With rapid increases in car use in many countries from the 1950s onwards, and better understanding of the health risks of airborne pollutants, there has understandably been growing interest in the EJ research regarding the relationship between area deprivation and the spatial variation in air pollution [
6,
7,
8,
9,
10]. However, although the findings of existing studies generally suggest a positive correlation, there have also been some indications in the literature that association between air pollution and deprivation might be more complex than that described by a simple linear relation. For example, with respect to the distribution of pollutants in the UK, the RCP report notes that “the relationship with deprivation is not straightforward” [
2] (p. 73) and [
5] found that, at the time of the 2001 census, the relationship between deprivation and air pollution was J-shaped. That is, both the most affluent areas and some of the most deprived areas experienced higher levels of air pollution, and areas of middling deprivation were exposed to least pollution on average.
Focus on a particular economic or social model of urban development can lead one to expect all cities to be converging towards a particular universal norm. A simple market sorting model, for example, might lead one to expect poor households to eventually be sorted into high pollution areas. If, however, one considers a wider set of theoretical perspectives, one might expect the relationship air pollution and deprivation to be rather more idiosyncratic. Not only might it not be universally positive and monotonic, but the relationship may also be spatially and temporally varying and persistently so due to path dependencies: local historical differences in development and topology lead to long term differences. Meanwhile, short term variations in demography and economic activity can have non-trivial impacts on the location of pollution and poverty, leading to fluidity over time in the relationship. Both spatial variations and temporal fluidity should therefore be viewed as potentially persistent features, rather than temporary states on the road to convergence to a uniform long-run equilibrium. Hence, the aims of this paper are first to encourage a broader theoretical outlook which supports the idea of spatially and temporally varying relationships; and second to demonstrate an approach to empirical modeling which captures such variations. We apply this approach to data for Scotland, showing substantial spatiotemporal variability in the observed pollution–deprivation relationship.
The structure of the rest of the paper is as follows. In the next section, we outline in more detail the various theoretical arguments which might lead to more complex relationships between air pollution and deprivation. We then outline our proposed method, and summarize the data used in the analysis. Finally, we present our findings and conclude with a brief summary, offering some tentative thoughts on the implications for policy and suggested avenues for future research.
A More Pluralistic Theoretical Framework
A range of theoretical perspectives lead us to question whether the relationship between environmental quality and deprivation will necessarily be positive, and to suggest that the relationship may vary over space and time in a persistent fashion. Different causal processes generating the relationship, with a focus on the discriminatory siting of environmental disamenities and post-siting population dynamics in the general environmental justice literature, have been reviewed in [
13]. The theoretical grounds on the spatiotemporal nonlinearity of the relationship is however not explicitly discussed and tested empirically, which is the focus of the study.
Why Not a Positive Relationship?
Two arguments can be mobilized to challenge the expectation of a positive relationship between air pollution and deprivation: the existence of location trade-offs, combined with heterogeneity of preferences; and lack of awareness of air pollution or understanding of its impacts.
Location Trade-Offs and Heterogeneous Preferences
Neighborhoods are composite commodities or “bundles of goods”, some private and some public, some desirable and some not, and some more attractive to some types of household than others [
12,
14]. There is an extensive literature using hedonic models to estimate the value of individual neighborhood characteristics such as proximity to a train station or a park [
15,
16,
17,
18,
19]. It is also clear that many neighborhood characteristics are clustered together in complex ways, with some generally seen as “bad” often positively associated with others seen as “good” neighborhood characteristics [
20,
21]. Households are therefore faced with complex trade-offs.
This paper is focused on one such association: that which exists between agglomeration economies and air pollution. Both arise broadly from density, although the basis for the relationship has altered as cities have undergone the shift from industrial to post-industrial economies. As industrial centers, the employment locations within cities were simultaneously the attractors for households and the generators of a substantial proportion of air pollution [
22,
23]. With relatively low levels of mobility, commuting distances were short so richer and poorer households lived close to these locations, adding to air pollution through domestic combustion of polluting fuels such as coal. With the shift to a post-industrial economic base in cities across the developed world, urban economic activities are no longer so locally polluting while domestic energy consumption produces far less pollution in cities, partly because of the switch to cleaner fuels, and partly because households have spread to suburban and ex-urban locations [
24,
25,
26,
27,
28]. Instead, the urban economy now produces a larger proportion of its damaging effects on air quality through commuting and other travel flows, particularly those made by the same cars which enabled suburbanization [
5].
Ongoing changes to urban structures, combined with complex interactions with local social and cultural trends and path dependencies, suggest a potentially fluid and complex relationship between air pollution and deprivation over time. For example, rising car ownership may have enabled more affluent groups in particular to remove themselves from the most polluted locations, but it is clear that not all have chosen to do so. Preferences are more heterogeneous. Indeed, various “Back to the City” counter-movements are led by some of the most affluent groups [
29,
30,
31,
32].
The processes of de-industrialization have created the opportunity spaces for gentrification in former industrial buildings and dwellings that would previously have housed industrial workers, as well as creating the social groups that drive gentrification, notably the younger, better-educated workers of the post-industrial economy. In many cities, a culture of urban living has been created, rediscovered or imported which is attractive at least to some with high levels of locational choice in the housing system. For these households, the opportunities afforded by living in dense urban locations are more than offset by the downsides, including air pollution.
Lack of Awareness or Understanding
The first point assumes that individuals are both aware of levels of air pollution in their locality and motivated by the risks which these pose. There is however an extensive literature on environmental risk perceptions which questions both assumptions (see Bickerstaff 2004 for a detailed review) and which leads us to further question whether there is likely to be a simple, stable positive relationship between air pollution and deprivation. First, there are psychological and behavioral economics literature which suggests we are poor judges of air pollution and similar environmental risks [
11]. We exhibit perceptual biases, notably the tendency to discount environmental risks in our own neighborhood (the “neighborhood halo” effect). Even when we identify risks, we tend to play down the potential for adverse effects to impact on ourselves (“personal invulnerability”). Furthermore, our perceptions of air quality are strongly influenced by visual stimuli (soot or dirt, or pollution “haze”, for example) even though these provide only a partial guide to the aspects of air quality which have most influence on health.
Second, there is a growing contribution from disciplines such as geography and sociology on the social and cultural influences on our perceptions. A place which had pollution in the past may retain a reputation or stigma which bears little relation to current risks. Conversely, we might expect positive reputations to last even once air quality has deteriorated. Looking at social differences, more powerful groups with a greater sense of personal or collective agency (crudely, white, male, or more affluent, for example) are more likely to discount risks and to demonstrate a greater sense of personal invulnerability. Thus, although these groups may have greater opportunity to avoid more polluted locations by virtue of their economic resources, they may be less motivated to do so. The “neighborhood halo” is also stronger for those who feel positively about their neighborhood, which is more likely to be the case for the more affluent groups in less deprived locations [
33]. Moreover, the process of updating our knowledge and the transition to more informed understandings of environmental risk are not necessarily gradual or convergent over time, but characterized variously by periods of inertia, myopia, sudden realization, and amnesia [
11]. The complexity and fluidity implied by our bounded rationality and knowledge acquisition implies similar potential characteristics for the relationship between deprivation and pollution.
Why Spatial Variation, Persistent Divergence and Fluidity?
We can also identify two sets of arguments for why the relationship between pollution and deprivation may vary over space in a stable fashion, as well as going through periodic shifts which cannot be described as simple convergence. These relate to issues of path dependency and the de/re-commodification of housing stocks.
Path Dependency
Cities and city regions are complex, historically contingent systems [
34,
35]. Physical development is relatively durable, and past forms and uses constrain current development possibilities. Cities and regions each developed in different physical environments, under different social and economic conditions, and adapted in response to different challenges and opportunities. This means that a common schedule of congestion and agglomeration effects as a function of the density of human activity, and the trade-offs or pay-offs that come with density, cannot be assumed to apply equally well, or to be at the same stage of evolution, for all city regions.
In addition, there is strong path dependence in neighborhood hierarchies. Reference [
36] shows how the reputation of a neighborhood, established when it is first constructed, may influence its status many decades later. Reference [
37] shows that deprived neighborhoods may find it difficult to shake off their stigmatized status even after substantial investment and social change. Localized externalities are another source of persistence. Positive amenities such as high quality public or private services serve to attract more affluent groups which, in turn, sustain those services either through their purchasing power or through active interventions in political processes—the “sharp-elbowed” middle classes [
38]. Neighborhoods can and do change but such path dependence means change may be slower or more discontinuous rather than a process of smooth adjustment. In other words, significant changes in levels of air pollution may not generate corresponding changes in residential status or mix, or may do so in different ways and at different speeds in different cities.
De-Commodification of Housing
In general, the argument from sorting models is that even altruistic planning decisions are ineffectual in re-distributing environmental disamenity in the longer term as market sorting serves to re-establish social gradients in parallel with disamenity [
4]. One significant exception here may be where housing is de-commodified so that market sorting processes no longer apply. In theory, this is the case with much social housing, since access is via bureaucratic allocations based on needs assessment, not economic resources [
39]. The consequences for the pollution–deprivation relationship will depend on the extent of de-commodification and the spatial distribution of the social housing stock, and will therefore vary between cities and over time. Scotland, the focus of the empirical material later in this paper, has a relatively large social rented sector which housed more than half the population at its peak at the start of the 1980s and which continues to be home to almost one in four. In England, by contrast, it never reached one in three, and has now fallen to one in six. Within Scotland, levels of social housing were much higher in urban than rural areas, and this may reinforce the link between deprivation and pollution. Within the cities, however, concentrations of social housing are found both in the urban core on the sites where slum clearance occurred and in large developments at the periphery where undeveloped land was available at the time.
Not only does the level of social housing vary between locations and over time, so too does the extent to which it is socially selective or targeted only on more deprived groups—compare, for example, the dual rental market system characteristic of Anglo-Saxon countries [
40] with unified rental systems which tend to be characterized by a much more diverse social housing population. Even within the one system, the degree of selectivity may vary over time depending on allocations policies [
39]. Policy changes, such as “Right to Buy”, and the transfer of the social housing stock from municipal ownership to housing associations, can cause major shifts in the patterns of housing tenure, adding to the complexity and temporal discontinuity in the pollution–deprivation relationship.
Summary
These various theoretical processes have the potential to generate changes over time and across space in the relationship between pollution and deprivation. The potential for interaction between these processes, including location trade-offs, heterogeneous preferences, bounded rationality, path dependencies and housing governance, will only exacerbate the potential complexity of outcomes. The corollary is that we should not assume that the relationship between deprivation and pollution will be linear, positive, monotonic or fixed in time or space. This presents us with an imperative to deploy empirical methodologies that are sufficiently flexible to cater for these features. It is these which we turn to now.