The case study is one of the qualitative research traditions discussed by Creswell (1998). Identify and discuss the underlying philosophy, assumptions, and components of case study research, using as examples three articles from peer-reviewed journals that employ case study methods in an educational setting. Discuss the philosophical and practical considerations that might lead to your choice of case study method in a study of universities with high degree completion rates among minority doctoral students
‘Case study’ is a term that has been used in a variety of different ways, not all of them clear, and some of them mutually inconsistent. It has been both a major category distinguishing complete alternative research styles, and a passing description meaning no more than that the study is of a single case. Its salience in the methodological literature has fluctuated over time, somewhat following the ebb and flow of quantitative versus qualitative emphases, though empirical research practice has not necessarily corresponded to that (Yin, 2002). Some relevant discussion has appeared under other banners, while ‘case study’ figures in the titles of some work which is not consciously working within a developed methodological tradition that distinguishes a case-study approach from alternatives. On the other hand, some studies which seem very plausible candidates to call themselves case studies, such as Erikson (1976) on the effects of a devastating flood, or Vaughan (1996) on the Challenger launch, do not do so, although each has a (very different) sophisticated discussion of the rationale for studying a single episode; some studies always discussed in other terms (e.g., Katz and Lazarsfeld, 1955) have in fact been based on limited local areas, and could be treated as case studies. Here, the focus is on the general methodological discussion of ‘case-study method’ and on the issues it raises (whether or not the discussion has used that name), and not on what is ‘really’ a case study (Creswell, 1998).
The idea has been used in a range of disciplines within the social sciences, not always in the same way or following the same historical course, and the associated literatures have been to some extent distinct; in the following discussion, therefore, the material is initially arranged in terms of the relatively coherent and distinct literatures which touch on relevant issues, though boundaries between them are sometimes crossed. The aim has been not to do full justice to the details of every position, but to bring out broad alternative arguments wherever they have appeared. The conclusion brings together key themes to summarize the issues raised.
It was in inter-war US sociology that ‘case study’ was particularly salient as a key category of mainstream methodological discussion. In [p. 103 ↓]this first phase, ‘case study’ represented the qualitative side of the quantitative/qualitative antithesis, and was opposed to ‘statistical method’. But the term had connotations which were not limited to the absence of statistics. Special emphasis was given to the case study's superior access to personal meanings, though other themes included the collection of data on many factors for each case and the placing of data on individual cases in a rich context (Yin, 2002). The case study was particularly associated with the life history and the ‘personal document’. Shaw's The Jack Roller (1930), a lengthy autobiographical account mostly written by its subject, could be taken as the key exemplar; few other empirical publications fitted the abstract conception so well, though ones of rather different character were also sometimes treated as examples. It was assumed, to degrees varying among authors, that case studies could in some way offer a basis for generalization.
The boundaries between these traditional methodological categories were eroded by attempts such as Angell's (1936) to systematize case method (using analytical induction), and by the development of richer and more meaningful quantitative techniques. By the 1950s, the idea of access to meanings was becoming more associated with the new category of ‘participant observation’, and the category ‘case study’ ceased to be of significance in the methodological literature. This was the period of hegemony of the modern sample survey which had emerged in the war, and case study, when mentioned, was seen in the mainstream texts as unimportant and suited only to preliminary stages of research, when it might suggest hypotheses but certainly provided no basis for prediction or generalization (Creswell, 1998).
The hypothetico-deductive model of scientific method, with a covering law model of explanation, became dominant in sociology and other disciplines. Social psychologist Donald Campbell's important work on experimental design (Campbell, 1957; Campbell and Stanley, 1966) was widely invoked in textbooks. In this, case study appeared only as the ‘one-shot case-study’ design, seen as unsatisfactory because it does not contain the before/after comparison which would justify the imputation of a cause. It also followed that case studies must show a bias towards confirmation of the researcher's hypotheses.
However, in the normal swing of reaction against dominant positions, the idea was to undergo a revival in Anglophone social science, and much of the later writing elaborates a critique of that orthodoxy. By the mid-1960s an increasing literature was developing which favored and elaborated new versions of qualitative method. This soon came to include fresh work on ‘case study’, not necessarily closely similar to the earlier work, and often apparently unaware of it. This is represented in works such as Eckstein (1975), Feagin et al. (1991), Gomm et al. (2000), Ragin and Becker (1992), Simons (1980), Stake (1995) and Yin (1984). (Hamel, 1992 and Hamel et al., 1993 show some parallel Francophone interest, though here Le Play is invoked as a key ancestor, and ‘monograph’ in his sense is equally salient as a category.)
This work, especially in the extremely influential ‘grounded theory’ of Glaser and Strauss (1967), moved away from the hypothetico-deductive model. Moreover, Campbell shifted his earlier view, proposing now that it was not true that interpretations of case studies must be inadequately tested, since the richness of data in them means that a theory is in effect tested with the degrees of freedom represented by its multiple implications. (Rosenblatt (1981: 195) points out that this represented a move from treating all case studies as analogous to experiments towards considering the specific characteristics of ethnographic studies. He also raises the question of whether it makes sense to apply to all those, as Campbell continued to do, epistemological criteria appropriate for work aiming to explain, when many of them have descriptive rather than explanatory goals.) It was also suggested that, in social-science case studies of one setting in one time period, the measurement of many variables at what [p. 104 ↓]could be seen as the post-test stage, combined with the existence of rich contextual knowledge and the possibility of making ‘intelligent presumptions about what this group would have been like without X’, could serve as logical equivalents of pretest measures and control groups (Cook and Campbell, 1979: 96). Some attention was given to this within sociology, sometimes as legitimation for positions held primarily on other grounds(Yin, 2002).
The growth of feminism often associated with the belief that only qualitative method was adequate to the experience of women (Oakley, 2000), supported the trend, as did the prevalence of versions of left-wing politics which saw statistical method as associated with capitalist rationality, though neither took positions specifically on case-study method as such. Despite such changes, mainstream textbook orthodoxy has continued to be that case studies are of very limited use for central social-scientific goals, while their proponents have continued active in advocacy and defense of alternative orthodoxies(Creswell, 1998).
Some of the discussion of case studies has come from authors who, if not themselves non-academic practitioners (in education, business, nursing, planning, social work, etc., or drawn from those fields), train them, may offer them consultancy services, and are oriented to the audiences they provide; this tends to be associated with some distinctive approaches, represented in this section.
Case studies by researchers are seen as contributing to such generalization by practitioners (in this context, teachers), and the model is a ‘democratic’ one in which the research results are not systematically analysed and are negotiated with the subjects. This conception of how to proceed clearly invokes ideological preferences which are not specific to research method, and would only be applicable to a limited range of research topics or settings. These authors get it for seriously that case studies are carried out as a root cause for sensible action. But Atkinson and Delamont (1985), also committed to ethnographic research methods and with experience in educational evaluation, have strongly criticized such methodological proposals, both because they abandon the possibility of a cumulative and theorized research tradition and because as method it is not clear what is proposed, only what is rejected.
Williams's chapter in Feagin et al. (1991) offers an interesting example which adds to the reasons for choosing a specific case-study approach. As a feminist especially concerned to retain the theoretical distinction between sex and gender, she felt that the usual quantitative methods’ use of biological sex as an isolated variable was not adequate. Her wish to study ‘the processes involved in maintaining and reproducing gender differences’ (Williams, 1991: 232) led her to undertake case studies of male nurses and female Marines, on the assumption that in gender-deviant occupations such processes would be more evident. (She indeed found that much effort went into creating conventional gender differences which participants did not spontaneously display.) Thus a deliberate choice of atypical cases was made, on the assumption that what was found there would represent processes typical of the wider society. How could this be checked? ‘The test of this, of course, is whether the findings strike a chord of recognition with those in other contexts’ (Williams, 1991: 239), but it is not proposed to undertake further research to find that out, so this could be seen as again invoking ‘naturalistic generalization’(Creswell, 1998).
The total research strategy thus defined is famed from four other possibilities experiment, survey, archival analysis and history seen as applicable to different research questions and under different practical circumstances. In this typology ‘case study’ is distinguished from ‘history’ only by the fact that it focuses on contemporary events, and so has access to some additional sources of data (Yin, 2003a: 5)
Basic definitions apart, a number of other points are made to elaborate the idea. The [p. 106 ↓]quantitative/qualitative distinction is not seen as relevant, and the style is much more quantitative. It is held that the goals of case-study research may be exploratory, descriptive or explanatory. The ‘problem’ of generalizability is boldly circumvented by saying that case studies may or may not set out to generalize, but that where generalization is attempted it should be to theory (which may later be tested) rather than to other particular cases (Yin, 2003a: 16, 38), so that it is neither necessary nor realistic to attempt to locate ‘representative’ cases.7 Studies using multiple as well as single cases are seen as applying case-study method, and there is useful discussion of types of design. The distinction is made between ‘holistic’ studies which focus on the case as a whole, and ‘embedded’ ones where sub-units within the larger case are treated as cases. For multiple case designs, it is held that a comparative or replicative rather than a sampling logic is appropriate, not a simple accumulation of more cases. These ideas are elaborated in discussion of modes of analysis of case data, which include ‘pattern matching’ of observations with one or more theoretical models, and time series analyses.
Yin frequently compares case studies with experiments, a comparison that is not salient to many other authors writing in this area, and the logic he follows has more in common with traditional discussions of ‘scientific method’ than do many of theirs in Ragin's terms, it is also more variable-oriented. His distinctive contribution may perhaps be understood as the interesting by-product of confrontation between psychological training and the contingencies of consultancy work where clients want to know about what applies to their organization, not about general theory. Although his work has been widely used, it does not seem to have been subject to critical commentary, despite its departure from some of the positions taken by other writers; perhaps this is because it serves another constituency, as well as being welcomed as the nearest approximation to a practical textbook on areas where there has been an unmet need.
It is not surprising that most academic psychologists have discussed case studies against the background of their discipline's distinctive methodological traditions, of which both the experimental and the clinical are relevant. Writing on the clinical tradition, Bolgar (1965) sketches a history in which clinical psychologists (such as Freud) followed the medical tradition of case histories and, since their prime concern was with understanding the uniqueness of the particular individual under treatment, were less concerned with the potential generalizability of what they found. She suggests, however, that in the somewhat separate field of human development psychologists (such as Piaget) commonly worked by intensive observation of very few children, often their own. The third area in which case histories of individuals have been used is the study of individuals chosen because they belong to special groups such as eminent scientists (e.g., Gruber, 1974), or are of unique historical importance such as Martin Luther (e.g., Erikson, 1959). She takes the line, conventional in other areas of the social sciences, that, while experimentation may be especially suitable to the testing of hypotheses, case histories are a valuable source for their discovery.
Kazdin (1982) goes further, arguing that ‘single-case designs’ have many uses, even in experimentation. He thinks of experimentation as evaluating the effects of an intervention—in this case, a treatment intervention. Study of the single case is appropriate because ‘The results of the average amount of change that serves as the basis for drawing conclusions in between-group research does not address the clinician's need to make decisions about treatments that will alter the individual client’ (Kazdin, 1982: 14). He suggests that the well-known weaknesses of some studies [p. 107 ↓]of single cases follow not from their study of only a single case but from the use on it of informal procedures, the lack of systematic observation and measurement, and the absence of steps to rule out the impact of extraneous factors; his book sets out to elaborate the logic and the procedures that are required to overcome these weaknesses.
In the experimental tradition, dominant in modern academic psychology, Danziger (1990) has sketched the emergence over the period up to the end of World War II of what he calls ‘the triumph of the aggregate’. He shows how in the reporting of psychological findings subjects were less and less often mentioned by name; group rather than individual data became dominant; and subjects’ extra-experimental social identities were less often mentioned when abstract knowledge (rather than application) was the aim (Danziger, 1990: 74, 82–3, 98–9), as psychologists became increasingly concerned to make universalistic knowledge claims which could be presented as the attributes of collective subjects. He suggests that one reason for the increasing dominance of statistical regularities as the basis of generalization was that individual behavior could vary markedly from one occasion to another, while if individual data were pooled regularities might emerge and be used for generalizing even if they corresponded to the behavior of no single person. In parallel with this, the nature of the subjects’ experience—and of the specific characteristics of the experimental situation—was increasingly defined as irrelevant, which he sees as necessary to the support of the political claim to societal expertise (Danziger, 1990: 153, 183–7). The result is the dominance of a statistical tradition which he regards as an inadequate scientific basis for psychological theory. Whether or not this critical analysis is agreed or fully applicable to present-day psychology, it brings out some important general points of methodology. In both clinical and experimental traditions the influence of biological and medical models is evident, and there has been disagreement about the utility of small-N methods with some of the characteristics of case study, though this has largely depended on the particular topics studied.
Anthropologists have also traditionally dealt with single cases, though these are societies rather than individuals. They have made intensive general studies of what became ‘their’ society, where ‘the unit of investigation had to be small enough for the researcher to get to know every individual, which was hardly feasible in a community larger than 500 students,’ and most ‘had an area or a tribe to study, not a problem’ (Barrett, 1996: 76). Given the linguistic and cultural problems that had to be overcome in carrying out fieldwork, there were strong practical as well as intellectual reasons for that. While this tradition held, it was so taken for granted as part of the defining character of the discipline that the issues were hardly discussed, and the ethnographies were not considered as case studies (Yin, 2002).
In the 1970s, however, what had become the standard social-science conception of scientific method started to be expounded by US anthropologists, in works such as Brim and Spain (1974), which assumed that hypothesis-testing and the use of representative samples and statistical tests were needed. In addition, the assumption that small isolated societies could be found (which it made sense to treat as independent units) was increasingly undermined by world developments (Barrett, 1996: 111). The felt need for comparative material led to suggestions that internal variation be sought within limited areas, or that a collective enterprise should be established to create comparative data (Levine, 1973: 187–8), and the advantages of team ethnography in more complex societies were discussed. More recently, the idea of ‘multi-site ethnography’ has been developed (Marcus, 1995). This is concerned to go beyond the traditional boundaries of the anthropological village study, especially [p. 108 ↓]under the circumstances of globalization which make it increasingly unhelpful to confine what is taken into account within narrow local geographical units, but the further ‘sites’ to be taken into account are not necessarily geographically defined, even if geographically separate. They are whatever is needed to locate the starting point within its global context:
It should be noted that this is not an approach associated with comparative method; it is concerned rather to understand the case studied fully by placing it in its wider context, and thereby also to understand the wider context, which may be as wide as the capitalist world system which is by definition unique, and so not open to (contemporary) comparative study.
Insofar as Yin's approach is one which emphasizes the need to take seriously the setting of the case which is the central focus of study, that has something in common with this discussion. Both are concerned with how boundaries can appropriately be drawn around the cases studied, although they start from totally different assumptions about the nature of the conventional pattern of research with which the approaches they propose contrast.
Such arguments have, however, probably been less acted on in practice than proposed in principle, except in fields such as migration or trade, where the value of following a flow is most easily evident.
A somewhat different line of discussion has been followed in some contributions from British anthropologists of the Manchester school led by Gluckman. He developed the idea of the ‘extended case study’ (Gluckman, 1961), suggesting that what is needed to improve on general statements about ritual and custom in a social system is study of specific cases of the way in which students behave in practice and, in particular, of the pattern of events in incidents involving the same students over time: this can show how social relations develop and change historically.Mitchell (1983), in an important article, goes beyond Gluckman's discussion to raise issues in epistemological terms. His key point is to distinguish between statistical inference and logical or scientific inference. The former is concerned with inferring the distribution of characteristics in a population from observation of a sample, which only provides a plausible basis for inference if it can claim to be a representative sample; the relationships between variables which it provides are correlational, showing which characteristics go together, but not why. For plausible imputation of causal relationships, it must be shown that there is a logical nexus between the factors considered, and for Mitchell case studies aim to explore just that.
[p. 109 ↓]It follows that the issue of whether one can generalize from the fact that a case study is a matter not of the number of cases studied but of the adequacy of the theory in relation to which it is interpreted, and the cogency of its theoretical interpretation against a background of knowledge of other cases.
Authors advocating case study method have not always confined themselves to the use of a single case (which might be the case of a large group such as a tribe, town or organization), and some have seen the use of more than one case as in general an improvement. Complex designs in which the number of cases varies with the level considered are quite common in practice. Probably commonest is the type of study where one town, organization or country is the ‘case’ within which data are collected on a sample of members or social units; this is not usually regarded as constituting a case study, though it can be used as such. The individuals’ characteristics may be used to characterize the group, or the group's to characterize the individuals, and the meaning of either procedure will depend on what the other cases are from which they are distinguished. Types of case-study research design which include the possibilities of comparison between cases have been distinguished, and these begin to approach some of the problems that have been more consistently and systematically considered under the rubric of comparative method; we shall now move on to consider some of the work and ideas more associated with that.
There is a long-standing tradition of publishing books which collect papers by authors, often from different countries, who have studied cases of the same kind. Their chapters are then presented as case studies from which comparative conclusions can be drawn, or are given an editorial framework which attempts to draw such conclusions, despite the lack of advance coordination to ensure close comparability in the questions raised and the data collected. That pattern leaves research ‘design’ as retrospective, but there have also been more prospective designs which bring together work by different authors, and these demonstrate some interesting possibilities.
Shavit and Blossfeld (1993) provide an impressive example of work, coordinated among members of a research committee of the International Sociological Association (ISA), which brings together comparable data from 13 countries on class and gender differences over time in educational attainment. Burawoy (2000) has edited a book where the substantive chapters are provided by research students working on topics individually chosen before they came together, but developed in collective work. It takes the theme of globalization, and uses their data—on groups which range from breast-cancer activists in San Francisco to Irish software developers—both to understand and theorize those groups in terms of the way in which globalization enters into them, and also to throw light on understandings of globalization. (Burawoy invokes Gluckman, and defines the method as that of the extended case, both extending out from micro processes to macro forces, and extending theory by improving existing theory in the light of fresh cases.) Lamont and Thévenot (2000) present an enterprise designed to create comparisons between French and American cultural differences in principles of evaluation. The authors worked on their own topics, in this case ranging from the rhetorics of working-class racism and anti-racism to styles of literary theory, but compared French and American data and worked in ways informed by the group discussions (Creswell, 1998).
The more conventional ‘comparative method’ literature relates to bodies of empirical work which are most often on topics related to national politics, such as elections, welfare policies or revolutions. These are macro-social topics in which the whole society (‘society’ here usually meaning a nation-state) is the normal unit (though the problems of logic are the same when smaller units are used). They have in this sense been mainly the concern of political scientists and sociologists. Ragin (2000: 25) points out that the empirical distribution of comparative studies in sociology and politics shows a dip in the range of numbers of cases used between one or two and more than 50. He suggests that this is the range where comparative method applies. That is the main territory of this literature relevant to case-study issues. A strong reason for the use of limited numbers of cases is that few of the units of interest exist. That problem is exacerbated by the varying levels of data available on the existing cases and their uneven distribution across types; there are intrinsic difficulties for a quantitative approach here.
There are also other practical issues with intellectual consequences. Where the unit for comparison is the whole society, considerable depth of expert knowledge (for which length of study and foreign-language acquisition are needed) is often required to give an adequate account of each case. (Those with the expertise are as likely to work within an ‘area studies’ paradigm as a comparative one.) This means that no single individual is likely to have expertise on many cases, while those who attempt to make secondary use of others’ data may be open to criticism for the shallowness of their knowledge of the cases, as by Goldthorpe (2000: 28–44).
Among those identifying themselves as comparativists, Ragin (1987: 70) distinguishes usefully between the more case-oriented area specialists and the more variable-oriented generalists. (The classic argument in favor of the variable-oriented approach was put by Przeworski and Teune (1970), who rejected the claim that national societies were unique and so could not meaningfully be compared; they proposed that, since comparability depends upon the availability of higher-level concepts which subsume historical particularities, the goal should be to replace the names of nations with the names of variables.) The case-oriented approach is strong in its access to complexity and historical specificity, and in its holistic grasp of the ways in which different factors are interrelated, but weak in its scope for plausible generalization because of the limits set by the small number of (commonly unrepresentative) cases that it draws on. The variable-oriented approach does not have the same problem of number of cases (though the cases it has may still be in practice an unrepresentative or narrow selection), and so can make use of sophisticated quantitative techniques, but it risks doing so at the cost of attention to the specific characteristics of individual societies, which may be concealed behind theoretical variable labels at such a high level of abstraction that they have little practical meaning. The case-oriented approach can address issues specific to relatively small [p. 111 ↓]numbers of societies, which the variable-oriented approach is compelled to ignore. Ragin develops a strategy that aims to overcome this distinction by providing methods which formalize qualitative comparison and can deal with configurations, not just separate variables, so that the strengths of the case-oriented approach are retained while some of the criticisms made of it are met. Such ideas are developed further in his later work on ‘fuzzy sets’ (Ragin, 2000).
Goldthorpe (2000: 49–53) argues that the key difficulty underlying the small-N problem is in fact one of data rather than method: insufficient information is available. But the information can be increased by adding more cases, either societal or of periods within the same societies, especially more diverse ones. Other issues he raises which proponents of case-study approaches have used against large-N work are the difficulty of identifying cases which are independent of each other, and the ‘black-box’ problem of establishing why it should be that certain inputs are associated with certain outputs. For each of these, he suggests that they are real, but as relevant to case-oriented as to variable-oriented research. Ebbinghaus (2005) reaches a similar conclusion, but treats it as implying more criticism of large-N research. He sees the critics of small-N work as tending to make the inappropriate assumption that comparative method can be reduced to the analysis of logical truth tables, though pointing out that Lieberson's criticism about the difficulties of dealing with probabilistic theories does not hold for conditions necessary to the outcome even if not sufficient, since those must be present whatever the intervention of chance events. He also points out that the cases used in large-N studies are in practice often far from a representative sample of the cases that exist, because of the difficulty of finding ones where the data required are actually available. In addition, where cases in large-N studies are relatively diverse, the extent to which they are effectively stratified violates the homogeneity assumptions of the inferential statistics used, while the path-dependency of some current historical situations means that the difficulty of introducing change when circumstances have changed (his examples are drawn from state welfare provision) produces an apparent lack of correlation with causal factors which is misleading.
It is clear that there is no single understanding of ‘case study’ or ‘case-study method’. Different disciplines and sub-fields have commonly proposed and discussed it in ways which take for granted different empirical types of cases as those for study, and define the method by contrast to alternatives specific to their intellectual traditions. However, they have also responded to some shared historical developments in such matters as the philosophy of science, quantitative methods and the public availability of data. There are general themes which cut across the varying approaches, whether or not all writers have treated them as such. We conclude, therefore, by drawing out key themes and making a few comments.
|The Case of Genetically Modified Food
|Recommendation for Technology Plan: High School Case Study
- Recommendation for Technology Plan: High School Case Study
- Benefits of Meditation
- The Case of Genetically Modified Food
- Barbie Generation