data analysis and interpretation: emergent issues in

This is a repository copy of Data analysis and interpretation: emergent issues in linking qualitative and quantitative evidence.

White Rose Research Online URL for this paper:http://eprints.whiterose.ac.uk/104662/

Version: Published Version

Book Section:

Irwin, S orcid.org/0000-0001-9591-147X (2008) Data analysis and interpretation: emergentissues in linking qualitative and quantitative evidence. In: Hesse-Biber, S and Leavy, P, (eds.) Handbook of emergent methods in social research. Guilford Publications , pp. 415-436. ISBN 978-1-59385-147-7

[email protected]://eprints.whiterose.ac.uk/

Reuse

Unless indicated otherwise, fulltext items are protected by copyright with all rights reserved. The copyright exception in section 29 of the Copyright, Designs and Patents Act 1988 allows the making of a single copy solely for the purpose of non-commercial research or private study within the limits of fair dealing. The publisher or other rights-holder may allow further reproduction and re-use of this version - refer to the White Rose Research Online record for this item. Where records identify the publisher as the copyright holder, users can verify any specific terms of use on the publisher’s website.

Takedown

If you consider content in White Rose Research Online to be in breach of UK law, please notify us by emailing [email protected] including the URL of the record and the reason for the withdrawal request.

mailto:[email protected]

https://eprints.whiterose.ac.uk/

CHAPTER 20

Data Analysis and InterpretationEmergent Issues in LinkingQualitative and Quantitative Evidence

Sarah Irwin

In this chapter I consider the value of usingdifferent kinds of evidence, from macroto micro, in exploring social processes.

Quantitative and qualitative data provideparticular lenses on these different “levels.”However, we need to acknowledge thatthese levels are often not distinct researchproblems but rather different dimensions ofunitary problems. For example, to under-stand an individual’s values and subjectivebeliefs, we need knowledge of him or her asa person, knowledge of his or her proximatecircumstances and experiences, and knowl-edge of the wider social structural contextsin which he or she is positioned. An inquiryinto well-being may reveal a poor person tobe more satisfied with his or her situa-tion than a wealthy person. It is only byunderstanding the social distribution ofwealth, the social organization of aspirationand constraint, and people’s diverse circum-stances that we can make sense of their per-

ceptions and of the differences betweenthem. Researching macro and micro is notjust about “linking data”; it is an issue of howwe conceptualize the phenomena we are in-vestigating.

In this chapter I explore some examples inwhich using different sources of evidencecan enhance our understanding and expla-nation of social processes. I argue that:

• Most of our research problems are com-plex and multifaceted. Different methodsand sources of evidence will reveal specificslices through the phenomena and pro-cesses under study, and we need to under-stand better precisely how the evidence re-veals a partial and particular picture. Forexample, in-depth interviews and survey re-sponses may provide different lenses onpeople’s perceptions of some particularevent or state of affairs, and so differentkinds of account are generated by the differ-

415

From Handbook of Emergent Methods, Edited by Sharlene Nagy Hesse-Bieber, PhD and Patricia Leavy, PhD.

Copyright 2008 by The Guilford Press. All rights reserved.

Copyright © 2008 The Guilford Press. All rights reserved under International Copyright Convention.

No part of this text may be reproduced, transmitted, downloaded, or stored in or introduced into any

information storage or retrieval system, in any form or by any means, whether electronic or

mechanical, now known or hereinafter invented, without the written permission of The Guilford Press.

Guilford Publications

370 Seventh Ave., Ste 1200

New York, NY 10001

212-431-9800

800-365-7006

www.guilford.com

ent media of data collection. Another illus-trative example here is the way people pro-vide accounts, in open-ended interviews, oftheir decisions about the best type of carefor their children in terms of a moral choice,yet it is only by understanding (say, throughsurvey evidence) that people in similar cir-cumstances make the same kinds of “moral”choices that we can see that morality pro-vides us with a partial account of the natureof their decision making.

• We need to keep under reflexive andcritical scrutiny the categories we use to or-ganize our thinking and order our data. Thisis consistent with seeking more expansiveand systematically adequate explanations ofthe processes under study. For example,gender, age, and ethnicity are standard vari-ables used to denote difference, yet thesecategories are not always adequately theo-rized in respect of the processes or patternsunder investigation. They may be effectivecategories for revealing structures of in-equality, for example, but qualitative re-search helps reveal that the salience of thesecategories of difference may vary across con-texts, in respect of material inequalities, sub-jective orientations, and so on.

• Developing adequate conceptualiza-tions of the phenomena and processesunder investigation must remain at the heartof social analysis. We use theory-laden cate-gories through which we interpret empiricalevidence, itself shaped by our tools of datacollection. Nevertheless, empirical data cansupply us with tools for reinterrogating, ex-panding, or changing our conceptualiza-tion. In the examples in this chapter Iconsider how drawing together data fromdifferent sources and “levels” of the socialcan contribute to enhancing social explana-tion.

The chapter is organized as follows. In thenext section I briefly explore some issuesthat emerge from recent debates on thevalue of mixing methods. I argue that weneed to understand how methods and evi-dence entail concepts about the nature of

what we are researching. Empirical evidencecarries with it the assumptions that went intoits making, but we need also attend to pre-cisely how it bears on the research problemat hand. It may give a partial picture in that itshows only one “part” of the complex phe-nomenon under study. Importantly, we alsoneed to be aware of the risk that a partial pic-ture may be a distorting one. It may misleadus as to the salient processes shaping thephenomena we are researching. Addi-tionally, it is inappropriate to treat macropatterns and knowledge of diversity as“background” context and qualitative evi-dence on meaning, interaction, value, andso on as holding a more direct line to “pro-cess.” I introduce some issues and examplesin considering how both macro- and micro-level lenses are important in building notjust a broader picture but a more adequateunderstanding of social processes. The thirdsection continues the theme of social expla-nation with reference to different models ofadequacy in connecting theory and data. Ilook at the historical example of explanatorymodels used to understand the incidence ofcholera in the 19th century. Critics haveseen in the different models lessons aboutadequacy and progress in social science re-search. Some argue that now, still, standardways of representing and modeling quantita-tive data are too abstracted from social phe-nomena and risk distorting the processes inwhich we are interested. We see how theform of data and modes of analysis shapeour understandings in particular ways. Nev-ertheless, there is scope for empirical dataanalysis not merely to confirm prior assump-tions but to contribute to theoretical expan-sion and transformation. The fourth sectionexplores and develops some of the themesthrough a consideration of empirical data.The examples all share a concern with thelink between people’s attitudes and percep-tions on the one hand and the social andeconomic structures in which they are em-bedded on the other. In recent research andtheory there is a tendency to treat individualperceptions and values as a “layer” of subjec-

416 INNOVATION IN DESIGN AND ANALYSIS









New York, NY 10001

212-431-9800

800-365-7006

www.guilford.com

tive and normative understanding that is dis-tinct from aggregate structures of distribu-tion, for example. This has been a source ofarguments that ideology prevents peoplefrom seeing the workings of an oppressivesocial system and of linked arguments thatin late modern society, values and choicesare more freed from social structural pro-cesses than they were in the past. However,people clearly hold diverse positions withinsuch structures, and it is to these contextsand positions that we need to relate theirsubjective orientations. We can understandthe links between micro and macro, and be-tween subjective and objective, only if wehave a sufficient understanding of socialstructural diversity. We need to move acrosslevels of evidence in seeking to adequatelyunderstand subjective orientations. The ex-amples I develop all reveal a connectednessof subjective orientations and social struc-ture. Additionally, they show that bringingtogether evidence, which is a theoretical is-sue more than a technical one, can help ustackle puzzles of explanation and transformunderstanding.

Issues in Linking Methods

There has been a recent surge of interest inmixed methods research and its potential.The idea of mixing methods is not new.Many writers have long advocated usingmixed methods, and many have done so intheir research practices. But there has beena renewed interest among social scientistsand funding agencies, in line with a percep-tion that mixing methods provides a way for-ward and perhaps a renewal of our re-sources for tackling social complexity andcontemporary social problems. Mixingmethods is sometimes defined in differentways, but in general it can be taken to referto bringing together qualitative and quanti-tative data collection and analysis (Teddlie &Tashakkori, 2003).

Many different rationales and schemahave been devised for elucidating the differ-

ent ways in which quantitative and qualita-tive strategies can be brought together.There are many ways to combine data andmany examples of good practice in this area.Brannen (2005), Bryman (2001, 2005),Hammond (2005), Mason (2006), Moran-Ellis and colleagues (2006), Teddlie andTashakkori (2003), and many others have ex-plored and classified ways in whichmultimethod and mixed methods researchproceeds. For example, there are sequentialmodels, in which one strategy follows fromthe other in the design of the research. Apreliminary qualitative (part of a) studymight sensibly precede a quantitative (partof) one to generate insider knowledge andinsights that would feed into the quantitativedesign. Alternatively, quantitative researchmight generate (among other things) a sam-pling frame from which key informant par-ticipants might be identified for a qualitativestudy. Another common reason for usingmixed methods is that, in tackling complexissues with different component parts, dif-ferent methods may be deemed most appro-priate to different parts of a study. The pat-tern of enhancement, in which data fromdifferent methods are seen to be supplemen-tary and adding value, or insight, was one ofthe most common uses of mixed methodsthat Bryman found in his content analysis ofU.K. social science articles published be-tween 1994 and 2003 (Bryman, 2005). An-other common claim for mixed quantitativeand qualitative research is that qualitativemethods allow us to interpret the relation-ship between variables. Thus we might haveevidence of associations at a macro level andinfer a causal relationship, but we need qual-itative research to develop and test out ourunderstandings of individual action andinteraction (cf. Goldthorpe, 2000). Anothercommon usage of mixed methods is triangu-lation, in which data from different sourcesare used to enhance understanding or to ex-plore validity by bringing different evidenceto bear on the same problem (Bryman,2001; Kelle, 2001). It may be, too, that differ-ent data sources allow for resolution of

Linking Qualitative and Quantitative Evidence 417









New York, NY 10001

212-431-9800

800-365-7006

www.guilford.com

some puzzle and in this way help advance ex-planation (Erzberger & Kelle, 2003). InBryman’s content analysis, this was shown tobe a relatively unusual rationale or outcomeof research (Bryman, 2005).

I want to say more about the logic of bring-ing together different methods and/or datasources and how they may help access socialprocess. Certainly it is widely recognized thatevidence is theory laden, not theory neutral.Although there has long been debate abouthow particular methods shape what we see(e.g., Becker & Geer, 1957; Trow, 1957), thisissue remains important. Linking differentmethods helps to crystallize some of the is-sues since we are more forced to confrontthe specificity of particular datasets.

One example of the specificity of data as alens on social process comes from Deaconand his colleagues’ (1998, as cited inBryman, 2000) study of researchers’ interac-tions with, and experience of, the mass me-dia in Britain. Here, diverse kinds of evi-dence reveal a seemingly single object ofanalysis to be multifaceted. In their re-search, qualitative and quantitative data gen-erated an apparent discrepancy, with theformer suggesting a relationship of conflictbetween researchers and journalists that wasabsent from the quantitative evidence. Whatshould we make of this discrepancy? It ishere that arguments typically start up abouthow effectively different methods tap intothe most important issues. But what is moreinteresting and productive is to considerhow the different kinds of evidence revealdifferent facets of social experience. Bothmay be valid so long as we understand thenature of the method, the context in whichthe data are created, and the precise waysuch data accesses the issues under investiga-tion. Deacon (1998, as cited in Bryman,2001) showed that in their survey responsesacademics gave an “average” rating of theirdealings with journalists, yet in semistruc-tured interviews they were oriented to mem-orable encounters. We might suggest thatatypical stories get played up in narrative ac-counts. They make a good story. They may

also have an impact on people’s lived experi-ence far greater than the “average” ratingreveals. Here we can see how differentmethods access (or “reconstruct”) differentfacets of the same experience.

Research into values provides a second ex-ample of how data from a particular microor macro perspective provide a specific andpotentially distorting lens on our researchquestions. Mason (2002) argues, with refer-ence to qualitative methods, that we shouldsee “asking, listening, and interpretation” astheoretical projects: “how we ask and listenare theoretical enactments of our assump-tions around where the phenomenon we areinterested in are located, and how the inter-viewee and interview can illuminate the is-sues” (Mason, 2002, pp. 233–234).

In her substantive interest in researchingvalues and morality in kinship relationships,Mason is concerned that people have soughtan understanding of values and moralsthrough abstract interview questions, argu-ing that these “direct attention to wrong or‘nonexistent’ locations” and, further, that“they miss the point about morality in thatthey assume it is a thing rather than a pro-cess or practice” (Mason, 2002, pp. 233–234). Following a more in-depth line of in-quiry to tackle this, we might invite people toprovide “real-life” stories, yet we need to beaware that people may describe their deci-sions and behaviors in moral terms. For ex-ample, in deciding on the best type of carefor their preschool child, or in making deci-sions about whether to work or care full timefor preschool-age children, intervieweesstress moral commitments and evaluativejudgments (Duncan & Edwards, 1999). How-ever, a wide picture reveals such moral ac-counts to be patterned in relation to socialand economic constraints and opportuni-ties, revealing “moral” judgements to be so-cially shaped (Duncan & Edwards, 1999;Duncan & Irwin, 2004). This is not to saythat people are making up a moral account.They may choose to emphasize certain fac-ets of their experience and choices, facetsthat may be important to them in their ac-










New York, NY 10001

212-431-9800

800-365-7006

www.guilford.com

counts to themselves, as well as to other peo-ple. Yet these may be partial accounts. Inconsequence, such accounts might lead us tooverestimate moral reflexivity as a driver ofbehavior. It is not that accounts of belief, val-ues, and choices are without value; far fromit. However, we do need to have a clear un-derstanding of how such data relate (or failto relate) to our research question.

In this example we may see that a breadthof evidence that looks at responses across di-verse contexts (whether through qualitativeor survey means) allows us to see moreclearly a patterning to the responses(Duncan & Edwards, 1999; Irwin, 2004).This patterning suggests that morality can-not be seen as an internal conversation inwhich individuals come to personal judg-ments about “doing the right thing,” nor canit be understood only as an outcome of prox-imate context and interaction. The struc-tured nature of such judgments revealsdiverse contexts that shape moral andevaluative judgments (e.g., which is the besttype of care for my preschool child?). “Upclose” (in-depth) evidence reveals the impor-tance to people of the moral content of theirchoices; “wide angle” (e.g., survey) evidencereveals the structured nature of their“choices.” Evidence from only one of thesesources provides only a part of a bigger pic-ture, but it may also lead us to misappre-hend the nature of “choice” as a singu-larly moral or social phenomenon. Evidencefrom both sources helps reveal the moraland social to be intertwined.

Methods, in part, create what we see. Thismust not lead us to relativism, in which weaccept the validity of multiple, and possiblyconflicting, accounts of the social world.Rather, it requires that we know more pre-cisely how we are tapping into the processesin which we are interested and how our dataoffer a particular construction of and lenson such processes. We need conceive of data(from different sources and different meth-ods) as offering specific kinds of evidence, asparticular rather than all-revealing slicesthrough our research problems.

The examples raise the question of howwe may best access social process and howdifferent data sources facilitate this. It is of-ten said that quantitative research allows usaccess to pattern and qualitative method al-lows us access to process. Bryman (2005)supplemented his content analysis withsemistructured interviews with 20 social sci-entists who have used mixed methods intheir research. We can note that his inter-viewees referred to qualitative evidence ac-cessing meaning and quantitative researchsupplying breadth (Bryman, 2005), and thisis a common enough observation. At onelevel this is clear-cut and not problematic.However, just as it is inappropriate to accepttoo clear-cut a distinction between quan-titative and qualitative strategies (e.g.,Hammersley, 1992), so, too, we need to becautious about dichotomizing pattern andprocess. They overlap and can usefully beseen as mutually made. Pattern and processare not distinct domains of social phenom-ena but, rather, different kinds of accountsof social phenomena.

Data on micro-level processes are oftendeemed to help illuminate pattern. Clearly,knowledge of micro-level beliefs, behaviors,interactions, and so on can help illuminateprocesses that may be hinted at by, butopaque to, quantitative research. However,social patterns are less often considered to il-luminate process at a micro level. Certainly,a core stock in trade of qualitative analysis,which often proceeds from patterns found“within the data,” is building understandingof patterns based on comparing cases. How-ever, “external” quantitative data are oftendeemed background or context for micro-level research and not connected in a directway. This is unfortunate. It may be essentialto adequate knowledge of the positioning ofindividuals being researched and of the con-tent of their beliefs. In the aforementionedexample of values in respect to child care,people’s values concerning “good mother-ing” (as a full-time care commitment or acombination of paid work and care) connectclosely with their circumstances in respect to










New York, NY 10001

212-431-9800

800-365-7006

www.guilford.com

employment-based opportunities and con-straints (Irwin, 2005). The constraints underwhich people act and perceive their experi-ence may not be articulated by them (or notalways reflected on). Nevertheless, we donot properly understand the nature of peo-ple’s values and subjective orientations if wedo not understand the contexts in whichthey hold meaning.

In short, it is important to acknowledgethat qualitative research does not have someprivileged direct line to process. As Kelleputs it, “structural nearsightedness clearlylimits the explanatory power of research re-sults derived exclusively from the qualitativeinvestigation of actors’ perspectives” (2001,p. 30). Often we cannot make proper ex-planatory sense of individual-level data ifwe do not have a handle on more aggre-gate structures. We may fail to understandcrucial meanings, motivations, and under-standings held by individuals, and accessiblethrough qualitative research, if we are un-able to locate them in the broader contextsand structures in which they are embeddedand take on shape. Furthermore, social di-versity means that we get not only a partialpicture from people but also a view from aspecific location within that diversity. Thestructured nature of social arrangementsnot only provides “context” for micro-levelbeliefs, behaviors, expectations, and so onbut also shapes their content in importantways.

Qualitative research is often charged witha need to better locate the specific as part ofthe general and to locate contexts of actionand belief as part of a wider social structure.Quantitative research is often charged with aneed to better access such contexts. Re-cently, there is a growing interest in moresufficiently connecting micro and macrolevels of evidence and analysis. However,clearly this is not just about supplying evi-dence pertaining to an “interconnecting”meso layer of context, although this wouldoften help. It is a conceptual issue. How webest bring together methods and data is notat heart a methodological question but one

that must be driven by tackling substantiveresearch questions and guided by criteria ofadequacy in how we connect theory anddata. In this section I have argued that par-ticular data sources offer a specific lens onmultifaceted problems, and by itself this maybe misleading. Additionally, I have arguedthat qualitative data may not necessarily ac-cess process, because we need an adequateunderstanding of structure and of diversityto adequately locate and interpret qualita-tive evidence. However, this is not simply acase of connecting qualitative to quantitativedata sources. In the next section, I exploresome issues in representing the “general pic-ture” through quantitative data analysis.

Issues in Researching Social Causality

In addressing issues in linking theory anddata, different writers have drawn on the fas-cinating historical example of the search forunderstanding the spread of cholera in the19th century (Freedman, 1991; Turner,1997). The example has been used to drawsome lessons in the use of different kinds ofevidence and modes of analysis for theoreti-cal development. I summarize it again in or-der to consider some of the lessons drawn byprevious writers and to add some observa-tions about the nature of evidence we canbring to our research questions.

Through his work in mid-19th centuryEngland, John Snow developed an under-standing of cholera as caused by a water-borne organism transmitted through hu-man waste. It was only in 1884 that thebacterium was isolated and observablethrough newly powerful microscopes. Be-fore that the nature of cholera and its inci-dence had to be deduced from an under-standing of extant patterns. Snow’sexplanation went against the grain of ac-cepted wisdom and understanding that thedisease was caused by miasma, or poisonousparticles carried in the air (Freedman, 1991;Turner, 1997). Scientific work that was inkeeping with the contemporary understand-










New York, NY 10001

212-431-9800

800-365-7006

www.guilford.com

ing was being developed by William Farr,the superintendent of the Statistical Depart-ment of the Registrar General’s Office.Turner describes the rival interpretations ofSnow and Farr and how their different meth-ods, assumptions, and questions shapedtheir very different understandings of thenature of cholera (Turner, 1997).

Farr, in line with miasma theory, had iden-tified a pattern, within a major epidemic, ofa strong inverse relation between the alti-tude of dwellings and the incidence of chol-era. He took this as strong evidence insupport of the prevailing theory. Turner em-phasizes the limits to causal reasoning basedon attempts to model cholera and its corre-lates without getting close enough to under-standing its patterning “on the ground.” Incontrast, Snow developed his radically newtheory through an approach that took himmuch closer to the transmission of choleraby an intensive, empirically based inquiry(hence the “shoe leather” in Freedman’s[1991] title). Snow sought a situated under-standing. He collected evidence surround-ing the incidence and outbreaks of cholera,exploring the details of people’s living ar-rangements and circumstances. He built evi-dence about the course of different out-breaks and found a strong clustering aroundwater sources that evidence showed to becontaminated (Freedman, 1991). Through aseries of naturally occurring experiments,Snow developed, elaborated, and tested outhis theory. Although Farr came to accept theplausibility of Snow’s conjectures, Turner ar-gues that he simply added these into his sta-tistical model and concluded that the keycausal mechanism of transmission (contami-nated water) simply held some additional ef-fect (Turner, 1997).

In particular, Turner argues that the as-sumptions embedded in Farr’s statisticalmodels and the nature of the process of sta-tistical modeling effectively blinded him tocountervailing evidence that should haveupended his theory. Additionally, Farr’smethod left him without effective means tofalsify the theory. In short, the efficacy of sta-

tistical models is bounded by the correctnessof assumptions that shape the model.Turner sees echoes in today’s modeling anda tendency still to wrongly equate correla-tion and causality. He is concerned that weare too quick to assume causality in the ab-sence of an understanding of underlyingmechanisms or processes that reveal the in-ternal workings of the causal process inwhich we are interested.

Turner argues that today, as in the 19thcentury, causal modelers risk being too dis-tant from their data and that, although wehave various tools for modeling associations,there remains the possibility that we are notcorrectly representing the mechanisms inwhich we are interested. Therefore “socialscientists are a bit like Farr before his com-plete conversion to Snow’s account of chol-era” (Turner, 1997, p. 43). Causal modelsare no better than the assumptions on whichthey are founded. Sound knowledge is built,rather, on intensive empirical work, whichholds qualitative insights and is available totesting and to falsification. Freedman, too,sees in Snow’s work a more scientific ap-proach to advancing explanation. He partic-ularly stresses the value of Snow’s develop-ment of questions and theory that could betested against the empirical evidence and ina wide variety of settings and sees this as themodel to emulate (Freedman, 1991). ForFreedman:

regression models are not a particularly goodway of doing empirical work in the social sci-ences today, because the techniques dependon knowledge that we do not have. Investiga-tors who use the technique are not paying ade-quate attention to the connection—if any—between the models and the phenomena theyare studying. (1991, p. 304)

Turner and Freedman both favor methodsand modes of data collection and analysisthat lie much closer to “internal” processesthan the relative abstraction of researchingaggregate patterns and associations andcausal modeling.










New York, NY 10001

212-431-9800

800-365-7006

www.guilford.com

Other writers have made similar argu-ments. Advocates of a realist program of re-search see some conventional approaches todata and explanation as entailing a “blackbox” approach to causal analysis. For exam-ple, experimental method follows an input–output model, measuring differences beforeand after the introduction of some man-ipulation but often failing to engage ade-quately with the actual processes engender-ing change (e.g., Pawson & Tilley, 1997). Inthe experimental method cause is seen as ex-ternal to that being measured, a force actingon an object (Goldthorpe, 2000; Pawson &Tilley, 1997). Similarly conventional ap-proaches to statistical modeling, andvariable-led analysis more widely, have beenchallenged for holding an inadequate repre-sentation of “internal mechanisms” and pro-cesses (Byrne, 2002; Pawson & Tilley, 1997).In their social realist perspectives, explana-tion needs to access internal processes, the“chemistry” of process, rather than simplydeducing it from “external” evidence. Cru-cially, “generative theory sees causationas acting internally as well as externally”(Pawson & Tilley, 1997, p. 34). Nevertheless,adequately representing processes remainsa challenge for social explanation.

For example, reflecting back on Turner’scritique of Farr, we need to recognize thatFarr believed that he had an appropriate andaccurate representation of causal process.He thought he was approximating the “inter-nal process.” He was working with the pre-vailing theoretical understanding, and theevidence available to him appeared to con-firm this theory (Turner, 1997). As Blalock(1991) says, it is only with the benefit of hind-sight that we can distinguish so straightfor-wardly “between the tactics of the very fewsuccessful detectives and those of the pre-sumably much greater number of failing de-tectives” (Blalock, 1991, p. 329). How can weknow when a particular understanding is thebest bet? Snow benefited from “natural ex-periments” in which he could develop andtest out his developing theoretical proposi-tions. An important principle, when we do

not benefit from “natural experiments,” is tocreate our own and make our assumptionsavailable to testing (e.g., Blalock, 1991;Lieberson, 1992; Pawson & Tilley, 1997).

Turner (1997) and Freedman (1991) advo-cate strategies that seek to “get close” to “in-ternal processes,” with movement betweenproximate circumstances and broader pat-terns, and see this as important to develop-ing theoretical inferences about causality.This movement across “levels” of evidence,between close up and wide angle, improvesour understanding of process through deriv-ing detailed empirical evidence, exploringgeneral patterns, generating propositions,and testing them out across different con-texts. It calls for working with all the avail-able data to develop a more adequate the-ory. Blalock points out that the choleraexample may be misleading as a metaphorfor social science given that in the latter weare usually dealing with multiple causalityand forms of contingency not evident in thecholera example (Blalock, 1991). Neverthe-less, the case for moving between levels andsubjecting theoretical propositions to test isevery bit as key to enhancing social scienceunderstanding.

The cholera example, as others, showshow prior assumptions and expectationsgovern the ways in which we approach andanalyze data. However, this is not to say thatwe are doomed to reproduce our prior con-ceptual frameworks. Empirical data can cer-tainly challenge and even lead to a renewalof theory. Before exploring this in relationto some concrete examples of research, laterin the chapter, I consider some recent criti-cisms of quantitative modeling. Again, wesee concerns with the level of abstractionfrom the source data entailed in causal mod-eling, and an argument has been made thatwe need to build macro-level datasets, evi-dence, and analyses that can more closelyrepresent social experiences, interactions,and patterns. Again, this is an argumentabout more adequately theorizing socialprocess through interrogating the links be-tween micro- and macro-level evidence.










New York, NY 10001

212-431-9800

800-365-7006

www.guilford.com

The critics of variable-centered, orvariable-led, quantitative modeling do not(generally) deny that the approach is neces-sary and valuable to representing patternsand regularities in social life (see Kemp &Holmwood, 2003, for a critique of thosewho do). After all, the knowledge of extantregularities is fundamental to exploring so-cial life and its stable reproduction overtime. However, in advocating a realist ap-proach, various writers have been critical ofvariable-led analysis in which variables, atleast at times, are seen to represent externalforces acting on people or on social systems.Variable analysis in causal modeling neces-sarily looks at average effects (Byrne, 2005)and risks failing to get at where the social ac-tion is. For Lieberson and Lynn (2002),causal modeling is part of an inappropriate(classical-physics-derived) model of good sci-ence in the linking of evidence and theory.They argue that this model needs a radicalreworking and that other models of scien-tific endeavor (such as the development ofknowledge in evolutionary biology) providebetter metaphors and guides for social sci-ence research (Lieberson & Lynn, 2002).Thus critics say that there is a tendency forvariable-led analysis to insufficiently accesscontext and to risk reifying variables as realforces. Byrne calls for death to the variable, ahumorous yet serious challenge to thosewho place too much store by variable-ledanalysis in resolving conceptual problems.He insists that there is a risk that such analy-ses mislead as to generative mechanisms, inpart through failing to engage with contextand contingency. Byrne (2002) argues that:

Variables describing complex systems are de-scriptions of properties of the system as awhole. We can consider them as the dimen-sions of a multi-dimensional state space withthe actual character of the system at any pointin time being represented by the set of val-ues on measured variables considered as co-ordinates in that state space. However, the co-ordinates are more of an address than adescription of causes: they tell us where—notwhy. (p. 7)

He desires an approach that is case basedrather than variable based, in which it is pos-sible to aggregate up from knowledge of in-dividuals in contexts and that additionallyrecognizes that systems are more than thesum of their parts and have emergent prop-erties; that is, they produce outcomes thatcould not be predicted on the basis of knowl-edge of the parts. He argues that there is animportant potential site for convergence be-tween qualitative research and method andquantitative method, should it take theroute he advocates, seeing “[the] key link be-tween the two [as] the focus on the caserather than some abstraction from the casereified and regarded as a variable” (Byrne,2002, p. 160).

Byrne may be overstating the extent towhich social scientists suppose that variableanalysis somehow does the theoretical workfor them. Many would see variables as pro-viding us with probabilistic descriptions ofsocial diversity. Social science researchersrarely insist that some cause determines someoutcome. One can think of many examplesin which researchers treat variable evidencein terms of “variate traces,” which Byrne rec-ommends: Variables provide a form of evi-dence, not a definitive account. Yet in prac-tice reliance on variables does push toward aparticular definition of the problem at hand.A variable-centered analysis offers a particu-lar reconstruction and representation of theprocesses in which we are interested.

It is useful to reflect further on Byrne’sconcern that we reify variables. The method-ological concern articulated by Byrne holdsechoes within recent subject-based debatesin sociologies of difference. Here some writ-ers have argued that there is a risk that socialdifferences are reified in research, as theirsalience is often assumed, where it needsto be explained or contextualized (e.g.,Anthias, 1998, 2001; Epstein, 1988; Young,1990, 1997). Some writers, including myself,have argued that categories of difference,such as ethnicity, gender, and class, aresometimes inappropriately treated ascauses, rather than outcomes, of wider social










New York, NY 10001

212-431-9800

800-365-7006

www.guilford.com

processes (Anthias, 1998, 2001; Bottero &Irwin, 2003; Irwin, 2005). However, it maybe crucial to research such divisions and nottreat them as starting points for analysis. Ata minimum we need to know when theycount and why. What processes shape differ-ences and give them salience (e.g., Bottero,2004; Brubaker, 2002; Irwin, 2005; Siltanen,1994)? In short, this requires an understand-ing of context and contingency and theshaping of diverse relevancies. There areparallels here with Byrne’s advocacy of“mapping coordinates” of diversity and oftaking a more taxonomic approach to order-ing quantitative data through which we canremain more true to social context. For ex-ample, later in the chapter I show howgendered differences are being reshapedand how it is more useful to locate gender asan outcome of social relations (which are un-dergoing change) than to treat it as a staticcategory or given social division.

In this section I have argued that we needto access underlying processes that shapethe phenomena in which we are interested, atask that requires knowledge of contexts.Furthermore, we need to acknowledge thatthe way we categorize evidence entails theo-retical assumptions. Neither qualitative norquantitative research provides a privileged“direct line” to underlying processes. Quali-tative evidence may speak more directly toprocess but will only do so where we can lo-cate its specificity. Standard forms of quanti-tative data modeling and analysis have beenchallenged for being at too far a removefrom specificity, for example, removed fromperceptions, expectations, beliefs, behav-iors, and modes of interaction as these relateto diverse contexts. Forms of evidence allcarry theoretical assumptions and provide aparticular, and theory-imbued, lens on ourresearch questions. To improve our bear-ings on the processes in which we are inter-ested, then, the use of different sources ofevidence can help us. How we connect thisevidence is a theoretical issue, and we needto reflect on how evidence relates to the so-cial processes and phenomena in which we

are interested. In the next section I focus onsome empirical examples, drawing on evi-dence from different “levels” of the social inseeking to tackle problems of explanationand to enhance our understanding of socialprocesses.

Subjectivity and Social Structure:Linking Data in Researching SocialDiversity and Social Change

In all the examples in this section, I exploreresearch areas in which some writers haveidentified what they see as a discrepancy ormisalignment between people’s social posi-tion on the one hand and their perceptionsand attitudes on the other. Some writers ar-gue that in the current era we have seen aloosening of the relationship between sub-jective orientations and social structural pro-cesses. I argue that such conclusions aremisplaced. Rather than accounting for dis-crepancies by reference to categories exter-nal to the empirical data (such as ideolo-gies), reinterrogating the data and exploringlinks between macro and micro evidencecontributes to a renewed understanding ofthe mutuality of subjective orientations andsocial structural processes.

Youth and the Life Course: Exploring Attitudesand Social Diversity

First I take two examples from studies ofyouth and early adulthood. One is fromqualitative and the other from quantitativeresearch. Both show interesting insights intothe link between position and disposition.We can draw out some general themes.

Various youth researchers have engagedin depth with the question of how values andchoices on the one hand relate to structuralprocesses on the other. One of the issueshere has been addressing the gap betweenmacro-level evidence that reveals clearlystructured patterns of inequality and its re-production and micro-level evidence of peo-










New York, NY 10001

212-431-9800

800-365-7006

www.guilford.com

ple’s perceptions of choice and ownership oftheir destinies. In qualitatively based inter-views, people will be likely to stress choiceand agency in the stories they tell aboutthemselves, in contrast to the quantitativeevidence that reveals significant class-relatedinequalities in opportunity and constraint(e.g., Furlong & Cartmel, 1997). How do weunderstand the seeming discrepancy be-tween the types of evidence? Some writershave posited ideological forces that obfus-cate reality and encourage compliance withunequal and unjust social arrangements (foran extended discussion, see Irwin, 2005).However, it is more productive to considerhow people’s positions within the socialstructure will tend to engender and normal-ize particular views. Furthermore, as Nilsenand Brannen (2002) say, people are not rou-tinely oriented to, nor typically particularlyaware of, the external and structured forcesthat shape their lives so, “When structuralforces and personal resources . . . supportone another there is a tendency for thestructural resources to take on an ‘invisible’quality” (Nilsen & Brannen, 2002, p. 42.)

An example of the link between positionand perception is revealed in a recent quali-tative research project by Gillies and her col-leagues (Gillies, Holland, & Ribbens McCar-thy, 2003). Here a generational dimensionis in evidence as young adults and theirparents describe their perceptions of theformers’ transitions to independence.Gillies and her colleagues stress the “embed-ded” nature of young adults’ accounts, par-ticularly the relational and interconnectednature of young people’s understandings.The researchers argue that for young peopledescribing their experiences, growing upwas a process of taking control of theirbehavior and accepting responsibility fortheir decisions. Young people saw them-selves as being at the center of their tran-sition, as agents or authors of their pro-gression to adult status. In contrast,interestingly, their parents emphasized theirchildren’s physical changes and the continu-ities they saw in their children’s personalities

as they progressed from childhood to adult-hood. Young adults highlighted the ways inwhich they had changed since their child-hoods, whereas parents reflected on consis-tencies.

We can see these differences as unsurpris-ing outcomes of the interviews, but it is per-tinent to remind ourselves that young peo-ple may emphasize agency and the “cult ofthe self” more than any other life-coursegroup. Gillies and her colleagues stress thatthe individualism expressed by the young-sters “was clearly contained within a widersocial context, characterised by interdepen-dent family relationships” (Gillies et al.,2003, p. 47).

I would suggest that we can also usefullydraw out something that remains implicitwithin their account—young adults and theirparents are positioned differently and mightbe seen as offering different “vantagepoints” on the question of transition to in-dependence. The young adults naturallyenough experience themselves as beingagents in a context in which boundaries arewidening and the scope for their action ex-pands as they seize greater autonomy and re-sponsibilities. Parents may have a more “so-ciological” understanding of this transition,having some social distance from it (and pos-sibly engaging in a fair degree of reflexiveanalysis about their children’s position andhow, as parents, to best relate to it). The van-tage points of youth and parent are verydifferent. Superficially they appear contra-dictory, but we can better see them asconsistent—an example of how diverse val-ues and perceptions are closely aligned withpeople’s diverse social locations. We seemore clearly the links between subjectiveorientations and objective structures if wedelineate the diverse contexts that shapepeople’s experiences and perceptions.

My second example draws on small-scalesurvey research that also points to a connect-edness of social position and subjective dis-positions. Within a survey exploring variousaspects of work and family life of 92 youngpeople, ages 16–34, all respondents were in-










New York, NY 10001

212-431-9800

800-365-7006

www.guilford.com

vited to rank claims to employment amongpeople in different household and life-course circumstances (see Irwin, 1995, fordetails). They were asked to imagine that sixpeople apply for a job and to assume thatthey are all equally qualified for the job.They were then asked “Who would you mostlike to see get the job? Who would you nextmost like to see get the job?” and so on. Thesix people in the vignette were described as:a young woman living at home; a young manliving at home; a young woman living awayfrom home; a married man with children,wife not working; a married woman, nochild at home, husband not working; and asingle mother with young children.

Of course the assumption about the per-ceived salience of household need is notburied very deeply, and one might see it as aself-corroborating exercise in which respon-dents rank in “need” order, merely repro-ducing the researchers’ assumptions aboutthe salience of need and obligations. In thiswe could see a clear example of the imposi-tion of meaning. The researcher establishes,through structured questions, a conceptualframework to which respondents obliginglyorient (regardless of its relevance to them).The researcher then mistakenly remainsconvinced of the value of the conceptualframework. We know from survey researchhow readily respondents engage in the taskwith which they are presented and rarelychallenge the framework in which questionsare asked, regardless of their perceived sa-lience in the eyes of the respondent (e.g.,Pawson, 1989). Yet in the responses to thequestion described previously is a pattern-ing that suggests that something rathermore interesting than “theory in, theoryout” is going on.

It should be noted that the question waspart of a small survey conducted in the con-text of an undermining of young adults’ sta-tus in the United Kingdom in the late 1980s.A reading of the contemporary literaturewould suggest that ideological, individualiz-ing processes prevented youths from seeingthe extent of their exploitation and deterio-

rating relative position. Additionally, a read-ing of the literature would suggest that, at aminimum, a self-interested age preferencewould prevail. This implied that respon-dents would commend youth first. Few did.The overwhelming majority favored theclaims of the single mother or the marriedman. This might at first seem to be a classicexample of a self-corroborating exercise, inwhich respondents reflect back the assump-tions embedded in the response categories.However, the patterning of responses sug-gests that we are accessing reflections on thestructure of resource distribution from dif-ferent vantage points within it. What is espe-cially notable is that the young adults, whowere themselves still dependent on parentsand/or without dependents of their own, fa-vored the claim of the single mother. A lackof financial obligations tallied with a likeli-hood of positive discrimination in favor ofthe single mother. Those who were them-selves married or cohabiting or had depen-dent children of their own were far morelikely to favor the claim of the married manwhose wife was not working. This held forwomen as well as for men.

The example shows that we can usefullymove away from age as the key variable in ex-ploring age-related patterns and explore po-sitions and attitudes as they relate to house-hold/family need and commitments. To doso reveals the connectedness of micro-levelperceptions and broader macro-level struc-tures. The respondents’ attitudes show aprioritizing of the claims of those with de-pendents, but within this structure their atti-tudes are patterned in relation to their ownlife-course position and circumstances inrespect to household resourcing commit-ments. We can see a connection between in-dividual orientations and the structures ofdistribution in which people are embedded,in which those with dependents are morelikely to favor the male breadwinner’s claim.Through considering diverse vantage pointsfrom within a variegated structure we can seemore clearly the links between micro-levelorientations and macro-level structures.










New York, NY 10001

212-431-9800

800-365-7006

www.guilford.com

Clearly, then, this was not a self-corroborating exercise. Rather, it revealed adiverse pattern of attitudes shaped in rela-tion to people’s own household circum-stances and a linked prioritizing of claims onwork in relation to household needs. Re-spondents’ attitudes reflected the structureof distribution and their position within it.The patterning of responses can be seen asan outcome of practical attitudes to distribu-tional exigencies. In this sense evaluativejudgments are shaped in relation to “whatis” and reflect people’s location within anasymmetrical pattern of distribution.

I have used both examples in the area ofyouth to argue that there is connectednessbetween people’s outlooks, attitudes, andtheir social positioning. This theme of co-herence between orientations and positionwithin the social structure runs through thenext two examples, which both relate to is-sues of gender and employment.

Work-Rich and Work-Poor Households:Using Data to Address Puzzles of Explanation

The details of my next example come fromresearch conducted in the early 1990s, butthe focus on puzzle solving and its value forsocial explanation retains its currency today.There is a parallel here with Erzberger andKelle’s (2003) advocacy of theoretical re-newal as a response to divergent conclusionsdrawn from qualitative and quantitative evi-dence. The example is based on researchinto social relationships and economicchange in the northeast of England, in thecontext of industrial restructuring throughthe 1980s (Morris, 1995). A concern waswith the concentration of employment andunemployment at the level of the house-hold. This showed a distinct patterning at anational level, a patterning paralleled in asurvey of 790 couples in Hartlepool under-taken in 1989 (Irwin & Morris, 1993). Thegeneral patterning, which is well known, canbe described as a division between work-richand work-poor households. The periodfrom the 1970s through the 1980s had mani-

fested a growing concentration. In 1986, at anational level, 67% of men in paid employ-ment had spouses also in paid work com-pared with 24% of unemployed men. Vari-ous studies revealed similar patterns, andresearchers sought to understand the pro-cesses shaping this concentration of employ-ment and unemployment at the householdlevel among married couples. Much of theresearch was framed by the question: Whatdo the wives of unemployed men do? A prin-cipal hypothesis of social policy researcherswas that the social security structure pro-vided a significant disincentive to workamong women with unemployed husbands.At the time, social security and benefits forthe unemployed carried very low earningsentitlements for dependents. Above a mini-mal earning allowance for the spouse, bene-fits were withdrawn, pound for pound.There was, therefore, a clear economic logicfor a married woman not to work if her hus-band was drawing unemployment benefit orincome support. This “social security” expla-nation remained a dominant understandingof causality in the patterning of work-richand work-poor households.

Interestingly, some alternative, althoughcomplementary, explanations followed asimilar theory of causality. For example,some argued the importance of a “bruisedmachismo” effect: that cultures of and be-liefs about masculinity and breadwinningworked against a wife being employed if herhusband was unemployed. However, in bothexplanations, there was a focus on the levelof the household, with women’s labor-forceparticipation understood in terms of their hus-bands’ labor-force status. We can note agendered assumption here about his inde-pendence and her dependence and the as-sumption of some external causal process(social security disincentive structure, cul-tural mores) having an impact on couplesand shaping unemployment outcomes.

A problem here lay with the failure to testout the assumptions embedded in the the-ory. Interestingly, though, this could bedone. In the 1980 Women and Employment










New York, NY 10001

212-431-9800

800-365-7006

www.guilford.com

Survey (WES), nonworking women marriedto unemployed men were asked to statetheir reasons for not themselves working.Eighty percent did not cite their husbands’employment status as a cause (Joshi, 1984).Of the 17% who said it was important, only14 out of 58, or 4% of the total, said they didnot work because their household benefitswould be cut. This evidence was echoed inthe Hartlepool survey. There was a concen-tration of employment and unemploymentin households, with 67% of households hav-ing an employed male with an employed fe-male partner, in contrast to 21% in whichthe male partner was unemployed. In open-ended questions within the survey, only 12%of nonworking women said they were un-available to work because their spouses’ ben-efits would be cut. This verbatim evidence,with its limited mention of benefits, appearsto directly contradict the assumptions of thesocial security model of causality. It is a niceexample of a research puzzle. To explain itsresolution, we can, here, consider one par-ticular finding (for more detail and discus-sion of further evidence, see Irwin & Morris,1993).

Other studies sought to control for possi-ble intervening factors to ensure that theyhad identified a “pure” causal effect. For ex-ample, they controlled by class, assumingthat to do so was to control socioeconomicstatus. Thus if, for each class, employed hus-bands have employed wives and unem-ployed husbands have nonworking wives, wecan be more confident that we have a purecausal effect: It is his labor-force status thataffects her labor-force status. However, if westop looking at women’s status under the as-sumption that it is caused by their husbands’status, a different picture emerges. A rangeof indicators revealed a direct associationbetween women’s own positions and theirhusbands’ labor-force status. For example,there was a strong association betweenwomen’s own occupational standing (fromtheir current or most recent jobs) andwhether or not their husbands were unem-ployed, even controlling for the husband’s

social class. To illustrate, among unskilledhusbands, 53% of those who were employedwere married to women whose most recentjobs were in low-status occupations, in con-trast to 90% of unemployed unskilled hus-bands. Broad class groupings are clearly veryinadequate as a measure of social disadvan-tage and advantage. The example is drawnfrom wider evidence in the dataset that theconcentration of employment and unem-ployment is more effectively explained by asimilarity within couples of employmentchances; the coincidence of spouses’ unem-ployment is closely linked to their similar, in-dependently held disadvantages in relationto employment opportunities. The new in-terpretation presents a direct challenge tothe social security explanation. Importantly,the new analysis of the aggregate dataprovides an explanation that is in linewith open-ended question data on women’sself-reported experience. In consequence,it enables an improved understanding ofthe social structuring of advantage anddisadvantage across households. Tacklingcontradictions arising from interpretationsof available data and their reinterpretationcan allow us to transform our understandingand develop a more inclusive explanatoryframework.

The next example maintains the theme ofgender relations and explores quantitativeand qualitative data to reveal the close linksbetween people’s social position and theirevaluations of the right thing to do in re-spect to work and care.

Reshaping Gender, Work, and Caregiving:Exploring the Connectedness of Attitudesand Social Position in a Context of Change

The example here is drawn from work in thearea of gender, work, and caregiving. In it Idraw together evidence in building a pictureof diversity and change in women’s commit-ments to child care and employment. Theanalysis offers an alternative account to theinfluential view that values are more autono-mous of social circumstances than they were










New York, NY 10001

212-431-9800

800-365-7006

www.guilford.com

in the past (e.g., Beck, 1992; Beck & Beck-Gernsheim, 2002; and, from a different per-spective, Hakim, 2000; see also Irwin, 2005,for the detailed account).

In the United Kingdom, full-time childcare among women with preschool childrenis quite common. However, the incidence offull-time care has fallen dramatically over re-cent decades. The increase in labor-forceparticipation rates has been most markedsince the 1980s. Among women with chil-dren ages 0–4, in the years 1949, 1959, and1969, overall employment participationrates stood at 14%, 15%, and 22%, respec-tively. The full- time employment participa-tion rate across these years was constant ataround 8%. In the years 1981, 1991, and2001, the overall employment participationrate of women with children ages 0–4 rosefrom 24 to 42 to 54%, respectively. The full-time employment rates across these yearsrose from 6 to 13 to 18%, respectively.

There is a wealth of research in the area.Through the 1980s and 1990s, many writersemphasized continuity in women’s positionof relative disadvantage, given the extent towhich the increase in participation was inpart-time, flexible, and often low-payingwork (e.g., Arber & Ginn, 1995; Hakim,1996). More recently, there has been recog-nition that the growth of women’s employ-ment participation in late-20th-century Brit-ain is bound up with important changes inthe economic and social positioning ofwomen and men (Bruegel & Perrons, 1998;Irwin, 1999, 2005; Walby, 1997). Severalwriters have presented evidence of an ero-sion of breadwinner patterns of householdresourcing over the past three decades, witha rise in the incidence of dual-earner house-holds and a growing importance of femaleearnings for household support. This doesnot betoken simply an improvement in theearnings of women, as it is also bound upwith a decline in the relative adequacy ofmale earnings among some men, particu-larly those in manual-labor jobs (Bruegel &Perrons, 1998; Egerton & Savage, 2000;Irwin, 1995, 1999). I have argued elsewhere

that these changes in women’s and men’s re-lations to employment, earning, and eachother are linked to changes in occupationalstructures but are also not separable fromchanging norms about women’s paid em-ployment through family building, fromwomen’s claims for independence, nor fromchanging perceptions of adequate standardsof living (Irwin, 2005).

A significant strand of recent research, inseeking to locate change, maintains that val-ues are more important than they were inthe past in shaping decisions about workand care. Some argue that attitudes andpreferences play a significant role in shapingbehaviors (e.g., Hakim, 2000; Hattery, 2001;Marks & Houston, 2002). However, al-though attitudes and preferences are clearlyimportant motivators, we need to be cau-tious about seeing them as newly “loosened”from social structural arrangements. Thedata explored here are part of a wider argu-ment that there is still a close alignment be-tween subjective orientations and social andeconomic circumstances. There is evidencefor this at different levels. We can see it inboth general and more targeted social attitu-dinal data and their association with circum-stance. We can see it in qualitative data thatexplore women’s circumstances and theirperceptions of the “right thing to do.” Whatis notable about the latter is the link betweenmany women’s values and their current posi-tions in a context of significant changes inwomen’s relations to paid work and childcare.

Attitudinal data provides a very particularlens, as do other kinds of data, on the pro-cesses in which we are interested (cf. Mason,2002). Attitudinal data are sometimestreated like a thermometer, an instrumentto measure the collective temperature, akind of average of the national outlook oncrucial issues. There are plenty of critics ofattitudinal surveys; in particular, many findfault with the superficial nature of attitudi-nal statements. As discussed briefly earlier,responses are not mere artifacts of imposedmeaning. In the examples here, they reveal a










New York, NY 10001

212-431-9800

800-365-7006

www.guilford.com

clear pattern of covariation with materialand situational factors.

General attitudes toward women’s rolesand appropriate patterns of behavior amongparents run broadly in parallel with actualchanges in women’s employment participa-tion rates (e.g., Crompton, Brockmann, &Wiggins, 2003; Dex, 1988). So, for example,when asked whether a married woman withchildren under school age ought to work orstay at home, in 1965 78% of female surveyrespondents thought she should stay athome. In 1980, 60% of respondents thoughtshe should do so (Hunt, 1968, and Martin &Roberts, 1984, as cited in Dex, 1988). In the2002 British Social Attitudes Survey (BSAS),46% of female respondents thought thatwomen with a preschool-age child shouldstay at home. (In the 2002 BSAS womenwere asked: “Do you think that womenshould work outside the home full time, parttime, or not at all under these circum-stances?” for different categories, includingwhen there is a child under school age.) No-tably, the 1965 and 2002 figures compareparticularly closely with actual participationrates at the time.

The preceding discussion of change is in-dicative only, but it gives a sense of a shiftacross the British population toward morepro-work attitudes on behalf of mothers ofyoung children. Generalized attitudes maybe of some interest in taking the nationalpulse, but they hold more limited use in un-derstanding the actions of different parts ofthe social body. We can usefully consider inmore depth the variable patterning of atti-tudes among those for whom the issues havea more direct salience. It is possible to seethe close links between people’s own cir-cumstances and their attitudes when we doso. For example, working mothers are in fa-vor of mothers working, and homemakermothers are not. In the 2002 BSAS, 16% ofhomemaker mothers of preschool childrenfelt that a woman in the same situationshould work, and 64% felt that she shouldstay at home. In contrast, 66% of workingmothers felt that such a woman should

work, whereas 16% felt that she should stayat home (a ratio of 4:1 homemaker womenfavor staying at home; a ratio of more than4:1 working women favor working). Clearlythere is little evidence of a dissonance be-tween experience and attitudes; rather, wesee a noteworthy consistency.

We need to consider a risk that the link be-tween attitudes and experience/situation istautological. It is possible that, if people donot feel particularly strongly about some-thing or have not much reflected on it, theirresponse to an attitude statement may sim-ply be based on their practical experience.What they are familiar with may simplytranslate for them, in giving a survey re-sponse, into the “right thing to do,” even,perhaps especially, if they have not given itmuch thought. This is a risk and a potentialproblem for attitudinal survey research.However, we can plausibly expect that thosefor whom the question has most direct sa-lience will have given it some thought—thatis, they will see the statement as tapping intosomething relevant to them. Additionally,we can note that the patterning of attitudesis closely aligned to social circumstancesmore widely and not just to their behavior.At this point we can be more confident thatthe patterning of responses is not simply anartifact of the mode of asking questions.

Different datasets yield evidence of an as-sociation between people’s circumstancesand their attitudes. I consider an example ofdata gathered within a small survey, con-ducted as part of the research by the Eco-nomic and Social Research Council Re-search Group for the Study of Care, Valuesand the Future of Welfare (CAVA). In this“Life as a Parent” survey, the distributions ofattitudes were explored. Women who weremost in favor of full-time maternal child care(based on three attitudinal variables and de-scribed as “pro-maternal care”) were identi-fied, with a view to comparison with Hakim’s(2000) argument that there is class-randomdiversity in their outlooks. It is notable thatin this dataset the women were all in similarsocioeconomic circumstances. The 14 out of










New York, NY 10001

212-431-9800

800-365-7006

www.guilford.com

96 women interviewed who were “pro-care”were all relatively constrained in their em-ployment options but were not among themost disadvantaged, and most of them wereliving with an employed partner. If thesewomen worked, it was part time, and in onlyone case did the woman describe herself asworking for “essentials,” whereas half of thewomen who were not in the “pro-maternalcare” group described themselves as work-ing for this reason. A similar analysis ofnational-level BSAS data, using the two par-allel attitudinal questions, also shows thatthose who are “pro-maternal care” fall with-in the lower half of household incomegroups but not among the lowest. This evi-dence runs counter to arguments that “val-ues are becoming more important determi-nants of behaviour, relative to . . . socialstructural factors” (Hakim, 2000, pp. 80–81). It suggests ongoing links between valuesand circumstances.

We can also explore perceptions of the ap-propriate commitments of mothers throughqualitative data. The Mothers, Care, and Em-ployment (MCE) project was a qualitativestudy also conducted as part of the CAVA re-search project. It was conducted across dif-ferent locales in Yorkshire and Lancashire,England. Parents (mostly mothers) of chil-dren age 14 and under were interviewed,with a particular focus on issues of value andpeople’s sense of “doing the right thing” inrespect to caring for their children (Duncan,Edwards, Reynolds, & Alldred, 2003;Duncan & Irwin, 2004; see also Duncan &Edwards, 1999). The data allow us to furtherreflect on general developments in the rela-tive position of women in particular. It seemslikely that the salience of work as a crucialcomponent of women’s identities has agreater spread across the population andthat it is growing among groups for whom ithas traditionally been a less definitive experi-ence or expectation. The data show an align-ment between circumstance and values, yetthese values reveal a significant work ethicamong women who are mothers of youngchildren. This is a feature of middle-class re-

spondents, many of whom see work as a corepart of their identities. But it is also a themefor many white working-class respondentswith more circumscribed opportunities andperhaps more circumscribed motivations forwork. Women hold work as more central totheir identities, and more white mothers, in-cluding working-class mothers, have a work-related identity, as well as a mother identity.This is consistent with the trends toward in-creased employment rates among womenover the past quarter century. Althoughsome groups of working-class mothers havealways worked, employment participationamong mothers of preschool children (0–4)is becoming more extensive. The MCE evi-dence reveals the very routine nature ofwork among women and suggests that itwould disrupt their sense of themselves ifthey were to stop work fully through thefamily-building period.

Even among the relatively few, typicallyworking-class, women defined by Duncanand colleagues (2003) as “primarilymother,” who express clearly their high levelof commitment to full-time parental care fortheir children, there is a clear sense of paidwork as a core part of their identities. For ex-ample, Theresa encapsulates what Duncanterms a “primarily mother” orientation:

“I believe if you have children you shouldfetch ’em up yourself rather than like youget your career mums who can go out towork and somebody else has fetched yourchild up and I don’t believe in that really.”

Nevertheless, this woman returned to workas a health care assistant when her child was10 months old. She has a job-share arrange-ment with her husband, and both work 25hours a week as care assistants. When asked“And you say that that is because you foundit difficult to be just at home?”, she replied:

“Yeah. Yeah I found it hard work, I neededto see other people and do other things aswell as be at home. I needed to be myselfas well as being a mum.”










New York, NY 10001

212-431-9800

800-365-7006

www.guilford.com

That is, although her commitment to care-giving may be paramount, she still sees workand its sociability as core aspects of her iden-tity.

Other interviewees expressed further di-mensions of the importance of work tothem. For example, Jessica said:

“I work so that I can give my son everythingthat I’ve never had and so that I can pro-vide for him and if he wants anything hecan have it, not to spoil him but to makesure that . . . we can provide a decent stan-dard of living.”

In discussing her return to work when herson was young, she said:

“I wanted to go back to work. I don’t knowwhy, but I did. I think it were, it were im-portant for me to get back to being thatperson, not just being me little boy’smum.”

Another respondent who encapsulatesthe “primarily mother” orientation wasChristine, who said:

“I couldn’t see t’point of having a child andleaving him with somebody else.”

Christine was from Barnsley, a traditionalcoal mining town and therefore a culturalcontext in which we might expect Duncan’s“primarily mother” orientation (Duncan etal., 2003) to be common. Christine has fivechildren and, despite her orientation, shehas worked fairly extensively in unskilled(factory and cleaning) jobs while buildingher family. Her desire for work is financial,although such a motivation needs to be un-derstood in the context of cultural expecta-tions about adequacy. It is also linked toother aspects of her identity, especially theexpectation of independence:

“My husband always, always wanted me tostop working yeah. Ye know, this were al-ways a bit of friction between me and[him] . . . ’cos he’d always say we’ll cope

and we’ll manage ye know but I were al-ways, I’ve always had money so I werealways scared of just relying on his wageand then I’d say yeah, but what happenswhen I want summat and what happens ifI want to do summat or I want to buy anew coat . . . do I ask you for money, Isays: ‘I don’t think it’ll work out like that’and he says ‘yeah yeah of course you askme’ but ye know, its not, I can’t. I’ve al-ways had a job, from 19 I’ve alwaysworked and I’ve always had me ownmoney.”

To work seems an important part of heridentity and the kind of role model shewants to be for her children (traditionally as-sociated with black women’s orientations towork and care) (Duncan & Edwards, 1999):

“I want my children to work, I want ‘em towork, I want em to do good at school, asgood as they can ye know, and try and tryand get on.”

So even among those who have few quali-fications and who express Duncan’s “primar-ily mother” orientation, it is notable thatstrongly expressed caregiving commitmentsare consistent with holding a significantwork ethic. A sense of paid work appears tobe a core component of the identity of awide spectrum of women who have youngchildren. It is common for this ethic to bebound up with women’s desire for indepen-dence and autonomy.

As well as the importance of work to thesewomen’s sense of themselves and their self-esteem, it is notable that their views werenot necessarily mirrored by their husbands,who, like Christine’s, tended to “fall in” withtheir wives’ plans following a position ofdoubt. The expressions here seem illustra-tive of the differential rate of change inwomen’s and men’s social positions. An-other example of a husband falling in withhis wife’s desires is evident in the responsesof Lisa, mother of five children. When askedif her husband was supportive when she re-turned to work, she said:










New York, NY 10001

212-431-9800

800-365-7006

www.guilford.com

“When I first started for t’first few weeks hedidn’t like it—and we did have a few argu-ments and I says ‘look, we either argueover t’fact that we don’t see each otherand you’re tired and you’re coming homeand seeing to t’kids or we argue overmoney’. I says ‘it’s like Hobson’s choice,which would you prefer? And he says ‘Iknow you’re right,’ he says, ‘carry on, we’llgive it a bit longer,’ and he’s fine now, gotused to it, the routine and there’s no prob-lem at all, he’s quite all right with it.”

The interview data is illustrative ofindividual-level experiences in a way thatquantitative data cannot be, but it is con-sistent with the themes that are revealedthrough the general-level numerical data.From the 1970s there has been a marked risein the employment rates of partnered moth-ers of young children. Work has become amore routinized experience within thefamily-building period. The quantitative evi-dence indicates that this is so for a larger sec-tion of the population. Qualitative data re-veal the importance of work as a more corecomponent of women’s identities throughthe family-building period, and this includesworking-class women who are relatively dis-advantaged in their employment prospects.A pattern of mutuality between norms andwomen’s (and men’s) social positioning is ev-ident in a period of significant changes inwomen’s employment patterns. This is im-portant. The different sources of evidenceallow us to build a picture of diversity andchange, and it is one that invites us to chal-lenge and reinterrogate arguments that weare witnessing a historically new kind of divi-sion between norms and social structuralprocesses. Different data sources give in-sights into social change, here notablychange in the circumstances and social iden-tities of women, and men, relating to em-ployment and child care. Additionally, theevidence reminds us that gender, taken as adescription of social division in those do-mains, is itself subject to change.

Across the examples I have given, I havedrawn on different sources of evidence in re-

examining the links between macro- and mi-cro-level processes. In the examples of youthand of gender, work, and caregiving, I ar-gued against accounts in which the subjec-tive and objective are treated as distinct so-cial “domains.” Such accounts are associatedwith theoretical arguments of a new separa-tion between social structural and subjectiveunderstandings. These arguments posit anew autonomy of values and choices, as inpreference and individualization theories,and claims about ideology, in which externalcategories are imported to explain the seem-ing noncorrespondence of subjective andobjective. However, a reinterrogation ofdata at different levels and an improved un-derstanding of diversity and the locatednessof different vantage points allow us to seemore clearly the connectedness of subjectiveand objective. It is by moving between levelsof evidence that we can better access pro-cesses that shape diversity and change ingendered commitments to work and care. Inthe example of unemployment patterns, Iaddressed a puzzle arising from differentdata sources and their interpretation andsought a more expansive explanationthrough tackling and resolving the puzzle.All the examples reveal the connectednessof micro and macro, the subjective and thesocial structural. And in all, the use of evi-dence from different “levels” adds to our ca-pacity to explain social pattern and process.

Conclusion

Linking methods is increasingly seen as away forward in advancing social explana-tion. In this chapter I have argued that link-ing data from different levels of “the social”is important for many social science re-search questions and is crucial in cases inwhich we are seeking to understand socialdiversity and social change. Through a seriesof empirical examples, I have argued thatbringing together macro- and micro-leveldata can contribute to better understandingof the connectedness of subjective orienta-tions and social structural arrangements and










New York, NY 10001

212-431-9800

800-365-7006

www.guilford.com

improve our understanding of important so-cial processes.

The linking of macro- and micro-level datais not first and foremost a technical issue buta conceptual one. Using a metaphor of ourresearch problems as multifaceted, I have ar-gued that different methods and differentsources of data provide a particular lens onthe social phenomena or processes being re-searched. Single sources of data can give us apartial picture, but it may also be one that isdistorting. We cannot simply build up an ad-equate picture by piecing together the dif-ferent components of evidence. Data them-selves are not theory neutral but carry withinthem assumptions about how they relate tothe phenomena under study. This does notmean that different data sources are incom-patible. Nor does it mean that we will onlyreproduce the assumptions embedded with-in the data-collection tools. Seeking to bringtogether data can help clarify the nature ofthe phenomena under study by forcing us toconfront the particularity and the theoreti-cally imbued nature of different kinds of evi-dence. Linking data from different levelsmay make us aware not simply that what wesee is part of a bigger story but that we maybe in the wrong story altogether. Corre-spondingly, it has the potential to contributeto theoretical expansion or transformation.

References

Anthias, F. (1998). Rethinking social divisions: Somenotes towards a theoretical framework. SociologicalReview, 46(3), 505–535.

Anthias, F. (2001). The material and the symbolic intheorizing social stratification: Issues of gender, eth-nicity and class. British Journal of Sociology, 52(3),367–390.

Arber, S., & Ginn, J. (1995). The mirage of genderequality: Occupational success in the labour marketand within marriage. British Journal of Sociology,46(1), 21–43.

Beck, U. (1992). Risk society: Towards a new modernity.London: Sage.

Beck, U., & Beck-Gernsheim, E. (2002). Individualiza-tion: Institutionalized individualism and its social andpolitical consequences. London: Sage.

Becker, H., & Geer, B. (1957). Participant observation

and interviewing: A comparison. Human Organiza-tion, 16(3), 28–32.

Blalock, H. M. (1991). Are there really any constructivealternatives to causal modelling? In V. Marsden(Ed.), Sociological methodology (Vol. 21). Washington,DC: Blackwell.

Bottero, W. (2004). Class identities and the identity ofclass. Sociology, 38(5), 979–997.

Bottero, W., & Irwin, S. (2003). Locating difference:Class, “race” and gender and the shaping of social in-equalities. Sociological Review, 51(4), 463–483.

Brannen, J. (2005). Mixing methods: The entry of quali-tative and quantitative approaches into the researchprocess. International Journal of Social Research Meth-odology, 8(3), 173–184.

Brubaker, R. (2002). Ethnicity without groups. Euro-pean Journal of Sociology, 43(2), 163–189.

Bruegel, I., & Perrons, D. (1998). Deregulation andwomen’s employment: The diverse experiences ofwomen in Britain. Feminist Economics, 4(1), 103–125.

Bryman, A. (2001). Social research methods. Oxford, UK:Oxford University Press.

Bryman, A. (2005, October). Why do we need mixed meth-ods? Should we differentiate integration versus mixedmethods? Paper presented at the Economic and SocialResearch Council Research Methods ProgramWorkshop on Mixed methods: Identifying the issues,University of Manchester, UK.

Byrne, D. (2002). Interpreting quantitative data. London:Sage.

Byrne, D. (2005). Complexity, configurations and cases.Theory, Culture and Society, 22(5), 95–111.

Crompton, R., Brockmann, M., & Wiggins, R. D. (2003).A woman’s place . . . Employment and family life formen and women. In A. Park, J. Curtice, K. Thomson,L. Jarvis, & C. Bromley (Eds.), British social attitudes:The 20th Report (pp. 161–188). London: Sage.

Dex, S. (1988). Women’s attitudes towards work.Basingstoke, UK: Macmillan.

Duncan, S., & Edwards, R. (1999). Lone mothers, paidwork and gendered moral rationalities. Houndmills,UK: Macmillan.

Duncan, S., Edwards, R., Reynolds, T., & Alldred, P.(2003). Motherhood, paid work and partnering:Values and theories. Work, Employment and Society,17(2), 309–330.

Duncan, S., & Irwin, S. (2004). The social patterning ofvalues and rationalities: Mothers’ choices in combin-ing caring and employment. Social Policy and Society,3(4), 391–399.

Egerton, M., & Savage, M. (2000). Age stratification andclass formation: A longitudinal study of the socialmobility of young men and women, 1971–1991.Work, Employment and Society, 14(1), 23–49.

Epstein, C. F. (1988). Deceptive distinctions: Sex, genderand the social order. New Haven, CT: Yale UniversityPress.

Erzberger, C., & Kelle, U. (2003). Making inferences in










New York, NY 10001

212-431-9800

800-365-7006

www.guilford.com

mixed methods: The rules of integration. In A.Tashakkori & C. B. Teddlie (Eds.), Handbook of mixedmethods in social and behavioral research (pp. 457–488).Thousand Oaks, CA: Sage.

Freedman, D. A. (1991). Statistical models and shoeleather. In P. V. Marsden (Ed.), Sociological methodol-ogy (Vol. 21). Washington, DC: Blackwell.

Furlong, A., & Cartmel, F. (1997). Young people and so-cial change: Individualization and risk in late modernity.Philadelphia: Open University Press.

Gillies, V., Holland, J., & Ribbens McCarthy, J. (2003).Past/present/future: Time and the meaning ofchange in the “family.” In G. Allan & G. Jones (Eds.),Social relations and the life course. Houndmills, UK:Palgrave Macmillan.

Goldthorpe, J. (2000). On sociology: Numbers, narrativesand the integration of research and theory. Oxford, UK:Oxford University Press.

Hakim, C. (1996). Key issues in women’s work: Female het-erogeneity and the polarisation of women’s employment.London: Athlone Press.

Hakim, C. (2000). Work–lifestyle choices in the 21st century:Preference theory. New York: Oxford University Press.

Hammersley, M. (1992). What’s wrong with ethnography?London: Routledge.

Hammond, C. (2005). The wider benefits of adult learn-ing: An illustration of the advantages of multi-method research. International Journal of Social Re-search Methodology, 8(3), 239–255.

Hattery, A. (2001). Women, work and family: Balancingand weaving. Thousand Oaks, CA: Sage.

Irwin, S. (1995). Rights of passage: Social change and thetransition from youth to adulthood. London: UCLPress.

Irwin, S. (1999). Resourcing the family: Genderedclaims and obligations and issues of explanation. InE. B. Silva & C. C. Smart (Eds.), The new family? Lon-don: Sage.

Irwin, S. (2004). Attitudes, care and commitment: Pat-tern and process. Sociological Research Online, 9(3). Re-trieved from www.socresonline.org.uk/9/3/irwin.html

Irwin, S. (2005). Reshaping social life. London: Rout-ledge.

Irwin, S., & Morris, L. (1993). Social security or eco-nomic insecurity? The concentration of unemploy-ment (and research) within households. Journal of So-cial Policy, 22(3), 349–372.

Joshi, H. (1984). Women’s participation in paid work: Fur-ther analysis of the Women and Employment Survey (De-partment of Employment RP 45). London: Depart-ment of Employment.

Kelle, U. (2001). Sociological explanations between mi-cro and macro and the integration of qualitative andquantitative methods. Forum: Qualitative Social Re-search, 2(1). Retrieved December 10, 2006, fromqualitative-research.net/fqs/fqs-eng.htm

Kemp, S., & Holmwood, J. (2003). Realism, regularityand social explanation. Journal for the Theory of SocialBehaviour, 33(2), 165–187.

Lieberson, S. (1992). Einstein, Renoir and Greeley:Some thoughts about evidence in sociology. Ameri-can Sociological Review, 57(1), 1–15.

Lieberson, S., & Lynn, F. B. (2002). Barking up thewrong branch: Scientific alternatives to the currentmodel of sociological science. Annual Review of Soci-ology, 28, 1–19.

Marks, G., & Houston, D. M. (2002). Attitudes towardswork and motherhood held by working and non-working mothers. Work, Employment and Society,16(3), 523–536.

Mason, J.(2002). Qualitative interviewing: Asking, lis-tening and interpreting. In T. May (Ed.), Qualitativeresearch in action. London: Sage.

Mason, J. (2006). Mixing methods in a qualitativelydriven way. Qualitative Research, 6(1), 9–25.

Moran-Ellis, J., Alexander, V. D., Cronin, A., Dickinson,M., Fielding, J., Sleney, J., et al. (2006). Triangulationand integration: Processes, claims and implications.Qualitative Research, 6(1), 45–59.

Morris, L. (1995). Social divisions: Economic decline andsocial structural change. London: UCL Press.

Nilsen, A., & Brannen, J. (2002). Theorising theindividual–structure dynamic. In J. Brannen, S.Lewis, A. Nilsen, & J. Smithson (Eds.), Young Europe-ans, work and family: Futures in transition. London:Routledge.

Pawson, R. (1989). A measure for measures: A manifesto forempirical sociology. London: Routledge.

Pawson, R., & Tilley, N. (1997). Realistic evaluation. Lon-don: Sage.

Siltanen, J. (1994). Locating gender. London: UCL Press.Teddlie, C., & Tashakkori, A. (2003). Major issues and

controversies in the use of mixed methods in the so-cial and behavioral sciences. In A. Tashakkori & C. B.Teddlie (Eds.), Handbook of mixed methods in socialand behavioral research (pp. 3–30). Thousand Oaks,CA: Sage.

Trow, M. (1957). Comment on “Participant observationand interviewing: A comparison.” Human Organiza-tion, 16(3), 33–35.

Turner, S. (1997). Net effects: A short history. In V. R.McKim & S. P. Turner (Eds.), Causality in crisis?: Sta-tistical methods and the search for causal knowledge in thesocial sciences. Notre Dame, IN: University of NotreDame Press.

Walby, S. (1997). Gender transformations. London: Rout-ledge.

Young, I. (1990). Justice and the politics of difference.Princeton, NJ: Princeton University Press.

Young, I. M. (1997). Intersecting voices: Dilemmas of gen-der, political philosophy and policy. Princeton, NJ:Princeton University Press.










New York, NY 10001

212-431-9800

800-365-7006

www.guilford.com

data analysis and interpretation: emergent issues in

Documents