Rul/t'lin de Mhhodoll1gie Sociologique lSi) rue I\lllchel, F 75017 Parts), (kloher 1999, N 64 - p.25

METHODOLOGICAL ISSUES IN SURVEY RESEARCH:

A HISTORICAL REVIEW (I)

by

Wim dt> Iit'er {Stallsllrs Ndherlandsl

Edith D. de Leeuw (M,thodlkAJ

,Johannes van del' Zouwcn (VrW Thl[Vf"r~lIf"itl (2)

R~sumt. Probl~mu mHhodolog1qu~. d..a.n. ta recherche par enquttu - Un bbtorique. Dills (;el .l111t"1r, n()l1 ... plrst"IlI11IIS WI survol hlstonque d("s ("nqueles sodaJes el c1ecnvons Ie developpemenl Ill ... lorlqllt" des slallstlqut"s ("I de la n1ethodologle {'I11pirique des enquetes. Les ongtnes de la recherche Jldf t'!Hllwtf"S dalt'nt du lklmt flu J0enle sleele: la premiere enquele emplnque a ele faHe en Angleterre ('n 1~}12 par Bowky. us llIt-fhodes modemf"s d'enquet(" trouv("nl leurs ortgtnes dans les annees 1930 q\lalld trois aspf"cts lIlaJnJrs dl'" I'enqut-le ont ete developpes et ~rlecl1onnes: technIques d'n harlUllorlIlagt", It''dllllqu(''s de roll("('le df" donnees et les melhodes slaUsUques d'analyse de donnees D,1I1'" II'S allfl('rs 1~}50, l'IJllervtew f"1I face-a.-lace s'esl devl"'loppe et des manuels ont ete publles. Au debuts drs allllt"es 1~}70, I'lnlf"rvtf"w par te-kphon(" est apparu et renqu~t~ postale s'est rependue. Dans Il'~ .lllllt"l'~ I !JHO, I !:}!JO, I'lntt"rvt("w assJst(" par ordlnateur et les euqlletes electronlques ont ele dC\"l·loJlJlt'('s. Ell 1IIf"llJe tc'mps, dt"'s prol.!;r("~ dans la mesurt" de I'erreur, claus I'ti'tud~ de la 110n~reponse et lidll ... it'" (,olllrolr' gloh<ll lit" Ia quallte des dOIHl("(,,S, ont de n~allses Dans Ie nouveau mtllti'natre, les plliH lilt'S df" basc ck ['(,llquelt" rllq)lrlquf" contlnll("ront d'avolr cours mats les staUsticlens d'enquele, \ UlllllW llowkv, :-.t"I..llrllt ~llrJlrls dt" voir la structure d'une enqutole en 2012, Sans doule, elle s'appukra '-Ill <II- 1l<HIVf"lk:-. tt'c-hllolog!t"s ("l COIlcf"lItrt"rall s\lr la rt"dllctlon de la l:i.che dll rt"pondanl. tOlit t"Il ,111ll'llol ;lilt 1.\ qllalltl' dcos c101l1\("("S Bowif"y appreclt"ralt de voIr que d("s methodes sophlstlque~s de J",lIle IID'1 Ill' la 1101\·rt"pOlISt' t"t dt" I'("rrellr dt" Illt"sure Incorporees dans Ia slructure d(" I'enquete pour .11111"1101 n Lt quall~(' llllalt'" des rt"sulla(s Metbode et .t.tl.tique. d'enquete., Enquete. 8OCIale., Enquete, eroptrlquea Nouvelle!l tecbnologies, QuaUtf: de. donn~ea, Non-r~ponae, TAcbe du n~polldrnt. TQM, CBM

Abstract. III thh paper. we prt"~t"nt a llistoriuli {l\'('rvt("w of social surv("}'S and dt"S(Tibe the historical 11"\'e'lppIIWIlI (,I ~nl'rlll/i(" sllrvry lIlf"lli(Jdolo~ and Sllrv("V slatls!lcs. The onglns of survey rt"searrh call 1)(' Ir.lll'li 1),1( k lei thl' t·;trly l~)tll C'l"lltliIY alld \lle first scklllillc survey was conducted In England In I')l',! ll" Huwlt-v Moderll slIrv("y methods l!;w(" Ihf"lr rOOls In t!l(" 1930's, w!J("n three major asperls of til(" ... \If','('y prOCl·s.........t"rt" df"vt"lop('d and r("nllt"d sampling tt"'dllliques. datd. cOll("CtlOIl tt"<:hnJques. alld ',t,lIl~!Il"~t1 Illdhods lor data analysIs In (!J(" nllies. Ihe sClenllfic fact"·to·lace intervtew was eSlablished, -I11l! 11.llHlbooks wt"rc pllhlhh("d In the earlv s("Vt'IlI!t's, lell'pholl(" \llt("rvtewlIlg was Introduced and mail· ... tll".·y Ilu'lho{!.'-o WI'H" llpdalt"d and r("fl11("d Tilt.... was Jollow("d by the dev~lopmt"nl of rompul("r·asslst~d

Illll'n'kwing aIltl t"lcctrollic survf"ys tn lh(" dghlles and nlndles, At the sam~ time, more empha.sls was L1,IVC'11 to COllCl'pb ~udl as measur("n!t"1l1 rrror. nonresponse, and lotal quality control. We are now at Ill(' hnrlk 01 a !lCW mlllt"llIlIlllll Although tlle basic prInciples of the sl.'tt:"lIl1nc survey stili hold. survey ... l:lllSllt 1,\[lS lrolll th(" last rnJllennlul1l, lIk(" Bowley, would be surprtsed to Sl"'(" the design of a survey in '1111). [!lle!elllllt{.'(lly, Illi .... clf"slgn will Incorporate n("w lechnologles and will focus on reducln~ lhe 1'· ... llllnc!l·1I1 ~ iJunlt"n, wllllt" Irnproyj,ng dala quallty Bowley would cerlalnl.v be pleasl"'d to see thai ~,['plll"'IIC\kd llletllOd~ lor ttl<' r("dllcllon of nOllr("sponse and measuremt"nt en-or have been incorporal("d III .I (e,[,tl fju.lhly survry dnlgn 1 Survey Metbods and Statistic., Sodal SUJ"Yey, Scientific Survey, New Technologlrs, Data Quality. Nonresponse, Respondent Burden. TQM. CBM

25

1'. 26 - DlIlletin de Me/hodoloKi(> Sociolol;ique ()9 rue POllchcl. f 7")()17 P.m ... ), (klohn 1()l)t), N. (,4

EARLY HISTORY

Throllgh tlH~ cenillries, administrators have been cOllnting ,md many eivilizalions have left us stalistics. These statistics had mostly ;111 cconomie 01' political origin, In ancient Egypt. in imperial China. ;11 III iu Ihe Roman empire, "heads" and their possessions W('\T cOlluted, alld delaill~d slatisties on YOllng men \\'ITI' collected for military plllvoses. The first ccnsus, or, at least the oldest sUI-vivilIL; ("'IISUS publication. is t he Doomsday Book. Ullder IIIc ordlTs of 'vVillialll the Conqueror, royal sel-vants eoulltl'd pro!,erty alld persolls Iwl\wcn 1085 and 1086. Early "bureaucrats" collected statist ieal data for taxation, r('enlitment, and to dclennilH' the power of tl1\' state and Ole wealf11 of newly-founded colonies. Examples are the ,'IlUIIH'rations of Peru by Spain in 1548 alld of Nova Scotia and Qllebec by France in 1660. Of course, Ihis infonn<ltion was a hi~hlv

.~uanled secret and privy of the nlling governments onlv. Statistics \\';\S (he work of the "secret bureaucrats" (llackillg, 1990)

Those who did publish were the "public amateurs." These ,lIl1atellrs used figures and statistics for quite different ~oals. sllch ,IS theological study, social refoml, or the gratification 01 (sci('n(ificj cllriosity. 111is is nicely illustrated by the work of two 1IH'llIhl'lS of tlIe British Royal Society. In 1713. Willialll Derham dl'slTilwd in Ids tlTatise, Physico 17 I cology, how populatioll statistics wen' a proof 01 tlH' "divine design". Half a century carlin'. ,Johll Graullt studied the welfare of the London Community. Craunt, a piolls and studious London tradesman, is considered the founding fat lIn of mOdeI'll demography. He ITduced a great number of unsystematic lists of deaths, and thcir causes, to several perspicuous and elear tables wllich were published in a short pamphlet in 16G2. Ilavin~ 110 formal training in mathematics, Graunt was not l"uIIIH'lI'd hy mathematical tradition and strict rules, and he ~rouped his fi~lIres

in new and original ways. In fact. he was the inventor of "survival all,dysis" and "life tables" (Boorst in. 19H5). From Gr;lIl11t's 1reat ise. we learn that, around 1660, only 25 0111 of 1'\Try 100 Lmldol J(TS rcached the age of 26, and a mere 6 of I'very !00 reached tIll' old aL;e of fiG (BoOl'stin. 1985).

III Ihe ei~htecnth cent ury. amail'urs and acadelllics deserilwd and puhlished demographic statistics Widely. Unfortunatl'1y. these statistics were not very systematically or scientifically colleeil'd. Still. tile enthusiasm of amateurs like Sussmilch (3), and his critic. Malthus. prepared the way for modern demo~raphy and statistics as \\'e know it today.

At tlIe end of the 181h century. the emerging new nations felt the IlCed for statistical agencies. In the COllstitution of the UnilI'd SLlks

26

/lul/l'I'" de Milhodologie Sociolol;iqlu (.'i t ) rile Pouchel. F 7)017 Pan~), OLlnher I(N9, N 64 r 27

is a provision calling for a national census every len years. and the 1I11itcd Stat<-s Census of 1790 is the start of the oldest continuous !HTiodic celisus (Boorstin. 19t15). The US ccnsus had. and still has, ~n'al political impol1ance. Representation of states in the US Senate h h;lsn! on the numbers of people counted In each stale. But new lid t iOllS necded 1110re than just "headeOlll1ts". They needed publicly d\',IILthk fi.L;IIITS (0 describe their nation and to proVide a '<,Iatistic,t!" identity (llcisl'l. 1996). In 1800, the Bureau dc la l,qJllli/;I/IIC was established in FI'anee. which was a predecessor of t!l(' .'-.;/rtlisliljlLC G{"w'"alc de la France. the French Statistical Officc Ifk MdSt. IDH7). I'nlssia was a trendsetter for Europe: ill 1805. I·'\cdcriek Wilhelm Ill, Kin~ of Prussia, decreed that a central <,l;ltistical officc should be established and dedicated to "thc ('oll('ctioll and iutegration of statistical tables". That was the elld of t!l(' SITITt blHeaulTats; the data collected were analyzed and published' The Prussian office had many tics with universities. thus c-lillllliatill.~ the development of statistical and economic scicnces in (;crlllany (Hacking. 1990).

After the fall of Napoleon and Ole Vienna Congress (1814), the ('l1Icrgill~ I~uropean states established offices to collect and publish sLltistics about all aspects of life, not only births, deaths, and IU;IITia~es. This resulted in a flood of printed numbers which stilllllbtcd s\'ientifie discussion. For instance, in Frallee. thc lallllLttioll of "deviants" was started which led to a controversy in I Hl:i Who are more SUicidal, citizens of Paris or London? 111is dehak was onlv settled 1Il00T than a decade later when the new sLIt ist ical inst itutes had relevant data available (4). Suicide became a ITI'uITin~ theme in medic"l statistics, and later in statistical so\'iO!O,L;V of the 19th century. as Ihe work of Durkheim testifics (I l;\('kin~. Iq~IO).

'1'111' I! It II century was Ihe age of the diseovl'ry of social laws and sl n II't ures. The Belgian mathematician-astronomer. Quetelct. began tf) collccl social facls and data about humankind. Following Laplace. IIC appli('(1 cstablished mcthods from the natural sciences to the polttit'al and moral sciences, as they were called al the time. StatIstics bccame social statistics. focussing on utilttarian knowledge: descript ive stat isties should SCIV\, progress, Quetde! Illadc two illlJlortant eontlibutions towards the method of social statistics: the concept of tlIe "avl'rage" man and the fitting of dist ribllt ions: the "nonnal" CUlve was born. lie analyzed such differcnt stat ist ies as deatii rates. soldiers' chest cxpansioll. lengt h alll! wt'i,L;ht (5). Qul'telct also acted as a "passer" between scientific c1isciplilll's: borrowin~ frOIII his baek~round in aslronomy. hI' introdlll'Cd tlIe I'olll:ept of "nICaSlllTlIlI'nt error" in statistics.

27

p.2X - 1IlllJel;n de Milhodolo1:ie Sociologiqur (.59 rue Pouch('l. r: 7.5017 Parl"-), Odnhc-r lq(}(), N (l--l

quetclet was an enthusiastic advocate of the new science of "social" statistics. He corresponded with scientists and politicians all over the world. sharing his results and convincing them of the utility of statistics. In 1834. he pleaded for the founding of thc l.AJndon Slatistieal Society with the mathematician and inventor of the arithmetical machine. I3abbage. The goal of this society was to he the "procUI'ing. arran,ging and publishing of facts calculal<'e! to illustrate the condition and prospects of society". '1'11<' London Statistical SOCiety soon became the now renowned f{oyal Statistical Society (Dc Mast. 1997). One of quctelct's cOITespondents was Prince Albert. whom he had tutored in mathematics. Whcn I'rill('(' Albert prornott'd the first World Exhibition in the Crystal I'alaee in 1851. Quetclet gl'abbed the opportunity for organizing inttTnational coopel'ation. It was on his initiative that. after consultation wit h many foreign delegates present at the Exhibition. it was decided to have an International Statistical Congress in Belgium in 1853 (Nixon. 1960). Eight more congresses followed and, in 1885. the successor organization. the International Statistical Institute (lSI), was founded (Kenessey. 1997). I3y 1900, the lSI reported that already 43 percent of the world's population was covered by sixty­eight censuses (I3oorstin. 1985). We had entered the age of world statistics.

TilE HISTORY OF SURVEY RESEARCH: FROM "SOCIAL SURVEY" TO "SAMPLE SURVEY" (1900.1930)

In 1912, I3owley, the first professor of statistics at the London School of Economics and Political Sciences. started ;l st udy of the working-class conditions in five British cities: I{eading, NOI'thampton. Stanley. Wan'ilIgtoll. and Bolton (5). Sa11lples of citizens were interviewed using a structured interview sclledule concel'ning their spending habits and financial situation (I)e Mast. 1997). I30wley was extremely care in selecting the sample, ;IIld paid serious attention to the potential for nonresponsc bias through refusal and non-contact (Moser. 1958). The first scient ific survey took place only 87 years ago! TI1e roots or thc survey arc much older. and go back to the idealistic "amateurs" in the 17th, 18th and 19t h cent uries who went beyond mere headcounts and reported on life and death. sickness and health. and povel'ty and prospcrity.

The first documcnted mail survey dates from J 788. Sir ,John Sinc:lair. a Scot tish ag;ricult ural reformlT, relurnerl home enthusiastic about his German travds. I>uI'ing his "grand tour" of the COllt iuent. he becallle convinced tha t facl sane! fi,gUITS were thI' ll<lndmaidC'n of progress, aud he w;ultcd to f(llmd a statistical a,~('llcy for Scot bud. Reality forced him to hecome a onCIll,lIl

2g

null"l;" d~ .Hhll()doloK;~ .1'iociolo~lqUl' (<;q rilL' Pnuchet, F 75017 PUI"L (klohcr I ()IN. N (14 p. 2 t )

statistical olliec, and he send out questionnaires with more than a IIIIJIdred questions to the ministers of all ~)38 parishes of the Church of Scotland. He pursued them rdentlessly with follow-up ktters. the last \Vl;tten in blood-red to suggest with "the draconian colour of his ink" what would happen to the nonrespondents. It took 23 (1) reminders but 11(' had a 100% response. This survey resulted in I}l(' Statistical ACCOlHlt of Scotland, published in 21 volumes ]wlwcen 1791 and 17q~). In his Account, Sinclair published what we \vOl dd now call II';Hlit ional stat ist ics like age distrihutions. life "XIJCl'l;llICY and estimaks of the total population and its rate or ChaIl~('. Bul he also published data on "social statistics" and lire styles. Thanks to Sindair, we now know that in the I Htl1 century, lilt' women of Inveresk organized football matches b(~tween the 11I;UTicd and IlnnlaITied WOlllcn, of whom the fomler invariably won. [Boorst in, I~lH5: llackin~. I ~190: I leiseI'. 1996).

Sinclair used key informants to collect his infonnation. The social n,ronuer and statistician. Charles Booth. used interviews with key informants and direct observation to collect systematic data on all Lllnilies with school children. Booth undertook this enterprise as a pI'ivate person and at his own expense. dividing his time between his work as a ship owner in Liverpool and his studtes of London. Two of his faithful assistants were Beatrice Potter-Webb and Seebohm l{ownt rec. who also played an important role in social reform In the 19th centllry. Booth's study into the Life and Labour of the People of London (7) st arted in 18GG and was published in 17 volumes in I ~)()2. Its ITsUltS drew attcntion to the f~xtent and sevC\;ty of poverty and had ('onsiderable political impact. It was also the inspiration for tlJ(' "social survcv movcmC'nt" in Europe and the US.

Booth's co-worker, the York merchant. l{owntree. added some jllll'0!'LlIIt mctllOdologieal innovations to the social survey. For his I~)()J study, POllcrty: J\ SllIrll/ of Town Life. Ill' obtained information direct/II rrom wae,e-earning families in York USlllg intervicwers. I(OWlIll'('C also o[JrmliorInlized his conce'pts with grC'at precision, II'SII!tiII~ in .~t(lrIdnrrlizf'd intf'm/elL' schedules (I lacking, I ~F)O: Dc M'ISt. I ~)~)7). !lawley. finally. made the last step towards the scienl jfic survey, anrl used snrnp1in9. including the study of pol cn ti,I! bias through uonrcsponse (Moser, 195H). Through his work on sampling (I !lOG. 1926), he cnds a discussion of more than a (Tnillry.

'I'll(' first to usc sampling; was Laplace in 1802. Laplace took a sv.slt'lllatic s<lmp1t' of ('onlmuuities in 30 departments spread oul C<i'I'lllv OVIT all of FraIHT. I Ie then used the birlh registlTs, which wert' considcred very accurate, to estimate the total number of births in France for the past year. J\lthoug;h Laplace Licked ,uly ('OIHTpt of random s;llllplin~. his sampling of (,(JIllllllIllities had

2')

p. 30 • Rlllldin de Mithodologie SodoloGiqru (5'J rue PouchCI. F 75017 Pan~), OClohcr 1'J9'J, N M

ckarly been desip;ned to achieve the goals we now ascribc to random sampling (Stigler, 1986: 164).

Quetelet was clearly attracted to this met hod and its potential for savings in time and money, and planned to use it to estimatc thc population of the Netherlands (at that time including Ilolland aud Iklgium) in 1824. Howevcr. the strong critieislll of Barou dc Keverberg made him change his opinion and Quclc1ct stalted planning a complete census instead. Quetelct was trained as an ;lstronomer and was well acquainted with the concept of "observation error": the prccision of an observat ion increascs with the square root of the number of observations. Therefore, he wonied about other sources of error and the inl1uellec of sampling on thc total error.

Quetc1et and his contemporaries did not have a satisfactory solution for this question and the general opinion was that one should count every member of the population, sampling was unscient ific (Hl. A sentiment that was heard only two years ago in the US Senate dUring the discussion of the coming census. At the end of tlIe 19th century. a debate took place at the International Statistieallnstitull' on statistical methods. and Andres Kiaer opened again the discussion on sampling. Kiaer, director of the Norwegi;lII Statistical Office, presented papers on his experience with syst "lIIat ic sampling and was a dedicated advocate of the "new" method. But it was only after the 1906 publication of Bowley's work Oil sampling that the method took off. Bowley's work was continued by Neyman, who further developed the concept of sampling elTor and published an inl1uential paper in the Journal of the Royal Statistical Society in 1934. This work was continued by other eminent statisticians, including Mahalanobis, Yates, Cochran. Hansen and Kish (O'Muircheartaigh. 1997: 21, who gave us the tools to usc' sampling in surveys as a scientific and accurate substitute for complc!t, cnumeration (Lessl{'1' and Kalsbeek, 1992).

TUE SURVEY INTERVIEW (l930-197())

Although the survey method was conceived and born in Europe. it grew up in the United Statf's. Inspired by Booth, Janc Adams started the Hull House prOject in Chicago. This was the beginning of the "social survey movement" in the US. ImpOltant benchmarks Weff' the founding of the Russel Sage Foundation in 1907. Paul l\.ellog's survey of Pittsburgh in 1908 and the Rockefeller Institute of Social and Heligious Research in Chicago in 1921 (for an ovel'View sec Dc Mast, 1997; Converse, 1986; Plat, 1996). The Depression. and the resuIting lack of funds, put an end to the "social sUl\rey

llrd!('llll de IHel},od()!ngi(' Soc;oloKiqu(' L'i0 rue I'tlUdH~1, F 75017 Pan"), 01 Lnh.:r 19<)9. N 64 p_ J I

1ll00'enwnt" in thc thiliies. However, at the same time, the .l!:ovcllllllCnt started Iising survev data. Also, consumci' market ITscarch ,111(1 newspapcr "polls" started using survey methods.

The slll'Vey method beeauw auothcr word for f<1ee-to-faee interview. TIJ(' first interviews werc short [around 5 minu\{'s) and simple. '!'cc!llliques for qucstion writin,,t(. interviewing, and <1nalysis wcre still r;IIIHT jJI'il1litive (Rossi et 01.. 1983). In the 1930s, the interview SllnTV developed rapidly. At the end of the fifties. the inteI'Vicw h<ld "OIlle of age and the fil'st handbooks were published. Examples ;Ire: Til(' Art of Asking questions (rayne, 1951), [nteruicwinq in Socia/ [<('s('(/fcll (Ilyman. 1954), 'l7l(' Dunamics of [nteruiewinq (Kahn ;uld Call1wil. 1~1571. and SWH;lj Sampling IKish, 1965).

Fronl 1930 to I~150, three major developments in the US wC!'e of great importance for the establishment of the modenl survey interview. At the US Bureau of Census, sampling techniques were tl'ied and sampling schemes developed. Market rese<1reh institutes, such as Gallup and Roper, were instnlmental in tIle implementation and development of data collection techniques. The groWing dem<lnd for valid infonnation on society created a need for standardized instnmlents to lIIeasure attitudes and opintons. Pioneers in social psychology and psychometrics, such as Thurstone. Likert, Guttman. Cantril. and Stevens, provided scientific measurement techniques and scales (Green. 1954; O'Muireheartaigh, 1997). With these instruments came statistical techniques.

Classic statistical techniques from psychology and "mental nICasurement" were adopted in other fields. A prime example is factor analysis. In addition, special techniques for survey data were developed (9). For instance, Stouffer and, in later years, Goodman f()('ussed on the analysis of (Toss-classified data. and the work of Lazersfcld was of great importance for the development of path analysis IBabbie, 1973). The rapid devclopment iu the sixtIes of dect nJllie computers. closely followed by the development of sl at ist ieal packages. such as SPSS, brought rapid data processing and 1II0re complicated analysis within tIll' reach or survey praetit ioners.

The inlTeasing need of the US federal govcnllllent for knowledge and iurol'JIIation 011 social trends and public opinion stimulated the rapid .,t(rowth of tlIe survey method. EXisting market and opinion ITscarch finns, sucii as Gallup, exp'lIIded. At the same tillie, new survey ITs"arch inst it utes were fOllndcd at llniversil ies. In 1940. the Office of Public Opinion Research was cstablished at Princeton by Cantril. Cantl-il, togethcr with Gallup, conduet"d hundrcds of split ballot cxpel-imcnls, and was olle of the first to study and improve Slll'Vey nH'tllOdology. Again with Galhip. he fOIllICkd PI/bile Opinion

1\ III

p.32 ltullnin de Milhodologie Socio[ogiqlu 159 rue POUChCl, F 75017 Pam.,), O((ohcr j()I)lJ, N (>-1

quarterly which is still one of the leading joumals in the field (Dc Mas!. 1997). In 1941, the National Opinion'Rcsearch Center (NORC) ill Chicago was established by Field, and supported by cminent social researchers such as Stouffer and Cantrill. It was followed in 1949 by the Survey Research Center (SRC-ISR) in Ann Arbor. Michigan, with Likert. Campbell, Cannell and Kish (Dc Mast. 1997; O'Muircheartaigh, 1997). Today, thesc institutes still playa major mle. For a thorough historical review of American survey rescalTh. see Converse (1986) and Platt (1996).

After thc Second World War, the new survey method was introduced in Europe. In many countries, market research finns were founded ,md were the first to use sample interviews. Statistical agencies wen' also interested in the new method. In the early sLxties, sUlVey research was introduced and used al European universities. Almost immcdiately, methodological research concerning the survey interview began. inspired by studies at NORC and SRC-ISR in the US. The goal was to improve the survey method and to find new and better techniques for question wording, interviewing and statistical analysis. Until the mid-seventies, the survey interview was the leading method and the number of face-to-face inlerviews grew exponentially. This was followed by a period of critical assessmcnt, and, in the eighties, the development and reevaluation of a\lernative methods.

CRITIQUE OF THE SURVEY INTERVIEW (I 970-I9!lO)

The dominance of the survey interview in the seventies is illustratl'd by Ihe following figures: Turner and Martin (1984, vol. I, p, :~O)

estimatecl that within the United States, dUring thc peliod between 1971 and 1976, a minimum of 100 million survey interviews were conducted; the gross revenue of the survey industry was cstimaled at $4 billion in 1978.

The popularity of the survey method was also rcfkcted in the ,grOWing number of aliicles in leading social science journals based on survey data. According to Presser (1984: 96), for sociology, till' perccntage of articles using survey data increased from 24 per ecnl in 1949/50, to 55 pCI' cent in 1964/65, and then to 56 pn cellt in 1979/80. For political science, the increase is evell sharper: from :~

pn cenl, to 19 per cent, and then 10 35 pCI' eellt.

In the seventies, the application of the survey interview reached its linlits. 't Hart (1972) has calculaled Ihat the probability of a Dutch household being included in a survey in a particular year was about one in three. As a result, a cedain "intelview fatigne" arose,

,12

lI11ltf'lin dl! Mtl!l()d(J[o/.:il! S(Jciologiqul! ()\) HIl'!'nudll"l, F 7';017 f\trI:-'l, lh.:lllhcr I()qq. N (H p. ,3

Moreover. til(' seventies wel'e characterized by distnlst of I,u'ge societal oq.;ilnizations, and especially the niltional govenlmenl. With the Onvcllian year 1984 coming, people wanted to safeguard their priv,lCy. They hesitated, or even objected, to giving confidential infonnation for purposes thaI were not clearly in the interest of the respondent, 1I0w can the prevention of misuse of personal illfonnalion be gllaranteed? Thesc. and similal' considerations (BulnH'r, 197~j), sccmed to have a serious effed on the willingness 01 rcspondcnts to l'oopel'ate, As a case in point, as a reslllt of many adi\'(' protests, the Nethcrlands National Census of I l)80 was post pOI1l'd indefinitely (Dijkstra and Van del' ZoUWell. I ~)H2: 3) S\ lIvCV nfJnresponse was becoming an iss lie.

TIll' main reason for thc popularity of [he standardized survey Inkn:iew iu Ihe seventies was its applicability. One can use sUivey inlnviews to coiled data about the bdlavior of people, to ITl'onstrucl the past of Ihc respondent. and even to assess values on val'i;lbles which are not directly obselvable (subjective phenomcna) Iikc opinions and attiludes. But this can only be adequalely done if ccrtaiu conditions arc fulfilled or. slatrd otherwise, if cel-tain allxiliary hypotheses are tnle. Oilly if the inlen:iewcrs ask the quest ions as Ihey arc worded in Ihe questionnaire, only if the rl'spondents lludrrstand them in Ihe same way as inlended by Ihc l]lll'stionnairc designer, only if rl~spondents' choices of a response categOlY arc affceted solely by Iheir value for the variable Ihe qucstion is Illr<lnt to measure, and similalities and diffel'elllTs 1)('lwr('u answers of different respondcnts lIIay validly be intcq)rcted ;IS similal-itirs 01' difTerenees between till' respondents with ITspec! to (lJe val'iables eOlleernrd. If these ronditions arc not fulfilled. rl'spons('s brcomr incomparable. and responsc bias occurs. In the sl'ventil'S. not only tlw application of survey interviews illlTrasrd. buI also critical ITsearch connTning (hc dcgrec 10 which 1Ill' conditions lIll'UtiOlll'd above an' fulfillrd (Sehaeffl~r. I ~j~j I).

hI slllVey intervil'ws, the stimuli, that is. the questions and items, slllluid be pn'sented to all respondents in exactly the same way. Bul do ;111 inIClYi['\\,ers follow thc lJllf'stilll1 wordings alld further ins In H·t ions provided by the ITscarclwr? To answer !hat quest ion. Michal'! lllT11l1C\', expanding on work started by I lyman (I ~)54) and Canncll and his co1\eaglles (1975). conducted a dclailed action-by­act ion ;Illalysis of audio lapcs of SUnT\' intelviews. 13ITI1I1l'r (I !)7H, IqH2) showed that a nonnal standardizcd survey intelvil'w is far less s(;llldardized than thc survey resl'arcI1l'1' Iikl's to admit: intl'!'vh'W(TS alt cr I11<' \Vordin~ of questions, respondents respond to ot her 'I' Il'SllllllS than in!l'IHled by the researclll'r, inll'lviewers sug,gesl ITsjlmlsl' options. c1c. In later years. his conclusions Wl'l,[, ('()ITo!Jorat('d in seve!'al countries (Billict and Loosvddt. I~mH;

I lijks( ra and Van lkr Zou\V('n, 1qHH; Syke~; and Collins, I ~)92:

.II

r ~4 RlIlll'tin tIe "'fithotIoloKie SocioloKique (59 rue PouchCI, r: 7:<:;017 Pans), Octobcr 1<)<)1), N. 64

t-!outkoop-Steenstra, 1996). For a statistical treatment of the (ktection of inteIViewer effects. sec FowleT (\ qq I), Grov(~s (I !1Wll. and 1I0x, Dc Leeuw and Krell (I991).

i\ second condition is that all respondents understand the questions l'xactly as intended by thl" questionnaire designer. But arc llw questions indeed understood in the same way? To anSWl"r that Cjuestion, Belson (1981) constructed a questionnaire Including seven or eight "test questions". This questionnaire was delivfTed to <I

sampk of survey respondents in til(' usual way by ordinary market research interviewers, Next day this interview was followed hy ,I

.<;eeond inteIView conducted by a different person, namely an intensive inteIViewer. The lalll"r spent an hour or more attempting to find Ollt how well or badly the rl"spondent understood the lest CjIICstions, and the nature of such misunderstandings as occurred ( 1~lH I: 22). i\lthough t he market research inteIViewers did not ohseIVe any signs of uncertainty or difficulty respondents had wit h answl"hng the questions, the intensive inteIViews showed major misunderstandings for each of the test questions. Belson's research was followed by a long series of research findings concerning tIl(' lingUistic and cognitive aspects of understanding and answering stIlvey Cjuestions; cognitive intervieWing, for short (Forsyt hand Lessler. 1991; Schwarz and Sudman. 1996; Sudman, Bradburn and Schwarz, 1996). For a state of t he ali oveIView of cognitive inteIViewing. sec the special issue of thc Blllletin de MdllOdologic Sociologique (BMS) on testing SUIVCy quest ions (guest editor, Pamcla Campanelli. 1997).

i\ thil'd auxiliary hypothesis behind the sUIVey intf'IView is (hat the respondent's answer is only dependent on the "(nle" valuf' of tIl(' variabk the question is meant to measure. Other valiab1l's. such as characteristics of the inteIViewer, or the place wlICre the interview is conducted, should not affect the respous(', or. as the Jil rgon ,~oes,

should not have responsl" cfff'ets. But arc these response effeels ahsent in sUIVey practice? 'nlat quest iOIl is askl,d in dozclls 01 methodological studics. and qUite often ;msw(Tcd in thc (H'g;ltivc,

For example, Phillips (\971. 1973) found that responses givcn by mcn. rcspectively women, to qUl"stions regarding menIal health were less infonnative about differences wit h respect to thdr men ta I hcalt 11 than they are infomlative abont differences betwl"en IlIen and women regarding the social desirability of endorsing certain items of a mental hl"alth questionnaire. Thf'se response effects induced Phillips to put a large question mark behind the crf'dibility of sUlvcy data, as the title of his 1971 book, Know!cdgefrorn What? testifies. In a book published two years latl"r, AbandoninfJ Method, he also places a question mark behind the usefulness of methodological resl"arch concelTIing response lin)validity. But. other methodologists

fiu{(di,. dl' MetlwdoloKie SocioJoKiqur I ~q rllC I'ouche!. F 7')()17 P,lfI\). (kloher I(JI)l). \l (10'1 P 15

llave not abandoned the surve\, method. They continue their split h;l!lo( expehnwnls concenling the effects of question wording, mode 1\1 ;I<ltI11nistratioll of tl\(' ques(ionllaire, style of intelvicwing, etc. '1'1](' 1""II1)('r of these studies has increased so rapidly that it has beCllI1H'

I'l\ssihle, ,mel necessary. to j}('r!oI1n some kind of meta-analysis. or I!l<"ordical synthesis, of tl\(' research results. i\lteslander and l\nelJlJ1lhler (I !17f11 constructed all ('ar1y theOly of measurCI1\('nt ,'ITors ill sun·eys. and Sudnlau amI Bladburn published a very Il1Iltwnli;d oV(Tview of response dkcts in SlIIVCyS. based on ovel' ,lilll' htlndn~d studies. The book by Sudlllan and Bnldburn (1974) sllOws thc salience of the infoI1nation reqnircd of the respon(knl, Ill<" I·o.~nitive difficulty of the task fOI" the respondcnt, the dcgnT to which the task involves problems of self-presentation, and how I}\('se arc I'elated to rcspondents' responses, or. in other words. ,let ennil){, t IJ(' occurrencc of response effects.

TIl(' l'litique of the slIIvey intnview. as it was clearly voiced in the seV('lIt ies. has opened new lines of met hodologieal research: ITseareh into the causes of nonrl'sponse due to refusals. intel'action <ll1<1lvsis of thc behavior of the intcIViewers and respondents. and c(),~nitive laboratory research concerning the processes by which ITspondents search for the proper meaning of a question and lor a response alternative that best reflects their opinion or situation,

NEW METHODS AND A NEW TECHNOLO(;Y (1980-1990)

Dcspite the critique of the sUlvey interview in the seventies, in the ('i,~htif's thc sluvey industlY was still booming. 111c'[e was a growing Iwed 1'01' fads ~Ind figures in ;';O\"Tnll1ent and the' cOIllmcTcial world. To flIifil this nced, new survey methods and techniques were dcvelollf'd. Mail and teIephOIw sUIVeys become increasingly popular wit 11 governlllent agcncies and rescarch firms. The increased costs of I;I("('-to-face intcIViewing make it vidually impossible, or, at least, ('xl rl'nlf'ly cos[Jy to utiliz:C' the faec-to-face sUIVey to its full potential wlH'lI national surveys or largc surveys in geographically dispersed alTas arc rf'quired. This has led to a rem'wed interest in altcnlatives, ;mel a rcnewed ITSf'arch errort to optimizc mail and telephone SUITCVS Ilk L('Cuw, I !J92).

Tlw publicat ion in 197H of Dillman's book did mueh to increase the rcspedability of I1Iaii slIIveys and contributf'll greatly to lhe reel lIel ion of nonrespollse and the enhancement of data quality in 1II;liI slIrvC'ys. III his book, Dillmall (I !J7tll gives an inspired account of mail slIIvey research mf'thodolor-;y with a clear and precise descriplioll of how to optimizc I1wil sllrveys by IIsing thf' Total Iksi~ll Ml,tlIOd or TDM. DillIn;Jn's 1"DM proved to hc ('(!cctive in

1) 14

r.H1 Rill/din dt l,-flthndQlogit SocioloK;qlle (59 rur POllchcl. r 75017 PilnS), Oc!oher I()(N.!'i (,0.1

both the US, Asia, and Europe (Dc Lecuw and !-iox. 1988), ,md is popular in academic research and official statistics.

Tc1('phone surveys started in market research in t hc US. but soon I)('callle inel-easingly popular in governnlCntal and academic f['scarch in the US. This was cncouraged by improved tec1111olol!,y. bv the development of random digit dialing as a sampling technique. and, above all, by the general public's increased availabilitv ali(I access to telephones (De Leeuw, 1992). Tf'lephone interviewing qUickly changed into Computer-Assisted Telephone Interviewing (CATI): an interviewer sits in front of a terminal and asks questions Ihat appear on the screen: the respondent's answer is then typed into the computer by the interviewer. Supervisors are present I()r quality control and to assist with specific problems. The first CAT] systems were developed in the US, primarily in market research. Soon CAl'I systems were developed at American univf'rsities. and finally introduced in government surveys (Nicholls, 1988). Europe followed later. In the mtd-eighties, telephone surveys were the maim­data collection mode in market research in the US, Canada and Australia. and increasingly popular in market research in Europe. Finally, telephone interviewing carne of age in 1987 with a large inteIT1ational conference (10) and a subsequent monograph published in 1988 (Groves et al.. 1988). This monograph soon became the standard textbook for telephone researeh.

Technological developments in Europe differed from those in North America. In Ule US. telephone interviewing stal"ted earlier and is rnore prominent than in Europe, Also, most computer-assisted "mail" surveys, including use of the Intenlet for data collection, arc 11l0re prominent in the US, especially in market research.

Computer-assisted mail surveys arc part of CASI (Computcr­Assisted Self Interviewing), also called Computerized Self­Administered Questionnaires or CSAQ. A characteristic of CASI is that the respondents read the questions on the screen and enter tlw answers. There is no interviewer: the interviewing progl-al11 gUides the respondent through the questionnaire. CASl can appcar as part of a CAPI session where the interviewer hands over the computer to the respondent for a short period, but remains available for instlLlctions and assistance. This is eqUivalent to the procedure used in traditional paper and pencil face-to-face interviews, where an interviewer might give the respondent ;} paper qlwstionnaire containing sensitive questions.

In the Netherlands, a special fonn of CASI was deYdoped: the tele­intclview. This is a foml of computer-assislcd palwl research (CAPAI{) where respondents fill in an electronic questionnaire about Ol1ce ;} week. DUl-ing the panel study. s('lected households receive a

1(,

Ilul/cfin de l,-filhodoloK;r Stlcitl/r'Kique (51) rur Pouchel, F 75017 P..tn~), Oclllhcr I{)(}(), N b4 r·n

microcomputer and a modern: questionnaires arc regularly sent to Ihc respomknts by modem and, when completed, sent back to the ITlllOk COlllputeL The administration and transmission is all done aulomatically hy the program: respondents only have to read the qucslions and type in the answers. TIle first tests were done in I DH;'. and, in 1~JR6. a commercial tclepanel was working at a large I )ulch m;u-ket f['s(';IlTh finn (Saris. 1991].

TIlt' lirsl complll<'r-assisted v(Tsion of the mail survey was the Disk­1\\' Mail survey (DUM) In DHM, a disk containing the interviewing progr,lm is sent bv post to the respondent who !'tillS the program on his or Iwr computer, and thcn retunlS the disk, with the responses, I)\' post. Disk- By- Mail surveys h,lve been successfully done in 1l1<lJkct rcsearch in tlw US. In tlIe Netherlands, ;} Disk-!3y-Mail .~un'('y in primalY schools was vClY successful; teaclHTs and pupils cl!Joycd the experience and produced high-quality data (Van liattum ,UH! De Leeuw, 1996). At stat istical agencies, such as Statistics Nctlwrlands, the electronic eqUivalent of a business survey is being (kYdoped: the EI('Ct nll1ic Dat a Interchange (ED]). l11C~ latest vCTsion of the computerized mail survey is the Intenlet survey. This is still cxp(Til1wutal, although the first surveys, using the net have o('cn pnJonIwr! (Witt, 19~J7: Dc Lceuw, 1997a). For an extensive review of ('ompulcr-assisted data collection and data quality sec Dc Leeuw, 11m; <Iud SUijkcTs (199;') ,md Dc Lceuw and Nicholls (19~)6).

FRO:',! DATA COLLECTION TO DATA QUALITY (I99tl-!99S)

TlIe 1~)HOs was tlIe dccade of the development of new data collection I(Thuiqlws. Moviug tow,lnl the new millennium, the focus is shifting low,lrd IlwaSIU'l'llwnt and data quality. An important contribution ('Ol1J('S fmln Grovcs (I DH9) who published a comprehensive rcvi('w of "1 J nTV lTrors and possibk canses, bascd on a synt hesis of 1\llo\\'I('d[~e from psye!IOIlwtl-ics. statistics, and the social sciel1ces. (;ro\'('s (I ~JH9: fi) emphasiz(~s ('ITOr estimation ami the iIH'OI-poration or SUI'Y('y ('!Tor and SIU'YCV cosls: "only by fonnally assessing the ('o"h or alternativ(' IlwtlIods (jointly (lssl'ssinrJ (jllfl/i(IJ) CUI the 'best' IIll't}]()d be identified." (Ill

Ninelcen hundred ninety was the year of two important confercnces. In I)ll' spilit of Gmves' book, au intenlational conferelwe in Tucson. Arizolla. was dedicated 10 l1leasurenlCnt errors in SUI-veys: both (~n'or

('sl illl;\t ion ,md error reduction werc emphasized. TI w result in,g Illollogr,lph was published ill I~)91 (Biel1ler eI uf.. I ~)!J I). Also in 1~l!)(). tlw nl-st "llIt(~nlational WOI'ksIIOP 011 Ilous('hold SUl-vey NOllrcspolIsc" was org,ll1izcd in Stockholm by Statistics Sweden. Tlw"c wod-lsllOps lWC;llllC a yearly ('vellt. At t Iw first workshops. the

17

p. .\H /J"l/ttin dt Melhod%git Sociologiqut (59 rue Pouchet. F 75017 Pari:>.). CklOl1cr jl)t)l), N 64

main interest was tile exchange of information, bul soon the emphasis shifted to the integration of knowledge and the initiation of new research. A large international conference was planned for 1999, and the resulting monograph will certainly be the handbook Oil survey nonresponse for the next decade. For a review of research on 1I0nrcsponse, see de Leeuw (1997b) and Groves and Couper (1998); for an international comparison of nonresponse figures over the last 15 years, see De Heer 11996, 1999) and Maas and De Hen ( l~l95).

Hecent research changed (lUI' view of meaSUf('l!lent enor alld the Ireatment of it. Going back to the early work of Deming (J 944, 1982). quality became the central issue in survey ll1cthodolo~. Of course, methodological studies and post-survey quality assessments, such as reliability, validity, and bias estimates, are important indicators of aceuraey, but may be of little value for iTllprovi1l9 sllrvcy data (Lybcrg el ell., 1997: Ill). Prevention is tlH~

betler' cure, and interest should shirt from post-survey quality cvalllation to controlling the survey processes, such as questionnaire construction, interviewing, data collect ion, coding, data capture. editing, and analysis. Given the importance of data quality and quality control. an international conference on "SlIrvey Measurement and Process Quality" was organized in 1994; in 1997, t he monograph of tilis conference was published ILyberg ct aI., 1997). The central concept was 'Tolal Quality Management" (TQM); the philosophy is that process quality generates product quality. As stated above, TQM is concerned with 1110nitOl'ing alld coutrolling the whole process, from the first formulation of the research qucstioll to the final publieat ion and dissemination of tire results.

VCI)' important in the TQM approach is the qUick implemeutatioll ill slllyey practice of recently-develope'll "stat e-of-t he-art" met hoels. New developments in survey methodology sho\lld be dOcUlllented and implemented as "Current Best Methods" (llox, I~)97). All excellent example of the successful implementation of a uew "state­of-the-al'! technology" in practical "eun'ent best methods" is tlIe "rognitive laboratory approach" for' testing and improVing questionnaires.

In the cognitive laboratory approach, rTcent developments frol1l sUlyey methodology and cognitive psychology are combine-d with more t radilional methods for questionnaire evaluation. Goal is to eleteet disturbances in the question-answlT process and find solutions to the problems identified. for an ovelyicw of tlIe techniques used, see the special issue of BMS of Jun(' 1997 (Campanelli, 1997). The first experiments with the cognitive laboratory approach took place in 1988 at the National Ce-nter for Health Statistics in I\menca (WilliS ct aI., 1991), in 1990. The first

lX

Ulllfl'litl de Mellwd.d(,~ie .\tJci("o~iqlH' \."\1) n.;t,: I'\)uchl't, F 75017 P·Hl~). Oltoher j\)I)\), N b4 [l. J\)

,",Tinv of t he methods used was presented at the Tucson conference (III sllrv('y ('ITOI', and was subsequently published in thl~ monograph (Forsy! h alld LessIer, 199 I ). Large survey agellcies adopt thcsc 'slall'-of-thc-art" Illethods qUickly; for instance, ill I ~l!l2, Statistics Nl'lllt'rlauds stalied a "cognitivc laboratory" to tesl and iIllpruvc ,\\1"S\ ionnain's. Although "cognitive laboratory methods" are still not SI;ll1,!<Ird prdctice in the sllrvey illdustry, they are among the (lllTcnt Ill'st methods of olficial statistics agcncies and laq~c survcy Ilistillllcs (Campallelli. 1997)

1"1 It' IT-c\,;1ll1al ion of lire survey process and the recent emphasis on T,,!,I! qllality Management has !('d to a new optimism in the surycy illllustlY. Problems, like nonn~sponse, are studied and solutions are \'xaminnl. At t he same time, new tcchnologies arc developed and !l'stl'd to improY\' survey quality (such as automated data capturc). III 11](' history of sci<:ncc, one onen can discern distinct stages in the silidy of scient ific phenomcna. The first stage is description and dassification as found in the work of Deming (1944) and Kish (I !J()5). The s('cond stage is a search for causes and theory IOl"lnal ion; good cxanlpics arc the books by Groves (1989) alld I\i('nHT cI al. (1991). We havc now moved into the third stage ('oping, controlling and imprOVing -- and Interest has shined to IU;l\1aginJ,( survey errors (Lyberg d al., 1997).

SlInn:YIN(; IN TIlE NEXT MILLENNIUM (200) AND BEYOND)

Ii 1 1!l5G, the BI-itisli "Astronomer Hoy'll" predicted that space travel would 1)(' tedmologically impossible for a long lime. A year later, the first ~plltllik was successfully launched, and, in 1968, lhc first man walker! on the moon. Predicting the future is hazilrdous. Still there ;1['(' some deady disccrnahl<' trends in SUrvl'y mcthodolog,y that shollid he mcntioned.

Slll\'ey nonrcsponse, in general. will remain an issue in tlIe next 1I1ilknniulIl. Changes in society make it more diflkull to reach po\('ut i,d respondents, tlIc groWing number of SUlYcy requests will makc respondents less enthusiastic and willing. The Utility uf telephonc surveys is being threatened espe-cially, and tdephonc interVieWS will become iC'ss efficicnt, particularly "cold-contact" interview.s. where respondents arc c;llkd without any prior contact or Iw(ilkation. "I-Iot-('ontacts", sueh ;1S in a telephone follow-up in a nuxed 1ll0'\(- (ksiJ,(n 01' a pauel SUlyey, will suffcr kss fn}Ju unwilling resj)()ndcllts.

The increase of unsolicited tckpholH' calls «-all-ITntcrs, I,' 1"III'II'k('! ing. inttTviews) nCJ,(at ivl'lv dT,'(( swill illl-',Ilt'SS (\) (ooperat t'.

1')

p. 40 Bull~tin de Mithodologie Sociologique (59 rue Pouchet. J-' 75017 Ptlfl~l. ()uoher ]<)()lJ. N 64

Also. households may start screenin~ incoming telephone' calls Ihrough answering machines to protp('l against unwanle'd calls. Tpchnologieal advances (such as CAT\). whie'h helppd tele'phoIH' inkrviewprs for a long time. are now being tUITle'd a~ainst the'll!. Sp(Tial telephones that show which numbe'r is calling amI that CilIl

bc blocked against certain numbers arc alreadv available in thc US. A lTcent (telephone) survey in the US shows that these "gadge'ts" are being weleomed (Simonetta, 1997). A vel)' large majority of pe['SOlIS (~n'Ju) believe they should have the right to block calls from tdpmarketeers, and a small majority (65'%) believp they should also bp allowed to block scientific surveys. Whether blocking will be' really used on such a large scale remaillS to be seen. V\Then asked whether they would be willin~ to pay $5 fOl' blocking, tIw percenta~es were more than hah-pd: 47');(, said yes for blocking against telemarketers, and 21 'Va still wanted blockillg against scientific research polls. In the US, people want the' .'ight to block unwanted calls. but a clear distinction is made in the type of calls the'Y wish to block. This is eneouragin~, at least for scientifie and government research. By emphaSiZing the legitimacy of the survey and its scientific nature, It should be possible to maillt aill adequate response rates. There Is sttll a future for telephone surve'Ys. but it is not as brilliant as it was in 1989.

In general, survey quality and process management will remain an issue in official statistics. Sophisticated statistical modelling has given us tools to model eITor sources. At the other end, advances ill data collection have given us tools for reducing eITor (O'Muircheartaigh. 1997). New technologies will undoubtedly influence future survey methodology further. Certainly fonns of computer-assisted data collection will come of age, and will proVide opportunities not feasible with paper-and-pen survf'ys. Examples arc "tailored" interviews, in which responses to earlicr lj\l('stions alT used to select the next question, resulting in individualized intelV'iews and adaptive conjoint measurcment (Sikkel, 19~16).

The power of advanced computcrs and advanced progralllming will make it possible to usc "background" knowledge in intCl'views in far IlIore sophisticated ways then at present. Interviewers in the field will have access to "expert systems" with instnIclions, automatic coding facilities, etc. Another valuable potential, especially for business surveys. is "dependent" interviewing, Using information from earlier surveys, figures are checked and specific questions about changes and reasons for changes al'e fornmlated. When the potential of computer-assisted data collection is fully developed, it will evolve from merely a useful tpchnology into an unique new method of data collection. Ilowevn. this evolution will only be successful if developers of new survey systems ke'ep the hUIllall user in mind.

R"llttin de "''''hodoloxie .",·ocioloKique {59 rue P\lu.:hct, F 750 17 rarL~l. (klohcr 19<)<). N 64 r 4 I

Successful human-machine interaction will be the key to success or lailure in computer-assisted intervieWing. Ergonomic interfaces should bp designed and evaluated using useability testing (Dumas and He'dish, 1994), In addition, more attention should be paid to social norms and how these inl1uence the usefulness of computer­,\ssistl'd surveys. A pl;me test case will be the Internet survey, a lluu word in thc late nineties; whether or not it will prove to be a "hVlw" or til(' data-collection method of the future is still unclear. Much dqJellds on a thorough knowledge of tllf' way pe'opk usc the Nd and the customs. social codes and "ncttique'tte" that will ,kvdop. Issues of seclll-ity, respondent burden aud re,H:tion to llnsolicite'd mail will be of great importance. Key aspects of ~ucccssful Intenlet surveys will be "respoI1(lP.nt-centcred", not ll'clmology-ce'ntered" (Witt, 1997; Dc Lceuw, 1997a).

At the samf' time', survPy researchers should also become more' "cliellt -o!"ien te'd". Advances in infonnation technology anel sophisticated statistics make users of survey data dependent upon inkrpn>ters to tell thnn what the rcsults actually m(~an. Knowledge of slIlV'ey I1l('thodology will not be enough; one should be' able to ([';)nslate raw data into relevant infonllation, using advaIHTd sLIt ist ics.

NOTES

(I) An earlier version of this paper was presented at the 141h World ('ongn'ss of Sociology. Montreal, July 1998. We thank Willem I kise'r. Olga ,Janssen, Klaas ten Boeve, and Jonathan VeIWeij for sharill~ their historical knowledge. Slwcial thallks arc due (0 Frans (k Mast, whose help with 1he history of censuses, surveys, and SLit ist iI'S Ne't llf'rlands was extrelllely valuable. I [is work was a great Illspiration. The Views expressed ill this paper arc those of the a ut hoI'S and do 1I0t necessarily reflect tIl(' policies of Sla tistics NdlHTlallds.

(2) In alphabetical order. Contact address: Dr. Edith D. de Leeuw, MI'l ho(likA, Plantage Doklaan 40, NL-I 0 18 CN Amsterdam, N,'t!ICrlands; td/fax 31 20 622343S: ('lliail edilhL@xs4aLL.nL

(:1) Sussmikh was a chaplain in the annies of Frederick The Grcat of I'nlssia. I-Ie was an admirer of G.'aunt and published, in 1741, I Jil 'ille Order in the Changes oj the IIurnan Uace Shown bLJ Its Birth, [)eu/h, and Propagation. Sussmilch was sevnely attaf'ked by a young British clergyman and mathematician, Malthus, 50 years later for his "naive generalizations ahout the differences between city and

40 41

p 42 Rulleh"n de MithodoloJ:ie Sociologique (59 rue PuuchCI, F 75017 Pan,,), Octohn jl)I)9, N t14

country" and the unsystematic collection of facts including "thc failure to include years of epidemic in his figures" (Boors tin, 1985: G71 J. Malthus published this criticism in An Essay on the Principle or Population in 1798. It is a mixture of quantitative social science ,mel moral philosophy, and had a very practical goal: to rdonn the Poor Laws of Britain. At thc same time. it had an immense influence Oil science and economy in thc 19th century. Malthus influcnced such divcrse thinkers as Marx, Keynes and Darwin (Boorstin. I~85:

'l74).

14) Then. as now. the citizens of Britain arc the least suicidal of Europe. Then. as now. Europeans of every nations were 1Il0re suicidal in summer than in winter (Hacking. [990: 75).

(5) The ,general public knows Quetelet not for his mathematics, but for the "Quetelet-index" which indicates whether the weight falls within the "nonnal" range. The Quetelet index is the weight in kilograms divided by the square of the length in mf'ters. [f t IW

resulting figure is between 20 and 25. one is considered tu bc "nornlal" ,

(6) This study. done in cooperation with Burnett-Hurst. was published in 1915.

(7) Booth and his co-workers produced very precisely colored maps of London. Each of the eight colors represented a soc:ial-eeonomie elass. ranging from "very poor and a danger to society" to "can afford servants" (De Mast, 1997). Using modern technology. this same met hod is now used in market research and mal'ket ing. Based on tlI<' postcode. geo-demographic coding schemes give detailed infonnaUon on inhabitants of cities. As a I·esull. colored maps of eitics. with different colors for each social-economic status, can be viewed on a computer monitor!

(8) At the congress of the [nternational Statistical Institute ill 18!l5, Gcorg von Mayer put it this way: "il faut restcr fenne et dire: pas de ('aleul la ou l'observation peut etre faite" [we should be finn and keep saying. no mathematics wherc observations can b(' dOIlcl (Dc Mast. [997J.

(9) This again has it roots in community refoml at the pnd of the [9th century, At Columbia University. a new statistical laboratory was started for social statistics. The English mathematician, William Ogburn, was one of its first directors. Pupils of Ogburn were, among ot hers. Burgess. one of the founders of the "Chicago School". and Stoufkr. head of the Department of the US AmlY Infomlation and Education Branch and author. in 1949. of The Americun SoldieT'.

~2

!iulldi" d~ M~tJlOdolo~ll' .\'oriolol:iquf' (')!) rue 1'11\Khet, f 750171'MI.'1. ()l-!llhl'r 11)1)()" N h~ r.43

(I OJ This was onc of thc first confclTncTs in a scrics o( international ('ollf('!TIHTS on sUr',"'Y ll1ethodolo/Zy, sponsorpd bv the Survey 1,('S('.1I'('h Method (SI,M) s('ction of the 1\uHT\('an Statistical /\ss()('\;llion (1\SA). Thcse l'Cmfercnces and the resulting monographs. which \\'('IT publishcd by Wi!<·y. contributed Illuch to the II1IPJ"{l\TI1Wllt of slirvey lIH'thodology.

II II ILllies .1ddcd 1J~' the duthors.

1{1':FFlHJ\;l'ES

/\Itl'sldlldc[', 1'., alld Kncubuhler. lL-U. (1~75) Verzernl11gen irn Illten'icll'; Zu Cin(T Pc/derlfleorie cler I3clrugun(j. Oplarkn: W('stdclltschn Verlag.

ILliJl)1e, E.!\. (I ~l7JI SllTlWlj Uesearell Metllods. Belmont: Wadsworth.

IlclsOIl, W.A. (1981) Tile Design and Understanding or Survcy qlll'stions. Aldcrshot: Gower.

Bi('nHT, 1'.1'., GWV('s, R.M .. Lyberg. L.E .. MaUliowetz. N.A.. and SlldlIl'lll. S. ([ 99 [). Measuremcnt Errors in SUTlJcys. New York: Wiley.

Billiet. ,J.. ,IrIt! Loosveldt. G. (1988) "[mprovement of th(' quality of H('sponses to Factual Survey Questions by [nterviewfT Training", ['uhlic Opinion Quarterly 52, 190-21 [.

!I()OI'still, Dallid J. (1985). TIle Di.~colJerers: A IIisloTlj or Man's ,'-;('urdt 10 Know his World uneI I Iirnsel{ New York: Handom [louse.

!lmvlcy, A.L. (1906J "Presidcntial Address to the Economic Section of til(' Bhtish Association for the Advancement of Scicn('c". JOlmwl or (/1(' UOljul Statistical SOciety. 69. 540-558.

Bowley, A.L. ([ 926) "Measurement of t h(' Precision Attained in Sdl1lpling". BILlIctin or tllc International Statistical Insututc. 22.

[lowky, A.L., and BunI<'tt-[Iurst. AR. ([915) LiuelillOod and l'olJCrtlj: A St udt} in the Economic Conditions or Working-CI(Jss IIol1scholds. LOlldon: Bell.

BH'I1IH'[', M. (I~l78) "[l1terviewillg: The Sucial !'lIenomcnoloiZY or a !\('s(',\!'c!I IlIstnll1H'l1t". [11: M. BI'f'llncr. P. Marsh ,ll1d M. I1ITI111f'I' (cds.) Tlw SoC'i(J/ Contexts of MethoeI. Lolldon: Croom Helm. pp. [£2, ITl.

~1

lIulletin de AI~/hodolo!:i( SocioloKique (Sq rue Pnuchet. F 75017 PMI'), Octohcr Iljl)l), N 64 p.45p.44 Bullt'lin d" Alhhodnln!:;f Sociolo/:iqllf (Sf) rue Pouchct. F 75017 P.lrl~L Ocloh...·r l\Jqq, j\,' (d

Brenner. M. (19H2) "Response Effl'C'ls of l{ok-I{,'striclcd Charactelistics of the Interviewer". In: Dijkstra and Van dl>I' Zouwell (1982), pp. 131- I05

Bulmer, M. (197!l). Censuses, Sun'c!js wnt I'riIlClCIj. London: Macmillan.

Campanelli, I'. (1997). "Testing Slln:(~y Questions. Spl'l'ial Issul' Oil

Cognitive InteIYiewing". l3ulletin de Mdhodologic SucioloCJiCJuc (BMS), 55, ,June 1997.

Cannell, C.1"., Lawson. S.A., and Hausser, D.L. (1975) A Tcchniquc Jar Evaluating Intemiewer Performance: A Manual Jor Coding and Analyzing Interviewer Behavior from Tape Recordings of House/wld Interviews. Ann Arbor, MI: Survey Research Center of the Institute for Social Hesearch. University of Michigan.

Converse, J.M. (1986) Survey Research in IIII' United States: Roois and Emergence I890-1960. Berkeley: University of California Press.

Deming, W.E. (1944) "On Errors in Surveys". Americall Sociological Review, 9, 359-369.

Deming, W.E. (1982) Quality. ProductiVity, and Competitive Position. Cambridge. MA: M.I.T.

Dijkstra, W.. and Van del' Zouwen, J. (I988) "Types of Inadequate Interviewer Behavior in Survey Interviews", In: W.E. Saris and LN. Gallhofer (eds.) Sociometric Research, Vol. 1: Data Colleclion alld Scaling. London: Macmillan, pp. 24-35.

Dijkstra, W., and Van del' Zouwen. J. (cds.) (1982) Ues!)()f1se Behavior in the Survey Intemiew. London: Academic l'ress.

Dillman, D.A. (1978) Mail and Telephone SunJCys: The Total Design Method. New York: Wiley.

Dumas, J .S., and Redish, J.c. (1994) A Practiml Gnide (0 [Jsahilitl/ Testing. Norwoocl: Ablex Publishing Corporation.

Forsyth, S.H., and Lcssler, J.T. (1991) "Cognitive LaboratOly Mdhods: A Taxonomy". In: Paul P. Biemer, Hobert M, Groves. LeU'S E. Lyberg. Nancy A. Mathiowetz amI Seymour Sudman (1991) Measurement Errors in SlImeys, New York: Wiley.

Fowler. F.J.Jr. (1991) "Reducing IntelYiewer-Related EtTol' Throll,gh Interviewer Training, Supervision, and Other Means". Ill: l'alll 1'.

H

niemer. [~obert M. Groves. Lars E. Lyberg, Nancy A. MaU1iowclz and SeVIlloul' Sunman (1991) MCUSllreT1lC1\( Ermrs in Surveys. New York: Wi1l'y.

Creell, IJ.F. (I D54) "Attitude Measurement". III HwuJ/)()ok of SociClI J'SI/c!IO!O(/I/, v()! J. Cambridge. MA: Addison· Weslev

(;rO\'('s. HM. (I !)8'l) SUnJCI/ ETTors WId SlIwe!/ Cosls. New York: Wilcv.

(;l'IlVl's. RM .. Bicll1er, P.l' .. Lyber,l2,. L.E., Massey, ,J.T.. NICholls. W.Lll ,111(1 Waksbeq..;. ,J. ((>cls) (l!lHHj Telephone SunJey Md/lOd%(/I/. N,'w York: Wiley

(;rovl's. H.M .. and Couper, M.l'. (199H) Nonresponse in /louse/wid lfl(cn,icll' Sun.'eys. New York: Wiley.

I LlI'king. I. (I !)~lO) DIe Twninf./ of Chancc. C,lI1lbridge; Cambridge UnivLTsity l'ress.

'I 11<11'1. II. (1 !)72) "lnterviewlIloeheid en IWijfcb over het cnqllete­onderzoek". Sociologische Gids, 19 (4), 249-257.

Ik I [(ocr. W, (1996) "International Response Trends: Development amI [{esults of an Intenlational Survey". Paper presented at the 4th Inlemalional ~ocial Srienee Methodology Conference, July 199G, EssI·x. cn. Forthcoming, JOllnwl of O[{iciClI Statistics, 1999.

1k I len. W. (1999) International Iksponse Trencls: Results ot an Illtcrtlational Slll'VCY. JOllnwl ofOI/icial StatL.,tics, 15. 129-142.

I Ieisn. W. (19!)G) "De probabiliselinl-( van het wcrel(I\J('c1d". Lezing llill-(I~sprokl'n op hel Symposium Il'r gelegenheid van het 25-jarige IlOol2,[(oraarschap van Ivo Molenaar. Gronin,l2,el1, 4 OctoLJCr 1996.

Iloulkoo[J-Stccnstra. II. (I ~)96) "Probing 13ehavior of InlelYiewcrs in thl' Stall(bnlized Semi-Open Research Interview". QIW/ity wHl (cll/(l1Iti(I/, :30. 205-2:\0.

I lox. J.J. (1997) "Er is nieuws onder de zan; Nieuwe oplosisngen voor otllie problemen". I~ede uilgcsprolH~u bij de aanvaarding van het ,lIubt vall hoogleraar "Illethoelen ell tcchniekn van socia;ll­wcl!'nsl'happl?Iijk onelerzoek" aan de Univnsileil Utrecht. I May l~l!l7.

Ilox. ,J,J .. De Lceuw, KD .. anel [{reft. I.G.G. (I!J!JI) 'The I<:flcct or Inll'l'\'ieWIT anel I{esponelcnt Characteristics on the Quality of SUI'Vey I),,!;t: A Multilevel Mocll'l". In: l'alll I'. Bi(,IlllT, l{oblTt M. Groves.

~,

r- 46 Bulletin de Methodologie Sociologiqu~ (59 rue Pnuchcl, F 75017 Pans), OLlnher 1999, N 64

Lars E. Lyberg. Nancy A. Mathiowetz and Seymour Sudman (1991) Measurement Errors in Surveys. New York: Wiley.

llyman. H.H. (1954) Interviewing in Social Researcll. Chicago: Chicago University Press.

Dc Jong-Gierw-Id. J .. and Van del' Zouwf'n. cJ. (1986) Dc Vru(JcnliJst in het sociaal onderzoek. Deventer: Van Loghum Slatnlls,

I\ahu. R.L.. alld Cannell, C.F. (1957) The [)ynwlIics oJ In (clI'iClI'illrj. T1won/, Techniques, and Cases. New York: Wilf'y.

I\enessey, Z.E. (1997) Adolphe Quetdet as Inlcmationalist. AIITIlwl [{eport on Intenwtional Statistics, Vol. 4- 1997. The Ha.t:;lIe-Vool'bul'g: International Statistical Institute.

Kish, L. (1965) Survey Sampling, New York: Wiley.

Lyberg. Lars. Biemer. Paul. ColIins. Maltin. de Lecuw. Edith. DipjJo, Cathryn. Schwarz. Norbert. and Trewln. Dennis (1997) Slln,1C!/ MeCL.'wrel11cnt and Process Quality. New York: Wiley,

De Leeuw. E,D. (1992) Data Quality in Mail. Tclcp/wrf(', Wid Fu(',' to Face Surveys, Amsterdam: TI-publikatics.

Dc Leeuw. E.D. (1997a) "Comment on Witt: Best I'ra cl iccs in Interviewing via Internet". Sawtooth Software Corl!crcncc Proceedings. SeqUim: Sawtooth.

Dc Lceuw. E.D. (1997b) "Nonrespons: Wat weten we en wat kunnen we eraan doen", In: A.E. Bronner et al. (red) Reccn/c on/wikkclingcn in het marktonderzoek. Jaarboek I997 PWI de N('(1crl(lndse Vereniging uoor Marktonderzoek en In(ormalienlClnagerncnl. lIaarlem: Dc Vrieseborch,

Dc Lceuw. E.D.. NieholIs. W.L.ll (1996) "Teelmological Innovations in Data ColIection: Acceptance. Data quality and Costs". Socioloqico/ Hesearch Online. I , 4 <http://www.soeresonline.org.uk/socresonlinf./1/4/kellw.html>.

De Leeuw. E,D .. and flox. J.J. (L988) "Hesponse Stimulating Factors in Mail Surveys". Joumal oJ01ficial Statislics, 4.24 I -249.

Dc Leeuw. E.D.. Hox. J .. and Snijkers. G. (1995) 'The Eitel'! of Computer-Assisted Interviewing on Data Quality: A Review", ,]ollnw/ o( tile Markel Research Socicly, 37. 325-344.

/lullellll de MctllOduloKIt· ,\-oci"I0I:It/IU' (C;;q fUl' l'ulIche!, F 7-"017 PJrl<,,), (}clnhcr I I)l)!) , N 64 P 47

Lessin.. J.T,. alld Kalsbeek. J.T. (1992) Nonsamplin(J Error in SUFIICI/S New York: Wiley.

M;\as. CY.. and ell' I leer. \\'.1' (1995) "Response Df'velopmcnts and J-'j"ld\\'ork Sl ral q.;y". Bul/ctin de Md/lOdoloqie Sociolor/ique, 41", :16­" I .

1)(' I\hst, F (I ~l97) "On'!' Statist iekcn. SlII-vevs cn het CBS: Eell I ~,\<.,i<.,p,lkkel \'all Bc~rippell. F('ilcll en Nanll'n". Intenl CBS rapport.

1\1o<,,(·r. (' ,i\. (I ~l;,K) SIIII'('1/ M<'Ihods in Suciul Irwesti<iution. Londoll.

11"II\('lllallll.

Nicllolls, W.L.II (198H) "Computer-Assisted Telephone Interviewing: /\ (;"I1<'I;t! Introduction", In H.. M. Groves el al. (cds) TeIephonc SIlT1WI/

,\[l'II lorlO/O(/I/. Ncw York: Wiley.

NiliiOlls, \V.L.lI (I !l!}G) "Chronology of CASIC Events", Intern l'apn. \ IS lllln'<Ill of I"c Census.

Nixon. ,J.W. (I !l(0) A IIistory CJ( tile Inlemational Statislical ,bsoci(llion. '111e Ilagllr: International Statistical Institute.

OMuirc"eartaigh. C, (1997) "Measurement Error in Surveys: A IlistOl'ieal Perspective", In: 1~'1rs Lyberg et at. (cds) Survcy /\!C'usuremcnl and Process Quality. New York: Wiley.

l'<Iync, S.L. (1951) 771('Ar/ o(AskingQuestions. Princeton: Princeton Univcrsity Press.

Phillips. D.L. (1971) KnolUlcdqc(rorn Wha/'! 771corics and Methods in Socinl [{<'scorch. Chicago: Rand McNally,

I'''illips, D.L. (1973) Ahrmdoninq Method: Sociological Studies in MC(/lOdoIO(J!J. San Francisco: Josscy-Bass.

I'Lit. ,J. (1996) A 11ist0I1/ o(Sociological l<escarcll Methods in America: 1U20- I ~J(j9. Cambridgc: Cambridge University Press.

l'n'ssn, S. (191"4) "Tile Usc of SUl-VCY Oata in Basic ResealTh in the Social Science'S". In: C.F. TUll'll'r and E, Maltin (cds.) Su[[)eying SIII)j('cIiL'C 1'/lCTlOfllcno. New York: Russell Sage Foundation,

Rossi. !'('In II., Wlight, Jall1f's 0 .. and Anderson. Andy 13. (198:~)

"Sample Slln;cys: IIistory. Current Practice and Future Prosjlects". Ill: I'l'tn I I. Hassi. ,lall1f's D. Wright and Andy B. Anckrson (cds) / IU/I(II}()o/( oJ StLrtW!J Hcsearrll, Ncw York: Academic I'rcss.

4(, 47

p. ..IX llllUetin de MhhodoloKie 5;ociologiqlle (59 rue Poucht'l, F 7S017 1',1l1~ 1, Ol!{1ht'r 1\)(){) , N (14

Sahs. W.E. (1991) Computer-Assisted Interviewing. Nl'wblllY Park: Sa~('.

Schaeffer. N.C. (199 I) "Conversation with a Purposl'- or Conversation? Interaction in the Standardized Interview" In: Paul P 13iC'Iner. Robert M. Groves. Lars E. Lyberg. Nancy A. Mathiowl'tz, and Seymour Sudman. Measurement Errors in SurvcIIS. New York: Wiley.

Schwarz. N.. Sudman. S.( 1996) Answering Qw!s(ions: Melhodologll Jor Determining Cognitive and Communicative Processes in Sun'('!1 Uesearch. San Francisco: Jossey-Bass.

Sikkel. D. (1996) "The Individualized Interview". Paper presCIllecl at the INTERCASIC'96 Conference in San Antonio. Texas.

Simonetta, L.G. (1997) "B1ockin~ Telephonl' Calls" . AAPORNET/WAPOHNET, June 1997.

Stigler, S.M. (1986). History oj Statistics. Cambliclge. MA: Harvard University Press.

Sudman. S.. and Bradburn, N.M. (1974) Response Effects in Surveys. Chicago: Aldine.

Slldman. S .. Bradburn, N.M.. and Schwarz, N. (1996) 111inking about Answers: The Application oj Cognitive Processes to Surve!) Methodology. San Francisco: Jossey-Bass.

Sykes. W .. and Collins. M. (1992) "Anatomy of til(' SlIIvey Intelview". Journal of o,[ficial Statistics, 8, 277-291.

Tumer. C.F.. and Martin, E. (1984) Surve!)ing Suqicctive Phenomena. New York: Russell Sage Foundation.

Van Hattulll. M., and de Lceuw. E. (1996) "A Disk by Mail SUlvey of Teachers and Pupils in Dutch Primary Schools: Logistics and Data quality". Paper presented at the INTERCASIC'96 Conference in San Antonio. Texas: MLS-publikatie n. 56. Universiteit van Amsterdam. Forthcoming. Joumal of o,[ficial Statistics (1999: 3).

Willis. G.B .. Royston. P., and Bercini. D. (1991) "The Use of Verbal Report Methods in the Development and Testing of Survey questionnaires". Applied Cognitive PS!Jcholog!). 5, 251-267.

Witt. K.,J. (1997) "Best Practices in Interviewing via thl' In(en1et". Sawtooth Software Coq{erencc Proceedings. Sequim: Sawtooth

48