Upload
others
View
6
Download
0
Embed Size (px)
Citation preview
AnexplorationofknowledgeproductionthroughGISinthecontext
oftheUKfooddesertdebate:ConstructingtheFoodAccessRadar
FHSinGeography2017
CandidateNumber:289123
WordCount:11994
2
TableofContents
TableofAbbreviations...........................................................................................2
Abstract.................................................................................................................3
Acknowledgements................................................................................................3
1.Introduction.......................................................................................................4
2.LiteratureReview...............................................................................................62.1.ScienceandTechnologyStudiesandCriticalGIS........................................................................6
2.2.FoodDeserts..............................................................................................................................................10
3.Methodology....................................................................................................13
4.TheFoodAccessRadarconstructionprocess....................................................16
5.TheinfluenceofGISonthefooddesertdebate................................................285.1Simplicityandrelevancetoevidence-basedpolicy-making..................................................30
5.2.GIS,fooddesertsand‘conventionalspace’...................................................................................31
5.3Objectifyingpowerofmapsandvalidationasascientifictechnology..............................33
6.QualitativeGIStore-envisiontheFoodAccessRadar.......................................36
7.Conclusion........................................................................................................47
8.References.......................................................................................................49
9.Appendix:interviewtranscripts........................................................................55
TableofAbbreviationsEBPM Evidence-BasedPolicy-Making
GIS GeographicalInformationSystem(s)
GUI GraphicalUserInterface
OS OrdinanceSurvey
STS ScienceandTechnologyStudies
3
AnexplorationofknowledgeproductionthroughGISinthecontextoftheUKfooddesertdebate:ConstructingtheFoodAccessRadar
Abstract
IntheUKinthelate1990sdebatesemergedovertheexistenceof‘fooddeserts’
inBritishcities.Bothgovernmentandacademicresearchintogeographicalfood
desertswas characterisedby theuseofGIS to ‘locate’ fooddeserts.Whilst the
partialandoversimplifiednatureofthefooddesertconcepthasbeencritiquedin
the literature, the effect of GIS to project food deserts in a UK government
contexthasnotbeenexplored.Asameansof interrogating theeffectofGISon
theUK fooddesert debate, I construct a government-designed fooddesertGIS
model,theFoodAccessRadar,applyingittoCambridgeshire.Departingfroman
understanding of food deserts as political constructions, the GIS model
constructionprocess is explored through the lensof criticalGISand isused to
supportbroaderclaimsabouttheeffectofGISontheUKfooddesertdebateasa
whole,particularly ingovernmentpolicy.Thedissertation first exploresGISas
bothaconvenientandconvincingtoolforportrayingfooddesertsandsuggests
several implications. Both the objectifying power of GIS and the UK political
contextintheearly2000sareimportantexplanatoryfactors.Havingshownthat
food desert models problematically conceptualise food access, I suggest a
possiblere-envisioningoftheFoodAccessRadar,adoptingtechniquesfromthe
emergingfieldofqualitativeGIS.Reworkingthefooddesertmodelinsuchaway
brings to the fore the situated and partial nature of food desert knowledge
silencedinconventionalGISprojections.
Acknowledgements
IamverygratefultoMichaelAthansonforgoingoutofhiswaytointroduceme
toArcGIS.IwouldliketoextendmythankstoThomasBurgoine,CentreforDiet
andActivityResearch, forhis insightful comments concerning theFoodAccess
Radar.ThankyoualsotoGruiaBadescuforhispassionandsupport.
4
1.IntroductionDuring the late 1990s and early 2000s the concept of the ‘food desert’ gained
significant purchase in UK government as an explanation for growing health
inequalities inUKcities (CumminsandMacintyre,2002a;Beaulacetal.,2009).
The term ‘food desert’ was used in government and academic spheres to
describe areas, specifically those already characterised by high levels of
deprivation,wherephysicalaccesstotheprovisionofhealthandaffordablefood
wasinadequate(Wrigleyetal,2003:151).Poorfoodaccess intheseareaswas
thoughttocause‘deprivationamplification’(Cummins,2007:355),wherelackof
accesscompoundedexistingsocialproblems.Eradicating‘fooddeserts’through
government policy was therefore seen as a way of reducing social exclusion
(Wrigley, 2002: 2031), in line with the 1997 New Labour government’s
commitmenttotacklesocialissuesintheUK(CumminsandMacintyre,2002b).
However, by the late 2000s, mounting evidence against the existence of food
deserts in UK epidemiology and public health research suggested that the
concept,whilst increasingly applicable in theUS,might not be appropriate for
theUK(Whiteetal.,2004;Beaulacetal,2009).Yet,despitescepticismoverthe
existenceinspaceoffooddesertsduringthis ‘fooddesertdebate’(Clarkeetal.,
2004), thesubstanceof thefooddesertconcept itselfreceived littlecritique. In
almost all public health debates concernedwith the existence of food deserts,
only empirical evidence was sought, using scientific methods, to prove or
disprove their existence. Only a small minority of academics critiqued the
construction of the food desert concept itself and its relevance to food access
(Cumminsetal.,2007).
Thisdissertationbuildson the fewcritiquesof theconceptduring theUK food
desert debate. Whilst these critiques questioned the applicability of the food
desertanalogy,withits‘conventionalviewofspace’,totherealitiesofthesocial
worldandsocialpolicy(CumminsandMacintyre,2002a;Cumminsetal.,2007),
whattheydidnotexplorewastheinfluenceofgeographicalinformationsystems
(GIS)invisualisingthefooddesertconcept.Departingfromanunderstandingof
fooddesertsassocialandpoliticalconstructions(Sismondo,2010),Iexplorethe
5
widespreaduseofGIStoproduceknowledgeaboutfooddeserts,informedbythe
“blackbox”motif fromScience andTechnologyStudies (Sismondo,2010:120)
andanunderstandingoftheobjectifyingpowerofGISfromcriticalGIS(Warren,
2011; Elwood et al., 2011). By constructing a government-designed GIS food
desert model, the Food Access Radar, for Cambridgeshire, I show how
projectionsoffooddesertsinGIScanhidesocialisedandpoliticisedknowledge
production processes behind a veneer of objectivity. In the context of
government food desert policy, I link the Food Access Radar’s objectivity-
reinforcing role with the ‘evidence-based policy-making’ paradigm of the late
1990s. In relation to the wider food desert debate, I suggest that GIS’s
associations with positivist science reinforced the ‘seeming objectivity’ of the
food desert concept (Pavlovskaya, 2002), contributing to the lack of critical
studies during the food desert debate, and perhaps explaining why the
problematic conceptentertainedUKacademicsandpolicymakers foralmosta
decade.
AsameansofinvestigatingtheuseofGIStovisualisefooddesertsIexploreda
particularGISmodeloffooddeserts.ThisGISmodel,theFoodAccessRadar,was
designedbytheUKnationalgovernmentin2003,intendedtolocatefooddeserts
at localauthority level. IconstructedandappliedthemodeltoCambridgeshire.
Cambridgeshire was a unique opportunity because, without a previous food
access study, the County Councilwelcomedmy application of the FoodAccess
Radarasapracticalresource.Formypart,constructingthemodelenabledmeto
illuminatethewaythattheFoodAccessRadar’sresultsrepresentfooddesertsas
something objective, whist masking the subjectivities of the knowledge
production process. The dissertation explores how this seeming objectivity
might have had policy implications by building onmy experience of the Food
AccessRadar constructionprocess and throughkey informant interviewswith
thosewhopilotedthemodel.
This dissertation begins by providing an overview of the concepts and
approaches used in the study, discussing the utility of critical GIS as a sub-
disciplineinwhichtobasemyanalysisandoutliningtheexistingcritiquesofthe
food desert concept. I then offer an account of the Food Access Radar
6
constructionprocessanditsapplicationtoCambridgeshire,beforediscussingthe
influenceofGISasapowerfulvisualisingmediumontheUKfooddesertideain
governmentpolicyandtheimplicationsofitsinfluence.Finally,havingdiscussed
theeffectof conventionalusesofGIS toproject fooddeserts, suchas theFood
AccessRadar,IsuggestqualitativeGISasameansofbetterconceptualisingfood
accessthroughGIS.
2.LiteratureReview
2.1.ScienceandTechnologyStudiesandCriticalGIS
7
AcriticalanalysisoftheFoodAccessRadar,andtheeffectofGIStoprojectfood
desertsmoregenerally,departs fromanunderstandingof scientificknowledge
as socially constructed. Since Kuhn’s (1962) initial theorisation of social
involvementintheconstructionofscientificfact,adepthofliteratureinScience
andTechnologyStudies(STS)hasscrutinisedtheroleofactors(researchersand
policy-makersinmycase)inknowledgeproduction(LatourandWoolgar,1986),
whoare influencedby their socio-political valuesandenvironment (Sismondo,
2010). By pointing to the influence of these actors, STS is critical of scientific
objectivityassociatedwiththeScientificMethod,especiallywhenappliedtothe
social world. Instead, STS research argues that ‘objectivity’ is something
developed in the right social circumstances (Sismondo, 2010: 140). Inmaking
claimstoobjectivity,aresearcherisseparatedandelevatedabovetheobjectof
study (Haraway, 1988). This “god trick”, Haraway argues (1988: 582), is
problematic because it enables the researcher to cede responsibility and
accountabilityfortheresearch,hidingthesituatednessofknowledgeproduction
(1988:581).Ausefulwaytoconceptualisetheobjectifyingpowerofscienceand
technology is through the “black box” motif (Sismondo, 2010: 120).
‘Blackboxing’, as described succinctly by Latour, is “the way scientific and
technical work is made invisible by its own success. When a machine runs
efficiently,whenamatteroffactissettled,oneneedfocusonlyonitsinputsand
outputsandnotonitsinternalcomplexity.Thus,paradoxically,themorescience
and technology succeed, themore opaque and obscure they become” (Latour,
1999).Theblackboxmotifisusefulforthisdissertationbecauseitcanbeusedto
understand how the social and political processes that create food desert
knowledgearehiddenbehindaveneerofobjectivitywhenconstructedinGIS.
BeforediscussingcriticalGISasasub-disciplineofSTS,itisimportanttobriefly
outline GIS. Geographical Information Systems are computer systems used to
capture, store, check and display data at particular positions on the earth’s
surface. They emerged out ofmilitary technology in the 1960s andwere first
used to record landuse information inNorthAmerica (Tomlinson, 1962). The
term‘GeographicalInformationSystem’wascoinedbyDaceyandMarble(1965).
By the early 1990s, GIS had become commercialised by the Environmental
Systems Research Institute (ESRI) and widely adopted by governments and
8
other organisations (Nyerges et al., 2011). GIS literature remained (and still
remains) highly technical, with textbooks focussing on tool development and
skillbuilding(StMartinandWing,2007).
CriticalGISemergedasareaction tomainstreamGIS,and isoneapplicationof
STS in geography. The sub-discipline critiques the widely held assumption in
‘conventional’ Geographical Information Science (Openshaw, 1991; Goodchild,
1992; Elwood and Cope, 2009) that GIS is an unproblematic, positivist tool
producingobjectiveknowledge(StMartinandWing,2007;Pavlovskaya,2009).
CriticalGISattributes thispositivistunderstandingofGIS-producedknowledge
to the visualizing power of maps and the association of GIS technology with
science (Pickles, 1995; Pavlovskaya, 2006). Firstly, aswith conventionalmaps
(Crampton, 2001), the visualizing power of GIS is at the root of its fetishized
objectivity.Mapsmakespatialdataimmediatelyaccessibletoourmind,andthis
seductive featuremeans they carry a discursive power as both an expert and
accuraterepresentationofreality(Pavlovskaya,2006:2012;Elwood,2009:17).
Secondly, GIS amplifies the power of maps by placing them into the realm of
informationtechnology,whichvalidatesGISasascientificconstruct(Sheppard,
1993; Lake, 1993; Pavlovskaya, 2006). GIS, therefore, “unveils worlds to
researchers, policymakers and the public,worldsmade ‘true’ by the assumed
legitimacyofdataandvisualdisplays”(Pavlovskaya,2009:13).AtthelimitGIS
can“evenproducetheplacesthataretheretoknow”(ElwoodandCope,2009:
10).AnunderstandingofhowGIScancreate ‘truth’makescriticalGISauseful
framework throughwhich toexplore theeffectof representing fooddeserts in
GIS.ThepowerofGISisparticularlysalientinpolicymaking,throughtoolssuch
as the Food Access Radar, where “administrators and lawmakers are naively
unaware of the limitations of maps and other geographic data” (Monmonier,
1998).
Critical GIS takes issuewith the understanding of GIS as a positivist tool, and
arguesthatallGIS-basedknowledgeissociallyconstructed,situatedandpartial
(Sheppard,2001;Kwan,2002;Warren,2011).Socialconstructivistcritiquesgo
right down to the development of the algorithms that make up the software
(Chrisman, 2001), far deeper than this dissertation ventures. Most directly
9
relevant,arecritiquesoftherelationshipbetweentheGISuserandtheGraphical
User Interface (GUI), or what we see on the screen. Critical GIS researchers
critique theassumptionheldbyconventionalGIS that theGISuser isapassive
actorintheknowledgeproductionprocess,whoputsdataintothemachineand
gets factsout(Warren,2011).Assumptionof thispassiverelationshipremoves
the effect of social and political factors in the knowledge production process.
ProponentsofcriticalGISarguethattherelationshipismuchmoredynamic(Sui
and Goodchild, 2001, 2003): “in GIS the user is actively engaged in the
knowledgeproductionprocess.Whathe/sheputs inaffects theGISasmuchas
what it represents on the GUI” (Warren, 2011: 72). Therefore, the knowledge
thatGISproducesisaffectedbythevaluesoftheresearcherandtheenvironment
inwhichhe/sheis.Forthisdissertation,theorizationsofthe‘active’GISuseras
socially implicated (Elwood, 2006) problematize the objectivity of the GIS
constructionprocessthatcreatesknowledgeaboutfooddeserts.
QualitativeGISbuildsoncriticalGIStoproblematizetheconstructionofGISasa
necessarilyquantitativetechnology.ProponentsofqualitativeGISarguethatGIS
hasalwaysbeen‘non-quantitative’:attributingitsquantitativeportrayaltohow
ithasbeenhistoricallyconstituted(Kwan,2002;KwanandKnigge,2006).Asa
result, there are “openings” in GIS that “enable” qualitative research, data and
analysis(Pavlovskaya,2006:2003).Bytargetingtheseopenings,qualitativeGIS
“intersects GIS and qualitative research with the goal of integrating multiple
forms of evidence or ways of knowing, in order to explain how spatial
knowledge, patterns, relationships, and interactions are produced, and with
what sorts of social and political impacts” (Elwood and Cope, 2009: 6).
Qualitative GIS does not deny that some fixity is inherent to GIS (in that GIS
projections ‘fix’ knowledge and meanings at particular moments). Rather, it
engages with this fixity through strategic deployments and iterations of fixed
representations, to account for ‘situated knoweldges’ (Harraway, 1988).
Qualitative GIS also accommodates an understanding “that any effort to fix
meanings (or todisrupt them) is inherentlypower-laden, inseparable fromthe
performative, representational, or analytical practices through which these
meaningsareproduced” (ElwoodandCope,2009:8).Byapproaching the food
desert(andwiderfoodaccess)issuethroughaqualitativeGISlens,Ican,having
10
shownthatthereareissuesassociatedwithconventionalGISrepresentationsof
food deserts, propose ways to re-envision the Food Access Radar so that it
incorporatesasubjectiveandheterogeneousunderstandingoffoodaccess.
2.2.FoodDesertsYet,despiteunderstandingsincriticalGISandqualitativeGISofthecomplexand
situated relationship between the GIS software, the user and representations,
muchofconventionalGIScontinuestotreatGISasatoolforpositivistscience(St
MartinandWing,2007;Pavlovskaya,2009).DuringtheUKfooddesertdebate,
bothinpublichealthresearchandpolicy-making,theconventionalviewofGISas
a positivist tool prevailed. Researchers and policy-makers alike used GIS,
seemingly unproblematically, for spatial analysis of food store distribution
across fixed areas (Shannon, 2015). This spatial analysis had the intention of
provingordisproving theexistenceof fooddeserts (Donkinet al., 1999,2000;
Cummins andMacintyre, 1999; Clarke et al., 2002;White et al., 2004; O’Neill,
2005;Pearceetal.,2006;ElinderandJansson,2008).Althoughslightlydiffering
in the way they used the GIS software, these studies assumed that their
representationsoftheurban‘foodscape’(Lakeetal.,2010,2012),aswellastheir
understanding of the choices that consumers make, reflected reality; they
thereforeassumedthatthefooddesertstheyrepresented(ordidnotrepresent)
were objective. Market-based food desert studies, as distinguished from
geographical food desert studies by Beaulac et al. (2009), also rested on an
oversimplified and rational understanding of human choice and activities (for
example:Mooney,1990;Soomanetal.,1993).
Steve Cummins and Sally Macintyre were the two main academics to offer a
sustainedtheoreticalcritiqueofthefooddesertidea,ratherthanjustinvestigate
itsexistencethroughempiricalresearch.Bothconsistentlyrefutedtheideaofthe
food desert, analysing it variously as a convenient political construction for
policy-makers (Cummins andMacintyre, 2002a) and as an oversimplified and
problematicunderstandingoffoodaccessibilityandlivedexperience(Cummins
andMacintyre,2005;Cumminsetal.,2007).Theysupportedtheiranalysiswith
qualitative and mixed-method evidence of the complex relationship between
11
geographicalaccessibilityandfoodchoice(Whiteetal.,2004;Clarkeetal.,2004;
Cummins et al, 2005). They repeatedly suggested dropping the concept in
academiaandpolicybecauseitwas“oflimitedfurtheruse”(Cummins,2005:3).
For Cummins et al. (2007), the food desert model is problematic because it
departsfroma‘conventional’understandingofspace.Inthisunderstanding,the
space between places is interpreted as physical Euclidian distance to fixed
points,withgeographicalboundariesdrawnata specific scale.Throughsucha
conceptualisation of space, focussing on the single spatial scale of the
“neighbourhood”,proponentsofthefooddesertarguethatvariationinphysical
distance from food stores causes health inequalities (Cummins et al., 2007:
1833).Cumminstakesissuewiththis“localtrap”(Cummins,2007:355),arguing
that it is an oversimplification of food access that “assumes that individuals
behaveinrathersimilarwaysandrangeovertheirneighbourhoodsinalimited
manner to purchase food items” (Cummins et al., 2007: 1833). The
oversimplification is problematic and unproductive, for Cummins and others,
becauseitcreatesabinarybetweentheroleofindividualsandtheroleofplace
(or context) as explanations for health inequalities, with food deserts
championing place. The concept therefore misunderstands the nuances and
complexityoffoodaccess.Theysuggestapproachingfooddesertstudiesthrough
a ‘relational’ approach that can capture the “tight interrelationships between
individuals and context” (Cummins et al., 2007: 1829). The theorisation of
‘conventional’ relationships across space is important for my dissertation
becausetheFoodAccessRadar,andotherfooddesertprojectionsinGIS,employ
thisunderstandingofspace.So,fromthisbasis,Icanarguethatconceptualising
fooddeserts inGISreinforcestheunderlyingunderstandingof foodaccessasa
function of ‘conventional’ space, giving furtherweight to the food desert idea.
Further, Cummins et al.’s ‘relational’ approach to food access is useful to re-
envisionaqualitativeGISrepresentationoffoodaccess.
CumminsandMacintyre(2002a)alsocritiquedtheevolutionofthefooddesert
metaphor in government policy. For them, the food desertwas a construction
arising out of a convenient political context: a “factoid” created in UK
governmentbecause“it fitted inwithbroaderpolicyobjectives”(Cumminsand
12
Macintyre,2002a:438). In the late1990sa shift occurred inhealth-promoting
activity towards spatial strategies (Lang and Caraher, 1998), reflecting wider
trends towards area-based anti-exclusion policies in theUK (Mohan, 2002). In
parallel,the1997NewLabourgovernmentmadeacommitmenttofightingsocial
exclusion (Social Exclusion Unit, 2004). In this political context, the
‘construction’ of food deserts, and a commitment to ‘eradicate’ them, was a
convenientwayofmeetingbothsocialexclusionpoliciesandmeetingarea-based
targets.
Thefooddesertwasalsoparticularlyappealingbecauseitwas“aplausibleand
attractive theory with seemingly straightforward solutions” (Cummins and
Macintyre, 2002a: 437). The simple causal relationship between food deserts
and health/exclusion outcomes meant the concept was well suited to New
Labour’s new policy paradigm of ‘evidence-based policy-making’ (EBPM) for
socialissues(Fafard,2015).Intendedtomodernisethedecision-makingprocess
for social policy (Wells, 2007), EPBM made policy-making dependent on
identifyingaprobleminordertosolveit.Forfooddeserts,policy-makers,armed
withnewGISsoftware(Monmonier,1998),couldprovideevidenceby‘locating’
areassufferingfromlowaccessbeforeimplementingstrategies.TheFoodAccess
Radar, aswill be shown, is a clear exampleof one suchgovernment-led effort.
However, the opening of a political space in which the food desert idea was
rapidly accepted is clearly problematic when food deserts are understood as
oversimplifiedconceptualisationsoffoodaccess.Whenreallivesareatstake,the
resting of a concept on a plausible and attractive theory is clearly not enough
(MacintyreandPetticrew,2000).
13
3.MethodologyThe main method I employ to explore GIS-produced knowledge about food
desertsisconstructingafooddesertmodelmyself.Idosoinordertounpackthe
technological ‘blackbox’ thatGISproduces.As theSTS theoryof ‘blackboxing’
(Sismondo,2010)appropriatelydescribes,GIShasatendencytoobscurepartial
andsituatedknowledgeproductionprocesses. In thecaseof fooddeserts,only
the inputs (data) and outputs (visualised food deserts) are clear, whilst the
esoteric GIS processes remain invisible (Sismondo, 2010: 120). Therefore, to
uncover the social and subjective process involved in the construction of food
desert knowledge inGIS, and understand theway this knowledge can become
treatedasobjective,itisnecessarytointerrogatehowGISmodelsarebuilt.The
GISblackboxinthisunderstandingistreatedlikeanaeroplane’sblackbox.Asa
14
technicianwoulddo,IlookinsidetheGISblackboxtoseewhattheproblemsare
withprojectionsofthefooddesertconcept.Insum,bycreatingtheFoodAccess
Radar,Iamabletotakeaparttheprocess,unpackingthe“blackbox”(Sismondo,
2010)inordertolookinsideit.
InChapter4IhaveproducedanaccountoftheFoodAccessRadarconstruction
process.Throughout theprocess Ikeptadiaryofall thesteps Imade towards
creatingtheoutput, frominitialsourcingofdatatoanalysingthefinalresults. I
recorded comments and reflections on any assumptions, amendments and
simplificationsmade tomake the software ‘work’ (Warren, 2011).Overall, the
construction process lasted two months, from July to September. Although it
wouldhavebeenpossibleforanexperiencedGISprofessionalworkingfulltime
toconstructthemodel inunderaweek,IhadtoteachmyselfGIS, familiarising
myself with the software, jargon and processes. This made the construction
processconvolutedandarduous.Locatingappropriatedataalsoprovedtobea
longprocess.
InthecontextofunveilingtheinnerworkingsoftheGIS ‘blackbox’,mylackof
previous GIS knowledge meant it was important to learn how to use the
software.AcommonthemerunningthroughthefieldofcriticalGISisthattogain
anuancedandcriticalunderstandingofthesubjectivitiesinvolvedinknowledge
production,onemustengagewiththeGIS itself (Kitcher,1998;Warren,2011):
look upon itwith an ‘insider-gaze’ (Elwood, 2006).Without the ‘vocabulary of
thetechnology’(Kwan,2002)itisimpossibletomakeaninformedcontribution
to any debate. With no previous GIS knowledge, it was therefore entirely
necessary forme to learnGISbycreating theFoodAccessRadar. In sodoing I
stepped away from being what Openshaw aptly termed a “technical cripple”
(1991:464),towardsbeingabletodevelopeffectivecritiquesofGIS’sroleinthe
food desert debate, on its own terms (Sheppard, 2001). As a result, all maps,
unlessotherwisestated,aremyown.
I chose to use the Food Access Radar model rather than other models for a
numberofspecificreasons.Firstly,CambridgeshireCountyCouncilinformedme
that no countywide food access study had been carried out so far,makingmy
15
studyoriginalanalysis.Secondly,asagovernment-createdmodel,theknowledge
thattheFoodAccessRadarproducedaboutfooddesertswouldhavehadamore
direct impact on policy than knowledge produced by academic models,
especially so if results were treated as evidence in an ‘evidence-based policy-
making’paradigm.Warren(2011:82)arguesthatexposingthesubjectivitiesof
“so called non-enlightened studies”, a category under which the Food Access
Radar falls, is an essential part of critical GIS. Thirdly, the different scales at
which the government model was designed and applied gave room for
misinterpretation, which merits further evaluation. Having been created by a
national body, but designed for local authority use, there was a risk that the
social context of results could bemisunderstood in local policy. Fourthly, on a
more pragmatic level, itwas possible forme, as a GIS novice, to construct the
FoodAccessRadarthankstotheavailabilityofconstructionguidelinesprovided
bythegovernment(ViningandDay,2005),whichhelpedmydatasearchesand
GISlearning.
To build upon the interrogation of the GIS model, I also conducted two key
informant interviews. These were with informants from Staffordshire and
OxfordshireCountyCouncil:theonlytwoCountyCouncilsforwhomIcouldfind
evidence of them having used the Food Access Radar. Both informants were
directlyinvolvedintheirrespectivefoodaccessprojectsin2005and2006.Given
thattheseprojectswereoveradecadeago,theinterviewsweredesignedtoadd
contexttomystudy,abouthowGIS,theresultsthatitproducedandtheconcept
of food deserts were understood at the time. They were carried out over
telephone, lasting approximately 25minutes andwere semi-structured.While
core interview questions were “predetermined” they remained “open ended”
(Ayres2008:810),allowingtherespondenttofullyexplorethequestionwhilst
allowing the researcher the flexibility to interrogate emerging themes
(Flowerdew and Martin, 2005). The interviews were recorded with verbal
consent and subsequently transcribed (See appendix). Information garnered
fromtheGISconstructionprocessandtheinterviewswasalsocomplementedby
analysis of the government-provided construction guidelines (Vining and Day,
2005), lookingat thewaycertainphrases couldbemisinterpreted to conveya
falsesenseofobjectivityintheGIS.
16
4.TheFoodAccessRadarconstructionprocess
In this section I explain the construction of the Food Access Radar in GIS.
Initially, Iprovideanoverviewof theFoodAccessRadarasagovernmenttool,
before explaining the construction process in a step-by-step manner. The
constructionprocessisakintounpackingSismondo’s‘blackbox’(2010:120).
TheFoodAccessRadarisastandardisedmethodforidentifyingfooddesertsata
local authority level, designed for construction in GIS software. The Food
StandardsAgencyandNationalConsumerCouncil(NCC)devisedtheGISmethod
in 2003, amid growing government interest in food deserts as a contributing
factortosocialexclusionandhealthinequalities.Itsaimwastocontributetothe
government’s commitment to eradicate food deserts, as specified in the Food
Poverty Eradication Bill in 2001 (O’Neill, 2005). The NCC, a national-level
organisation,producedconstructionguidelinesforlocalauthoritiestofollow,to
17
encourage them tocarryFoodAccessRadar studiesata county leveland then
useresultstoinformlocalpolicy.
TheFoodAccessRadar labelsparticularOutputAreas, the lowestgeographical
levelatwhichgovernmentdata iscollected,as ‘fooddeserts’ if they fulfilgiven
accessibility criteria (referred to as ‘food-access-poor’ areas in the guidelines).
Three parameters are used tomeasure physical accessibility to food: distance
fromfoodoutlets,distance frompublic transportandtheprevalenceofaccess-
inhibiting characteristics in each Output Area. By overlaying these three
variables in GIS, the guidelines claim that the model can locate Output Areas
(OAs) in a local authority that suffer from accessibility issues resulting from
geographical distance to food outlets and public transport, and where a high
occurrenceofindividualsexhibitaccess-inhibitingcharacteristics.
Forthisdissertation,IdecidedtoconstructtheFoodAccessRadarforthecounty
of Cambridgeshire, as a basis fromwhich to interrogate ‘objective’ knowledge
productionthroughGISinthecontextoffooddeserts.IappliedtheFoodAccess
RadarmethodtoCambridgeshireforseveralreasons.Asnotedinchapter3,no
FoodAccessRadarstudyhasbeenundertakenforthecounty,makingmystudy
original research. I am also familiar with the county, making it easier to find
contacts fordataandhelpingme interpret themapsanddata Iwas lookingat.
Finally,asshowninfigure1a,althoughCambridgeshireisagenerallyprosperous
county, there are pockets of deprivation (Figure 1b) and it experiences a
significantnorth/southdivide(Figure1c).Itwasinterestingtoseewhetherthe
Food Access Radar also identified these areas as food-access-poor, supporting
‘deprivationamplification’hypotheses(Cummins,2007:355).
18
IwillnowdescribeandexplaintheFoodAccessRadarconstructionprocessina
step-by-stepmanner,recordedasaflowchartinFigure2.Significantly,although
theflowchartpresentsthestepsasdiscreteandtheprocessaslinear,thisvastly
oversimplifiestheconstructionprocess.
Step8:Identifying‘food-access-poor’outputareas
Step7:Identifying‘highconcentration’outputareas
Step6:Amalgamationoflayersintoonemapdocument
Step5:Masstransitdataprojectionandnetworkanalysis
Step4:Foodoutletdataprojectionandnetworkanalysis
Step3:IntegratedTransportNetworkprojectionandgeodatabase/networkdatasetcreation
Step2:CensusdataprojectioninGIS
Step1:Datacollection(Census,foodoutlet,publictransport,IntegratedTransportNetwork)
Figure1Threemaps(1a,1b,1cfromlefttoright)showingdeprivationlevelsacrossCambridgeshireusing2015IMDdataattheLSOAlevel.Figures1band1crepresent
theredsquaresinFigure1a.Figure1bshowspocketsofdeprivationinnortheast
CambridgeCity.Figure1cshowshighratesofdeprivation,atnationallevel,inNorth
Fenland(Source:Author)
Figure2FlowchartsummarisingtheFoodAccessRadarconstructionprocess(Source:Author)
19
Toprojectthethreekeyvariables(Censusdata,foodoutletsandpublictransport
stops)inaGIS,IfirsthadtocollectappropriatedataforCambridgeshire(Step1
in Figure 2). The FoodAccessRadar construction guidelines provided a list of
datasets needed, but only limited information onwhere these could be found.
Dataonaccess-inhibitingcharacteristicsattheOutputArealevelwasextracted
froma governmentdatabaseof2011Censusdata (InfuseData).Theguidelines
suggestedusingsixsocio-economicanddemographic indicatorsofaccessibility
issues: lack of car ownership, lone pensioners, lone parents, being sick or
disabled, BME and income level. With no freely accessible sources of income
data,Iused‘highestlevelofqualifications’asaroughproxyforincome.Secondly,
I created a dataset containing the location of all 3461 bus stops in
Cambridgeshirebyeditinganationaldatasetofbusstops,downloadedfromthe
Government open database (data.gov.uk). The final dataset required was one
containingallfoodoutletssellingfoodforhomepreparationandconsumptionin
Cambridgeshire. The most comprehensive source of this data is the ‘Food
Premises Register’ held by local authority Environmental Health (EH)
departments.Dataprovedparticularlyhardtoobtain. Icontactedemployees in
EH departments at the five District Councils in Cambridgeshire, but was only
able to obtain three of the five datasets, despite filing Freedomof Information
requests.CambridgeCityCouncil,forexample,askedmetopay£800,whichwas
notfeasible.Instead,inneedofuniformdatasetsacrossthecounty,Idecidedto
usedatafromtheFoodStandardsAgencyFoodHygieneRatingScheme(FHRS).
TheFHRSisapubliclyaccessibleregisteroffoodpremisesandrepresentedthe
next best alternative to theEHFoodPremisesRegister.However,manyof the
addressesintheFHRSwereincompleteornon-existent,sobeforegeocodingthe
data,significantcleansingwasrequired.
Once the datasetswere cleansed and geocoded, Steps 2-5 involved visualising
dataintheGISsoftware.IusedESRI’sArcGISsoftware(ArcMap10.4),despiteit
beingless‘user-friendly’thanothersoftware,becauseitistheindustry-standard,
usedbymostotherfoodaccessstudies(Charreireetal.,2010).Inowexplainthe
constructionprocesswiththehelpofscreenshotsfromtheArcGISGUI.Asshown
20
bytheredsquareinFigure3,thescreenshotsrepresentaparticularsubsection
ofCambridgeshireasadataframeatthescaleof1:100,000.
After having visually represented the county, district and Output Area
boundaries,sourcedfromgovernmentopendata,Step2involvedprojectingthe
CensusdataintoGIS(Figure4).
Figure3MapoflocalauthorityboundarylinesinCambridgeshire.Theredsquareshowsthedataframeusedinthenext5screenshotstovisuallyexplainthe
constructionprocess(Source:Author)
21
Data for each Census characteristic likely to cause food access problems, as
suggested by the guidelines,was visualised in a choroplethmap. Each dataset
wasmappedproportionatelyinto10deciles,meaningeachcolourinthegradient
represented a 10th of Output Areas. So, for example, Figure 4 shows the
percentage of the population in each Output Area who do not own a car,
separatedintotendeciles,witheachdecilerepresentinga10thofOutputAreas.
Theareasof interest fortheFoodAccessRadarwerethose inthetopdecileas
they indicated areaswith high concentrations of peoplewith access-inhibiting
characteristics.
Toprojectandmanipulate the twoothervariables, foodoutletdataandpublic
transport data, into useful map layers, the Ordinance Survey (OS) Integrated
TransportNetwork(ITN)wasrequired(Figure5).
Figure4ArcGISscreenshotshowingCensusdatamappedattheOutputArealevel,classifiedintodeciles,witheachdecilerepresentinga10thofOAs.Inthiscase:
highesttolowestcarownershipfromyellowtoblue(Source:Author)
22
The ITN is anOSdataset containingdetails of the transportnetwork forGreat
Britain and, when projected, provides a virtual representation of the UK road
network,constructedasaseriesofpoints,andlinesbetweenpoints.Importantly
fortheFoodAccessRadar,theITNprovidesmeterlevelaccuracyforroadlength,
sodistancesandtimestakentotraveltofoodoutletandpublictransportstops
couldbecalculated.
The846foodoutletssellingfoodforhomepreparationwereprojectedontopof
the ITN in a point shapefile (Figure 6). Several technical adjustments to the
dataset were required in ArcMap to ensure that the food outlet data
superimposeddirectlyontopoftheITN.Todifferentiatebetweentypesoffood
stores,supermarkets(113points)wererepresentedbyabluepointwhilstgreen
pointsrepresentedallotherfoodoutlets.
Figure5ArcGISscreenshotshowingIntegratedTransportNetwork(ITN)shapefile.ITNdownloadedfromOSresourceindigimap.com(Source:Author)
23
Since the purpose of the Food Access Radar is to identify areas that are
geographically distant from food outlets, and are thus at risk of being food-
access-poor,walkingdistancesfromeachfoodstorewerecalculated.Asoutlined
bytheguidelines, ten-minutewalking-timepolygonswerecreatedaroundeach
foodstoreusingtheServiceAreafunctionoftheNetworkAnalystextensiontool
(shownbyredpolygonsinFigure6).Thesepolygonsshowhowfaranindividual
can travel along the road network in 10minutes walking at an average pace,
nationallydefinedas81metersperminute(Knoblaughetal.,1996).Fromthisit
is possible to locate pockets of population that live outside of these polygons,
who are therefore geographically distant from food outlets. The Food Access
Radarmethodthencombinesthisinformationongeographicaldistancewiththe
accessibilityinformationfromtheCensusandpublictransportdata.
Thesameprocessofwalking-timepolygoncreationwasemployedforbusstop
andtrainstationdata(Figure7).LearninghowtousetheNetworkAnalysttool
and creating an appropriate ‘network dataset’ from my data was the most
complexandtime-consumingpartoftheresearch.
Figure6ArcGISscreenshotshowingfoodoutletpointdatasetanditsassociatedwalking-timepolygondataset,overlayingtheITN(Source:Author)
24
With maps for all three variables projected in ArcGIS in accordance with the
construction guidelines, Step 6 involved overlaying them in the same GIS
document(Figure8).
Having completed the layer construction process, I subsequently carried out
analysis to locate Output Areas in Cambridgeshire that could be classified as
‘food-access-poor’ (Step 7 and 8 in Figure 2). The construction guidelines
Figure7ArcGISscreenshotshowingpublictransportpointdatasetanditsassociatedwalking-timepolygondataset,overlayingtheITN(Source:Author)
Figure8ArcGISscreenshotshowingfoodoutletpointdatasetandpublictransportpointdataset,andtheirassociatedwalking-timepolygondatasets,overlayingthe
ITNandCensusdatamappedatOAlevel(Source:Author)
25
suggestedthat‘food-access-poor’areasshouldbedefinedasOutputAreaswitha
high concentration of people with access-inhibiting characteristics, which are
alsonotwithinfoodstorepolygonsorbusstoppolygons.
Step 7 involved identifying the ‘high concentration’ Output Areas in
Cambridgeshire.TodothisIusedStructuredQueryLanguage(Figure9)toselect
the Output Areas that, as recommended by the guidelines, had at least two
Censusindicatorsintheworstperforming10%ofallOutputAreasinthecounty.
From this Iwas able to create a newdata layer of ‘high concentration’Output
Areas,reducingthenumberoftargetOutputAreasfrom1937to214.
I then (Step 8), manually, but systematically on the GUI, selected those ‘high
concentration’OutputAreasthatalsohadasignificantproportionofhouseholds
fallingoutside the foodoutletpolygons. I achieved thisbyprojectingOSraster
maps at the 1:25,000 and 1:50,000 levels as basemaps so I could distinguish
households. I selected those Output Areas meeting both criteria, as shown in
Figure 10, and from these made a new GIS layer: ‘food-access-poor’ areas
(Example in Figure 11). In total I identified 8OutputAreas in Cambridgeshire
thatwere‘food-access-poor’andtherefore,ingovernmentdiscourse,were‘food
deserts’.
Figure9ScreenshotofattributetableforthefoodoutletdatasetinArcGIS,showingequationinqueryboxusedtoidentify‘high-concentration’OAs(Source:Author)
26
Theconstructionofthefooddesertmodeldescribedabovewasanecessarystep
towards explaining the influence of GIS on the food desert debate. By
constructing the Food Access Radar I could get inside Sismondo’s ‘black box’
(2010:120):enablinganobservationoftheeffectofGISonfooddesertsthrough
an‘insidergaze’.Thepositionoftheresearcherrelativetothetechnologyisakey
Figure10ArcGISscreenshotshowing‘highconcentration’OAdatasetasbluepolygonsandfoodoutletwalking-timepolygonsinred,overlyinga1:50,000raster
OSmap(Source:Author)
Figure11ArcGISscreenshotshowingpurple‘food-access-poor’OAcircled,alongwithfoodoutletpointsandpolygons,overlyinga1:25,000rasterOSmap(Source:
Author)
27
focusofcriticalGIS(Elwoodetal.,2011).Positioningmyselfasan‘insider’tothe
technology, therefore, allows me to critique GIS on its own terms (Sheppard,
2001:549),usingthelanguageofthediscipline(Schuurman,2000:569).
Explanation of the process demonstrated for me that the Food Access Radar
construction is not such a linear, scientific process, as mainstream GIS would
have it (Elwood and Cope, 2009). The Food Access Radar was constructed
throughaseriesofinteractionsbetweentheGISsoftwareandmyself,astheuser,
wheresubjectivestepswerenecessarytomaketheGIS ‘work’(Warren,2011):
for example, Step 8 involved manual selection of ‘food-access-poor’ Output
Areas.Eventheaboveaccountissimplifiedandmakestheconstructionprocess
appearsequential,withaninitialaimandafinaloutputthatisworkedtowards
throughastep-by-stepprocess.Therealitywasverydifferentandquite‘messy’.
Manytimesthereweresetbacks,datahadtobereworkedto fit thepurposeof
the model and the process did not proceed in stages, but in an iterative and
piecemealfashiondependingonavailabilityofdata.Youcouldarguethiswould
beasmootherprocessifdonebyagovernmentbodyorlocalauthoritywithall
theinformationathand,buttherealityisthatdatagetsshapedandfilteredtofit
theprocess.
However,apartfromtheimperfectionsoftheprocess,thebiggerlessonIlearnt
from being ‘on the inside’ was how selective and simplified the inputs are.
Despite the impressive advances of digital cartography and the power of the
softwaretoassimilateandtransposedenseinformationintoengagingmaps,the
outputswere all basedon a very limited rangeof data inputs and some crude
assumptions. I followed the guidelines, but was left with a realisation of the
partialityofthedata,thearbitrarinessofquantitativecut-offsandthesimplistic
assumptionsunderlyingthealgorithmsthatdrivetheprocess,whichIdevelopin
thenextchapter.
Yet, in thecontextof theFoodAccessRadar’sapplication ingovernment, these
inconsistencies are not what the policy-makers see. In an interview with a
member of the GIS research team at the Staffordshire County Council, for
example, I was told that the research team that constructed the Food Access
28
Radarwerenotinvolvedinthesubsequentpolicyteamthatusedtheresults:“I
usedtogetstuckinadarkroomattheresearchunittodoallthisstuff,thatother
peopleusedtotakeon.Itmaybethatcolleaguesofminearemorewellplacedto
saywhatactuallyhappened[withtheresults]”.Thus,forthosepolicy-makersthat
used the GIS-created knowledge, the process remained within the ‘black box’.
Thesepolicy-makersseeonlyapowerful representationof fooddeserts,which
canmake the food desert concept seem like something quite real. In the next
section,Isuggestthat,inthecontextofUKfooddesertpolicyandresearch,food
desert GIS models such as the Food Access Radar had an influence on the
longevityandcredibilityofthefooddesertidea.
5.TheinfluenceofGISonthefooddesertdebate
29
AsCumminsandMacintyrehaveconvincinglyargued(CumminsandMacintyre,
2002a; Cummins et al., 2007; Cummins, 2007), the food desert concept lacks
nuance when portraying food access. It reduces the complexity of social
behaviour down to an assumption that humans act homogeneously and
rationallywiththeirfoodchoices.Prioritisinga‘conventionalview’ofspaceover
a ‘relational view’, it silences the socio-cultural factors that determine food
accessbyrepresentingonlyphysicalaccessoverEuclideandistance.Repeatedly,
CumminsandMacintyresuggestedthatthefooddesertconceptwasof“limited
further use” (Cummins et al., 2005: 2) and an “idea whose time had come”
(CumminsandMacintyre,2002a)
Despite the flaws identified in the concept, the food desert idea remained a
legitimatemeansofdirectingUKgovernmentsocialexclusionandhealthpolicy
foraroundadecade.In1995theLowIncomeProjectTeamoftheNutritionTask
Forcefirstcoinedtheterm(Wrigley,2002:2030),whilst, in2005, fooddeserts
were still considered significant enough an issue to recommend that all local
authoritiesshouldcarryoutaFoodAccessRadarstudy(ViningandDay,2005:
49).Betweentheseperiods,apolicyteamhadthespecificmandateofdeveloping
astrategyforimprovingaccesstoshoppinginpoorneighbourhoods(Wrigleyet
al.,2003)andtheFoodPovertyEradicationBill,withclauseson foodaccessat
the neighbourhood level, was passed through the Commons (Cummins and
Macintyre, 2002a). Even as late as2013, theBristol CityCouncil FoodPoverty
ReportsuggestedtheuseofFoodAccessRadarasaGIStool forassessing food
poverty(Malsenetal.,2013).Furthermore,duringthedecade-longperiodfrom
1995therewerefewacademicorpolicyresearchersthatquestionedthevalidity
of the fooddesertconceptand those thatdidrepresented “more theexception
thantherule”(Shannon,2014:256).
This section seeks to address why the misguided food desert idea gained
purchaseinUKgovernmentandacademiaforsolong.Althoughthefactthatfood
desert policies alignedwith the social exclusion interests of the 1990s Labour
government (Mohan, 2002; Cummins and Macintyre, 2002a) is significant, I
suggest that the ability to represent food deserts through GIS played a role. I
argue that the convenience of GIS representations for policy-makers in the
30
context of an evidence-based policy-making paradigm, in combination with
objectifyingpowerofGIS technology, gave the fooddesert conceptaveneerof
objectivity in policy-making spheres. Throughout this chapter, I use examples
fromtheFoodAccessRadarandevidencefrommykeyinformantinterviewsto
arguemycase. I suggest that the interplayof three factorsmake the fetishized
objectivity of GIS projections of food deserts convincing in UK government
policy,despitetheheterogeneityofthesocialworld,wherefoodaccessismuch
more complicated than physical distance alone. Ramsey’s quote succinctly
capturestheeffectofGISonUKfooddesertpolicydiscussedbelow:“GISisnot
merelyanobjectivetoolfortheunbiasedanalysisandrepresentationofspatial
phenomena.Rather, it isasystemoftechnologyandsocialpracticesthat isnot
onlyshapedby itssocial,cultural,political,and institutionalcontext,but italso
activelyshapesthatcontext”(Ramsey,2009:2348,myemphasis).
5.1Simplicityandrelevancetoevidence-basedpolicy-making
ThefirstreasonGISwasimportantinthefooddesertdebateisthatitfacilitated
theuseof thesimple fooddesertmechanismwithinan ‘evidence-basedpolicy-
making’ (EBPM) paradigm. EBPMwas a significant government reform in the
1990s. The 1997 New Labour government introduced EBPM in order to
‘modernise’thedecision-makingprocess(Wells,2015).Byextendingthealready
established concept of ‘evidence-basedmedicine’ to all policy areas, including
socialpolicy,decisionswere tobemadeon thebasisof rigorouslyestablished,
‘objective’evidenceandwerethereforeconsideredtobemoreaccurate(Fafard,
2015). However, critics have attacked the apolitical, ‘scientised’ way evidence
informs social policy. Sanderson has branded EBPM as a return to positivism
(2002:6).Nolan(2015)andothershavecritiquedtheassociationsofEBPMwith
neoliberalism, in that EBPMenables the government to cede responsibility for
theirdecisionsbymakingclaimstothelegitimacyofpositivistscience.
In the context of the simple food desertmechanism, EBPM created a space in
which theuseofGISas ‘evidence’of fooddesertexistence legitimised the food
desertconcept.Thesupposedcausalmechanismforresolvingfooddesertissues
31
wasalreadyattractiveasaresultofitssimplicity(AdaptedfromCumminsetal.,
2007:1833):
1. Lack of healthy, affordable food affects food consumption, which has
healthanddeprivationimpacts
2. Peoplelivetoofarfromhealthy,affordablefood
3. Therefore,increasingaccesstofoodcanreducefoodconsumptionissues
andsocialexclusion.
Within the context of an EBPM paradigm this mechanism is particularly
appealing.This isbecause, forpolicytobe implemented,onehasonlyto locate
theseareasofpoorfoodaccesstoprovideevidenceoftheirexistence,andthen
implementstrategieswithinthemtoincreaseaccess.Locatingtheseareaslends
itself to GIS. The Food Access Radar is a case in point. In my study of
Cambridgeshire, for example, through the process of creating the Food Access
Radarmodel I reduced the number of possible food-access-poor output areas
from 1937 to 8. From a government perspective, this is a highly attractive
prospect because it reduces the number of areas that policy makers need to
focus on and, resting on the assumption that the Food Access Radar is an
accurate representation of reality, provides evidence of food desert existence.
Therefore, theFoodAccessRadar epitomiseshow the growthof democratized
GIS software facilitated the uptake of the food desert idea within a specific
political context. GIS provided a convenientmeans to collect evidence of food
deserts,makingthefooddesertideaparticularlyattractiveingovernmentwithin
anEBPMparadigm.
5.2.GIS,fooddesertsand‘conventionalspace’
Building on the convenience of GIS software for providing evidence of food
desertsinaUKpoliticalcontext,GIScanalsobeunderstoodasaconvenientway
ofportrayingfooddesertsatthesoftwarelevel,whichisthesecondreasonwhy
GISwasimportantforthefooddesertdebate.Thisisbecausebothfooddeserts
and GIS software are grounded in the same understanding of ‘conventional
space’.InmainstreamunderstandingsofGISsoftware(Pavlovskaya,2006),data
is plotted as Cartesian coordinates and the space between places is
conceptualised inEuclidianterms. Inotherwords,onamapwithmanypoints,
32
the two points that are ‘nearest’ each other will have the smallest physical
straight-linedistancebetweenthem.
The food desert concept starts from the same fundamental understanding of
space:termeda‘conventionalview’ofspaceandplacebyCumminsetal.(2007:
1827). In thisviewof space, theproximityof thehousehold toa foodoutlet is
determinedentirelybythephysicaldistancebetweenthetwopoints,measured
throughCartesiancoordinatesinEuclidianterms.IntheFoodAccessRadar,for
example, access is conceptualised through thewalking-time polygons: a direct
measurementofthedistancewalkedintenminutesfromashopalongtheroad
network. In themodel, a fundamental determinant of food-access-poor output
areas are that they fall outside ofwalking-time polygons. In so doing, through
quantification of food access in spatial terms, food deserts create a binary
betweenthosewhohaveaccessandthosewhodonot.IntheFoodAccessRadar,
forexample,thelimitofthefood-access-poorareaistheboundaryoftheoutput
area.Thiseffectivelysilencesmorecomplexunderstandingsof foodaccessthat
relyona ‘relational’understandingofspace(Cumminsetal.,2007:1833). Ina
relational understanding, the distance between places is understood in socio-
cultural terms, where the social power relations imbued in each place are as
muchadeterminantofaccessasphysicaldistance.Conceptualisingthisnatureof
foodaccessrelationshipreliesonamorecomplexandnuancedinterpretationof
GIS,discussedinChapter7.
By reducing food access to physical distance in such a way, the food desert
concept is clearlyvery compatiblewith traditional representation throughGIS,
asbothareunderlainby‘conventionalspace’.GISthereforeinherentlyfacilitates
the projection of physical distance between points as the determining
characteristicof foodaccess.TheappropriatenessofGISprojectionsmakesGIS
an even more appealing and legitimate tool in the eyes of policy-makers
representing food deserts. GIS being more appealing leads to more people
representing fooddeserts throughGIS.Asmorepeople represent fooddeserts
through GIS, themore the idea that food access can be simplified to the food
desertconceptisreproduced,givingthefooddesertcredibilitybysilencingother
interpretations of food access. Reproduction as a result of a common
33
understandingofconventionalspacethereforepresentsonemechanismthrough
whichaveneerofobjectivitycouldhavebeencastoverthefooddesertconcept.
ThepowerofmapsandthevalidationofGISasascientifictechnology,discussed
inthenextsection,cementthiscommonunderstandingconventionalofspace.
5.3Objectifyingpowerofmapsandvalidationasascientifictechnology
InthelasttwosectionsithasbeenarguedthatGISlegitimisedandincreasedthe
appeal of food deserts in UK government policy because of the appropriate
politicalcontextandsoftware-levelcompatibility.Thethirdmainreasonforthe
purchase of GIS in the food desert debate was to do with appearances and
perception; I argue the power of GIS representations and the software’s
validationastoolsofscientificprogressgavethealreadyconvenientfooddesert
concept a veneer of objectivity. This veneer of objectivity legitimises the food
desert concept, silencing other interpretations of food access and hiding the
‘messy’blackboxofGISfooddesertcreation.
GIS gets much of its seductive power from the power of representation
associated with maps (Crampton, 2001; Pavlovskaya, 2006). Elwood explains
succinctlythat,“asmapsarepresentedandperformed,theyshapethemeanings,
identities, and characteristics that individuals and groups may assign to
individualplaces,andevenproduce theplacesthataretheretoknow”(Elwood
andCope,2009:10,myemphasis).ElwoodandCopealludetothewaythatmaps
cancreatetherealitiesthattheyseektorepresent.Inthecontextoffooddeserts,
representingthemthroughGIScantherefore‘bringthemtolife’,asitwere.GIS
takesthe inherentlysubjectiveconceptof foodaccessand, throughsilencingof
the partiality and situatedness of data, creates a realitywhere the food desert
becomesanobjectiverepresentationoffoodaccess.TheFoodAccessRadarisa
caseinpoint.Thefinaloutputlayeroffood-access-poorareas(seeFigure11in
previouschapter)is,totheuninformedobserver,aconvincingrepresentationof
those areas that are likely to suffer from food access problems. The complex
knowledge production process involved in the Food Access Radar, including
subjective steps taken by the researcher, partial data, and the setting of
quantitativecut-offs,is“blackboxed”:lostinthefinalrepresentation.Therefore,
34
inanenvironmentwherethefooddesertconceptisalreadyattractivebecauseit
meetsparticulartargets,thefinalfooddesertmapproducedbytheGIScouldbe
interpretedasanentirelyobjectiverepresentationoffoodaccess.
To compound the power of maps, the widespread faith in GIS as a scientific
technology(Pavlovskaya,2009;StMartinandWing,2007)thickens theveneer
ofobjectivityoverfooddeserts.Theinitialquantitativedatathatisinputintothe
model “has enormous cachet amongst policy makers” (McGuirk and O’Neill,
2012)totheextentthatpolicymakersdonotconsiderwhatis insidetheblack
box.Theyassumethattheoutput, fooddesertknowledge, is legitimatebecause
the input is quantitative data from ‘reliable’ sources: Census data and
government databases. Indeed, in an interview with a representative of the
OxfordshireCountyCouncilIwastoldthatCensusdata“iswhatitis.It’snational
statistics.Youcan’targuewiththat…”.Additionally,theteamusingthedatainthe
OxfordshireFoodAccessRadar,asintheStaffordshirecase,wereseparatefrom
the team creating the GIS. The interviewee, part of the policy team using the
results, described theGISprocess as “wizzy” and “technical”, in almostmagical
terms.TheinterviewsvalidateSchuurman’sclaimthat“mostuserstreatdataas
ifittheywerethetruthabouttheworld”(Schuurman,2009:42),atleastinthe
context of the Food Access Radar. Furthermore, even at a semantic level, the
FoodAccessRadarappearstobeassociatedwithpositivistscience. ‘Radar’ isa
scientific technology used to survey the physical environment. Application of
radar toahumancontext appears to imply that the resultsof theFoodAccess
Radar will also be objective. Therefore, the association of GIS with scientific
technology, combined with trust in quantitative data, enhances the veneer of
objectivity that veils the subjectivityof the fooddesert concept, especially in a
policyenvironmentwherequantitativeevidenceisrequiredforEBPM.
In sum, the argument above, collating evidence from my Food Access Radar
studyandcriticalGIS literature,suggests thatGISmodelsof fooddesertscould
have plausibly given the food desert concept a veneer of objectivity in UK
government in the decade from 1995. Firstly, the conceptualisation of food
deserts through GIS fitted in with political context of EBPM, and food desert
conceptualisations were compatible with GIS at the software level. These two
35
characteristics of GIS suggest a possible explanation as to why food desert
eradication was such an attractive government strategy. To add to this, the
associationsofGISwithpositivistscienceandtheseductivepowerofmapsare
likely tohavecastaveneerofobjectivityover the fooddesertconcept. Indeed,
my interview with a former policy-team member from Oxfordshire County
Council suggests that, at least in her case, GIS representationswere treated as
‘real’. Following from this argument it is possible to suggest two speculative
conclusionsrelatingtoGIS’sroleinthefooddesertdebate.Firstly,Iproposefood
desert projections in GIS as a possible reason explaining why the food desert
idealastedsolonginUKgovernmentpolicy.ForoveradecadeUKgovernment
toyedwith theproblematic fooddesert concept.The legitimacyand credibility
that GIS gave the food desert concept made it a productive use of time and
resourcesforpolicymakerstryingtosolve‘real’issues.Secondly,followingfrom
its appeal in policy, I suggest that the veneer of objectivity cast by GIS is a
contributingfactortothelackoffooddesertcritiquesduringtheUKfooddesert
debate. Belief in the objectivity of the food desert, fostered throughGIS, could
have crowded out other explanations of food access issues. Indeed, in my
experience,onlybylookingattheFoodAccessRadarthroughacriticalGISlens
wasitpossibletoseethroughtheveneerofobjectivitycastbyfooddeserts.
However,theaboveargumentconcerningGISwouldbeinsignificantifthefood
desertconceptwasitselfconcreteandreal.Yet,ascriticshaveshown(Cummins
and Macintyre, 2002a; Cummins et al., 2005; Cummins et al., 2007) the food
desert is an oversimplified and politically constructed explanation for food
access issues.Theclear implication, then, is thatprojecting fooddeserts inGIS
may have led to misguided policy initiatives, which could have failed to
incorporate important ‘relational’ aspects of food access. McGuirk and O’Neill
(2012),resonatingRamsey(2009),capturetheeffectofconventionalGISonthe
fooddesertconceptwell:“GISmakesrealitiesasmuchasitrepresentsthem:by
naming,coding,andfixingtheidentitiesandimaginariesofplacesandpeople,all
fromanimaginedpoliticallyneutralknowledgecreationposition”(McGuirkand
O’Neill,2012:1385).InthenextsectionIaddresstheissueofconventionalGIS
representations of food deserts by proposing qualitative GIS as a means re-
representingtheFoodAccessRadarinrelationalterms.
36
6.QualitativeGIStore-envisiontheFoodAccessRadar
InthepreviouschapterIhavesuggestedthattheconventionalviewofthefood
desertduringtheUKfooddesertdebatewasreinforcedandreproducedbythe
representation of food deserts in GIS. The Food Access Radar presents a
particularly salient example of reproducing the fooddesert idea because of its
political context (Monmonier, 1998). When food deserts are understood as
oversimplified social and political constructs (Cummins and Macintyre, 2002;
Cummins et al., 2007), their representation and reproduction through
37
conventional formsofGISbecomesproblematic.TheFoodAccessRadarmodel
conceptualisesfooddesertsinthisconventionalway,wherefoodaccessbecomes
a function of only physical accessibility from fixed locations within a distinct
neighbourhood, in ‘absolute’, Cartesian and Euclidean terms (Cummins et al.,
2007).
Yet, despite the deficiencies of conventional GIS as a means of understanding
food access, GIS can still be an effective tool if alternative representation
methodologiesareadopted.QualitativeGIS,departingfromanunderstandingof
GISashavingalwaysbeen ‘non-quantitative’ (Pavlovskaya,2009:13),presents
anopportunitytoincorporatethequalitativeandsituatedaspectsoffoodaccess
thatconventionalGISisunabletocapture.InthissectionIattempttore-envision
theFoodAccessRadarbyadoptingmethodsfromqualitativeGISandbuildingon
Cumminsetal.’s theorisationofa“relationalviewofcontextandspace”(2007:
1826).
‘Relational space’, a concept developed in human geography (Massey, 1999;
2005),isamorenuancedframeworkthroughwhichtoconceptualisefoodaccess
because it isdirectly focussedon the socio-cultural aspectsof foodaccess that
conventionalunderstandingsoffooddesertssilence.Placesareviewedasnodes
innetworksratherthanasdiscreteandautonomousboundedspatialunits,and
areimbuedwithsocialpowerrelationsandculturalmeaningsthatchangewith
timeandbetween individuals (Cumminset al., 2007:1827).Rather thanbeing
separatedbyphysicaldistance,placesareseparatedbysocio-relationaldistance.
Interactions are not bounded within a fixed locality; they take place within a
relativelyfluidanddynamicallydefinedarea.Thus,inadoptingarelationalview,
accesstohealthyandaffordablefoodcanbeunderstoodasdependenton“social
networks and social power, interventions of various ‘actors’ and degrees of
regulationwhichproduce‘layers’ofresourcesaccessibletodifferentmembersof
localpopulationsindifferentways”(Cumminsetal.,2007:1827).Itispossibleto
integrate this relational understanding of food access into GIS by adopting
methodsfromtheemergingfieldofqualitativeGISresearch.
38
Researchers in the sub-field of qualitative GIS have proposed methods for
integrating qualitative data into GIS software (Elwood and Cope, 2009). For
them, integration isnecessarybecause conventionalGIS representations, using
quantitative data, lack reflexivity and silence the situated, partial and socially
constructed nature of GIS knowledge (Pavlovskaya, 2009; Elwood, 2006).
Integrating qualitative data to expose silences is possible because “geographic
phenomena,theirrelationships,andtheirmeaningsareproducedandnegotiated
atmany differentmoments in GIS development and application” (Elwood and
Cope, 2009: 2). Qualitative GIS engages with these different moments of
production and negotiation of meaning to both qualify contexts and
epistemologies and to integrate new forms of qualitative data. Schuurman
(2009), for example, engages with spatial databases, whilst Kwan (2004)
engageswithspatialanalysistechniques.
Here I draw on several attempts to integrate qualitative data into GIS and
speculate how they couldbeuseful as ameansof re-representing food access,
thereby re-envisioning a Food Access Radar to incorporate situatedmeanings
andsilences.Inthisprocess,Itrytorepresentlivedexperienceandhowpeople
differentiallyengagewithandmakeuseof their foodenvironment,ratherthan
simply mapping spatial distance from locations where food is available, as
traditional food desert studies have done (Shannon, 2015). I propose using
Schuurman’s ‘ontology-based metadata’ technique (2009: 42) to enrich GIS
analysis of food access, and Pavlovskaya (2002, 2009),Matthews et al. (2005)
and Pain et al.’s (2006) techniques for incorporating qualitative data on lived
experience intoGIS. I thenpropose integrating theseperspectives intoGISata
software level through Jung’s Computer-assisted Qualitative GIS (CAQ-GIS)
(2009:115)andKwan’sSpace-TimeAquaria(2004:267).
One way that qualitative data can be incorporated into GIS is by expanding
existing metadata structures used in spatial databases (Schuurman, 2009).
Metadata(informationaboutadataset)canbecomearepositoryforqualitative
information, enablingGIS users to understand the distinct ontology associated
withaspecificdataset.SchuurmancallsthisqualitativeGIStechnique‘ontology-
basedmetadata’ (2009:42). ‘Ontology’ in critical/qualitativeGIS isunderstood
39
throughitsinformationsciencedefinitionratherthanthatofsocialtheory,inline
withcallsfromcriticalGIStoengagewiththesoftwareonitsownterms(Kwan,
2002;Sheppard,2005).Ininformationscience,eachdatasetandeachdatafield
inalegendorattributetablehasitsownseparateontology,reflectingadistinct
way of seeing the world (or epistemology). By including more thorough
metadata,adeeper,contextualisedunderstandingofeachdistinctontologycan
be obtained. Ontology-based metadata is therefore a way of reintegrating the
socialcontextofadatasetbyincludingnon-spatialattributesofthedataintoGIS.
In the caseof theFoodAccessRadar, for example, themetadata for thedata I
usedwasoriginally limitedtospatialattributes(Figure12showing foodoutlet
dataset). To turn this into ontology-based metadata, more information was
added using ArcCatalog, ESRI’s programme formanaging and organizing data.
This included the process of collecting, cleansing and representing food outlet
dataaswellasidentifyingsilencescreatedbythedataset(seewithinblackboxin
Figure 13). In my application of the Food Access Radar, food outlet data was
takenfromtheFSA’sFoodHygieneRatingSchemeandusedforpurposesbeyond
theoriginalintention.However,whilstexpandingmetadatastructuresisauseful
wayofre-contextualisingtheFoodAccessRadar,onitsownitisonlyoflimited
value for integrating the lived experience andnuancedunderstandingsof food
access that a ‘relational’ food access model would require. For this, other
qualitativeGIStechniquescouldbeused.
Figure12ScreenshotshowingtheoriginalfoodoutletshapefileinArcCatalogwithnometadataatall(Source:Author)
40
Painetal.(2006),Pavlovskaya(2002,2009)andMatthewsetal.(2005)provide
accountsofhowtousequalitativedataonlivedexperience,primarilyintheform
ofkeyinformantinterviews,withGIS.Qualitativedataisintegratedtodifferent
extentsintotheGIS.Painetal.useinterviewdataasseparatefromGIS,toqualify
theoutputofGISmapping.PavlovskayaandMatthewsetal.inputthequalitative
datatheyobtainintotheirGISrepresentations.Painetal.(2006),analysingthe
effect of street lighting on crime rates through GIS, use interviews with key
informants to compare their GIS results with lived experience. This technique
hasclearapplicationtotheFoodAccessRadar.Uponcreationandanalysisofthe
fooddesertmodel, interviews couldbe soughtwith individuals living in ‘food-
access-poor’areas,aswellaswithindividualsinareasnotclassifiedassuchfor
comparison.Questionssuchas‘doyouconsideryourselftoliveinafooddesert?’
or‘doyoufeellikeyournearestsupermarkettoofaraway?’couldrevealsocio-
cultural characteristics silenced by the Food Access Radar’s focus on physical
access. Following Pain et al. (2006), spatial analysis of food access can be
qualified by qualitative data on lived experience. In the process, interview
evidencewouldalsotestthevalidityofthefooddesertmodel itself.Yet, inthis
applicationofqualitativeGIS,analysisisclosertomixed-methodanalysis.
Figure13ScreenshotshowingtheeditedfoodoutletshapefileinArcCatalogwithmetadata.Textwithintheblackboxshowsanexampleofontology-basedmetadata
usingtechniquessuggestedbySchuurman(2009)(Source:Author)
41
Pavlovskaya (2002), inherstudymappingurban transformation inpost-Soviet
Moscow, and Matthews et al. (2005), mapping neighbourhood mobility of US
welfare recipients, use qualitative research to uncover silences in traditional
information sources before including their data in GIS. Pavlovskaya uses 45
interviewstorevealthemultipleeconomiesinMoscowhouseholds:formaland
informal, monetary and non-monetary, and includes these in her GIS
representations. Similarly Matthews et al. create GIS layers visualising the
community resources that key informants visit (Figure 14). Matthew’s
conclusions are highly relevant to the Food Access Radar because, by
representingdataonlivedexperience,theyrevealthatindividualsdonotalways
gototheirnearestcommunityresource.Livedexperiencethereforerevealsthat
socialandculturalvaluesareimportant.Bycontrast,theunderlyingassumption
of the Food Access Radar, and other food desertmodels (Donkin et al., 2000;
Pearceetal.,2006),isthatrationalconsumersvisitthenearestshopwithintheir
neighbourhood. Mapping lived experience, garnered through qualitative
research, therefore complicates conventional GIS food desert models by
revealing silences, and creates a space in which a ‘relational’ model of food
access (Cummins et al., 2007) might be a more accurate representation.
Qualitative GIS, in this respect, by plotting data garnered through interviews,
provides amechanism tomap food accessmore accurately. However, plotting
information collected from interviews is only innovative in a content sense:
traditionalGISmechanismsare stillused to representdata,whichcanobscure
thesocialcontextofdata.
42
AsawayofintegratingqualitativedatadirectlyintoGISwithoutlosingthedata’s
richness or social context, Matthews et al. (2005) also propose ‘geo-
ethnography’. They suggest integrating audio, video, image and text files,
collectedthroughethnographicresearchmethods,bycreating“hotlinks”(2005:
76)within the GIS software to external databases. This way the ethnographic
material isnotquantified,so is integratedwhilstmaintaining itssocialcontext.
Applying this technique to food access models could be an important way of
integrating written and photographic accounts of lived experience. After
establishing an appropriate sample of theoretically food-access-rich and food-
access-poorindividuals,eitherfromtheFoodAccessRadarresultsorresearcher
knowledge, individuals’ lived experience of their food environment could be
garnered through qualitative methods. Participants could be asked, through
ethnographicinterviews,aboutthesocio-culturalmeaningsimbuedinfoodretail
infrastructure and about access as a determinant of their food consumption
patterns.Tocomplimentinterviews,theycouldbeaskedtodrawsketchmapsof
their foodenvironment, therebyaccounting fortheirsituatedknowledgeof the
Figure14ThediagramproducedfromtheethnographyresearchconductedbyMatthewsetal.(2005)revealsthatmanyofthecommunityresourcesusedbythe
samplefamilylieoutsidetheirhouseholdcensustract
43
foodenvironment,highlightingtheoutletstheyusemostfrequently.Theycould
alsorecordthefoodstuffspurchasedfromparticularshopsthroughphotographs,
andalsophotographtheseshopstocaptureaground-levelviewofthesesites.All
thesequalitativedatasourcescouldbe“hot-linked”(Matthewsetal.,2005:76)
into the GIS, building a rich picture of food access patterns, and generating a
‘relational’understandingoffoodaccessthroughqualitativeGIS(Cumminsetal.,
2007).Throughthismethod,individuals,otherwisesilencedinconventionalGIS
accessmodels,aregivenagencyintheGISrepresentations.
Taking ‘geo-ethnography’ (Matthews et al., 2005) techniques to the limit, Jung
(2009) proposes integrating qualitative data directly into GIS data structures
through ‘Computer-aided Qualitative GIS’ (CAQ-GIS) (2009: 115). Directly
encodedqualitativedataismoreusefulbecauseitcanbeanalysedinGISthrough
‘queries’and‘selections’,forexample.Jungproposestointegratequalitativedata
intoarastergrid,superimposedasanewlayerinGIS.Eachlineintherastergrid
isdefinedbygeographicalcoordinates.Qualitativedata,havinggonethroughan
encodingprocessinComputerAssistedQualitativeDataAnalysissoftwaresuch
as ATLAS.ti, can be given a geographical coordinate in the raster layer
corresponding to the location forwhich itwascollected.Therastergridstores
theencodedqualitativedata in anattribute table compatiblewithGIS analysis
(Figure15providesa summaryof theCAQ-GISprocess).Theprocess couldbe
appliedoveraprojectionofCambridgeshire,andethnographicaldataproduced
aboutfoodaccesscouldbeintegrated.Thisabilitytointegratemultipleformsof
evidence and multiple ways of knowing into GIS in this way resonates with
Knigge and Cope’s use of “grounded visualisation” (2006: 2021) to analyse
iterativelybothquantitativeandqualitativedatainGIS.BytakingqualitativeGIS
to its limit, by integrating qualitative data into GIS at a software level, well-
grounded,richdescriptionsandexplanationsoffoodaccessprocesseswouldbe
availabletoresearchersandpolicymakers(Jung,2009:116).
44
One final application of critical GIS, whilst not qualitative GIS per se, can
integratefurtherdetailsoflivedexperiencedirectlyintoGISsoftware.Kwanhas
created a ‘space-time aquaria’ (Kwan, 2004: 267) to conceptualise how the
interactionofspaceandtimecanstructurehumanactivitypatternsinparticular
localities.Space-timeaquariacanbedirectlyappliedtofoodaccessibilityissues,
addingfurthercomplexityandnuancetofoodaccess.A‘space-timeaquaria’isa
three-dimensional representation in GIS of an individual’smovement over the
course of a day, between different nodes (Figure 16). It therefore represents
movement as a function of space-time, directly advancing an understanding of
therelationalgeographiesoffoodaccess.Thesignificanceofthisforfoodaccess
isthatitaddsgreaterflexibilitytothelocationfromwhichpeoplepurchasefood
thanconventional foodaccessmodels.Conventionalaccessmodels,suchas the
Figure15ThisimagedescribesthequalitativedataintegrationprocessinCAQ-GIS.Inthetopleft-handcornertherastergridisoverlainontothedesiredspatial
area.Thebottomtwotablesshow(1)thespatialdatabasefortherastergridcontainingqualitativeinformation(animageinthiscase),and(2)thesocialdatabasecontainingfurtherqualitativedataassociatedwiththatimage(Source:
Jung,2009)
45
Food Access Radar, consider the home as the only basing point from which
people access food, and thus consider food-access-poverty as a function of
distance from food stores to households. This does not consider that the
workplace,school,church,socialfacilities,oranypointalongtheroutebackfrom
these locations to home, are also places where an individual can make food
purchases(Burgoineetal.,2014).WhilstNetworkAnalyst,asused in theFood
AccessRadar,isoneofthefewinstrumentsinGISthancananalysespace-time,it
doesnotgofarenough.Constructinga3Dspace-timeaquariumfromindividual
mobility accounts (the qualitative element of the aquariawhen garnered from
ethnographic interviews) offers a richer perspective on food access
opportunities across a range of sites. Space-time aquaria can be created using
complex algorithms, as detailed by Kwan (2004), in ArcGIS using ARC Macro
Language (AML). Space-time aquaria have additional value for food access
studiesbecause foodoutletopeninghours,which restrictaccess in space-time,
could also be integrated into 3Dmodels. Space-time aquaria therefore offer a
richer perspective into relational food access conceptualisation, integrating
mobilityintoqualitativeGISandnewparametersforaccessthroughspace-time.
46
AsIhavesuggestedwiththeabovestrategiesforintegratingqualitativedatainto
GIS analysis, significant potential exists within qualitative GIS to respond to
Cummins et al.’s (2007) call for a relational approach tomapping food access.
Unlike conventional approaches to spatial analysis of food deserts such as the
Food Access Radar, that explore only physical, Euclidean proximity as a
determinant of food access, qualitative GIS offers potential to integrate lived
experience,situatedknowledges,andreflexivityintoGISanalysis.
Figure16AdiagramofKwan’sspace-timeaquaria.Thediagramshowsadailyspace-timepathforanindividualbetweentheworkplace,churchandhomeacross
FranklinCounty,Ohio.Thisindividualpathwaycanbeinsertedintospace-time
aquariain3DGISalongwithotherspace-timepathways(Source:Kwan,2004)
47
7.Conclusion
Inthisdissertation,theFoodAccessRadarhasbeenusedintwoprincipalways.
First,ithasprovidedanavenuethroughwhichto‘unpack’theintricaciesofGIS-
projectedfooddeserts.CriticalGISnecessitatesan‘insiderview’(Elwood,2006),
andonlybyconstructingtheFoodAccessRadarcouldIpeerinsidetheGIS‘black
box’(Sismondo,2010).TheseconduseoftheFoodAccessRadarhasbeenasan
evidence base. Together with a review of relevant literature and qualitative
information garnered from interviews, I have assembled a critique of the food
desertconcept,whichfocusesontheinfluenceofGISinthefooddesertdebate.
TheresearchinthisdissertationchallengestheobjectivityofGISprojectionsin
thecontextofUKfooddeserts.Previouscritiqueshaveunderstoodfooddeserts
asoversimplified,partialexplanationsof foodaccess (CumminsandMacintyre,
2002a; Cummins, 2005), which do not take into account the relational
geographies of resource access (Cummins et al., 2007). This dissertation has
extended such arguments, by suggesting that the use of GIS, specifically,
transformedthepartialfooddesertideaintoatotalisingandseeminglyobjective
concept. Combining knowledge garnered from the Food Access Radar
constructionprocesswithinterviewevidence,informedbyCriticalGISliterature,
IhavearguedthattheprojectionoffooddesertsinGISatleastpartiallyexplains
thepoliticalpurchaseheldby the fooddesert idea inUKgovernmentpolicy in
the late 1990s and early 2000s. GIS both facilitated the identification of food
deserts within an evidence-based policy-making paradigm and reinforced the
concept.GISsoftwarestartsfromthesameunderstandingofconventionalspace
as the food desert concept, framing a more complex and nuanced issue as a
simplified geographical reality based on a limited number of quantitative
variablesmapped in space.GIS also casts a veneerof objectivityover the food
desert concept through the power of maps and its association with positivist
science. From the argument that GIS was both convincing and convenient,
tentativeconclusionscanbedrawn.Firstly,IwouldarguethatGISprojectionsat
leastcontributedtothepersistenceofthefooddesertideaforalmostadecadein
UK government policy. Secondly, and running in parallel, the convincing and
convenient nature of GIS food desert projections possibly explains why there
48
were so few critiques of the food desert concept in policy and academic
literature.
With the Food Access Radar, and other ‘conventional’ food desert GISmodels,
problematizedastotalisingandobjectifyingconstructions,thisdissertationthen
proposes an alternative way of conceptualising food access through GIS.
RespondingtoCumminsetal.’scalltoengagewitha‘relational’approachtofood
access(2007),IinvestigatethepotentialitiesofqualitativeGIS,afieldofresearch
that has been growing in recent years, for projecting food access in GIS.
QualitativeGISisappropriatebecauseitenablesaresearchertoincorporatethe
subtletiesoffoodaccess,silencedbytraditionalfooddesertmodels,directlyinto
theGISsoftware.SeveralqualitativeGISmethodshavebeenproposedandmade
relevant to food access modelling, such as geo-ethnography and CAQ-GIS. To
implement these qualitative GIS techniques would require greater scope and
resourcesthanthisdissertationpermits;however,ithasshowntheneedfor,and
potentialityof,aqualitativeGISapproachtofoodaccessmodelling.Insodoing,it
hasopeneduparesearchavenueforfuturefoodaccessstudiestobuilduponthe
qualitative GIS techniques suggested here. As the field of qualitative GIS
continuestogrow,doubtlessothertechniquesapplicabletofoodaccessresearch
willalsoemerge.
In UK food accessibility research today, there is growing interest in the
proliferation in other aspects of food availability, such as the proliferation of
takeawayoutlets (Burgoine et al., 2014).Perhaps it is in this bodyof research
that qualitative GIS may be applicable. Alternatively, in the on-going US food
desert context, selected critiquesof the fooddesert concept arenowemerging
(notably Shannon, 2014). In these debates too, this dissertation couldmake a
number of contributions. Firstly, American academics and policy-makers alike
shouldbemadeawareofthepowerfuleffectsthatvisualisingfooddesertsinGIS
canhave,inordertobetterunderstandthepartialityofthefooddesertconcept.
Second,theuseofqualitativeGISasameansofrepresentingfoodaccesscouldbe
further advanced in the US context. A qualitative GIS approach to food access
may destabilise current understandings of food deserts and, importantly,
uncoverhiddengeographiesoffoodaccess.
49
8.References
50
Ayres,L.(2008)Semi-StructuredInterview.In:Given,L.M.TheSageencyclopaediaofqualitativeresearchmethods.LosAngeles,Calif:SAGE.pp.811-812.
Beaulac,J.,Kristjansson,E.andCummins,S.(2009)‘ASystematicReviewofFood
Deserts,1966-2007’,PrevChronicDis,6(3),pp.1–10.
Burgoine,T.,Forouhi,N.G.,Griffin,S.J.,Wareham,N.J.andMonsivais,P.(2014)
‘Associationsbetweenexposuretotakeawayfoodoutlets,takeawayfood
consumption,andbodyweightinCambridgeshire,UK:Populationbased,cross
sectionalstudy’,BMJ,348(mar135),pp.g1464–g1464.
Charreire,H.,Casey,R.,Salze,P.,Simon,C.,Chaix,B.,Banos,A.,Badariotti,D.,
Weber,C.andOppert,J.-M.(2010)‘Measuringthefoodenvironmentusing
geographicalinformationsystems:Amethodologicalreview’,PublicHealthNutrition,13(11),pp.1773–1785.
Chrisman,N.(2001)Configuringtheuser:SocialdivisionsoflaborinGISsoftware.Availableat:
https://pdfs.semanticscholar.org/2018/32f504cebeba1203927769bb97cc99ea
27fb.pdf(Accessed:2016).
Clarke,G.,Eyre,H.andGuy,C.(2002)‘Derivingindicatorsofaccesstofoodretail
provisioninBritishcities:StudiesofCardiff,LeedsandBradford’,UrbanStudies,39(11),pp.2041–2060.
Clarke,I.,Hallsworth,A.,Jackson,P.,deKervenoael,R.,Perez-del-AguilaR.and
Kirkup,M.(2004)‘Retailcompetitionandconsumerchoice:Contextualisingthe
“fooddeserts”debate’,InternationalJournalofRetail&DistributionManagement,32(2),pp.89–99.
Cope,M.andElwood,S.(2009)QualitativeGIS.SAGEPublications.
Crampton,J.W.(2001)‘Mapsassocialconstructions:Power,communicationand
visualization’,ProgressinHumanGeography,25(2),pp.235–252.
Cummins,S.(2005a)‘Foodenvironmentsandobesity--neighbourhoodor
nation?’,InternationalJournalofEpidemiology,35(1),pp.100–104.
Cummins,S.(2005b)‘Largescalefoodretailinterventionsanddiet’,BMJ,330(7493),pp.683–684.
Cummins,S.(2007)‘Commentary:Investigatingneighbourhoodeffectson
health--avoidingthe“localtrap”’,InternationalJournalofEpidemiology,36(2),pp.355–357.
Cummins,S.,Curtis,S.,Diez-Roux,A.V.andMacintyre,S.(2007)‘Understanding
andrepresenting“place”inhealthresearch:Arelationalapproach’,SocialScience&Medicine,65(9),pp.1825–1838.
Cummins,S.andMacintyre,S.(1999)‘Thelocationoffoodstoresinurbanareas:
AcasestudyinGlasgow’,BritishFoodJournal,101(7),pp.545–553.
51
Cummins,S.andMacintyre,S.(2002a)‘“Fooddeserts”---evidenceand
assumptioninhealthpolicymaking’,BMJ,325(7361),pp.436–438.
Cummins,S.andMacintyre,S.(2002b)‘Asystematicstudyofanurban
Foodscape:ThepriceandavailabilityoffoodingreaterGlasgow’,UrbanStudies,39(11),pp.2115–2130.
Dacey,M.andMarble,D.(1965)Somecommentsoncertainaspectsofgeograhicalinformationsystems.DepartmentofGeography,NorthwesternUniversity,Evanston,Illinois.
Donkin,A.J.,Dowler,E.A.,Stevenson,S.J.andTurner,S.A.(2000)‘Mappingaccess
tofoodinadeprivedarea:Thedevelopmentofpriceandavailabilityindices’,
PublicHealthNutrition,3(01).
Donkin,A.J.M.,Dowler,E.A.,Stevenson,S.J.andTurner,S.A.(1999)‘Mapping
accesstofoodatalocallevel’,BritishFoodJournal,101(7),pp.554–564.
Elinder,L.andJansson,M.(2008)‘Obesogenicenvironments–aspectson
measurementandindicators’,PublicHealthNutrition,,p.1.
Elwood,S.(2006)‘Negotiatingknowledgeproduction:Theeverydayinclusions,
exclusions,andcontradictionsofparticipatoryGISResearch∗’,TheProfessionalGeographer,58(2),pp.197–208.
Elwood,S.andCope,M.(2009)‘Introduction:qualitativegis:forgingmixed
methodsthroughrepresentations,analyticalinnovations,andconceptual
engagements’,inQualitativeGIS.SAGEPublications,pp.1–12.
Elwood,S.,Schuurman,N.andWilson,M.W.(2011)‘CriticalGIS’,inTheSAGEhandbookofGISandsocietyresearch.ThousandOaks,CA:SAGEPublications,pp.87–106.
Fafard,P.(2015)‘Beyondtheusualsuspects:Usingpoliticalsciencetoenhance
publichealthpolicymaking’,JournalofEpidemiologyandCommunityHealth,69(11),pp.1129–1132.
Flowerdew,R.andMartin,D.(2005)Methodsinhumangeography:aguideforstudentsdoingaresearchproject.2nded.Harlow:PearsonPrenticeHall.
Goodchild,M.(1992)‘GeographicalInformationScience’,InternationalJournalofGeographicalInformationSystems,6(1):31-45.
Haraway,D.(1988)‘SituatedKnowledges:Thesciencequestioninfeminismand
theprivilegeofpartialperspective’,FeministStudies,14(3),p.575.
Jung,J.(2009)‘Computer-aidedqualitativegis:asoftware-levelintegrationof
qualitativeresearchandgis’,inQualitativeGIS.SAGEPublications,pp.115–136.
Kitcher,P.(1998)‘Apleaforsciencestudies’inKoertge,N.Ahousebuiltonsand:exposingpostmodernistmythsaboutscience,NewYork:OxfordUniversityPress,
52
32–52.
Knigge,L.andCope,M.(2006)‘Groundedvisualization:Integratingtheanalysis
ofqualitativeandquantitativedatathroughgroundedtheoryandvisualization’,
EnvironmentandPlanningA,38(11),pp.2021–2037.
Knoblaugh,R.L.,PietruchaM.T.andNitzburgM.(1996)FieldStudiesof
PedestrianWalkingSpeedandStart-upTime.TransportationResearchRecord
1538;pp27-38
Kuhn,T.S.(1962)Thestructureofscientificrevolutions.Chicago,IL:UniversityofChicagoPress.
Kwan,M.-P.(2002)‘Feministvisualization:Re-envisioningGISasamethodin
feministgeographicresearch’,AnnalsoftheAssociationofAmericanGeographers,92(4),pp.645–661.
Kwan,M.-P.(2004)‘GISmethodsintime-geographicresearch:Geocomputation
andGeovisualizationofhumanactivitypatterns’,GeografiskaAnnaler,SeriesB:HumanGeography,86(4),pp.267–280.
Kwan,M.-P.andKnigge,L.(2006)‘DoingqualitativeresearchusingGIS:An
oxymoronicendeavor?’,EnvironmentandPlanningA,38(11),pp.1999–2002.
Lake,A.A.,Burgoine,T.,Greenhalgh,F.,Stamp,E.andTyrrell,R.(2010)‘The
foodscape:Classificationandfieldvalidationofsecondarydatasources’,Health&Place,16(4),pp.666–673.
Lake,A.A.,Burgoine,T.,Stamp,E.andGrieve,R.(2012)‘Thefoodscape:
Classificationandfieldvalidationofsecondarydatasourcesacrossurban/rural
andsocio-economicclassificationsinEngland’,InternationalJournalofBehavioralNutritionandPhysicalActivity,9(1),p.37.
Lang,T.andCaraher,M.(1998).Accesstohealthyfoods:PartI.Barriersto
accessinghealthyfoods:Differentialsbygender,socialclass,incomeandmodeof
transport.HealthEducationJournal,57(3),pp.191-201.
Latour,B.(1999)Pandora’shope:Anessayontherealityofsciencestudies.2ndedn.Cambridge,MA:HarvardUniversityPress.
Latour,B.andWoolgar,S.(1986)Laboratorylife:Theconstructionofscientificfacts.EditedbyJonasSalk.UnitedStates:PrincetonUniversityPress.
Macintyre,S.andPetticrew,M.(2000)‘Goodintentionsandreceivedwisdomare
notenough’,JournalofEpidemiology&CommunityHealth,54(11),pp.802–803.
Malsen,C.,Raffle,A.,Marriot,S.andSmith,N.(2013)FoodPoverty:Whatdoestheevidencetellus?BristolCityCouncil.
Massey,D.(1999)Power-geometriesandthepoliticsofspacetime:Hettnerlecture1998.Heidelberg:Dep.ofGeography,Univ.
53
Massey,D.(2005)Forspace.London:SAGE. Matthews,S.A.,Detwiler,J.E.andBurton,L.M.(2005)‘Geo-ethnography:
CouplinggeographicinformationanalysistechniqueswithEthnographic
methodsinurbanresearch’,Cartographica:TheInternationalJournalforGeographicInformationandGeovisualization,40(4),pp.75–90.
McGuirk,P.andO’Neill,P.(2012)‘CriticalGeographieswiththestate:The
problemofsocialvulnerabilityandthepoliticsofengagedresearch’,Antipode,44(4),pp.1374–1394.
Mohan,J.(2002)‘Geographiesofwelfareandsocialexclusion:Dimensions,
consequencesandmethods’,ProgressinHumanGeography,26(1),pp.65–75.
Monmonier,M.(1998).‘Chapter13TheThreeR'sofGis-BasedSiteSelection:
Representation,Resistance,andRidicule’ModernCartographySeries,3(C),233-247.
Mooney,C.(1990)‘CostandavailabilityofhealthyfoodchoicesinaLondon
healthdistrict’,JournalofHumanNutritionandDietetics,3(2),pp.111–120.
Nolan,K.(2015)‘Neoliberalcommonsenseandrace-neutraldiscourses:A
critiqueof“evidence-based”policy-makinginschoolpolicing’,Discourse:StudiesintheCulturalPoliticsofEducation,36(6),pp.894–907.
Nyerges,T.L.,Couclelis,H.andMcMaster,R.B.(eds.)(2011)TheSAGEhandbookofGISandsocietyresearch.ThousandOaks,CA:SAGEPublications.
O’Neill,M.(2005)PuttingFoodAccessontheRadar.NationalConsumerCouncil.
Openshaw,S.(1991)‘AviewontheGIScrisisingeography,or,usingGIStoput
Humpty-Dumptybacktogetheragain’,EnvironmentandPlanningA,23(5),pp.621–628.
Pain,R.,MacFarlane,R.,Turner,K.andGill,S.(2006)‘“When,where,if,andbut”:
QualifyingGISandtheeffectofstreetlightingoncrimeandfear’,EnvironmentandPlanningA,38(11),pp.2055–2074.
Pavlovskaya,M.(2006)‘TheorizingwithGIS:Atoolforcriticalgeographies?’,
EnvironmentandPlanningA,38(11),pp.2003–2020.
Pavlovskaya,M.(2009)‘Non-quantitativeGIS’,inQualitativeGIS.SAGEPublications,pp.13–38.
Pavlovskaya,M.E.(2002)‘MappingurbanchangeandchangingGIS:Otherviews
ofeconomicrestructuring’,Gender,Place&Culture,9(3),pp.281–289.
Pearce,J.(2006)‘Neighbourhoodsandhealth:AGISapproachtomeasuring
communityresourceaccessibility’,JournalofEpidemiology&CommunityHealth,60(5),pp.389–395.
54
Pickles,J.(ed.)(1995)Groundtruth:Thesocialimplicationsofgeographicinformationsystems.NewYork:GuilfordPublications.
Ramsey,K.(2009)Acallforagonism:GISandthepoliticsofcollaboration.
EnvironmentandPlanningA40:2346–2363
Sanderson,I.(2002)‘Evaluation,policylearningandevidence-basedpolicy
making’,PublicAdministration,80(1),pp.1–22.
Schuurman,N.(2000)‘Troubleintheheartland:GISanditscriticsinthe1990s’,
ProgressinHumanGeography,24(4),pp.569–590.
Shannon,J.(2014)‘Fooddeserts:Governingobesityintheneoliberalcity’,
ProgressinHumanGeography,38(2),pp.248–266.
Shannon,J.(2015)‘RethinkingFoodDesertsUsingMixed-MethodsGIS’,
Cityscape,17(15),pp.85–96.
Sheppard,E.(1993)‘AutomatedGeography:Whatkindofgeographyforwhat
kindofsociety?’,TheProfessionalGeographer,45(4):pp.457-60.
Sheppard,E.(2001)‘Quantitativegeography:Representations,practices,and
possibilities’,EnvironmentandPlanningD:SocietyandSpace,19(5),pp.535–554.
Sheppard,E.(2005)‘KnowledgeproductionthroughcriticalGIS:Genealogyand
prospects’,Cartographica:TheInternationalJournalforGeographicInformationandGeovisualization,40(4),pp.5–21.
Shuurman,N.(2009)‘Metadataasasiteforimbuinggiswithqualitative
information’,inQualitativeGIS.SAGEPublications,pp.41–56.
Sismondo,S.(2010)Anintroductiontoscienceandtechnologystudies.2ndedn.Malden,MA:Wiley-Blackwell(animprintofJohnWiley&SonsLtd).
SocialExclusionUnit(2004)TheSocialExclusionUnit:InformationLeaflet
Availableat:
http://webarchive.nationalarchives.gov.uk/+/http:/www.cabinetoffice.gov.uk/
media/cabinetoffice/social_exclusion_task_force/assets/publications_1997_to_2
006/seu_leaflet.pdf(Accessed:2016).
SoomanA,MacintyreS,AndersonA.(1993)Scotland’shealth-amoredifficult
challengeforsome?Thepriceandavailabilityofhealthyfoodsinsocially
contrastinglocalitiesinthewestofScotland.HealthBull;51(5):276-84.
St.Martin,K.andWing,J.(2007)‘ThediscourseanddisciplineofGIS’,
Cartographica:TheInternationalJournalforGeographicInformationandGeovisualization,42(3),pp.235–248.
Sui,D.Z.andGoodchild,M.F.(2001)‘GISasMedia?’,InternationalJournalofGeographicalInformationScience,15(5):387-90.
55
Sui,D.Z.andGoodchild,M.F.(2003)‘AtetradicanalysisofGISandsocietyusing
McLuhan’slawofhemedia’,CanadianGeographer,47(1):5-17.
Tomlinson,R.(1962)AnIntroductiontotheUseofElectronicComputersintheStorage,CompilationandAssessmentofNaturalandEconomicDatafortheEvaluationofMarginalLands.CanadaDepartmentofAgriculture.
Vining,J.andDay,N.(2005)FoodAccessRadar.StaffordshireCountyCouncilfortheFoodStandardsAgency.
Warren,S.(2011)‘TheSocialPotentialofGIS’,inTheSAGEHandbookofGISandSociety.ThousandOaks,CA:SAGEPublications,pp.67–86.
Wells,P.(2007)‘Newlabourandevidencebasedpolicymaking:1997-2007’,
People,PlaceandPolicyOnline,pp.22–29.
White,M.,Bunting,J.,Williams,L.,Raybould,S.,Adamson,A.andMathers,J.
(2004)Do‘fooddeserts’exist?Amulti-level,geographicalanalysisoftherelationshipbetweenretailfoodaccess,socioeconomicpositionanddietaryintake.Newcastle-on-Tyne(UK):UniversityofNewcastle.
Wrigley,N.(2002)‘“Fooddeserts”inBritishcities:Policycontextandresearch
priorities’,UrbanStudies,39(11),pp.2029–2040.
Wrigley,N.,Warm,D.andMargetts,B.(2003)‘Deprivation,diet,andfood-retail
access:FindingsfromtheLeeds“fooddeserts”study’,EnvironmentandPlanningA,35(1),pp.151–188.
9.Appendix:interviewtranscripts
TelephoneinterviewwithJonathanVining,StaffordshireCountyCouncil,GIS
practitionerfortheFoodAccessRadarProjectinStaffordshire,byLouis
Rawlings(L),05/09/16
56
[19:32mins]
Louis:Hello?
Jonathan:HiisthatLouis
L:ThatisLouishello!
J:Hiit’sJonViningfromStaffordshireCountyCouncil
L:HiJon,thankssomuchforagreeingtospeaktome
J:SorryI’vejustbeenonareallylongphonecall
L:Noproblematall,I’lltrynottokeepyoufortoolongthen
J:HowcanIhelpthenLouis?
L:SoifIgiveyouabriefoverviewofwhatmyprojectconsists.EssentiallyI’ma
thirdyeargeographystudentanddecidedthatIwantedtodoadissertationon
socialinequalityandGIS,andIcameacrossthefoodaccessradar,thatwasabout
threemonthsago.SincethenI’vecreatedthefoodaccessradarfor
CambridgeshireandthenI’mmakingitintoanacademicdissertationby
analysingthecontributionofthefoodaccessradartopolicydebatesintheUK
andthewideracademicdebateaboutfooddesserts.Aswellaslookingatits
applicationinAmericatodaywherethefooddeserthastakenoff.Socouldyou
describetomeyourroleinStaffordshire?
J:Yeasure.InmycurrentroleI’minvolvedineconomicregeneration.Lookingat
businessparksandinfrastructure,employmentland.ButrewindingtheclockI
wasintheresearchunit.Thiswasverymuchfocussedonprovidingresearchfor
ourplanningandeconomicdevelopmentteams.Butalsotothecorporate
functionsasaCountyCouncil.Sooneofourserviceareasisinconsumerhealth,
socolleaguesinthatdepartmentapproachedustoaskforhelpwiththeirfood
accesspieceofwork.Myroleintheprojectpredominantlywasinthedatasideof
it.Soassemblingthedata,doingtheGISworkanddoingallthemagicwiththe
datasets.
L:Sointermsoftheproject,youweren’tinvolvedincomingupwiththefood
accessprojectthen?
J:Asin,asanidea?
L:Yeahthedirection
J:Theideaofproducingthefoodaccessradarwasinitiatedbythehealthteam
whosatwithintradingstandardsatthetime.Theywereawarethatweworked
withGISandcouldhelp.Obviouslythetechnologyhasmovedonsignificantly
sincethen.WhatIwasdoingwaslotsofdrivetimeanalysis.Particularlythings
57
like,populationslivingwithinacertaindrivetimeofabusinessparkora
motorwayjunction.Sotranslatedthatreallyintoawalkingdistancefromvarious
shopsandfoodoutlets.That’swheremyrolecameinbecausetheywereaware
thatIdidthat.AndIcameupwiththemethodologyandhowtomakeitwork
withthedatareally.
L:Sodoyouknow,intermsofthewholefoodaccessidea,doyouknowwhere
thepressurescamefrom?
J:TherewasanorganisationinvolvedcalledtheNationalConsumerCouncil.
Theywereleadingonitatanationallevel.Somefundingwasmadeavailableto
helpStaffordshiretakethispieceofworkforward.Bythesoundsofitthey’ve
longgoneunfortunately.Butintermsofmyrole,Icantellyouthatwehadthe
FLAREdatabase,whichessentiallyhadallthedetailsofthepremisesthatwere
licensedtosellfoodacrossStaffordshire.Iessentiallygotallthepostcodesand
mappedthoseandgroupedthemintodifferentcategoriesofstress.Off-license,
supermarkets,cornershops,thattypeofthing,andstratifiedthatontheGISfile,
andthencreatedthewalkingtimeisochrones.
L:yeahI’vehadfundoingthat
J:Areyoudoingthatatthemoment?
L:I’vefinisheditnow.IwaslearningGISatthesametimeasworkingonthis
thing.Usingthenetworkanalyst
J:Yeahthatwastheoldfavourite.ArcGIS
L:WasthismodeldesignedspecificallyforStaffordshire?
J:Iwouldn’tsayitcouldonlyworkinStaffordshire.Ithinkitcouldbereplicated
elsewhere.Imeanthehardestpartformeinmanywayswasgettingaclean
databaseoffoodstores.Idon’tknowifyou’vehadthisissue?
L:YeahIhave,it’sclearedupalittlebitsince.TheEHteamskeeparecordof
theirfoodstoressoit’sabitmorestandardised.
J:Sothatwasthebiggestchallengetostartoffwith,ittookmequiteawhileto
produceadatasetthatwasworthyofdoingapieceofworkwith.Andthenwe
werequiteluckybecausewereabletopullonthe2001census,sowegotdown
tooutputareageography.Wecoulddoallthepopulationdemographictrends
withthat.Andalsowedidabitofworkoverlayingpublictransportroutes.We
hadcolleaguesinourpassengertransportteamwhowereabletoprovideus
withbusroutesandsoon,andtheygaveussomeisochrones.Sowecould
interchangetimesonvariousareas.Soifyougotabusfromyourhouseyou
couldgettothesupermarketinxminutes.Butthatwasabitfurtherdownthe
linereally.
L:ItsinterestingthatI’veseensofar,alotofthemorerecentstudieshave
measureddistancefromthehouseholdtotheshop.TheFoodAccessRadarwas
58
fromtheshoptothehouseholds.Didyoudothatdeliberatelyorwasit
somethingmoreconvenienttodo?
J:Ithinkitwaseasiertodoitthatway.Becauseinsomewhereyou’vegot
somethinglike400,000addresspoints,whereasthenumberofpremisesis
gettingonsomethinglinkathousand,oracoupleofthousand.SoIcan’t
rememberoffthetopofmyhead,butthecomputinginvolvedinthecoupleof
thousandpremiseswasabitmorestraightforward.Weworkedonlooking
wherethegapswere.Wecouldseethatthecoveragewasgoodbutitwasabout
identifyingwherethegapswere.Soobviouslyyouhadthatinruralareas,
becausewe’vegotsomefairlysubstantialrurallocalities.Butalsoinpartsof
someofourlargertowns,wherebythetenminuteshowedusthatyoucould
correlatequitecloselywithareasofdeprivation.Ithinkthatweoverlaiditwith
theIndexofMultipleDeprivation2004andwithotherhealthindicatorse.g.
debilitatinglong-termillness,theusualsuspectsthatyou’renodoubtawareof.
L:DidthatrevealanythingsurprisingaboutStaffordshirewhenyouplottedit?
J:Can’trememberI’msorry.I’velivedalifeofmapssinceIdidthis.Ididhavea
documentonitbutIhaven’tbeenabletofindit.
L:Doyourememberhowyouselectedtheproblemareas?I’velookedatthe
documentthatyouhad.Downnearthebottomittellsyouabouthowtoidentify
thesepotentialproblemareas.Doyourememberselectingthempurelybecause
ofwhatthemapshowedorwereyouselectingareasalsofromyourknowledge
ofStaffordshireanddeprivation?
J:Ithinkitwasprobablyalittlebitofboth.Themapwillhaveshownthe
geographicareaswherethereweregapsfromtheisochrones.Butwedidfocusin
onthedeprivedcommunitiestoseeifthefoodaccessreinforcedorconflicted
againstthoseideas.Inthelargesenseitwasthosedeprivedareasthattendedto
exhibitafoodaccessissuethanthemoreaffluentareas.
L:I’mawarethatI’mkeepingyoualittlebitlongerthanIsaid,butcouldIjustask
youafewmorequestions?
J:Ofcourseyoucan
L:I’vebeenlookingtocritiquethemodel,lookingattheassumptionsthatlimit
thefoodaccessradarasatool.Doyourememberatthetimewhetheryouwere
awareoftheassumptionsandlimitationsatthetime?
J:Yeatherewasprobablyoneortwo.Thefirstonewasthequalityofthedata
thatwewereworkingwith.Iwasn’ttotallyconvincedthatitwasafullyvalid
dataset.Butwehadtolivewiththat,becauseitwasthebestwehadatthetime.I
thinktheGISsoftwarewasfairlygood,butbackinthedayweusedthenetwork
analyst,andIthinktheunderlyingnetworkdatasetwasn’tnecessarilyasgreatat
pedestrianconnectivityasitcouldhavebeen.Sotherewaspossiblysomeareas
whereyouhavecutthroughthatwouldn’thavebeenidentifiedinthenetwork,
socouldhavebeenoverexaggeratingaprobleminthoselocations.
59
L:Finalareaofquestioning.Doyourememberwhattheresultswereinthe
Staffordshirestudy?DidtheGIShaveanymaterialresultsonStaffordshire?
Policychanges?
J:It’sabitofashame,Iusedtogetstuckinadarkroomattheresearchunittodo
allthisstuff,thatotherpeopleusedtotakeon.Itmaybethatcolleaguesofmine
aremorewellplacestosaywhatactuallyhappened.There’saladycalledNicola
DaywhowasmoreintothisthanI.
L:That’severythingIreallywantedtoknow.ThanksJon.OnafinalnotecanI
justaskifIcanusethisconversationrecording?
J:Yeahofcourse,you’rewelcometousemynameifyouwant,Idon’tmindthat.
Ifthere’sanythingyoucanthinkofjustletmeknow.
L:ThankssomuchforthatJon.Andthankssomuchforyourtime,that’sbeen
reallyuseful.
J:Okay,noproblem.Goodluckwithyourdissertation!Byefornow
L:Bye!
TelephoneinterviewwithKateAustin(K),OxfordshireCountyCouncil,directly
involvedintheFoodAccessRadarProjectinOxfordshire,byLouisRawlings(L),
08/09/16
[26:58mins]
Louis:Hello?
Kate:HiLouis,itsKate
L:HelloKate,thanksforagreeingtospeaktome.
60
Kate:Noproblem,whatcanIhelpyouwith.Itsfoodaccessisn’tit?
L:Itisyeah.InthelastfewmonthsIhavebeenconstructingthefoodacessradar,
puttingtogetherthedatasetsthatI’vefoundandthenI’veputthemintotheGIS.
I’vecreatedtheGIStoolandamnowanalysingitforCambridgeshire.Itlookslike
therecouldhavebesomesortoffoodpoverty/foodaccessibilityissue
K:Okayyeah
L:ButonthewholeinCambridgeIfoundthat…maybeitsbecausestudieslike
thishavebeengoingonforthelastdecade,thatactuallyonthewholefood
accessibilityinCambridgedidappeartobequitegood
K:BecauseactuallyifIremember,oneofthefirstfunctionsoffoodbankswasin
Cambridge.Ithinkthat’swherealotofitstarted.SoIthinkyou’rerightthere’s
probablybeenloadsofactivityinCambridge
L:Essentiallyformydissertation,asmuchasIneedtoconstructthefoodaccess
radar,it’sgoingtobequiteusefultohaveabitaboutthecontext,ofhowthisidea
offooddesertswasdrivenandhowitcameaboutintheOxfordCountyCouncil.
AndthenIhadafewfinalquestionsaboutGISingeneralinthecouncil.
Qualitativevs.quantitativedata.I’dbereallygratefulifyoucouldanswerafew
ofthesequestionsforme.CouldIrecordthisanduseitbytheway?
K:Yeahnoproblem,I’lldomybest.ItwasalittlewhileagoandIwasoneof
manypartnersinvolvedintheproject.TheGISstuff,Iwasn’tsoinvolvedinatthe
time.Butintermsofthefirstquestion,whereitcamefromforOxford:therehad
beenalotteryfundedproject,thatstartedoffastheEastOxfordhealthyliving
initiative,andthenitbecameknowasthehealthylivingpartnership,anditwasa
lotteryfundedprojectthatlookedathealthandwellbeingwithinareasthathad
beenhighlightedasdeprivedwithinthecity,wheretherewerehealth
inequalities.MainlyfocusedonhealthinequalitiesinBlackbirdLeysandEast
Oxford.Oneofthemainpartsoftheprojectwasaboutfoodworkandabout
healthyeating.Andsotherewereawholeseriesofprogrammesandworkshops
withinthecommunityaroundhealthyeating.Sotherewasquitealotofcontact
withdifferentgroups,goingoninthebackground.
Andthenwealsohadvariouspartnershipssetup.Thereusedtoexistan
Oxfordshirehealthalliance,andthereusedtobetheOxfordcommunitycaterers
network,wherecommunityorganisationssuchasthecouncilandtheNHS
representativeswouldgoto,butitwasmainlyintermsofsoupkitchens,and
providefoodforhomelessnessorganisations.Sotherewasthatelementaswell.
SoawarenessthatfoodwasquiteanissueinOxford
Andthen,coincidently,someonenewjoinedtradingstandardswithintheCounty
CouncilwhowasfromStaffordshire.
L:Okaythatwouldmakesense.
61
K:AndtheyhaddonesomeworkinStaffordshireandwantedtoseewhether
theycouldreplicateitinOxfordshiretodevelopitfurther.Soitallthencombined
togetherreally.Andthenwesetupapartnership.TheCityCouncilhadaccessto
lotsofgroupsandunderstoodtheactualontheground,qualitativesituation.We
thenhadtheNHSguyswhohadthatstatutoryviewonhealthyeatingand
pickingupthingsthroughthecriticalsideofthings.Therewassomeoneelse
fromtheCityCouncilinvolvedwitholderpeoplethatalotofthequalitativestuff
wascomingoutof:oldpeopleandyoungfamilies.ThenTradingStandards
joinedinandtheyhavealltheexpertisewiththeGISandthemapping.They
knewhowithadbeendoneinStaffordshire.Andthenwedevisedabitofaplan
abouthowwewoulddoit.E.g.foodbasket,groups,andthenTradingStandards
cameupwithallthetechnical,wizzystuff,andpulleditalltogetherifyouget
whatImean.Andthenwewentouttovisitoutgroups.AndthenTrading
Standardslayeredupwhereallthebusroutesmightbe,andwheretheshops
are.Probablysimilartowhatyou’vedoneforCambridge.
L:andthenIguessthiswholeprojectwasspecificallyaboutfoodaccess,andI
guessthat’swhatI’vebeenlookingataswell,butattheCouncilwasthereany
distinctiondrawnbetweenfoodaccessandsocialvaluesthatinfluencedfood
eating?
K:That’swherethethinkingaroundfoodaccessandfoodpovertyhasreally
movedonnow.Weonlyreallygottospeaktoacoupleofgroups,onlyreallythe
oldpeople’sgroups.ButcertainlytheworkthatGoodFoodOxforddidrecently
haspickeduptheculturalissues.Wenowhaveinroadsine.g.Asianwomen’s
group.E.g.intheseyouwouldhearaboutpeopleonlywantingtoshopincertain
placesandthewholesortofhalalelementsaswell.Theperceptionofwhatwas
healthyandwhatwasnotvaried.Healthyforsomegroupsmeantagoodportion
offoodandhavingmeatinit.Orquiteabitoffatinit,becausethatwasgood.The
womenwereexpectedbythementoservethemeatwiththefatonit.Those
sortsofthingswereveryinterestingbuttherewasnoscopetoincludeitinthe
accessproject.
L:Soatthetimeaccesswasthemajordrive?
K:Itwas,butbecausewewereworkingwitholderpeopleinparticulartherewas
acertainamountofsocialelementaroundreducingisolation.Forexamplethe
socialgroupthatweworkedwithupinCuttselowe,wherewemanagedtoget
theRedCrossbustotakethemshopping,actuallyaccesswasthedriverinitially
butitbecameveryapparentthatpeoplewantedenoughtimetograbacoffee,at
theSainsbury’s,andsomeofthebenefitswereenjoyed,Theythenbuildupa
relationshipwiththecaféstaff,forexample.Peoplegottoknoweachotheron
thebusbutitwasseenasasocialeventaswellasforfoodaccess.Buttheinitial
intentionwasabouttheaccessandthesocialpartofitwasabonus.
L:Yousaidatthestartthattheemphasishasshifted.Whendidtheshiftaway
fromfoodaccesshappen?
K:Idon’tknowbecauseIhadabitofagap.Istoppedworkingatthecounciland
thencameback.
62
NowIamnoticingthatthereisadifference.E.g.thepaperbyGoodFoodOxford
showsthattheyhadamuchmoreindepthprocessinthefocusgroups.Theyhad
awiderrangeofdemographicstheywereworkingwithanditpulledoutmore
stuff.
SoinmymindI’mthinkinggoshtherehasactuallybeenashift,cultural
preferencesnowaswell.Ofcoursethey’vealwaysbeenthere,butwiththe
limitedscopeoftheveryearlyprojectitwasn’twideenoughtopickupthatsort
ofstuff.
L:That’sveryinteresting.Partofthisdissertationinvolvesavastliteratureofthe
academicworkonfoodaccessandI’vefoundthatthereseemstobeashift
aroundabout2006/2007towardssayingthatthefoodaccessideaisnotgoingto
causecausation.Todothisyouhavetoconsiderawiderrangeoffactors.
Cultural,socialeconomic…Soit’sveryinterestingtoseethatfromyourside.
K:Andwhatyou’resayingwouldabsolutelymakesense.Itwasaroundthattime
withtheCuttselowegroupwhenwehadfinishedthemapping,thatyouknow
thiswasamechanismfor…Actuallywefoundthatpeoplemightnothavegoton
abusattheendoftheroadtogosomewhere.People’selementsofchoicewe
noticedwereimportant.Theychosetokeeptheircaranddrivemilesfor
example.Peoplewerechoosingwheretheywereaccessingthingsforcultural
reasons.Theycouldhaveaskedforleanmeatatthebutchersbuttheychosefatty
meatbecauseofwhatitrepresented.
L:Iwasgoingtothenaskyouabouttheresultsbutwe’vediscussedthatalready.
Itbeingtheminibus.Elderlypeopledirectlytargeted.Onequestionmightbe:
obviouslyyoutargetedelderlypeopleasatargetgroup,butwherethereany
solutionsthatyouhadinmindbeforeyoustartedite.g.wecouldputanewshop
inplaceorcreateanewbusroute?
K:Ithinkwecameupwithafewsolutionsasweweregoingthroughtheproject.
Werealisedthatoneoftheissuesthatpeoplecouldn’tgettoashopthatsolda
goodrangeofproductsatareasonableprice,Andiftheycouldgetthereitwasn’t
veryeasy.Sotheyweregettingtaxisyouknow.Sotheycouldn’tactuallyget
thereindependently.Sowethoughtthroughafewoptions.Theshopwas
considered,butactuallywethoughtthatwouldbelimitedbecauseitwouldstill
notstocktherangeofgoodsneeded.Andyouknowthewholesocialaspectwas
comingout.Anditwasaroundthetimethatonlineshoppingcameout.Wewent
backtothegroupaboutonlineshopping.Buttheydidn’tlikethisideabecause
theythoughttheycouldn’tgooutandchoosee.g.whatapplestheycouldget.
Theywantedthatwholeelementofchoiceandthatelementofgoingout.Soit
wasatthistimethatwerealisedthatthesocialelementwascomingout.
L:Thanksfortalkingabouttheresults.Iguessmylastminisectionintermsof
questioningisonGISinhealthpolicymoregenerally,andqualitativevs.
quantitativeinhealthpolicy.Inthatstudyspecifically,didyouseetheGISasa
complimenttothequalitativestuffordidyouhaveconcernsaboutit?
63
K:IfIrememberitdidcomplimentquitewell.InCuttesloweforexampleitdid
showthattherewasalimitationintermsofwherefoodoutletswerepositioned.
Transportwasabitdifficult.Ithinktogetabustoamainshoppingplace
involvedwalkingagoodhalfmileatleasttogettothebusstop.Soitdid
compliment.InoneofthegroupsdownCowleyroaditshowedthatfoodpoverty
wasn’tasmuchofanissueasitmightbe.Yep,there’sasupermarketonthathigh
street.Therearealsoexcellentbusservices.Soactually,whatpeopleweresaying
wasmatchedbytheGIS.ButthenagaintheGISwasn’tasadvancedasitistoday.
L:andintermsofmoregenerally,intermsofwantingtocreateapolicyand
convincingsomeoneatahighleveltodosomething,whereisthebalance
betweenquantitativeandqualitativedata?
K:Ithinkthatifyouaretryingtogetfundingandbuyin,theshifthastogo
towardsthequantitative.Withfundingseriouslyrestrictednow,reducedtoa
fewspheres,itdoescomedowntofigurestoprovethatthereisanissue.Prove
thereisanissue.Okaytenpeoplearetellingyouthereis,butneedingthat
analysis.Butthequalitativeisnecessarybecauseitbringsitalive.Iwouldsay
thatthatthebalanceshouldbeshiftedtowardsthequantitative,sothereisthat
strongevidenceofwhatmightbecausingtheissues,thenbackingthatupwith
whatpeoplearesayingtobringittolife.SoyeahIthinkthathavingthat
evidencethereisabsolutelynecessary,gettingitasrobustaspossible.Case
studiesarealsouseful;Mrxdoesthisandthisandtheimpactuponthisisthis
andthis.Thatkindofthingsbringsittolifeaswell.
L:InmygeographydegreeIamtaughttotreatquantitativeevidencewithabitof
scepticism.IsittreatedasobjectiveintheCouncil?
K:Ithinkitdependonthesource.Ifyouhavedataontransportroutes,itiswhat
itis.Ifyouarelayeringupthingslikeindicesofmultipledeprivation,itiswhatit
is.It’snationalstatistics,youcan’targuewiththat…Youknowwhatthough,any
dataneedstobetreatedwithcaution.SoIwouldn’ttreatanydatadifferentlyto
anyotherdata.Alldatahasthecaveatthatit’sonlyasgoodastheperson
recordingandpublishingit.Andyouknowsometimesitdoesn’tgodownto
smallenoughlevelsandallthatkindofstatisticalstuff.Butifyouweregoingto
chooseyourdata,certainlytherearesomethingsthatyoucan’targuewith.Bus
routes,indicesofdeprivationandstufflikethat,therobuststuff.Thereissome
stufffromtheONSaboutpeoplebuyinghabitsandthingslikethatthatyouhave
totreatwithcautionbecausethat’speoplereportingit.Andpeoplearen’t
necessarilycorrectinwhattheywritedown.Theirperceptionofrealitymight
notberight,theymightthinkthattheyonlybuythreechocolatebarsaweekbut
actuallytheybuyoneaday.Sotreatthingswithcautionandbereallyselective
abouttheevidencethatyouarecollecting.
L:Wellthankssomuchforthat,you’veansweredthequestionsIhad.Justone
quickquestion:canIuseourconversationformydissertation?
K:Yesnoproblem,gladIcouldhelp!
L:WouldyouminditifIsentyouanemailaboutanythingthatcomesup?
64
K:Yeahnoproblemofcourse.Wellgoodluck.Byebye.
L:Thanks!Bye