SPSSasan‘inscriptiondevice’:

fromcausalitytodescription?

EmmaUprichard, RogerBurrows and

David Byrne

Abstract

Thispaper examines thedevelopmentofSPSSfrom1968to2008,andthemannerinwhichithasbeenusedinteaching andresearch inBritishSociology.Wedothisin order torevealsomeofthechangesthathavetakenplaceinstatistical reasoningas an inscription device in the discipline over this period. We conclude that tocharacterisethesechanges asashiftfrom‘causal’tomore ‘descriptive’ modes ofanalysisistoosimplistic.Suchashiftiscertainlyapparent,butitmeshesincomplex wayswith arange ofother –just asimportant–changes, that together mark aphase-shiftinthefunctioningofsociologicalquantification.

Introduction

From the late 1960sonwards, British sociologists have had accesstoalarge numberofdifferent statistical software packages.Itisdifficulttoestimate the exactnumber thathavebeenused,butthere haveprobably been50ormore systems that have been utilised forteaching and research atdifferent times overthepast40yearsorso. Some,suchasGLIM(GeneralizedLinear Inter- activeModelling)developedbytheRoyalStatistical Society’sWorking Party onStatistical Computing,werepopular foratime,butthenfelloutoffavour. Minitab,waspopular forteaching inthe1980s andremains soinsomeinsti- tutions.Inaddition,arecent‘roughandready’auditofquantitativelyinclined colleaguesgeneratedthefollowinglistofpackagesusedinasustained manner inrecent decades:LISREL;MLwiN;R;SAS;andStata;somecolleagues also noted howwidelyExcelisnowusedforbasicquantitative analysis.Readers, undoubtedly,willbe able to add to this list.There has been one package, however, that hasnot onlyenduredbut hasalsoremainedthe most widely known andused:theseemingly ubiquitousSPSS –theStatisticalPackagefor theSocialSciences.Socentral tothe experienceof‘becoming’and‘being’a sociologist inBritain overthelastfewdecades,neitherthematerialandthe semioticfunctioningofthisparticularpieceofsoftwarenortheparaphernalia

.

surroundingit(manuals, textbooks,courses and the like),canbeignored if wearefullytounderstandthepanoply ofinscription devicesthat constitute sociologicalformsofknowledge.

For those familiar with using SPSS in their sociological teaching and research overanumberofyears,butwhohavenotpaidmuchattentiontothe changing context of its development and dissemination,a visit to www. spss.comin2008 mightcomeassomething ofasurprise.1 First,thecorporate history section2 makes clear that, althoughSPSS‘stood for the Statistical Package for the Social Sciences’(emphasisadded),itisnow no longer an acronymbutanamed softwarebrand.Indeed,onehastodrilldowndeepinto thesitetorecover thisetymology.Second,forthosewhomight,atonetimeor another,havefeltsomesortofvaguesociological‘responsibility’ forthesoft- ware,thistenuous disciplinary connectionwithoneofthecore‘toolsofour trade’isquicklyobliteratedwhenitbecomes apparentthatthesitefunctions tointerpellatenot‘socialscientists’,butthoseseeking‘businesssolutions’.One quickly feelsnaïve tohave imagined that inanera of‘knowing capitalism’ (Thrift,2005) someofthecoretoolsofthesocialscienceswouldnothavebeen fully ‘commercialised’ and ‘globally branded’ in this way. Naïve or not, the profoundand stark transformationofSPSSfrom a tool for empirical socialresearch toacorporate behemoth primarily concernedwithsomething called‘predictive analytics’ (aboutwhich more below) has been fast and dramatic.

Our storyofthechange inBritish quantitativesociologyisthereforeone thatis toldalongside changestoSPSSanditsimpactonwaysofknowingthe socialworld.After all,ifitisthecase–ashasrecently been argued (Savage and Burrows, 2007)–that weare atthe cuspofacrisisof‘causal’formsof empirical analysis in sociology, which results in a need to recover more

‘descriptive’formsofthediscipline,thenwemightsuspectthatthiswillmani- festitselfemblematicallywithinthealgorithms, interfaces, visualisationtools and other formsofinscription device that SPSSoffersup.Indeed,weargue that SPSSInc.andthenumerousproductsthat theynowproduceunder the auspicesoftheSPSSbrandrepresentnotjustanearlyinstantiationofthisshift insociological orientation,but ratheraprefigurativecatalyst inbringing it about.

Underpinning both the changes in quantitative sociology and SPSS, however,aretheprocesses ofdigitizationandassociatedchangesintheavail- abilityofdata.Indeed,weconsider thisaspectofthewidersocialworldtobe anessentialdriverof thevarioustransformationsdescribedhere.Althoughthe processesof digitizationremainalargelyimplicitstrandof ourstoryof change, theyarefundamentalastheyhavebeenquietlymakingtheir presencefeltin quantitativeresearch more generally,sotheycannot beignored here either.

Thisoverallargumentemergesfromtheresearch conductedtopreparethis paper.Attheoutset,weworked outacrudeschemaforperiodisingtherecent history ofquantitativemethodsinBritish sociology drawing on numerous historical sources,ranging fromgeneral accounts ofchange,suchasAbrams

607

(1968),Hindess (1973),Irvineet al.(1979),Kent(1981),Bulmer (1985)andso on,rightthrough tosomeoftheInstitute ofSociology’sreports onteaching acrossthesocialsciences.Because itbecame clearthat theeffectsofthePC wereimportanttothechangesvisibleinsociology,wethen‘mapped’thisonto developmentsinbothhardwareandsoftwareforstatistical analysis.Thistook ustothe corporatehistory section ofthe SPSSInc.website, whichoffers a reflexiveaccount ofhowSPSSInc.cametopossessawideportfolioofsoft- ware,functions andcustomers.Thewebsiteoffersbothanarrativehistoryand averyinteresting attempttoperiodisetheinstitutionalhistory ofthebrand. Ironically,however,ourownperiodisation ofsociologyanddevelopmentsin statisticalanalysisandthenarrativeaccountofferedupbySPSSInc.provedto bealmost exactlyhomologous!Inwhatfollows,therefore, wedescribe SPSS Inc.’s self-periodisationand how it meshes with the post-1968 history of British quantitativesociology.

SPSSisborn

Before SPSS and some other packages became available, researchershad nochoice but learn tousethe highlevelprogramminglanguage Fortranto write their own programmesand relied on the likes of Veldman’s (1967) FortranProgrammingfortheBehavioral Sciencestohelp them doso.SPSS wasdevelopedin1968byNormanNie, TexHullandDale Bent –allthen at StanfordUniversity –asan alternativeFortranbased program,specifically designedtoanalysequicklythelargeamountsofquantitativesocialdatabeing gatheredby Faculty and Graduatestudents. Initially, therefore,SPSS was primarily aimedatallowingresearcherstodobasicdescriptivestatistics,cross tabulationsandregression.Otherpackages wereusedforother purposes;for example clusteranalysistendedtobecarried outusingCLUSTAN.However, oncenewsthatSPSS was availableforbasicstatisticalanalysis,whatwas colloquiallyknownas‘theStanfordPackage’wassoonindemandinother US institutions.

In 1969,Nieand Hull moved to the University ofChicago –Nieto the NationalOpinionResearchCenterandHulltotheUniversity’sComputation Center. Ayearlater,the publishersMcGraw-Hillrepackagedthe documen- tation that had been producedtoaccompanythe software asthe firstSPSS User’s Manual. This sparkeda huge demandfor the programmeand the incomegeneratedfromtheroyalties fromsalesofthemanual wassubstantial enoughtothreatenthe non-profitstatus ofthe University ofChicago;soin

1975,SPSSbecame incorporatedasabusinesswiththetwofounders,Nieand

Hull,asthecompany’sexecutives.

Thefirstversions ofSPSSavailable intheUKwerewritten inFortran,so usersstill requiredquitehighlevelprogrammingskills.Theyranonmainframe computersystemsstoredinlargeairconditionedrooms,oftenlargeenoughto besmalllecture theatres.Themainstreamingofsuchcomputerizationmadeit

608

possible for quantitativesocial researchersto radically speed up analytical procedures.Nevertheless, throughoutthe 1970smost researchersinBritain stillusedvariationsoftheHollerithPunchCardsystem(Grier,2005).Thiswas oneofthefirstmassdata storage andsortingsystems,anditemployedindex cardsmarkedwithholes,whichactedasaformofcode.Later counter-sorters became available which facilitatedthe use of cards that could be multi- punched.SaraArber, currentlyaProfessorofSociologyattheUniversity of Surrey,andoneofthesociologicalpioneersofsecondaryanalysis(Dale et al.,

1988)wasaveryearlyuserofthesystem,andinapersonalcommunicationto thesecond named author, sherelates someofherrecollections. Importantly, theyareremarkablyresonantofanumberofothersweheardwhilstpreparing thispaper:3

IfirstusedSPSSinthe US,whenIwasagraduatestudentin1973–74at Michigan, usingpunch cards (ofcourse). Igotaone year Lectureshipat Surrey starting inSeptember1974toteach methodology,and one ofthe thingsthatAsher Tropp [theFoundingHead ofSociologyatSurrey anda keyfigureinthedevelopmentofsociologicalresearch methodsintheUK] wasparticularlykeenonwasteaching studentstoanalyse‘real’surveydata usingSPSS.So,in1974–75,Istartedteaching undergraduatestudentshow touseSPSSandweusedtheUSGeneralSocialSurvey. . . Atthistime,all theSurreycomputingwasvia anovernightlinktothecomputerinGuilford Street, University ofLondon.Studentsused to punch their cards for an SPSSrun(inadedicatedpunchroominthecomputercentreatSurrey)and then the cardswere submittedtoLondon–the following dayyouwould receivethepaper output (either showinganerror orsomeanalyses,ifyou were lucky).So,itwasveryfrustratingforstudentsand everyoneelseto havetowaitovernightforanyresults...Atthistime,Iwasunawareofany UKsurveydatasetsthat couldbeusedforanalysisusingSPSS.

These personalrecollectionshighlight a numberof significant issues that are importantwhen trying tounderstandthe history ofBritish quantitative sociology,whichareworth spellingoutmore explicitly.

First,itisworthnotingthesignificantrolethattheDepartmentofSociology attheUniversity ofSurrey played inthedirectionquantitativesociologyhas taken. ItsearlyMScinSocialResearch,alongwithitsshort training courses, placed Surrey asoneofthefirstprofessionaltraining grounds insociological research.Surrey,alongside theUniversity ofEssexSummer SchoolsinSocial ScienceDataAnalysisCollection,wasthemajorconduit forthepromulga- tion ofdevelopmentsinstatistical analysis inthe discipline.Surrey wasalso crucialinmaking available large-scale officialdata setsineasytouseSPSS versions.The work that SaraArber, NigelGilbertandAngela Dale didon converting GeneralHouseholdSurvey(GHS)datafilesintofully documented SPSSsystems files(Gilbertet al.,1983),4 wascatalytic intransforming both teaching andresearch intheUK.

609

Second, the process of doing quantitativeresearch has changed signifi- cantly.Both data and programminginstructionshad tobeenteredoncards whichwerepunchedbyoperatorsfromprintedsheets whichresearchershad to fillinpainstakinglyfor themselves. One card wasone row ofdata; one mistake onthecardmeant throwing itawayandre-punching theentire card. Data oftenhadtobecopiedfromhundredsofquestionnairesorfromprinted sources.Thiswasallsimplytoproducethe data set,asthere were then, of course, noonline resources, althoughsometimes data could beobtainedas cardsetsor,lateron,as magneticdatatapes(the1971UKCensuswasamongst thefirstdatasetstobeavailableinthisformat)thatcouldbemountedonlarge tape readers.

AsSaraArbernoted, themode ofcommandanddata entry andtherela- tivelyslowspeedofcomputing atthattimemeantthatturnaroundwas atleast overnightorlonger;more frequently,itwould often take much longer than anticipated,sincewaiting foralineof‘errors’ waspart and parcel ofdoing statisticalanalysis.Thethirdnamedauthorremembershavingseveralabortive runsuntilhehadworked outthat hehadusedexclamationmarksinstead of thenumberone.Mistakessuchasthesewereeasytomakegiventhenecessary precisioninvolved inenteringdata.Forexample, caseswhichmightgoonto multiple linesofacardif there weremanyvariables hadtobespecifiedina mannerthat nowappearsarcane: 10(F8.0)/4(F8.0,F6.2,F5.2,F1.0)...Output cameonprintedsheets fromlineprinters–hundredsofthem.5

AlthoughtheHollerithpunchcardsystemandcounter-sorterstransformed quantitativecollectionandanalysis,comparedtotoday’scomputing standards, itwasincrediblyslow.Italsodemandedareasonablelevelofstatisticalunder- standing,alongwithpractical skillsindata entry andcommandlanguage.As obvious asitmayseem,italsorequiredtheabilitytotype.Althoughbythe

1970stypingwaspretty wellauniversal skillamongAmericans, British aca- demicsstilloftenwrotewithpenandpaper andhadsecretariestypeuptheir work,soinfactmanyacademics,especiallymaleones,couldnottype.Indeed, ifone can read beyond the acerbic critique ofthe discipline containedin Malcolm Bradbury’s(1975) The History Man, much of the descriptionof work-a-daylifecontainedwithinitdoesincludeafairlyaccurateportrayalof thetechnologiesanddivisionsoflabour thatpertained inBritishsociologyin theearly1970s.Allinall,doinganykindofquantitativeworkwasincredibly labour intensive.Thishadimplications onwhohadtheresourcestoconduct largescalesurveys,whichinturn placed governmentasthe solelarge-scale surveyprovider.

Finally,asechoedinArber’srecollections,bythelate1970sandearly1980s, althoughSPSSwasstillforUKacademics primarily adenizen oftheirownor otherUniversitiesmainframecomputers,accesstothesecomputerswasoften done remotely, largelythroughUNIX commandlanguage initiallyviadumb- terminals,butlater fromdesktopPCsabletoactasterminal emulators.Now commandfilesetscould bewritten electronicallyand stored forcorrection. Similarly, large electronicdata sets, such as the Surrey GHS files,slowly

610

became available.Operationswerefaster,eventhough theywerestillinitially baseduponbatchprocessing;commandsanddatawouldbeenteredbutwere thenoftenqueuedandsubmittedovernight.Later still,mainframeversionsof SPSS(andlaterSPSSX)became interactive;onecouldsubmitcommandsand dataand,afteratime,outputcouldbedeliveredbacktothescreen.Atthetime this was a revelation;it representeda step-changein the speed at which analyses couldbecarried out.

The‘PCturn’

Itwasonlyinthemid-1980sthat SPSSintroducedthefirstmainframestatis- ticalpackage abletorunonaPC(versions wereshipped abletorunonIBM PCs,PS/2running OS/2,Apple Macintosh,UNIX and VAX/VMSworksta- tions,minicomputersandlargersystems,as well as mainframes).SPSS/PC+ran underMicrosoftDOSbutstillrequireddataandcommandfilestobecarefully enteredusing,whatnowappearstobe,esoteric SPSSsyntaxbutwhich,atthe time,became acentral inscription device familiar to allthose undertaking quantitativesocialresearch.ButtheshifttothePCwasincredibly important, andduring theperiod between around1984and1992,SPSSfurtherconsoli- dated itsposition asthe‘industry standard’.

Thisperiod alsomarksthebeginnings ofanimportant qualitativeshift–a phase shift even –that occurredinthe development,pedagogy and use of quantitativemethodsinBritishsociology.Thisiswhenthecomputerisationof statistical teaching inHigherEducationreallytookoffandtheteaching space ofthe ‘PClab’began to emerge. Note, however, that the pedagogical shift involvedmore significant,albeit subtle,changesinemphasis onhowtointer- pret, analyse andunderstandquantitativedata.Asweimplythroughoutthe remainderofthepaper,thisshiftnotonlyresultedinadifferent wayofdoing quantitativeresearch,butinadifferent kind ofquantitativeresearch aswell. Ineffect,prior tothe arrival ofSPSSonthe PC,there had been alotmore concentrationon both the philosophicalnature of data –qualitativeand quantitative–and with it,the role of the researcherin interpretingand constructingquantitativedata.Incontrast, post-1980squantitativepedagogy placesmoretimeandfocusontheoutputthanthelabours oftheresearcher.

Tosomeextent,ashiftinpedagogy wastobeexpected–studentsandtheir teachersneededto learn how to conduct statistics using a computerand related software.Inturn,there wasarapidgrowthintheproductionofSPSS (and,toalesserextent,Minitab) guidesandworkbooks on‘howtodo’quan- titative analysisusingthesoftware packages.Butwhataccompanied thisshift in doing quantitative analysis with a particular software package was a perhapssurprising substantive shift inhow to do it.Textbooksand student guides shifted tohaving much lessdiscussion onthe processes and theories involved ininterpreting andconstructing quantitativerepresentationsofthe socialworld than theyhad done before the widespreaduseofthe software.

611

Evenwhentheemphasis wasnoton,say,SPSSperse,thefocuswasstillmuch more about ‘techniques’ than itwasabout the‘construction’ ofquantitative socialdata.

The shift inhowquantitativemethodswere taught isanimportantone. The‘blackbox’natureofSPSSquietly transformedtheuserastheybecame increasingly dependenton the software and computertechnologyto think statisticallyandtoperformanalysis.Studentswere,asamatterofcourse,taken outofthedialogic spaceoftheclassroom andintothemore individualised, taskperformingspacesofthecomputerlab.Here,theylearnt whichbuttons to click,which menus to use and which parts of the outputto focus on. The absolute physicality ofthe task involving,asitdid,working withcards, machines,tapesandhugeamountsofpaper asmaterialinstantiationsofcases, variables andsoon,became absent andwasreplacedinstead withspeedand theapparentdematerializationofbothdataandthemechanicsofanalysis.This becameevenmorethecasepost-1992asSPSSmigratedtoMicrosoftWindows andtheuseofSPSSXonmainframesdwindled.

Thistendency to‘doquantitativeresearch withSPSS’hasbeen buttressed by the publicationof a plethoraof textbooksand guidebookson how to conduct traditionalstatistical analysesusingSPSS,wherethefocusisfirstand foremost onthebasicsofnavigating thestudentthrough SPSSwindows,data inputtingandediting,throughtohowtoconductvariousstatisticalprocedures and howtointerpretthem.The task of‘managing the software’ haslargely been attheexpense ofdiscussingmore philosophicalconcerns andthedeci- sionmaking processes involved inmaking substantiveinterpretationsabout theanalytical findings.

This isnot to saythat contemporarystatistical textbooksare devoid of theory;manyareofcoursestilltheoreticallydriven.However,itispossibleto thinkofquantitativesociologybooks appearingafter the‘PCturn’,asfalling intooneofthree camps:those that focusonunderstandingandinterpreting quantitativedata atthe expense ofalmost anyexplicitSPSSguidance, even though itmightbereferredtointhetext(egMarsh,1988;Byrne,2002);6 then 1 there arethose attheother extreme, that arealmost entirely SPSSdriven to thepointthattheyendupbeingfirstandforemostguidebooksonhowtouse SPSSandthenaccounts ofhowtoconduct quantitativeresearch (egField,

2007;Pallant, 2007);and finally,those that lie somewherein the middle, attemptingtocombinetheinterpretationofquantitativedatawithsomebrief guidance onwhattoclickandwhatparts oftheoutputtofocuson(Fielding andGilbert,2006;Bryman andCramer, 2008).

Interestingly,while insociology teaching and learning broadenedthem- selvesoutbydevelopingthese three alternativepedagogic pathways, wesee somethingsimilarintheway thatSPSSdevelopedas well.Upuntilabout1994, SPSShadremainedfocused ontheproductionofitsownstatistical products. TheWindowsversionofSPSS,forexample,generatedhugeincomeforSPSS Inc.–reaching some$84mby1996 according totheSPSSwebsite.However, assisted bythe commonalityofthe Windowsinterface, itbegan to acquire

612

other statistical productsfrom other companies whichitattemptedtointe- grateintotheoverarching SPSSbrand.Fromacommercial pointofview,this wasasuccess:againaccording totheirwebsite,by1997SPSShadrevenuesof

$110m.Clearly,thiswasnolongeranacademic smallbusiness;ithadbecome aglobalcorporation.

Towards‘predictiveanalytics’

Theprocessofacquisitionacceleratedpost-1997asSPSStookoveranumber ofdifferent software companies. Insteadofacquiring productssimilartothe ones that had long been at the core ofSPSSasapiece of‘socialscience’ software,theproductsthatwerenowbeingsoughtwereonesthatoffered‘data mining’,‘business intelligence’, ‘webanalytics’,‘online analytical processing’ and‘textmining’.

On the one hand, then, SPSS isstill concernedwith some things most sociologists willrecognise: marketresearch, survey research,public health, administrationand institutionalresearch,statistics,structuralequationmod- elling,education,governmentandhealth. Ontheother hand,there areother items,many foregroundedover those justmentioned,withwhichmost soci- ologists may be less familiar: marketingeffectiveness, fraud detectionand prevention,risk management, enterprise feedback management, business intelligence, data mining,text mining,webanalytics,financialservices,insur- ance,retail,telecommunicationsand,atthetopofthelist,predictiveanalytics. According toSPSS,thisshiftinsoftwaredevelopmentwastomeetan‘expand- ingneed forunderstandingever-increasingvolumes ofdata,andtosupport-

. . . the widespreaduseofdata indecision-making.’ By2002,revenueshad almost doubledto$209m.

Our sense,asroutineusers ofSPSS,isthat within British sociology this changeindirectionhas not really registered.Most social researchershave continuedtouseSPSSinthemannertowhichtheyhavebecomeaccustomed, primarilyfortheanalysisofprimarysurveydataandthesecondaryanalysisof large data sets using a fairly routineset of statistical procedures. Some approacheshave gained inpopularitywithimprovementsincomputational power–logisticandprobitregressionanalysesforexample(reliantastheyare oniterativealgorithms)–and some methodologicaladvances –multi-level models forexample –havebecome mainstreamed. FewintheBritish socio- logicalheartlands, however,appeartohavetaken cognisance oftheimplica- tionsforthedisciplineoftheprofoundprocesses ofsocialdigitizationthatare occurring (McCue,2007;SavageandBurrows,2007;Thrift,2005).Yetthisshift hasbeen fundamentaltodrivingthestrategic shiftthat SPSSInc.havemade towards, what they articulateas,the‘ageofpredictiveanalytics’.AsMcCue neatly sumsup:

Whetherit iscalled data mining, predictiveanalytics, sense making, or knowledgediscovery,therapid developmentandincreasedavailability of

advancedcomputationaltechniqueshavechanged theworldinmanyways. There are few, if any, electronic transactions that are not monitored, collected, aggregated,analyzed and modelled.Data are collected about everything from our financial activities to our shopping habits. (McCue,

2007:xxv)

Notsurprisingly perhaps, SPSSInc.hasquicklycaught ontotheimplications that the increaseddigitizationofdata implies,and ithassuccessfully estab- lished amarketsegment inpredictiveanalytics. Indicativeofthis shift was their acquisitionofthe Netherlands-basedDataDistilleriestowards the end of2003.

Sofar,the turn towards predictiveanalytics inquantitativeresearch has taken holdprimarily withinthebusinessandpolicysectors,whilsttheimpact onBritishsociologyhasbeenmuted.7 Ontheonehand,asasubstantivetopic, ithasbeen acentral featureofwhathascometobeknown as‘surveillance studies’.Sociologistshavevoicedconcernsabout theimpactoftheimplemen- tationofsuchtechnologiesconceptualisedas‘softwaresorting’devicesableto producea‘phenetic fix’on society (Graham,2004;2005;Lyon, 2002;2003; Phillips and Curry, 2002).On the other hand, however, the methodological implications ofthesetechnologiesforsociologicalpractice haveamountedto littlemore than vaguerumblings (SavageandBurrows,2007)about thepos- sibilitiesthattheyafford.Ofcourse,theserumblingshavenotbeenarticulated intermsof‘predictiveanalytics’–thiswouldsmackofacommercialsensibility manywithinBritishsociologywouldfind objectionable–butthemethodologi- callogicbehind the approachhas certainly been explicitly referredto.For example,somehavesuggested that geodemographicandsocio-spatialprofil- ingoffersome newsubstantiveand methodologicalopportunities(Burrows andGane,2006;Parkeret al.,2007;SavageandBurrows,2007)inrelation,for example,todebatesaboutthespatialisationofsocialclass(Savageet al.,2005). Othershavepointedtowardstheimportanceofdescribingandexploring data toidentifyparticulartypesofcases–akeyelementofpredictiveanalytics.For example, ChapterSixofByrne’s (2002)Understanding Quantitative Datais precisely about ‘exploring,describing andclassifying’which,itisargued, are keytostudying thecomplex socialworld.

Although‘predictive analytics’ might seem like an uneventfulturn of events,wesuggestthattheyareemblematicofanewmethodologicalethosin quantitativemethods. Indeed,weargue that whereas the mid-to-late1980s markedafirstphase-shiftrelating totheincreaseduseandavailability ofthe PC and relatedsoftware, today the course of British quantitative social research iswitnessing asecond radical shift.Thisiscertainly beingdriven by theincreasingubiquity ofdigitizationprocesses but,relatedly,alsobychanges indomain assumptionsabout contemporary socialontology.Forthebusiness sector,thisnewmethodologicalethosisdriven,primarily,bythecoldlogicof theprofitmotive;insociology,itcomesviatherecent‘complexityturn’(Byrne,

1998;Urry,2003),whichinvolvesaquitefundamentalreappraisalofanumber

614

ofdomain assumptionsthat,potentially,involvenothinglessthana ‘paradigm shift’.

In what follows,weoffer anover-characterisation ofthismethodological shift and what isat stake here.We do this for two reasons. First, to keep engaged those fewreaderswithnoinherentinterestinSPSS,but anactive interestinsomeoftheconceptualaspectsofthepaper hitherto.Second,asa methodologicaldevice,topresentastark ideal-typical contrasttowhichwe hope colleagues willreact, inafurtherattempt(Savage and Burrows, 2007; Byrne,2002)tojoltthediscipline outofamethodologicalcomplacency that the‘comingcrisis’will exploitmercilesslytothedetrimentofusall,unlesswe confront,in a pro-activemanner, the implications of the new realities of

‘knowingcapitalism’(SavageandBurrows,2007;Thrift,2005).

The‘newface’ofquantitativeresearch?

MostsociologistsinBritainhavenotexplicitlyconnectedwhatwearegoingto characteriseastheemergent‘newface’ofquantitativeresearch withtherise ofthe digitizationofroutinelyconstructedtransactionaland administrative data (Savage andBurrows,2007)orthe‘everyday life’data banks that‘Web

2.0’technologiessupply(Beer andBurrows,2007).Yetthis‘digitalturn’inthe availability ofdata isprecisely what makes the SPSSrhetoricofthe‘ageof predictiveanalytics’ both possible, and some of the methodological shifts associatedwithit,potentially obligatory.Wesummarisesomeofwhatfollows inhighlyschematic forminTable1.

The emergent,or‘newface’,ofBritish quantitativesociology operatesat twolevels.Thefirstrelatestotheactualtechniquesthathavebeen‘invogue’at variouspointsoverthepost-warperiod.Thisisarelativelyeasytasktonarrate and,indeed, wehavealready made astart onthisintheaccount ofSPSSwe have already provided.Over time some statistical techniqueshave become favouredover others. This isnot surprising –things change, fashions come and go,new methodsemerge and new technologiesoffer new affordances. However, SPSSandother quantitativeanalysissoftware packages areprime examplesofasetofinscription devicesthatnotoriouslysetsometechniquesas

‘default’options,whichmayormaynoteasilybealteredtoallowothers tobe foregrounded.Letustake,by wayofasimpleexample,thestem-and-leafplot.8

ThisisoneofTukey’s(1977)manygraphsusedtodescribeasinglecontinuous variablethatis,inmanyrespects,moreinformative thanitshistogramcousin. Likethehistogram, itdisplaystheoverallshapeofthedistribution,butitalso providesaveryprecisedisplayofthevaluesthatgivethedistributionits shape. Thestem-and-leafplotisavailable inSPSS, butunlikethehistogram, itisnot readilyavailable viatheupfront‘graph’menu.Instead,itishidden awayasa tickboxunder ‘options’ in‘descriptive statistics’.What issetas‘default’ in SPSS,andindeed thewayitisprovidedtotheuser,then,affectstheinterpre- tation and use ofeach and every technique.The stem-and-leafplot –asis

615

Table1Aschematic overviewofchangesinBritish quantitativesociologicalresearch

Shift‘OldFace’Emerging‘NewFace’?

Technique•Parsimoniousmodelbuilding,egmultipleregressionanalysis,factor analysis,multi-levelanalysis,etc.

•Describing andexploring groups,egclusteranalysis,simulation (micro-simulationandagent-basedsimulation especially),correspondenceanalysis,QCA,etc.

Generalisation•Generalisationaspossible,reliable anddesirable•Moveto‘moderatumgeneralisation’whichare somewhatreliable,butremain more orlessenduringrelativetofuture research andfindings

Causality•Singlecauses

•Linear causalmodels

• Generalcausal‘laws’–outcomeconsistently explainedthroughparticularvariableinteractions

•Faith in‘finding’causes

Prediction•Onus onpredictingtheultimatesinglefuture

•Consideredpossible

Sampling•Probabilitysampleasbestpossibleformofknowingpopulation

•Sample usedtostatistically infersamplefindingstopopulation

•Statistical significancetesting askeytounderstandingpopulation

•Multiple,contingentcausality

•Configurationalcausality –sameoutcomepossible throughdifferent variableconfigurations;different outcomespossiblethroughsamevariable outcome

•Complex,nonlinearandinflux

•Lessfaithin‘finding’causes

•Onus ondescribing multiple possiblefuturesinstead ofdeterminingonesimplefuture

•Possibilityofpredictionbecomes questionable

•Populationdata widelyavailable throughincreaseddigitizationofdata

•Statistical inference isseenasunnecessary

•Probabilitysampleusedtoconfirmdescriptionsofpopulationratherthaninferring tothem

Interpretation•Focusonexplanationandconfirmation•Focusondescription,exploration, classification,caseprofilingandvisualisation.

VariablesandCases •Focusonthevariable•Focusonthecaseanddescribing typesofcases.

portrayedthroughSPSS–isseeminglylessusefulthanthoseother inscription devicesthatarereadilyavailable viatheGraphmenu.

Of course SPSShas not evolved inavacuum. Nor has itbroadenedits analytical capacities randomly. Inscribed inSPSSitself isalong history of

‘analyticalmemory’whichis visibleaseachnewversionbringswithititsown storyofwhatusersmightnowwantandneed,andofcourse whatistechno- logicallypossible.Some techniques,suchassocialnetworkanalysis,are still absentfromSPSSwhereasothers,suchascorrespondenceanalysisandcluster analysis, have become ‘mainstreamed’.Conversely, basic new mechanisms become available as SPSS becomes more ‘interactive’. Thus, for example, version12.0 camewith(among other things)anew‘visualbander’ tool(now called‘visualbinning’–theargotitselfisnotunimportanthere,andelsewhere, of course).Thisallowsforacontinuousvariabletobedisplayedas ahistogram, whichcanthenbeinteractively‘cut’intogroupstoproduce amoremeaning- fullydata driven ordinal variable.