Surveys, Statisticians and Sociology: A History of (a Lack of) Quantitative Methods

Abstract

While respecting the intellectual rigour of social constructionism, it is problematic that much of current UK sociology remains fixed in a qualitative research methods fugue. Despite programmes like Q-Step and its ESRC predecessors challenging this narrow approach, most sociologists are trapped by their lack of numeracy, projecting it onto new generations of undergraduates and new entrants to the profession. However, to explain why this situation has arisen, and why an innovatory programme like Q-Step cannot on its own neatly solve all of our difficulties, needs a better understanding of the discipline’s history and the social context which determines the form that sociology takes.

This article seeks to demonstrate how the framework of higher education social institutions has shaped sociology’s evolution before, during, and since the ‘great expansion’ of the discipline in the 1960s, from a tiny cadre to a substantial professional group. The rapidity of growth from small beginnings is emphasized, when the initial teaching of ‘social research methods’ actually consisted almost exclusively of mathematical statistics and survey methods, often taught by non-sociologists. This turned new generations of young sociologists against quantitative methods, leaving them open to the attractions of new intellectual schools like social constructionism and feminism. Whilst acknowledging other contributory causes, such as the cost of survey research, this review of early sociology’s approach to research methods prompts the lesson that today’s Q-Step and similar reforms should aim to assist the great majority of undergraduates to acquire quantitative skills, rather than producing a small number of statistical experts.

Keywords:

• social research methods
• quantitative methods
• qualitative methods
• teaching sociology
• history of sociology

Introduction

British sociology’s ‘crisis of number’ (Williams et al. 2004) has both generic and discipline-specific roots. On the one hand, the generic causes, shared with other social sciences, lie in the English education system where ‘Arts’ students give up quantification after GCSE, arriving at university not having practised number work for over two years. Indeed, the conservative teaching of GCSE Maths, oriented towards the internal logic of that subject and the production of future mathematicians, rather than equipping young people to analyse their social worlds, is poor preparation for learning quantitative methods as social science undergraduates. On the other hand, the discipline-specific causes lie buried in sociology’s history of professional development and undergraduate teaching. This article attempts to excavate some lessons from the historical record and apply them to the current welcome debate about enhancing quantitative methods, and the Q-Step programme. In showing how the discipline has evolved, and giving greater emphasis to the historical and institutional context, a case can be made that enhancing basic quantitative skills for the majority of students is both more beneficial, and a more feasible policy, for sociology, than seeking to promote high levels of statistical sophistication among a few specialists.

A key feature of UK sociology’s history is how research methods became routinized when this tiny new discipline grew abruptly during what we might call the ‘Great Expansion’ of the 1960s. The consequences of the combination of being so small with the speed and scale of expansion, at a particular period when new ideas were making an international impact, set off the implosion of quantitative methods (for convenience of exposition this article retains the arbitrary quantitative–qualitative dichotomy). This fostered the myth of sociology’s quantitative/‘positivist’ past that would subsequently be invoked to protect the self-interest of later generations of non-numerate sociologists.

‘Smallest, least popular, and least well-endowed’

By the First World War there were still only 19 British universities (Platt 2003, p13) in which the social sciences continued to be ‘the smallest, least popular, and least well-endowed’ subjects during the inter-war years (Burns 1935, p231). The year the Second World War started, Britain produced only 33 graduates in sociology, social anthropology and social administration combined (Payne et al. 1981: separate figures for sociology are unavailable due to the records’ format). Even two decades later, ‘only 12 universities were listed as offering degrees in sociology’. But within four years this had risen to 19 (Banks 1967, p2), plus the Colleges of Advanced Technology being upgraded to university status (among them Aston, Bath, Bradford, Brunel, City, Loughborough, Salford, Strathclyde and Surrey which would all go on to offer degrees in sociology), and the 30 emerging polytechnics denied university status until 1992. By the early 1970s, the annual output of sociology and social anthropology graduates (i.e. excluding social administration) topped 1,700 (Smith 1975).

This 5000 per cent increase in student numbers was matched by a massive expansion in staff. Just after the war, the London School of Economics (LSE) with two chairs and five lecturers, and Oxford with one lectureship, were the only institutions employing full-time sociologists. Even adding in ‘social studies’ and ‘social anthropology’, the total was meagre: this over-generous enumeration would credit LSE with 19 staff, Bedford College with five, and University College six, while Belfast, Birmingham, Edinburgh, Glasgow, Leeds, Liverpool, Manchester and Nottingham could be included as having ‘social studies’, plus Oxford and Cambridge’s 11 anthropologists (Commonwealth Universities Yearbook 1949–50, quoted in Savage 2010, p122). By the early 1970s, these barely 60 staff had grown to 1200 sociologists teaching in higher education and another 900 employed as researchers (Smith 1975). Inevitably this expansion could not consist of staff trained exclusively in sociology, or with higher degrees in the discipline (Platt 2000, Platt 2012).

During its early period, sociology had something of a schizophrenic character. On the one hand it propagated a wide, comparative and largely theoretical perspective, in which Ginsberg and others grandiloquently regarded the other social sciences as mere hand-maidens of Sociology, the Queen of the Social Sciences. On the other hand, sociology owed its survival to being initially taught in vocational social administration degrees with a practical political orientation to social policy and the amelioration of social disadvantage, i.e. as part of an applied programme. These courses drew on religion and morals, history, law and psychology. Sociology was unable to produce sufficient graduates to form a professional core: permeable boundaries between psychology, economics, politics, education, social work and social administration permitted cross-recruitment and collaboration. However, there was dispute as late as the mid-1960s over whether this multi-faceted subject should be seen as ‘sociology’ or ‘social science’ (Platt 2003, p14, Oakley 2011).

The BSA’s founding proposal in 1950 illustrates this diversity:

The Association should deliberately be made very broad, embracing such fields as contemporary, historical and comparative studies of social structure, morals and religion; sociological aspects of Law; social philosophy; social psychology; social–biological aspects of mankind; social aspects of urban and rural settlement; human geography; and methodological aspects of social investigations.

(Banks 1967)

While this reflects a wish to recruit wide support, its breadth demonstrates sociology’s dependence on sister subjects before the Great Expansion.

Statistical methods in social investigation

The concentration of sociologists at LSE and the University of London meant these institutions subsequently had a profound influence on how sociology as degree programmes was established elsewhere. Nearly half of the growing number of UK chairs in sociology were filled by former LSE sociologists, who took with them the London University model of an undergraduate social science degree (Payne et al. 1981, p24). The broad social science content of most ‘sociology’ programmes until the 1970s resulted not only from the inter-disciplinary, ‘social science’ intellectual inheritance, but also from practical considerations: combining sociology with sister disciplines meant fewer sociologists were needed to deliver a degree programme. This also applied to the London University BSc (Hons) Sociology – and Economics – external degrees, taught by university colleges (e.g. Southampton before 1952, Exeter before 1955 etc.), and most of the pre-polytechnic colleges in the 1960s.

The contribution of the polytechnics to UK sociology has been underestimated in patrician histories of sociology. By 1974 there were more polytechnic teachers of sociology than in the universities (Platt 2003, p148) and by the mid-1980s between one-third and one-half of BSA members were in the polytechnics (personal survey of membership listing). Initially, polytechnic sociology ‘staff members had lower average qualifications than those in universities’ (Platt 2003, p148), as well as much poorer funding, and being tasked with a narrowly vocational and ‘teaching’ mission. However, a number of the former polytechnics out-performed some older universities in the 1992 Research Assessment Exercise, the first to which they were admitted (HEFCE 2009).

Despite the polytechnics’ overall lower mean score, the five top polytechnics – Anglia, East London, Middlesex, Plymouth and Teesside – were ranked higher than seven older universities, and on a par with 14 others. They achieved this typically not with expensive survey work, but using other methods requiring less institutional funding. However, with limited staffing, many of the new polytechnics adopted the universities’ earlier solution to teaching research methods, often out-sourcing it to social statisticians.

The 1964–65 calendar at LSE shows five courses in methods, four of which were taught by statisticians:

The main research methods course at this period was entitled ‘Survey methods in social investigation’ and taught by Claus Moser . . . ‘research methods’ was more or less equated to ‘survey methods’, although the existence of other methods was recognised.

(Bulmer 2011, pp80–81)

‘Statistical Methods of Social Investigation’ (also the ‘methods paper’ in the external degrees) interpreted research methods as survey research, with very limited consideration given to epistemology, methodology or alternative methods. ‘Methods’, with other social sciences, were chiefly completed before Part II, whereas most sociology papers were taken in Finals. There were no Stage Three dissertations which could have connected methods to other sociological content.

The first half of its three-hour exam paper (the only assessment) required calculation of t-tests, χ² values, and correlation and regression coefficients by hand in the exam room, and from memorized equations. The second half covered data collection in official statistics, sampling and non-sampling error, non-response, interviewer bias, and types of sample. The only qualitative technique regularly included before 1970 was participant observation. Apart from the rubric to ‘illustrate your answers by reference to relevant studies’, there was no requirement to connect methods to sociology.

The implicit but incorrect message to undergraduates (who of course included the next cohort of postgraduates and academic sociologists) was that research methods consisted essentially of social surveys, and by implication, earlier sociology had been essentially quantitative – and even ‘positivist’.

Moser’s (1958) eponymous book, not surprisingly the pre-eminent text for nearly two decades in its original form and then as Moser & Kalton (1971), provides further clues to the Statistical Methods in Social Investigation syllabus. Documentary sources, participant observation and ‘informal interviewing’ were the only alternative methods included in the first edition, occupying 17 (5.3 per cent) of the 320 pages of text (Moser 1958, pp162–173 and 204–209). Documents were briefly discussed as contributing to sampling for surveys (approved); similar treatment was given to unreliable and non-comparable case studies by non-statisticians (not approved) and un-representative and difficult to access ‘personal’ statements (not approved). Participant observation and community studies were patronized in a slightly longer section. Informal interviews are accepted as:

Invaluable at the pilot stage of even formal surveys . . . They do, however, require more skill and alertness to the danger of personal bias than formal methods: they are also relatively slow and expensive .

(Moser 1958, p209)

This representation of ‘methods’ as survey methods is central to the failure of quantitative methods in UK sociology, and the subsequent confusion between techniques and epistemological questions (Marsh 1979, Platt 1981). Many research methods courses were taught by statistics lecturers with little sociological knowledge, who were reluctant contributors (and sometimes less skilled teachers, allocated to do low level ‘service’ work). Sociology students had little interest in product-testing in ball-bearing manufacture, or elaborate mathematical proofs for probability theory. Undergraduates from Arts backgrounds encountered assessment in methods as the numerical challenge of working out statistical coefficients by hand in the pressured atmosphere of the exam.

This institutional framework almost inevitably devalued research methods and particularly quantitative sociology. When the expansion in sociology began to accommodate new degree designs and staffing, methods teaching was reclaimed from the statisticians, only to be re-isolated in a new ghetto, usually as one module rather than integrated into the broader curriculum (Payne et al. 1989), and taught by a single ‘quants specialist’ (Williams et al. 2004) – or in an abrupt change-over, by lecturers whose first-hand expertise was limited to qualitative methods.

Two legs of a three-legged stool: survey research and social theory

One could argue that despite the pedagogical style or format of the emerging sociology curriculum, a focus on survey research was, at that stage, justified. Early ‘surveys’ were often less narrowly based on random sampling, questionnaires and formal interviews than are today’s ‘social survey’ format. For example, they combined interview data with administrative data on crime, morbidity, housing stocks and unemployment rates and observations of family life. University sociology drew on Booth and Rowntree’s survey tradition (Kent 1985), although much of this work came from social statisticians and human geographers assisted by amateur volunteers outwith the universities. Led by statistical researchers (e.g. Carr-Saunders et al. 1927), these studies ‘combined the techniques of Le Play with the methodologies worked out by demographers and vital statisticians’ (Carpenter 1935, p680). Examples include the New Survey of London Life and Labour (Llewellyn Smith 1930–1935), studies of Merseyside (Jones 1934), and numerous local and regional surveys by amateurs in organisations like the National Association for the Promotion of Social Science, the Sociological Society, the Charity Organisation Society, the Eugenics Society, and the statistical societies of London and Manchester (Scott & Husbands 2007, Scott 2008).

As Wells (1939, pp424–7) notes, two approaches in survey work emerged, one focusing on the local, and the other on the regional or national. Apart from the a-political Statistical Societies, the impetus for both came from a concern (shared with sociology) as much for the problems of poverty, unemployment, housing, and health – ‘the insistent questions that are in all men’s minds’ (Thomas 1936, p461) – as from purely quantitative impulses. The increasing influence of what we would recognize as more modern, post-war survey methods in the style of Lazarfeld or Moser happened in part because the subject of the research required extensive and national, rather than intensive and local, data collection to support a case for modifying social policies.

In the 1940s and 1950s, outside of the small university core of social scientists, ‘social research’ increasingly became equated with the emerging forms of modern survey research (Bulmer 1985). Ironically, this tended to divert expertise away from universities, apart from a few London-based specialists (Bulmer 2011): in addition, those who conducted most of this work were not sociologists. Efforts by the SSRC and its successor ESRC in the 1970s and 1980s to promote survey methods were unproductive (Platt 2012, pp692–3), unlike more recent ESRC programmes (ESRC 2008, MacInnes 2009, Payne & Williams 2011, ESRC 2012). Survey research was thus never well-embedded in sociological research, let alone its dominant activity.

While post-war sociology inherited a concern for social inequality, early sociologists were not over-concerned with practical research at all. Spencer, Hobhouse, Westermarck, and to a lesser extent, Ginsberg addressed broad inter-disciplinary understandings of ‘civilisation’, rather than the detail of empirical research or social policy. These gentlemanly pursuits of ‘society’ and ‘world history’, to borrow Savage’s (2010) term for the intellectual climate up to the mid-1950s, reflect an elite university system and the dominance of a small professional class.

During the Great Expansion, social theory continued to flourish, sustained into the 1970s not by comparative history, social evolutionism, or 1960s Parsonian structural functionalism, but by Marxian debates (Payne et al. 1981, pp70–86), in which there was little attempt to deploy empirical evidence. The few exceptions like Westergaard & Resler (1975) do not alter Bottomore’s Marxist Sociology observation of a ‘notable absence’ of ‘thorough historical and sociological investigations’ even of social class, let alone other social institutions in Marxian sociology:

Marxist sociology has largely failed to develop empirical studies of particular social phenomena. There have not been significant and extensive Marxist contributions to the study of crime and delinquency, bureaucracy, political parties, the family, or to a great number of specialised fields of inquiry.

(Bottomore 1975, p73)

This tendency to abstraction has since fed on post-modernism and cultural studies.

This theoretical emphasis contributed to the absence of a canon of technical research expertise, essential to the practice of the discipline. The National Subject Benchmarking Statement for Sociology (QAA 2007) does not require sociology graduates to complete even one piece of empirical research, using any method. Today’s continuing resistance to the suggestion that undergraduate dissertations should combine theory with empirical research is another manifestation of a discipline still hooked on abstract theory.

The third leg of the three-legged stool: other methods of research

Despite the survey tradition’s visibility, other methods of research had been practised at the same time, due not least to the diversity of views of what the discipline should be. For example, the Webbs had made heavy use of participant – or ‘personal’ – observation as Beatrice called it, in their study of trade unions and local councils (Webb and Webb, 1932, pp158–201). Social anthropology was developing fieldwork techniques including flexibility in interviewing (RAI 1929, Richards 1939). Even enthusiasts for surveys like Wells (1939, pp429–430) spoke approvingly of the methods deployed by the Chicago School, and the Lynds’ Middletown. As Savage (2010, pp1–11) demonstrates, modern routinized and highly structured survey interviews are different from the more personal and discursive practices of the 1950s and early 1960s.

The importance of alternative methods shows in early research methods textbooks. The Webbs’ Methods of Social Study, the first such book in Britain, focuses on analysing how social institutions operate. Its authors claim to:

Give in detail the methods of investigation used by us in our successive studies . . . we have been unable, through lack of qualification, to use, and perhaps even adequately to appreciate, the Statistical Method.

(Webb & Webb 1932, ppv–vi)

Precision in systematic recording of information, and exploration of as many alternative explanations as possible, were prioritized. Chapters on ‘The Written Word’, ‘The Spoken Word’ and ‘Watching the Institution at Work’ drew heavily on Beatrice’s field notes.

The second textbook to appear came just at the start of World War II, Bartlett et al.’s The Study of Society (1939), which similarly did not focus on social surveys or quantitative methods. It contained three chapters on ‘Some Methods of Social Anthropology’ and five on ‘Some Methods of Sociology’. Only one of the five dealt with social surveys (Wells 1939, pp424–435). The limits of survey analysis were identified by one contributor who, in citing Boas, observed:

Anyone reading The New Survey of London Life and Labour, or The Social Survey of Merseyside, for example inevitably reflects that it would be illuminating and important to have parallel studies . . . of the behaviour related to the variously statistically analysed conditions.

(Oeser 1939, p402, p406)

The Study of Society reflects both the variety of research techniques and the diversity of sociology/social science between the wars.

However, foreign imports dominated the post-war market for methods textbooks. Jahoda et al. (1951), Goode & Hatt (1952), Festinger & Katz (1953), Selltiz (1959) and Duverger (translated from the French in 1959) were supported by more specialist texts and readers such as Payne (1951) on questionnaires, Lazarsfeld & Rosenberg (1955) on survey analysis, and Hyman (1954, 1955) on interviewing and survey design. The first home-grown textbook was Madge’s (1953) The Tools of Social Science, which covers surveys (‘The Mass Interview’) and experiments, but allocates nearly two-thirds of its pages to documents, observation, interviewing, and the logic of research work. Apart from Abrams (1951) and Moser (1958), British research methods books adopted a similar broad perspective; e.g. Stacey’s Methods of Social Research (1969) and the BSA’s social research series (BSA 1969, Gittus 1972) – the rather rudimentary tone of the latter revealing how under-developed empirical research was at the time. General textbooks might have offered more guidance on methods, but there were few available, and like MacIver & Page (1949), and Cotgrove (1967), said little about doing research beyond brief statements of epistemological issues (see also Platt 2008).

These British methods textbooks (Moser (1953) excepted) demonstrate that UK sociologists did not concentrate exclusively on a ‘positivist’, social survey, or quantitative approach. The period of using American texts more oriented to quantification reflects an initial absence of home-grown methods books, itself a product of low levels of interest in empirical research among the still small community of UK scholars, rather than an active interest in quantitative methods. Before the appearance of more specialist contributions – Cicourel (1964), Webb et al. (1966), Glaser & Strauss (1967), Denzin (1970a, 1970b), Geertz (1973) and the expansion of the qualitative methods literature – methods books covered a broad range of methods.

Given sociologists’ earlier openness to different methods, and justifiable student resentment at how Methods was taught, one might almost expect qualitative methods to have been even more dominant. At LSE, ethnography had been available as an alternative model for some time, courtesy of Malinowski and Radcliffe-Browne, while the post-war, post-colonial ‘coming home’ of social anthropologists briefly offered a new approach. However, despite a number of successful studies of urban areas and work settings, ethnographic and participant observation work failed to take over, perhaps because anthropology itself was also a small discipline (Mills 2008, p1), paying little attention to research technique. Notes and Queries on Anthropology had only 27 out of 366 pages dealing directly with ‘Methods’ (RAI 1951/1971, p36–62). Until the late 1970s, learning ethnographic techniques relied on word of mouth and ‘experience’, with little codification of skills or cumulation of knowledge.

Ethnography had also been set back by the failure of the Mass Observation, which collapsed as a movement in 1947, splitting into a market research firm and an under-used archive until its re-discovery in the last decade. The publication of Malinowski’s diaries which queried what actually went on in fieldwork (Malinowski 1967/1989), and the perceived decline in community studies (Payne 1996) discouraged serious ethnographic work in the late 1960s and 1970s (although see Crow & Allan 1994, and Qualidata’s ‘Pioneers of Qualitative Social Research Project’ (ESDS 2012) for a more optimistic interpretation). American books on qualitative techniques which had begun to appear a decade earlier were either strong on theory but less use to a young researcher seeking practical help (Glaser & Strauss 1967, Denzin 1970a), or relatively un-integrated collections of essays (Denzin 1970b, Filstead 1970).

The qualitative turn

The abrupt growth of British sociology in the Great Expansion coincided with an international revolution in sociological ideas, as sociology re-focused on the question of how individuals deal with their social encounters. The 1960s brought in symbolic interactionism, phenomenology, ethnomethodology, and social constructionism, while the next decade saw Marxian disputation supplanted by feminist sociology. This was followed by post-modernism in which discourse shifted to the cultural contextualisation of words and ideas. Powerful though the new ideas were, their impact also results from UK sociology’s weak empirical research tradition. These new perspectives were seen as triggering an epistemological crisis: there were ‘no longer any obvious criteria for adequately certifying knowledge, nor for judging the appropriateness of methodological procedures, nor even for what are suitable topics for investigation’ (Bell & Newby 1977, p9: emphasis added).

Sociology’s Great Expansion had created a younger generation of academics, with different interests, in competition with an older generation of (male) sociologists. Traditional ‘political’ topics like class, power and material inequality were replaced by an interest in social interaction and process, a change in levels of analysis from the broad brush and extensive to the narrowly focused and intensive. Worsley’s critique of the new perspectives was less about technique, than about studying the wrong topics: for

. . . applying great ingenuity and sophistication to the personally sometimes problematic, but societally insignificant difficulties of managing the first five seconds of conversations, walking down the street, or exchanging “Good Mornings” in a world agonised by infinitely greater social problems . . . The abjuration of societal concerns has bred a new generation of sociologists, who . . . actually know little about society because they do not wish to know.

(Worsley 1974, pp15–16: emphasis added)

Feminist sociologists were not a-political but rather political in a new way: their personal politics embraced sexual harassment and violence, the body, emotion, domestic labour and family relationships, etc. The development of a more reflexive approach, and the ideological need for an alternative model of researcher/researched relationship, reflected institutional pressures:

Feminism needed a research method, a different methodology, in order to occupy a distinctive place in the academy and acquire social status and moral legitimacy. Opposition to ‘traditional’ research methods as much as innovation of alternative ones thus provided an organising platform.

(Oakley 1998, p725)

Without wishing to imply that this was a conscious and deliberate policy, and so to devalue the intellectual rigour of the newer perspectives, a similar argument can be applied to other members of the new generation of sociologists, who found in ethnomethodology, social constructionism and cultural studies a means of establishing social status and moral legitimacy in the academy as a largely unintended consequence.

Other contributing factors

Although we do not want to lose sight of the core point that the experience of being taught statistics for survey methods, instead of sociological research methods or even quantitative sociology, lies at the heart of the flight from numbers, the institutional dimension of the Great Expansion also extends to important issues of resource availability. As a young and relatively non-empirical discipline in the UK, sociology took its educational and intellectual model from the humanities, rather than the sciences or the applied arts. Unlike science and engineering, with their laboratory training, or fine art with its studio work, or even human geography with its field-trips and social psychology with its small group experiments, sociology set its students to read in the library, talk in tutorials and write essays.

The rapid growth of sociology as a whole was not simply a rise in the demand among young people to study society and social relations, but also an increase in the supply of opportunities at low cost to the HEIs. Not only was the university sector expanding as a whole, but sociology offered a quick and, crucially, cheap means of filling the newly created places. The new discipline was willing to function with no specialist labs, no expensive equipment, no teams of technicians, no subsidized fieldwork, no staffing to cover 20 plus hours of class contact, nor even library stocks of hideously expensive journals. Staff could be hired cheaply because more of the new generation were on lower grades.

On anecdotal evidence, the inexperience of sociologists in university politics during the Great Expansion exacerbated this situation. Few (inevitably junior) sociologists sat on senior committees where budgets were allocated, or indeed wanted to participate in administration: this was a 1960s and 1970s generation more inclined to reject hierarchy and conservatism, unsympathetic towards university ‘bureaucracy’, and opting for the purity of ideas in preference to engagement with pragmatic deals and practical solutions. As a result, there was little understanding among increasingly less-numerate sociologists of how university incomes could be manipulated to sustain the generous spending habits of higher status subjects like Science and Engineering (subjects less popular among student applicants, but well-established in university power structures) at the expense of the naïve new kids on the block. Institutional manipulations of income re-distributed resources in favour of more established subjects by means of baseline budgeting and subject weighting factors (such as in staff–student ratio calculations) resulting in under-funding of sociology. The discipline could only afford to develop as a Humanities subject with no practical work, thus reinforcing its weak research traditions, but of course making the subject acceptable to the planners in the new universities and polytechnics.

The resource issue also played directly into the practice of research. Large-scale social surveys were expensive and so required a high level of success in the competition for research grants, but funding bodies showed an understandable reluctance to entrust large sums to the care of relatively junior investigators with little track record of successfully completed projects (or to institutions like polytechnics). Survey research also takes a long time to organize, and to analyse the data collected. Qualitative research methods in a small-scale project require much less funding and tend to be easier and quicker to set up, albeit possibly needing a longer analysis stage. This is not to imply that one method is intellectually easier than another, or can always be completed more quickly, but simply to note that survey research costs more. For example, one criticism of the major national social mobility studies of the early 1970s was that they absorbed too high a proportion of the research council’s annual budgets. It is not surprising that where quantitative methods survived, it was largely outside of the universities, in independent institutes and government agencies (Bulmer 1985).

While not wishing to descend into technological determinism, it is also important to remember the timing and sequence of innovations. Small portable, reel-to-reel tape recorders were becoming available and affordable in the mid-1960s, replacing paper, pencil and memory, and so facilitating recording of semi-structured interviews. At that point, computing was still based on punch cards, counter-sorters and main frame installations. Clumsy counter-sorter machines helped speed up analysis (even the classic ‘community study’, Family and Kinship in East London (Young & Willmott 1957) acknowledges these innovative contributions to its data processing.

Although the number and processing capacities of computers were increasing, access was still typically limited to single, overnight ‘batch’ runs: early versions of SPSS becoming available in the 1970s placed heavy demand on processing power. The ESRC Data Archive was still in its infancy: access to the documentation and data-sets it held was a laborious and decidedly user-unfriendly process of ordering spools of tape and getting them read onto a mainframe, and struggling with opaque code-books. Desk-top computers would not begin to provide opportunities for flexible and swift data analysis until more than a decade later. Whatever the philosophical or intellectual merits of qualitative methods, their practical advantages and their resonance with an older tradition which had eschewed the formalities of structured data collection, analysis and theorising, gave them an advantage during this period.

As successive reports of the time documented (Carter 1968, Peel 1968, Wakeford 1979, Payne et al. 1988), the lack of quantitative output and formal methods training provided a poor base for development. Even allowing that sociologists also publish in a very wide range of journals (Kelly & Burrows 2012), one later manifestation of this was the well-documented dominance of qualitative methods in mainstream UK sociology journals (Payne et al. 2004, Crothers 2011, MacInnes et al. 2013). In the early 1960s, one-tenth of research-based articles relied on qualitative methods (Platt 1981), compared with six in ten in the last decade (purely theoretical articles consistently comprise about one-third). By the turn of the century, on a strict definition of quantitative work, 95 per cent of empirical articles in the mainstream journals by UK-based sociologists (admittedly not a complete representation of our national sociological output) relied on qualitative methods (Payne et al. 2004). Platt (2012, p.691), using a different definition, suggests a lower figure.

It would of course be wrong to leave the impression that UK sociology has been purely qualitative in nature: generalising about a whole discipline is a risky business (Payne 2007), especially one as diverse as sociology (Scott 2005, Stanley 2005). Several university departments have fought the good fight for quantitative analysis, producing not just admirable research output but also talented young numerate sociologists: among others, obvious examples are Cambridge, Cardiff, Edinburgh, Essex, Manchester, Stirling and Surrey, together with formerly Newcastle Polytechnic, and more recently Plymouth. Nuffield College too has championed quantitative methods, but as Crompton (2008) has argued, its impact has been diminished by the aggressively negative stance taken by some of its members towards other methods and indeed colleagues. An indication of the flavour of this is Erikson’s claimed intention to enhance the reputation of British sociology, but which is actually a coded message about his own distaste for qualitative research and the democratisation of sociology. He asserts that the BSA is not a learned society, dismissing the journal Sociology out of hand:

Sociology has in recent years, on the whole, been not worth following . . . British sociology is in a bad shape if this issue of the journal is representative of the prevailing standards . . . If the BSA wishes to live up to the expectation of being a learned society it will certainly have to put much work into bringing its general sociological journal up to international standards.

(Erikson 2005, p127, p129)

This tone is hardly likely to win converts within the profession to Nuffield’s preferred quantitative method.

Erikson’s attack on the British Sociological Association is misinformed. Necessarily reflecting the variety of approaches taken by its members, the BSA has at various times minimalized the importance of practical research training (e.g. in its original input to the QAA national benchmarks), advocated qualitative methods (in consultations with ESRC), and promoted quantitative methods. In the latter, it produced the BSA’s social research series mentioned earlier, established the Quantitative Sociology Group which operated in the late 1960s and 1970s (Platt 2012, pp698–9) and sponsored a series of projects, such as those led by Gubbay on the curriculum in the 1980s and by Williams and Payne on the teaching of quantitative research methods in the last decade. The association’s journal Sociology has carried a number of articles on the methods curriculum, not least those by Becchofer, Marsh, Platt or Wakeford, in addition to the influential special issue edited by Burgess and Bulmer in 1981. Association members have been at the forefront in making the case for better teaching of research methods, increasing basic research training for postgraduates, taking leading roles in the National Centre for Research Methods, and making significant contributions more recently to the revival of interest which has led to the Q-Step programme.

Conclusion

The purpose of this historical review has been to show how what we practise as social research is not, as we commonly assume, simply a result of free-floating knowledge or the logical superiority of our intellectual frameworks. Rather, practice has to be seen as dependent on social processes, processes which are much more basic and mundane: the size of the discipline, the point at which it developed, the growth and control of higher education, the curriculum as taught, the availability of resources. The demographics and institutional context of sociology’s emergence as a university discipline constrained the teaching of degree programmes, and particularly social research methods. The neglect of this aspect of the curriculum inflicted poor statistical teaching on sociology students, who found it irrelevant and distasteful. The lesson for contemporary curriculum reform is that quantitative methods must be controlled by sociologists themselves, and embedded in the core of the discipline. If not, we shall see a repeat of the mistakes of the 1960s, because the current Coalition Government’s austerity programme imposes similar financial constraints on sociology’s development.

A case can be made for the production of more social statisticians per se (one of the thrusts of current ESRC and indeed Government policy), but the priority for sociology should be enhancing undergraduates’ basic number skills, their capacity to reason with quantities as part of an expansion of research methods training as a whole. Pace the good intentions of the Q-Step programme, what we really need is a new cohort of undergraduates who have greater confidence in, and skills to apply, core quantitative analysis and perspectives where appropriate, as one of several research tools, rather than trying to enforce sophisticated statistical techniques that our students understand only at a superficial level, and will probably never use. A generation of more numerate students would provide new, more numerate postgraduates and young professionals, able to choose a wider range of research methods and topics, and appreciate the centrality of how research is done in creating sociological knowledge. If they then wish to develop greater statistical competence, they will be in a stronger position to do so. Unless they have come to see quantitative approaches as integral to sociology and the exploration of the sociological topics that interest them, they will continue to be reluctant to embrace work with number. And that in turn depends on a substantial change in attitudes and skills among the current older generation of sociologists.

The intellectual ferment of the Great Expansion was not simply about the introduction of non-quantitative methods. The new perspectives shared a move towards a closer, more detailed analysis of parts of the social realm, the exploration of less deterministic processes, and a shift of interest from macro-sociological issues like class and power, towards the micro-worlds of identity and feelings. Research topics helped to drive the selection of research methods away from quantitative techniques and into small-scale, ‘softer’ methods. Ironically, the key tenet of methodological pluralism, that ‘topic should determine method’, normally invoked to protect quantitative methods, has indeed been applied to opposite effect: early sociologists bred new generations whose choice of topics pointed them towards qualitative research methods.

The prospects for success in stimulating quantitative methods are constrained by what sociologists want to study, not just by their current skills set. If students are to see quantitative methods embedded across their degree programmes, degree content needs to be broadened. Work in progress in Q-Step projects, for instance at Cardiff, Edinburgh and Manchester Metropolitan, are beginning to show a way forward. The availability of archived data-sets for secondary analysis, some with ‘teaching versions’ and others in preparation, continues to grow, although the problem of user-friendly interfaces remains.

From its early days, sociology drew on a variety of research methods: there never was a golden age of quantitative methods or social survey work. Indeed, there has never been even a silver age of research methods: a strong tradition of theoretical work produced cohorts of sociologists whose own empirical research skills were at best little more than adequate, and who cared insufficiently about inculcating the skills needed for effective research practice. It follows that recent welcome initiatives to promote quantitative methods should aim to improve research, not just quantitative techniques, and to re-orient undergraduate sociology towards active, ‘studio-based’ learning. We are not dealing with a technique but rather the core practice and content of our discipline.

Acknowledgements

I am grateful for the helpful comments of the Editor and reviewers on my initial draft.