ESRC RESEARCH GROUP ON CARE, VALUES AND THE FUTURE OF WELFARE

University of Leeds

Workshop Paper No 18
Prepared for Workshop Four
Methodologies for Researching Moral Agency
Friday 17 March 2000

Simon Duncan

UNDERTAKING FIELDWORK AND GETTING DATA - A FOUR-PRONGED RESEARCH DESIGN

1 Building on the 'feminist sociology norm' - introduction and summary

How are we going to undertake the field work for the empirical projects in Strand 3? To some extent, I get the feeling that there is an assumption that we simply carry out in-depth, semi-structured interviews. We might call this the 'feminist sociology norm'. There are many strengths to this norm, particularly in comparison to the 'male positivist norm' of questionnaire and survey based empirical sociology, and it is one that forms a central part of my own research with Ros Edwards on lone mothers and paid work (Duncan and Edwards 1999). However, this 'feminist sociology' norm is not without its problems, and in some ways can be seen as just as one-sided (if a better side!) as the 'male positivist norm'. In this paper I argue for:

(1) a multiple qualitative research strategy, with three prongs of

a) in-depth semi-structured interviews,
(b) focus groups, and
(c) interpretive biographies.

These would respectively provide information on (a) behaviour, attitudes, beliefs and motives, (b) collective local discourses and group interaction and (c) the development of feelings about social moralities. ( This scheme follows discussion with Wendy Hollway and Gillian Burgess - see Hollway and Jefferson 2000 and Burgess 2000).

(2) 'grounded interview analysis

This would provide a means of analysing the interview material from the the in-depth semi-structured interviews, so as to provide generalised and categorical information which remains grounded in the accounts of the respondents.

(3) a mixed-methods approach, using qualitative and quantitative information.

Quantitative information gives us a 'fourth prong', which allows us to assess the generality of the qualitative information about social relations and understandings by using representative quantitative information as a check. Conversely, the qualitative results provide process explanations for the more descriptive information on characteristics found in quantitative data. This allows us to better link evidence on social process and cause (on how and why things happen) with evidence about social patterns (on what has happened).

The claim is that these four prongs will together provide a better means of accessing motives, beliefs and feelings about motherhood and paid work, and of evaluating their significance. The four prongs will also give a high level of triangulation. This research design would in addition allow better presentation of the results as properly 'evidence-based', where the prejudice can be that a small number of qualitative interviews are just anecdotal or unrepresentative. (Such critics should read Sayer 1984 or 1992 on the advantages of intensive over extensive work in accessing social process and cause, but they usually haven't!).

This then forms a proposal for carrying out the field work for the Mothers and Employment project in Strand 3.

2 Semi-structured interviews and grounded interview analysis - the first prong

2.1 Interview method, selecting groups, and getting respondents

Following the 'feminist sociology norm', this was the central method used by Ros Edwards and myself in our lone mothers research, and indeed in current joint research, based at South Bank University, on partnered mothers and gendered moral rationalities (where the pilot research for the Cava Mothers and Employment project forms my field work contribution to the SBU project).

I will not go into into much detail on how Ros and myself carried out the semi-structured interviews for the lone mothers project (interview date 1994), and the partnered mothers project (interview dates 1998 -Simon and 2000 - SBU team). This was according to the 'feminist sociology norm', and as such is well known. We used semi-structured interviews using an aide-memoire with a list of questions, prompts and pursues, carried out over 1- 1.5 hours, which sought to discover beliefs, attitudes and motivations as much as behaviour and characteristics. Interviews were with a few (10 was the target) members of selected social groups (10 for the lone mothers project including 3 abroad, 6 for the partnered mothers project), which were thought to be theoretically interesting (eg young African-Caribbean lone mothers, alternative partnered White mothers). Adding to the 'feminist sociology norm', interviews took place in comparative case study localities which were seen to encapsulate social contexts in which these groups lived. (In these cases Inner south London and Hebden Bridge respectively). Interviews were carried out by interviewers already in relevant social networks, using snowballing. Interviews were transcribed, but material felt to be extraneous (eg talk about pets), was summarised rather than recorded verbatim.

This design can also be justified by reference to Sayer's realist account of explanation in social science, where intensive work of this type is superior in accessing social process and in establishing social cause. This interview design is, however, weaker in terms of representaveness, and I will return to this when discussing the advantages of combining this sort of information with quantitative information. More unusual, perhaps, is the method of analysing the interview material, which I have called 'grounded interview analysis'. I discuss this at some length in the next section, and would appreciate some comments on this.

2.2 Grounded interview analysis

'Grounded' refers to the attempt to use the interview data inductively, so that the production of abstracted analytical categories comes from the respondents' accounts. However, another name might be 'interpretive interview analysis' as the method does seem to depend quite a bit on the analysts' understandings in carrying this out.

Briefly, the procedure was as follows:

Stage A - Creating analytical categories

1. Start with one particular case group, probably a 'less conventional group' so that less conventional ideas are not 'drowned out' at the beginning.

2. Work through all the transcripts identifying all talk relating to the issue of analysis (eg mothering and employment, or partnering, or childcare).

3. 'Semi-abstract' the selected talk into generalised statements that reflect the meaning of the respondent's words, but aren't verbatim quotes.

As we are not interested at this stage in the position of the individual respondent, but with a list of possible statements, these can be taken out of context. For example, in the current SBU partnered mothers project, a lesbian mother who held values about joint control over income, housing etc, talked about how her own mother, in contrast, was supported by her father and consequently had no control over money. We interpreted this to provide a general, abstracted statement that ' Supported mothers don't have control over spending money'.

The interpretation of statements can also be tenuous at times. For example a working-class White partnered mother in Hebden Bridge was complaining about coming home from long part-time work to find the place a tip and piles of ironing. This was abstracted to ' Male partners don't carry out domestic work'.

4. Create a list of statements abstracted from the interviews. (Because we are creating categories here, not analysing the position of individual respondents - which is Stage B - these do not need to be repeated more than once).

5. Sort the statements into similar / connected groups, which then form analytical categories to use in the individual interview analysis. We can than give these categories inductive / conceptual names, eg for the lone mothers project 'primarily mother', 'primarily worker', 'mother/ worker integral'.

One difficulty here is getting statements in the most appropriate category. For example, should the abstracted statement ' Male partners don't carry out domestic work' be part of a 'self-sufficient individual' group of statements (as a negative), or a 'homemakers and carers' category (where an alternative positions of 'sharing partners' would be inappropriate').

Stage B - Analysing individual interviews and locating individual / group positions

1. Go through each individual interview and record the number of times where the respondent makes talk which fits into one of the categories.

2. Locate the respondent's position, using this quantitative information, with respect to the categories identified in Stage A. This allows for transitional / contradictory locations eg an individual lone mother may be located 2/3rds of the way along the ''primarily mother', 'primarily worker' axis, and 1/3rd up towards the 'mother/ worker integral' position..

3. Diagrammatically represent the location of the individuals in each group, and subsequently summarise and contrast group positions diagrammatically.

See Figures 1 and 2 for examples of the result of all this from the lone mothers project. The claim is that this 'first prong' of semi--structured interviews, using grounded interview analysis, will provide categorised information on attitudes, beliefs and motivations.

2.3 Limitations of the 'feminist norm'

The semi-structured interview has a number of limitations, despite its relative superiority over much 'hard' research in accessing process and so establishing cause. These limitations are:

1. It accesses information at the individual level, while some of the information we want refers as much to collectivities (eg neighbourhood social networks) as much as the individual.

2. There is potential for the respondent to reply in terms of a discourse she feels is expected, and this becomes all the more severe as we travel down the path from 'facts' to 'feelings'.

3. It is weak in accessing general patterns.

The next three prongs are incorporated in the research design so as to compensate for these weaknesses.

3. Two more qualitative prongs

3.1 Focus groups

The prime role here would be to access dominant local, neighbourhood, social network collective discourse. Individual respondents may have qualified or even quite different views, but the idea here would be to find out what dominant, expected, approved views are See Wilkinson 1998, Rahman 1996). The interactions observed in focus groups may also be valuable in pointing to areas of different interpretation. Vignettes may be an apropriate technique for the study of normative and moral beliefs and attitudes (Rahman 1996).

3.2 Interpretive biography

The role here would be to access individual belief systems (subjectivities) and how these have emerged in relation to respondents' social contexts and experiences. How do mothers feel about good motherhood and paid work, and how do they relate these ideas to their biogrphical experience.? This is possible is more 'sensitive' semi-structured interviewing, but interpretive biography can better relate this to respondent's biographical development. Technically, and related to this, interpretive biography can better discover important issues not recognised by the interviewer, either practically in the semi-structured schedule or indeed theoretically. The method also allows superior access in the technical sense of better avoiding 'discourse determination' by the interviewer (cf Hollway and Jefferson, 2000).

The operational model I have here is Terry Allen's 'narrowed down', but 'extended' version of interpretive biography(Allen, 1998, 2000, see also Hollway and Jefferson 2000). Noting that most versions of biographical research in fact say very little about how practically to do it, he develops a method inthe course of his research on housing renewal and well-being. First of all, Allen narrows down the discussion from 'tell me about your life' (the life story approach) to 'tell me about your experiences of housing renewal'. Maybe we should call this 'guided interpretive biography'. Secondly, Allen extends the method operationally. All interviewees received a follow-up visit where they were invited to comment reflexively upon a tape, delivered to them a week earlier, of their first interview. At this stage interviwer's prompts could also be made. This, then, was more of a semi-structured interview, but partly structured by their own previous account.

4. The fourth prong - Quantitative analysis

4.1 Mixing methods

One of the criticisms of the 'feminist sociology' approach is that, because it is generally based on a small number of in-depth interviews, it is ungeneralisable and unrepresentative, and therefore inadequate in explanatory terms. This assertion is incorrect, and in itself reflects a misunderstanding of cause and process. However, it does remain the case that such research is weak in terms of its ability to discover general patterns, and hence to establish representaveness, however good it is at accessing social process and thereby establishing cause. This is where mixing quantitative and qualitative research can strengthen the research. This is then the the fourth prong.

The distinction drawn by Andrew Sayer in his Explanation in Social Science (1984, 1992) between 'extensive' and 'intensive' research is useful here. The former refers to research which aims to describe overall patterns and distinguishing features, for example the characteristics of a population. Taxonomic groups (eg 'lone mothers') are the type of group studied usually (though not necessarily) using quantitative data from large scale surveys including official statistics. While producing representative description, this design is weak on explanatory power, that is on how something happens. Intensive research in contrast seeks to find out how a process happened by focussing on what agents actually do. It focusses on substantive process connections in causal, social groups (eg 'alternative lone mothers living in a gentrifying area of Brighton) rather than taxonomic groups.. Taking a close-up look, it can better identify processes and mechanisms, going beyond simple association. It is often 'local' in that it deals with the complexities of social action in context, usually (although again not necessarily) employing in-depth case study and qualitative methods.

Each research design has different strengths and weaknesses, and are therefore more or less appropriate to different research questions. In this way they should be seen as complimentary. Unfortunately, and unhelpfully, they are usually regarded as being in opposition to one another. This is particularly evident in the strict division often made between 'hard' quantitative, and 'soft' qualitative, data and methods (see Brannen, 1992). There are three problems here. First of all, the complimentarity of the two research designs is lost. Secondly, the reduction of research design (what the research is able to do) to techniques and methods (how the research is undertaken) in itself exacerbates the division. There is no necessity that a particular research technique or method is limited to either intensive or extensive research. Thirdly, and most unfortunately, there have been strong tendencies to privilege one design over another. Usually this has taken the form of seeing extensive research and quantitative methods as somehow superior, partly because of its associations with economic science, figures and machines, and men. Ironically, this is despite its weaknesses for explanation. A reaction by the excluded has been to privilege intensive, qualitative research, just because it has been particularly associated with feminist analysis, personal contact and women. Either sort of privileging further exacerbates the tendency for researchers to cut themselves off from complimentary research designs and to make unwarranted claims for their own.

In contrast a combination of intensive and extensive research designs allows us to assess the generality of the qualitative interview information about social relations and understandings by using representative quantitative information as a check. Conversely, the interviews provide process explanations for the more descriptive information on characteristics found in quantitative data. This allows us to better link evidence on social process and cause (on how and why things happen) with evidence about social patterns (on what has happened). It also functions as another triangulation method. In both these ways the strength of the research is enhanced. So too (as the male, positivist slang implies) its perceived 'robustness' and presentational strength. The research is all the more 'evidence-based'. Certainly Rosalind Edwards and myself found this combination very useful in seminars on our work (Duncan and Edwards 1999) in both scientific and presentational terms.

4.2 Sources

One source for extensive research is the 1991 British census. For information on local labour market areas, and on case study areas and neighbourhoods within them, the Local Base Statistics (LBS, providing census variables at District Council level) and the Small Area Statistics (SAS, giving the same information at ward and the even smaller enumeration district levels). A major source for socio-economic information on different social groups of lone mothers was the Sample of Annonymised Records (SARs) and the Longitudinal Survey (LS). The former provides a 1% sample of UK households in 1991 (the Household SAR) and 2% of individuals (the individual SAR) and the latter follows a one per cent sample of individuals through census dates. The SARs and LS give two major advantages over conventional, area based census sources, such as the LBS and SAS. First, analysis at the level of individuals avoids the 'ecological fallacy' of extrapolating individual social associations from aggregate spatial averages . Second, the SARs and LS allow the creation of census categories, and the production of statistical outputs from them, in line with the needs of research (subject, of course, to what was asked in the census), rather than having to rely on those predetermined by the Office of Population, Census and Surveys in their published output. While the SARs provide 'snapshot information for one census date in 1991, the LS provides longitudinal data between 10 year census periods.

A major problem is that the 1991 census is ten years out of date. Unfortunately necessarily detailed results from the 2001 census will not be available in the lifetime of Cava's empirical projects. Nor does the census provide attitudinal data. There are alternatives, such as the General Household Survey, although these do not usually have enough detail to match quantitative data to substantive groups, and hence tend to be used taxonomically. One source which might get round both these problems is the British Household Panel Survey.

5. Conclusion

These four-prongs of: 1. semi structured interviews with grounded interview analysis, 2. focus groups, 3. interpretive biography, and 4. mixing qualitative and quantitative methods are proposed as a the research design for the mothers and employment study.

The claim is that this design will be superior to relying on semi-structured interviews alone in that it will be more efficient in accessing feelings, beliefs, attitudes and motivations, it will better combine information about substantive relations and representative patterns, and it will allow much greater triangulation. The result should therefore be better description and explanation which can claim to be 'evidence based'.

References

Allen, T. (1998) 'Interpretive biography as a method: researching tenants' experiences of housing renewal' International Journal of Social Resaearch Methodology, 1, 3, 231-49.

Allen, T. (2000) Housing, health and well-being: uraban renewal in Allerton, Bradford. Unpublished PhD thesis, Department of Applied Social Sciences, University of Bradford.

Brannen, J. (ed.) (1992) Mixing Methods: Qualitative and Quantitative Research, Aldershot, Avebury.

Burgess, G (2000) 'Finding motives and beliefs: a three-pronged qualitative research design' paper presented to Cava workshop 4, 17 March 2000

Duncan S. and Edwards R. (1999) Lone Mothers, Paid Work and Gendered Moral Rationalities. London, Macmillan.

Hollway, W. and Jefferson, T. (2000) Doing Qualitative research Differently: Free Association, Narrative and Interview Methodology, London, Sage.

Rahman, N. (1996) 'Caregivers' Sensitivity to Conflict: The Use of the
Vignette Methodology', Journal of Elder Abuse and Neglect, Vol. 8(1): 35-47.

Sayer, R.A. (1992) (2cd.edition) Method in Social Science: a Realist Approach, London, Routledge.

Wilkinson, S. (1998) 'Focus Group Methodology: A Review', International Journal of Social Research Methodology, Theory and Practice, Vol. 1 (3):
181-203.

BACK