One step forward but two steps back to the proper appreciation of spatial science

Spatial science: What it was and what it is

The main response to the first part of our paper is Cresswell’s; ² when writing our original paper his book was the most recent exemplar of the problems we identified as too common in contemporary texts. We found much to admire in his book. Moreover, lest it be seen that we are overly preoccupied by one author’s writings, we emphasise that his mis-representations of spatial science as largely mired in the 1960s are not unique. Indeed, the third edition of a bestseller (Cloke et al., 2014) makes exactly the same claims regarding spatial science’s marginality (pp. xxii–xxiii and 940), not having changed from its 1960s origins of ‘formulating and testing theories of spatial organisation, interaction and distribution’. That book has three goals (p. xv): to map out ‘the big, foundational ideas that have shaped the discipline past and present’; to explore ‘key research themes [now] being pursued in Human Geography’s various sub-disciplines’; and to identify ‘some of the current research foci that are shaping the horizons of the subject’. In the editors’ view spatial science certainly plays no part in the second and third of those, so students need not be introduced to its work (in stark contrast, Cox, 2014, sees it as fundamental). A book written for ‘students new to university degree courses’ basically ignores contemporary spatial science. Its editors claim that human geographers ‘argue both that human life is shaped by “where it happens” and that “where it happens” is socially shaped’ (p. xvii) – which is exactly what much contemporary spatial science demonstrates. Alongside Cresswell, however, they see quantitative work as descriptive only (p. xxii: Little, 2014, 26 – implies that quantitative analysis can describe but not explain and equates spatial science with ‘a kind of spatial determinism in which spatial difference caused social inequality’ – p. 25, her emphasis – and a ‘belief that space was passive’)! Explanation and causation are very much on quantitative social science’s contemporary agenda with considerable interest and progress in doing this with observational not experimental data – witness the 1600+ pages of Davis (2014) and also Gangl (2010).

Cloke et al.’s book (2014) is not entirely lacking reference to more recent work in spatial science – witness Kitchin’s (2014) careful discussion of the difference between explanation and understanding, clearly identifying the methodological and philosophical distinctions between quantitative and qualitative work but without (understandably) detailed illustrations of the types of analysis conducted in contemporary spatial science; Conradson (2014) briefly mentions, but without exemplars, the epidemiological tradition in the geography of health and well-being. A chapter on GIS asserts that it facilitates data collection, analysis and modelling but provides no examples of the latter two (Haklay, 2014); it seems more concerned to advance engagement between GIS and ‘critical human geography’ than with spatial science and its illustrations of GIS’s utility refer to business applications only. Nowhere in a book of over 900 pages is the full range of spatial science’s powerful contemporary arsenal engaged with and there are no illustrations of what can be achieved through rigorous quantitative data analysis. There is an acceptance that students should have some introductory training in numerical methods but no willingness to address why. Books like Cresswell’s and Cloke et al.’s – plus others and, perhaps, many lectures, seminars and tutorials – risk ending-up normalising the discipline (Amariglio et al., 1993; Klein, 1993; Lenoir, 1993; Luke, 1999) according to their authors’ conceptions of what it should and should not contain rather than presenting introductory students with a conspectus of its diversity of practices and letting them decide.

Castree et al. illustrate the same issue in their two definitions of spatial science (2013: 486). The first – ‘An approach to human geography centred on the analysis of spatial patterns and processes through quantitative methods, with the ultimate aim of establishing spatial laws’ – is very similar to Cresswell’s and Cloke et al.’s, especially regarding laws; a subsequent discussion clearly identifies such work as concentrated in the 1950s–1970s. Their second definition – ‘A collective term for GIS, cartography, remote sensing, photogrammetry, surveying, geodesy, and related disciplines concerned with scientific spatial analysis, sometimes also termed “spatial sciences” – they suggest is “often used to imply a set of academic and technical interests separate from but also complementary to geography”’. Their definition of locational analysis is synonymous with that for spatial science (p. 291) and spatial analysis is defined as (pp. 480–481) ‘The mapping and analysis of spatial properties and patterns’, apparently incorporating much of what we include within contemporary spatial science, but their brief definition concludes that ‘With the development of GIS, spatial analysis has become a mainstream policy tool for making sense of spatial data and for aiding companies to plan their activities’ – much narrower than our definition of spatial science. They do, however, note that ‘While quantitative geography is not as dominant as it once was, it is still a potent and vital part of the discipline’ (p. 406) – exactly our point: with Cox (2014), we very much regret any ‘marginalization of quantitative methods in human geography’ (p. 255). ³

What we, following Cresswell (2013), call spatial science may therefore be associated by some with pre-1980s practices, whereas more contemporary work, because it distances itself from logical positivism, is simply embraced by the collective title of quantitative geography. However, Cresswell’s commentary suggests that he does not make this distinction – nor, we believe, do Cloke et al. (2014) and others; they associate spatial science with the logical positivist search for laws of spatial patterns and behaviour rather than – as we stressed when initiating this dialogue – the identification and analysis of spatial variations and the search for explanatory accounts in the realist not the positivist tradition. Hence this response to Cresswell’s commentary, in which he extends his critique of such work. ⁴

Place in spatial science

… place – especially as manifested in neighbourhoods – is a fundamental context that has widespread effects on crime, perceptions of order and disorder, well-being and much more, including the social organization of the contemporary metropolis. (Sampson, 2013: 1)

Cresswell’s view that the ‘contemporary focus in spatial science on local variability sounds like a numerical version of regional geography’ further illustrates his misinterpretation of what he terms a wider ‘culture of numbers’. Regional geography as practised in the United Kingdom up to the 1970s suffers from a poor reputation (Gould, 1979). Nevertheless at its best it ensured that students knew the basic lineaments of the wider world and gave some appreciation of ‘place’ as Cresswell defines it (can one understand the world unless one can first describe it?). He believes that our deployment of place is simply a description of ‘what things gather where’. It is much more than that, however, because analyses show that which people gather, and in many cases interact, where can influence what happens there and may be transmitted elsewhere (near or far), of which there is no better illustration than decades of meticulous quantitative research in the epidemiological tradition (e.g. Cliff et al., 2000). Place to spatial scientists is a context in which things happen, because places are constituted by people who interact there: a recent overview of American research in this area (see e.g. Sampson, 2012) concluded that not only ‘the American system of stratification is organized, in part, along spatial lines’ but also ‘the spatial dimension of American inequality plays an important role in the maintenance and reproduction of inequality across multiple dimensions’ (Sharkey and Faber, 2014). With increasing inequalities at all spatial scales (Wilkinson and Pickett, 2009), it seems perverse not to focus on the role of places as contexts in the reproduction and exacerbation of those inequalities.

Counting the individual

One recent major change in spatial science, indeed social science more generally, has been its increasing collection and use of data at the individual rather than the aggregate scale, avoiding the ecological and other fallacies that so many earlier studies encountered as well as the modifiable areal unit problem. ⁵ Cresswell is ‘far from sure that I want to celebrate the entry of the individual into the domain of calculability’, citing issues of privacy and surveillance plus ‘a more existential set of issues about the reduction of human subjectivity’. He asks ‘Do we want individuals to be quantified in this way? Do we want to encourage it by using the data?’ Our answer is a resounding yes – we want students to be savvy users of useful data that may well be the product of capitalist or ‘surveillance’ societies but can be used to inform, critique and reveal the shortcomings and social disparities within those societies. Are we unwitting co-producers of a society driven by cold quantitative and economic logics? Perhaps, but so also are those who are unwilling (unable?) to turn the numbers against themselves and engage in social and other public policy relevant critique. We fully recognise the problems of privacy and confidentiality, and believe that students should be trained to do so too – issues addressed in Elvin Wyly’s commentary – as well as ensure that collection and analysis of data compromise neither, but we firmly believe that without such data many important issues could not be fully addressed and compelling problems would remain unresolved. This is clearly illustrated in much spatial science research showing, for example, that without individual data it is impossible to elucidate the relative importance of genetic and environmental factors (let alone their interactions) on disease causation (Sabel et al., 2009) and appreciate the mortality impacts of environmental events (Pearson et al., 2013). To illustrate the potential problems when groups are used to make inferences because individual data are unavailable, consider claimed gender bias in admissions to the University of California, Berkeley (Bickel et al., 1975). Overall, male applicants were considerably more likely than women to be admitted, but when individual data were examined by department there was no evidence of bias – because males tended to apply to less-competitive departments. Students need to be fully aware of such paradoxes and their geographical equivalent, the ecology fallacy, in making inferences with aggregate data.

We have never claimed that the approaches we promote ‘even begin to approach our subjectivity’, nor would we argue that research in cultural geography focusing on subjectivity (whose ‘data’ must presumably be collected from individuals?) should not occupy a prominent place within geography. We simply contend that there is more to studying and trying to change the world than subjectivity.

We live in a capitalist society in which the state is necessarily important, where more big data are being ‘pumped out by corporations, governments and the media’. We agree with Cresswell that students should ‘know enough to question the stream of numbers’ but disagree with his wish that they should not be ‘part of the making of a comprehensively calculable world’. They are going to be, and we see it necessary that they are prepared to be citizens of such a world (however undesirable we might think that is idealistically). Those data are involved in the ‘production of profit and the manipulation of populations’; ensuring students are aware of that means our teaching must involve much more than them knowing ‘what correlation means (even just to know that it is still not causation)’. Our argument is represented by him as ‘Big businesses, governments and an assortment of ill intentioned people know about numbers and use them therefore we (well intentioned) people must do so too’. But that does not make us the lackeys of corporate capitalism and governments; we are social scientists ensuring that students become citizens who can evaluate how those (often well intentioned) people collect and use data as well as, as far as possible, that they are used for the greatest good of the greatest number. As Thompson (2010: 381) expressed the case for retaining Canada’s decennial census, high-quality data are needed not only to inform decision-making but also ‘to allow the society to question and judge whether or not the government is acting in its best interests’.

Nevertheless, we share Cresswell’s (and Wyly’s) concerns about the intrusiveness of much corporate big data production. There is, however, a difference between data obtained through covert surveillance and that given either freely (credit or store loyalty card use or via leaving your mobile phone on permanently) or as part of one’s civic duty, as with data collected by democratically accountable governments to pursue their policies (as with censuses: Hannah, 2010 discusses citizen opposition to their collection and particular forms of use). The UK government, through its open data initiatives, is increasingly making such data available for public use if there is no conflict with individual and group confidentiality and privacy, enabling both critical analysis and well founded contributions to policy development. ⁶ Without access to such data, carefully balanced against the individual rights to geoprivacy and guards against abuse and misuse, there is the risk of significantly impeding advances in the understanding of and responses to morbidity and mortality patterns (Exeter et al., 2014).

We too do not favour a society where, as Wyly expresses it, ‘Human trust is under siege, replaced by an expanding universe of surveillance’. But a civilised society surely needs some form of welfare state to tackle Beveridge’s five ‘Giant Evils’ – want, disease, ignorance, squalor and idleness (Timmins, 1995). Although policies designed to tackle each of those five are ultimately aimed at individuals, many are delivered to them within collectivities (e.g. schools and hospitals). Analyses conducted by academic social scientists and those in the public sector trained in their methods – using place as context, whether in studies of unemployment, morbidity and mortality, poverty or a wide range other topics – are fundamental to appreciating where the problems that the welfare state exists to solve are concentrated and thus the where of its policy directions (Bastow et al., 2014).

There is, however, one element of the ‘domain of calculability’ not addressed by Cresswell, that requires caution. Counting and, especially, reporting what we have counted frequently involves categorising individuals, which structures society in ways that become part of how it is represented and reproduced. The earliest censuses did this by not only counting individuals in places but also what categories – such as occupational classes – they were members of (Levitan, 2011). Some categories are (relatively) uncontestable, such as age and sex; many others are highly contested, such as the much-debated ethnicity categories deployed by the UK’s Office of National Statistics (Berthoud, 1998). This involves what has been termed the ‘mutual construction of statistics and society’ (Sætnan et al., 2011), much of it undertaken (especially by social scientists) by and for the state to promote its governance objectives – and, some would argue, the exercise of power by those controlling the state bureaucracies (Desrosières, 2011; Kullenberg, 2011). Those categorisations can be arranged along an objective–subjective scale: as they approach the latter pole, they should be treated with far more caution because of the difficulty of measuring many concepts – as with the UK Prime Minister’s ‘happiness index’. ⁷ All quantitative spatial science curricula must ensure student awareness of the social construction of many of the categories analysed and the role of such analysis in how socio-spatial structures are represented. They must also appreciate that measurement and categorization are not simple and straightforward interpretive acts and subjectivity brings its own problems. Oakley’s (1990: 177) polemic ‘Who’s afraid of the randomized control trial’ argues that ‘random numbers have the edge over human intuition because human beings are not always right in the judgements they make’. Indeed she claims that quantitative experimental methods are a key part of emancipatory social science lest the ideology of researchers and practitioners is imposed on those being researched and the status quo is thereby maintained.