel that would otherwise be required for these.
•	Exploring alternative ways of representing the spatial patterning of settlement. Because they could be readily implemented using the Individual CAMS, two dimensions of urban–rural context were chosen for the above analysis, so it would be useful to test whether any other representations of space might be more relevant. Even if choosing to stick with the district basis, there are other classifications of districts that might be more powerful than the 6-fold typology used above. As regards the settlement size variable, it may be worth using other sources to check whether it is indeed the case that the likelihood of a long commute varies so little across the spectrum between urban areas and the most dispersed settlement, and to seek an alternative aspect of settlement morphology that might be more discriminating.
•	Analysing further the effect of rural in-migration on commuting behaviour. The modelling results in Appendix 1 have confirmed the results of previous studies that showed a change of address in the year before a census being positively associated with higher odds of longer-distance commuting. According to Champion et al. (2008), this is especially the case for people moving into rural England from a Major Urban district. On the other hand, it was also found that longer-term residents (including the many in-migrants of more than one year's standing) had much lower odds than the recent migrants, suggesting that quite soon the new arrivals change their workplace to give a shorter commute. Given the large scale of residential movement from urban to rural England, it is important to try and measure the overall contribution made by migration to the above-average commuting distances observed in rural areas, allowing for whatever adjustments are made in subsequent years—something that is not possible from the census.
•	Assessing the implications of commuting behaviour varying by other socio-demographic characteristics. The modelling results have shown that the odds of residents undertaking a long commute depend on much else besides being a recent in-migrant to a locality. In particular, females, older workers, the self-employed, and those in part-time jobs or low-skill occupations all tend to have much lower odds of a commute of 20 km or more, when standardizing for other factors including place of residence. It would be useful to explore whether increasing the representation of these groups in the rural population could allow growth without a proportionate rise in total commuting travel. There is also the question of how far potential improvements in the provision of transport and service infrastructure in rural communities might reduce the number and distance of trips by residents. Finally, research might usefully examine the carbon emissions savings to be achieved through encouraging more of those working in rural jobs to live in the countryside rather than commuting from urban areas.

Finally, given that this study has identified a direct association between distance to work and degree of rurality irrespective of whether or not people's characteristics are controlled for, it is important to widen the discussion to consider alternative ways of mitigating the ‘carbon footprint’ impact besides imposing some form of moratorium on rural development. Indeed, according to Banister (2006), the latter approach is just one option within one of three societal and strategic solutions. Also under Banister's ‘land use and development solution’ is the adoption of the nodal approach where growth is directed towards the ‘beads on a string’ along a transport corridor between larger settlements. Support for affordable housing and appropriate services in smaller settlements more generally could also reduce the amount of in-commuting to rural jobs from the larger towns and cities. The other two ‘solutions’—technological ones such as developing greater fuel efficiency in cars, and transport ones such as getting the car to take on more of a community role—are more generic, although arguably having a greater relevance in more rural settings.

Conclusion

The stimulus for the research undertaken for this paper originated in the disjuncture between the evidence of a substantial and growing shortage of affordable housing in rural England and the steps being taken to reinforce the controls on building there in the wake of rising concerns about climate change. Fears have been expressed that rural development prospects will be weakened by recent claims that rural communities are intrinsically unsustainable owing to their limited service and transport infrastructure and the need for their residents to travel longer distances for employment, shopping and other facilities than urban populations. In particular, in its efforts to rebut what it sees as ‘myths’ surrounding this issue, the CRC (2007a, p. 6) was unable to find sufficient evidence to pronounce one way or another on the question of carbon emissions. Focusing on commuting as one of the most important contributors to the latter, the paper has reviewed the findings of more recent studies of commuting in England that have included a rural dimension. It has then presented the results of new analyses of people's journeys to work in order to gauge the relationship between commuting distance and two urban–rural dimensions.

From the material reported in the paper, it would seem that the CRC's view is vindicated, at least with respect to this aspect of people's travel behaviour. The evidence provided by previous studies is rather fragmentary, no doubt partly because they have varied in their research aims. Their findings on the effect of rurality are also somewhat contradictory, this being partly because of the different ways in which the rural has been defined. The latter was certainly evident in the results of analysing the 2001 Census Individual CAMS, which indicate remarkably little role in differentiating commuting distances for the finer-grained settlement size dimension but a strong and regular one for the broader district-based urban–rural typology.

What both these lines of evidence do clearly reveal, however, is that there is no justification for a simplistic distinction to be made between urban and rural England, such that residence in the one is good in terms of work–travel sustainability and residence in the other is bad. The journey-to-work trips by residents of both parts of the country generate their carbon footprints. Yet, when the debate is more sensibly translated into relative terms, the issues need much more careful treatment, most notably acknowledging that the differences in carbon footprint would seem to go well beyond the simple duality of the average urban resident generating a smaller one than the average rural resident. The analyses above have shown substantial differences within the urban and within the rural parts of England as well as between them, while some of the previous research has revealed greater labour market self-containment in more remote rural areas than in those situated nearer to larger towns and cities.

Building on these observations of the importance of relativities rather than a black-and-white conclusion, the main message that now needs to be stressed for the future of the rural sustainability debate is the need for a balanced and integrated approach. The proportion of new development that should take place in the different types of rural (and urban) areas should be decided not merely on the basis of commuting trips but also in the light of information about the carbon footprint of all personal travel and delivery services. Along with other factors affecting environmental sustainability, there is also a need to take account of the relative economic and social benefits of development in different places. Inevitably, if such a balanced approach is to be implemented in an objective and accurate way, there is a massive amount of information needed on the pros and cons of alternatives relating to the scale, type and geographical patterning of settlement and economic activity. At least as far as commuting is concerned, the evidence currently available for inputting to this wider exercise of weighing up and trading off options would seem to be woefully inadequate, and possibly the same applies to the many other factors that also need to be fed into it. Yet, without such efforts being made to improve and extend the evidence base, it is hard to see how planning decisions designed to achieve the most sustainable futures for the residents of both rural and urban areas can be made with confidence.

Acknowledgements

The support of the Office for National Statistics, the Centre for Census and Survey Research and the Economic and Social Research Council/Joint Information Systems Committee Census of Population Programme in helping to provide access to the 2001 Census Individual CAMS is gratefully acknowledged. Census output is Crown copyright and is reproduced by permission of the Controller of Her Majesty's Stationery Office and the Queen's Printer for Scotland. The author alone is responsible for the interpretation of the data.

Appendix 1. Results of modelling the propensity of England's working residents to commute 20 km or more, 2001

Characteristic	Odds
Constant	0.058***
Personal
Aged 30–44
16–29	0.897***
45–59	0.819***
60–74	0.685***
Male
Female	0.590***
Household Reference Person (HRP)
Spouse or partner of HRP	0.901***
Child of HRP	0.798***
Other relation of HRP or unrelated	0.735***
1 earner in household
2 earners in household	0.882***
3+ earners in household	0.747***
Not female HRP with dependent child
Female HRP with dependent child	0.773***
No car available to household
1 car available to household	1.614***
2 cars available to household	2.421***
Stayed in locality
Moved 5 km + in past year	1.584***
With a first degree or equivalent
Does not have first degree or equivalent	0.799***
Full-time employee
Full-time self-employed	0.315***
Part-time employed & self-employed	0.406***
Higher professional/managerial
Lower professional/managerial	0.741***
Intermediate occupation	0.611***
Low skill occupation	0.372***
Working in non-primary sectors
Primary sector	0.881***
Geographical
Living in south-eastern England
Not in south-eastern England	0.755***
Living in urban area with population of 10K +
Town/fringe	1.012
Village	1.009
Hamlet and isolated dwelling	0.979
Living in Major Urban LA
Large Urban LA	1.017***
Other Urban LA	1.385***
Significantly Rural LA	1.853***
Rural-50 LA	2.080***
Rural-80 LA	2.292***
Nagelkerke R Square	0.156
−2 log likelihood	446279
Notes: N = 646,439 people living in households in England at the 2001 Census who had a workplace address and were living at a known address in England 1 year ago. Table shows the odds of commuting 20 km or more compared with the reference case for each variable (in italics, odds = 1.000). South-eastern England comprises London, South East, and East of England Government Office Regions. Significance levels: ***0.001, **0.01, *0.05.
Source: 2001 Census, Individual CAMS, © Crown.