Design and Methods
Second Edition

Robert K. Yin

Applied Social Research Methods Series

SAGE Publications
~~ International Educational and Professional Publisher
Thousand Oaks London New Deihl


The case study is but one of several ways of doing social science research.
Other ways include experiments, surveys, histories, and the analysis of ar-
chival information (as in economic studies). Each strategy has peculiar ad-
vantages and disadvantages, depending upon three conditions: (a) the type of
research question, (b) the control an investigator has over actual behavioral
events, and (c) the focus on contemporary as opposed to historical phenomena.
In general, case studies are the preferred strategy when "how" or "why"
questions are being posed, when the investigator has littlecontrol over events,
and when the focus is on a contemporary phenomenon within some real-life
context. Such "explanatory" case studies also can be complemented by two
other types-"exploratory" and "descriptive" case studies. Regardless of the
type of case study, investigators must exercise great care in designing and
doing case studies to overcome the traditional criticisms of the method.


This book is about the design and conduct of case studies for research
purposes. As a research strategy, the case study is used in many situations,

• Policy, political science, and public administration research
• Community psychology and sociology
• Organizational and management studies
• City and regional planning research, sucb as studies of plans, neighborhoods, or
public agencies
• The conduct of dissertations and theses in the social sciences-the academic
disciplines as well as professional fields such as business administration, man-
agement science, and social work

This book covers the distinctive characteristi~s of the case study strategy,
compared with other types of research. Importantly, the book deals with

In all of these situations. 1969. The more appropriate view of these different strategies is a pluralistic one. such as medicine. variety-are far different than those for doing research (e. ogy. TIris book. And each strategy has its Own Similarly. for instance." offering few guides on how by the subfield known as historiography. Decision: Explaining the Cuban Missile Crisis. So case studies arises out of the desire to understand complex social phenomena. A common misconception is that the various research strategies should be or social work. 1992a). or the economy of a city or a region. WITH OTHER RESEARCH STRATEGIES ences. Towl. the case study contributes uniquely to our knowl. social work. Society. data collection. research issue. (The fol- Dowdy. empirical evidence. often.. In addition. Case studies are even found in economics. however. 194311955. However. rather. descriptive. answers these questions. For teaching purposes. For example. business. it is not intended to help those who might use case studies as teaChing devices. We were once taught to believe that case studies were different than those for designing case studies for research. and that experiments used as a research tool (e.2 CASE STUDY RESEARCH INTRODUCTION 3 design. law. you consider doing an experiment instead? A survey? A history? A computer- ing good cases for teaching-usually of the single. it is hoped. research case studies need to do exactly all of them. popularized in the fields of law. Some of the best and most famous Case problems of composing the case study report. neighborhood change.and not multiple-case based analysis of archival records such as student records? . analysis. following its own logic. Stein. however. Whyte's Street Corner phases of design. that surveys and In contrast. Finally. may be investigated by The overall goal of this book is to help investigators deal with some of the using a case study design. the case study allows an investigation to retain the holistic and who has asked (a) how to define the case being studied. this book does not cover all uses of case studies.) Each is a different way of collecting and analyzing that. and reporting. Medical records. or even how to minimize the being only an exploratory strategy. medicine. 1986) and that were the only way of doing explanatory or causal inquiries. social work files. Again. 1952. the distinctive need for more difficult questions commonly neglected by available research texts. business. organizational and managerial processes. 1983) but COMPARING CASE STUDIES now prevalent in virtually every academic field.g. To get the most out of using the case study are used as a form of record keeping. the rationale for this book is that case studies are increasingly histories were appropriate for the descriptive phase. the development of evidence. sociology. relations. Perry & Kraemer. or (c) what should be done with the data. you need to know these differences. in which the structure of a focus on data collection or fieldwork. Hamel. surprisingly. see BOX 2). Experiments with an explora- and analyzing case studies and is not merely a guide to collecting case study tory motive have certainly always existed. When and why would you want to do case studies on some topic? Should work for discussion and debate among students.g. social. including the natural sci. is incorrect. 1987). and the maturation of industries. international once collected. Teaching case studies need not be concerned with the rigorous lowing discussion focuses only on five choices and does not attempt to catalog and fair presentation of empirical data.or multiple-case studies to investigate a could not be used to describe or test propositions (Platt. 1948. political science. or . The criteria for develop. Caulley & These and other choices represent different research strategies. case studies are far from to start a case study. or public policy (see Llewellyn. other case records are used to facilitate some practice. given industry. This book concentrates heavily on the problem of designing This hierarchical view. The hierarchical you-who may be a seasoned or budding social scientist-would like to know view reinforced the idea that case studies were only an exploratory tool and how to design and conduct single. organizational. appropriate for the exploratory phase of an investigation. Each strategy can be used for all three purposes---exploratory. Windsor & Greanias. the relevant data to be collected. this book is not intended to cover those situations in which cases advantages and disadvantages. Not title]. (b) how to determine meaningful characteristics of real-life events-such as individual life cycles. the case study has been a common research strategy in psychol. and planning (Yin. the criteria for developing good cases for practice are arrayed hierarchically. how to analyze the data. 1971 [emphasis added to edge of individual. causal explanations has long been a serious concern of historians. the author has been confronted by a student or colleague In brief. This book covers all of the studies have been both descriptive (for example. its purpose is to establish a frame. and political phenomena. reflected which is dominated by texts on "field methods. 1992. and reporting issues--and not merely the more traditional 1983). a case study need not contain a complete or accurate rendition of actual events. analysis. the book fills a void in social science methodology. and strategy. see BOX 1) and explanatory (see Allison's Essence of As a research endeavor.. In this sense.

an exploratory survey. Allison develops the best explanation for this type of crisis. Thus it traces the sequence of interpersonal be to pose competing explanations for the same set of events and to indicate events over time. Even though each strategy has its distinctive characteristics. 198Ib).'' "what. However. There also may be explora- tory experiments. descriptive experiments. why the United States responded to the missile deployment with a neighborhoods. paradoxically. First. group structure. sic categorization scheme for the types of questions is the familiar series: there are large areas of overlap among them (e. What distinguishes the strategies is not this hierarchy but three other condi- tions. strategy but another is really more advantageous. previous study. column 1). Thus Allison cites the U. The single case is the confrontation ties. such as this one: "What are the ways of making schools effective?" This type When to Use Each Strategy of question is a justifiable rationale for conducting an exploratory study. its generaIizability to Union placed offensive (and not merely defensive) missiles in Cuba in the first issues on individual performance. some types of ''what'' questions are exploratory. missiles in Cuba. have been the outcomes from a particular managerial reorganization?" Iden- analysis. and why the Soviet Union even- nerville in their work. in tifying such outcomes is more likely to favor surveyor archival strategies distinguishing among the five strategies. blockade (and not an air strike or invasion). nuclear confrontation more generally. by William F. involvement in Vietnam. 1993). (b) inquiry. the goal being to develop pertinent hypotheses and propositions for further The three conditions consist of (a) the type of research question posed. there- by extending the usefulness of his single-case study. useful explanation.g. There may be exploratory case studies. For example. A ba- clear and sharp. The second type of "whaf' question is actually I. descriptive case studies. and the social struCbJre of place. or explanatory case studies (Yin. Sieber. purpose." ''where. possibilities arises. and explanatory experiments. 1973). Later investigators have repeatedly found remnants of Cor. this does not imply that the boundaries Types of research questions (Figure 1. either of two. is discussed below. The importance of each condition. Figure or an exploratory case study. By comparing each theory with the actual course and different time periods. and been recommended reading in community sociology. 1981. Nevertheless. and case studies. I displays these three conditions and shows how each is related to five major a form of a "how many" or "how much" line of inquiry-for example. including answers to three key questions: why the Soviet 50 years old. and the termi- nation of wars by nations as other situations for which the theory can offer explanatory. Bickman. as an exploratory study. and discovers key phenomena-such as the career advancement This strategy was followed by Graham Allison in Essence of Decision: of lower income youths and their ability (or inability) to break neighborhood Explaining the Cuban Missile Crisis (1971). between the United States and the Soviet Union over the pIacementof offensive The study bas been highly regarded despite its being a single-case study." and "why. and can be used-for example. (c) the degree of focus on contemporary as opposed to historical events. The goal is ''who. than others. of events. a survey can be readily designed to enumerate the .'' "how. an exploratory experiment. histories. & Rog. Allison posits three competing theories or models to explain covering one neighborhood ("ComerviIle") and a time period now more than the course of events. discussed below. any of the five research strategies the extent of control an investigator has over actual behavioral events. The book is a classic notmereIy exploratory (or descriptive). The analyst'sobjective should example of a descriptive case study. surveys.S.1. describes a subculture that had rarely been the topic of how such expIanations may apply to other situations. has for decades Even a single-case study can often be used to pursue an explanatory.4 CASE STUDY RESEARCH INTRODUCTION 5 BOXl BOX 2 A Famous Descriptive Case Study An Explanatory Case Study Street Corner Society (194311955). archival ." to avoid gross misfits-that is. Whyte. "What research strategies in the social sciences: experiments. Allison suggests that this explanation is applicable to other situations.. when you are planning to use one type of If research questions focus mainly on "what" questions. The value of the book is. even though they have studied different neighborhoods tually withdrew the missiles. The first condition between the strategies--<>r the occasions when each is to be used-are always covers your research question(s) (Hedrick..

in addition to conducting interviews..stI_ how. These strategies are advantageous when the research goal is to economy in denying support for the Bush~Quayle ticket as incumbents. needing to be traced over time. study. 1971). if you wanted to know "why" riots - occurred... and experiments as the preferred which surveys or the analysis of archival records would be favored).. you would lean toward either a case study or a field experiment. as in economic favor. if you wanted to know "what" the outcomes of a new govern- "'OWID. you might survey residents. SOURCE: COSMOS Corporation.. ''How'' research strategies. you could answer this frequency question by doing ~ how." Uwhat. you could Am I asking a "who. histories. bystanders fail to report emergencies under certain conditions. research. or conduct a ''windshield survey" of the riot area In contrast. question being asked.IICk mental program had been.'' "why.... The key is to understand that research questions have both doing a history or a case study. rather than mere frequencies or incidence.- lmoa 01 . 1969).. if you wanted to know how a community successfully thwarted a Defining the research questions is probably the most important step to be proposed highway (see Lupo et al.. why DO a surveyor by examining economic data. "How" and "why" Bill Clinton got elected in 1992 can be studied by either a surveyor a case study. The survey might examine voting patterns. you would probably be doing a multiple-case howmada . showing ''whats. if you wanted to know why substance-for example.S. h_.. the first and most important condition for differentiating spread of a disease like AIDS (in which an analysis of health statistics might among the various research strategies is to identify the type of research be the favored strategy) would be typical examples... if you focused on the "why" question in more than one city. But if you needed to know "how" or ''why'' the program had worked Figure 1. ." or "how" question? Others design and conduct a series of experiments (see Latane & Darley. If you were studying "who" participated in riots... Relevant Situations for Different Research Strategies (or not). ''what'' questions may either be exploratory In contrast.. Similarly. how_ DO mation. "who" and ''where'' contrast. or histories.t. . Some "how" and "why" questions are ambivalent and need clarification...'' whereas a case study would not be an advantageous strategy in this that voters for Ross Perot drew largely from supporters of then President situation... have focused on some of the substantively important issues (see Campbell.1. -. the case study might examine how Clinton conducted his campaign questions (or their derivatives-"how many" and "how much") are likely to to achieve the necessary nomination and to manipulate public opinion in his favor survey strategies or the analysis of archival records... depending upon the type of program - DO involved." ''where.s... What is my study about?-andform-for example...ta how_ DO -'DO Similarly. what.6 CASE STUDY RESEARCH INTRODUCTION 7 : esperbneIlt ... you would have to draw upon a wider array of documentary infor- oaney . In general.. This is because such questions deal with operational links and "why" questions are likely to favor the use of case studies..where. ''liow'' and "why" questions are more explanatory and likely (in which case any of the strategies 'could be used) or about prevalence (in to lead to the use of case studies.. Bush. you would be less likely to rely taken in a research study. and "how much" damage had been done. why Let us take two more examples. This describe the incidence or prevalence of a phenomenon or when it is to be approach also would be an acceptable way of addressing the "how" and "why" predictive about certain outcomes. like this second type of ''whaf' question.. so patience and sufficient time should be allowed on a survey or an examination of archival records and might be better off for this task.""" DO program serve? What kinds of benefits were received? How often were different benefits produced? These could all be answered without doing a case study. Thus consider such questions as these: How many clients did the . The study would cover the potentially helpful role of the weak V. examine business records (an archival analysis). and this could satisfactorily address the how and why questions. experiments.... The investigation of prevalent political questions but would be different than the survey study. attitudes (in which a surveyor a poll might be the favored strategy) or of the To summarize. -.-. Thus. In Similarly.

students think-an end in itself. of course. the of a literature review is to detennine the answers about what is known on a case study's unique strength is its ability to deal with a full variety of evi. Too many times. the relevant behaviors cannot be manipulated. an investigator may be interested in studying fonn of the study question best matching the strategy you were inclined to race riots or lynchings (see Spilennan. the point of the preceding discussion is that the fonn includes those situations in which the experimenter cannot manipUlate be- of the question provides an important clue regarding the appropriate research havior (see Blalock. To this extent. Budding investigators think that the purpose atic interviewing. column 2) and degree Summary. even retrospectivelY. case study. and when an investigator must rely on primary documents. the case study investigator has been sloppy and providing them with different kinds of vouchers (Boruch. artifacts. If so. requires much preparation. The full range of experimental science also the findings and conclusions. column gies might be relevant (such as exploratory research). such as research. 1961. a survey within a case study or a case study wi~ Histories are the preferred strategy when there is virtually no access or a survey). for some questions. mental design because no control over the behavioral event was possible. Again. which two strategies might be considered equally attractive (such as how a further distinction among history. infonnal manipulation can occur. in which. in which an experiment may focus on one or two isolated variables research investigators nevertheless have disdain for the strategy. precisely. in some situations. control. about contemporary events. the various strategies are not mutually exclUSive. We can identify some situations in which all research strate- offocus on contemporary as opposed to historical events (Figure 1. and cultural and physical • a ''how'' or "why" question is being asked about a contemporary set of events artifacts as the main sources of evidence. topic. the methods overlap. although case studies and histories can overlap. in contrast. 1971) and may use a quasi-experi- pursue in the first place. and systematically. and observations-beyond what develop sharper and more insightful questions about the topic. where the tenn social experiment has emerged to cover research in Perhaps the greatest concern has been over the lack of rigor of case study which investigators ''treat'' whole groups of people in different ways. interviews. To detennine the questions that ate most significant for a topic. 1982). that you may be predisposed to pursue a situations. Remember. and not-as most usually included in the historian's repertoire: direct observation and system. 1984). Extent ofcontrol over behavioral events (Figure 1. experiments are done when an investigator can manipulate behav- ior directly. . Histories can. forthcoming). strategies. be done over which the investigator has little or no control. might be available in the conventional historical study. The quasi-experimental approach can even be used in a historical particular strategy regardless of the study question. Thus the distinctive contribution of the historical method is in But we can also identify some situations in which a specific strategy has a dealing with the "dead" past-that is. In other (and presumes that the laboratory environment can "control" for all the words as a research endeavor. but when some precision in fonnulating these questions. or it can be done in a field desirable fonn of inquiry than either experiments or surveys. a choice among strategies might These situations have been commonly regarded as "quasi-experimental" actually exist. the large areas of overlap among the 1979) but in which the logic of experimental design may still be applied. the strategy begins to overlap with that of the case study. and to gain The case study is preferred in examining contemporary events. what occurred. Campbell & Stanley. Note that of the same techniques as a history. Cook & Campbell. The case study relies on many One way is to review the literature on the topic (Cooper. but it adds two sources of evidence not such a literature review is therefore a means to an end. and experiment is the extent and why Clinton got elected). given study (for example. too. Why is this? setting. We also can use more than one strategy in any of the investigator's control over and access to actual behavioral events. 1966. finally.8 CASE STUDY RESEARCH INTRODUCTION 9 Daft. in this situation. Remember. this is when report. Moreover. secondary documents. and other situations in 3). so that. This can occur in a laboratory Although the case study is a distinctive fonn of empirical inquiry. Assuming that "how" and "why" questions are to be the focus of study. experienced investigators review previous research to dence---<locuments. & Hulin. For the case study. many setting. case studies have been viewed as a less remaining variables beyond the scope of interest). for instance. when no relevant persons are alive to distinct advantage.1. be sure to create the setting. strategy to be used.1. such as participant-observation. has allowed equivocal evidence or biased views to influence the direction of Again. Traditional Prejudices Against the Case Study Strategy Finally.

such as ethnography or participant-observation. unanswered. Since neither view is well founded. 1976). Orum. writing the case study-including ones in which the traditional. observational evidence. 1956. and this book will help him or her to ing upon the topic being studied. Or. For detailed.. skills. Somehow. One could even do a valid and such step would be strictly forbidden.g. Mosteller. Every case study investigator must high-quality case study without leaving the library and the telephone. is to expand and generalize theories (analytic generalization) and not to BUT A COMMON DEFINITION enumerate frequencies (statistical generalization). even from another field. like experiments. like the experiment. the skills for doing A second common concern about case studies is that they provide little good case studies have not yet been defined. used with multiple-case studies but requires a different concept of the appro. pp. given the way case studies have been done in the past (e. one experiments (see Rosenthal. The same approach can be This quotation is from a book by five prominent statisticians. and nearly all of us believe nedy. as above. the goal is to do a "generalizing" and not a "particularizing" analysis (Lipset. commonly asked questions about case studies have still been A third frequent complaint about case studies is that they take too long. However. 419-420). experiment?" In fact. scientific facts are rarely based on single experiments. 1982. In this sense. case study materials may be deliberately contrast. do so. 1991). For example. the case study. People know when they cannot play music. & Coleman. 134) they are usually based on a multiple set of experiments. unreadable documents. and they can be tesred for other encountered and less frequently overcome. depend- work hard to report all evidence fairly. most people feel that they can prepare a case study. does not represent a "sample. Light. Definition of the Case Study as a Research Strategy This incorrectly confuses the case study strategy with a specific method of data collection. the case study been asked about an experiment: "How can you generalize from a single receives a good deal of approbation it does not deserve. & Sjoberg. which have replicated the same phenomenon under different conditions. McPeek. The problem such as designing questionnaires for surveys (Sudman & Bradburn. (Hoaglin. The discussion has progressed without a formal definition of case studies. case studies are a form of inquiry that does not depend solely on altered to demonstrate a particular point more effectively. Nor need case studies take a long time. In with case study research. . Surprisingly. they recognize the challenge of doing good case priate research designs. lengthy narrative can be avoided altogether. and as a result. this is discussed in detail in Chapter 2. "How can you generalize from a single case?" is a frequently heard question. as three notable social scientists describe in their single case study. major lesson is still that good case studies are very difficult to do. Participant-observation may not require the example. 10 CASE STUDY RESEARCH INTRODUCTION 11 The possibility also exists that people have confused case study teaching same length of time but still assumes a hefty investment of field efforts. Trow. This complaint may be included in the same study? Do case studies preclude the use of quantitative appropriate." and the investigator's goal DIFFERENT TYPES OF CASE STUDIES. but this is not necessarily the way case journalistic accounts? Let us now attempt to define the case study strategy studies must be done in the future. basis for sciel1tific generalization. is it still a case study when more than one case is and they result in massive. evidence? Can case studies be used to do evaluations? Can case studies include Feagin. any ethnographic or participant-observer data. in the words of one observer. p. The most frequently encountered definitions of case studies have merely phies usually require long periods of time in the "field" and emphasize repeated the types of topics to which case studies have been applied. In research. Moreover. Ethnogra. consider for the moment that the same question had we can understand one. The problems are not to do good case studies. 1966) and the use of other research strategies. The answer is not a simple one (Ken. they also different. but in case study research. Chapter 6 discusses alternative ways of and answer these questions. The short studies. 1968). answer is that case studies. In teaching. such as by the bar examination in law. & Stoto. What is often forgotten is that bias also can enter into the conduct of Despite the fact that these common concerns can be allayed. they may have been more frequently know when they cannot do mathematics. are generalizable to theoretical propositions and not to populations or universes. 1982) or is that we have little way of screening or testing for an investigator's ability conducting historical research (Gottschalk.

citing the topic is surely insufficient for estab. in fact. Fmally. a strategy to be collection and analysis. for instance. with data needing to converge in a definitively dissociated the case study strategy from the limited perspective triangulating fashion. the technical definition begins 1991) but a comprehensive research strategy." wanted to cover contextual conditions-believing that they might be highly and even "events. especially added) when • the boundaries between phenomenon and context are not clearly evident This definition thus cites the topic of "decisions" as the major focus of case studies. but now may be restated in two ways. begins with "a logic of design . Second because phenomenon and context are not always distinguishable work" only as a data collection technique and omit any further discussion of in real-lif~ situations. now become the second part of In a historical overview of the case study in American methodological our technical definition: thought. The most popular contemporary texts (e. 1989) or with participant-observation (Jorgensen." "neighborhoods. 1992a. constantly a textbook's presumed discussion of case studies was in reality a description struggles to limit the number of variables to be analy~ (and hence the either of the ethnographic method or of participant-observation as a data number of questions that can be asked) to fall safely wlthm the number of collection technique. A history. and as one result in suspension. and casework in social work. leaving the further definition of any distinctive case study strategy · variables of interest than data points. She then shows how ''participant-observation'' emerged as a data collection • copes with the technically distinctive situation in which there wiII he many more technique. 2. 1989). She traces the practice of doing case studies back to the conduct of life histories. a whole set of other technical characteristics.:"een phenomenon and itself was only mentioned in a line or two of text. 1986. encompassing method-with the logic of design incorporating specific ap- And just what is this logic of design? The technically critical features had proaches to data collection and to data analysis. In this sense. and the case study comparison. pertinent to your phenomenon of study. other topics have been listed. and as another result of doing participant-observation (or any type of fieldwork). A case study is an empirical inquiry that is that it tries to illuminate a decision or set of decisions: why they were taken. 1982)." "institutions. preferred when circumstances and research problems are appropriate rather than an ideological commitment to be followed whatever the citcumstances" In other words. how they were implemented. This first part ~f ~ur logic . As context. in her words. you would use the case study method because you deliberately "organizations. 19810. p. the central tendency among all types of case study. by exploratory stage of some other type of research strategy. still cover "field. Finally. emphasis • investigates a contemporary phenomenon within its real-life context. so that attention can be focused on only a few vanables (typICally.. so that context is extremely limited. First." However. Similarly. surveys can try Another common flaw has been to confuse case studies with ethnographies to deal with phenomenon and context. discussed earlier. 46). context. The case study • benefits from the prior development of theoretical propositions to guide data strategy. Jennifer Platt (1992a) explains the reasons for these treatments. but usually with noncontemporary events. The survey designer.12 CASE STUDY RESEARCH INTRODUCTION 13 the essence of a case study. (Schramm.'' "programs." ''processes. does deal with the entangled situation bet. 1992)..of des~gn lishing the needed definition. the case study been worked out prior to the first edition of this book (Yin. and with what result. for instance. the case study as a research strategy comprises an all- (Platt. 1971. one common flaw was to consider the case study as the the context is "controlled" by the laboratory environment). she explains how the first edition of this book (1984) • re1i~ on multiple sources of evidence. . deliberately divorces a phen?menon fr?m its by five statisticians from Harvard University-Hoaglin et aI. Kidder respondents that can be surveyed.g." In other words. 1981b) is not either a data collection tactic or merely a design feature alone (Stoecker. data collection and data analysis strategies. therefore helps us to understand case studies by contmumg to distinguish Alternatively. including "individuals.. case study a formal research strategy at all (the major exception is the hook An experiment. including case studies. but their ability to investigate the (Fetterman.. & Judd. I How the strategy is defined with the scope of a case study: and implemented is the topic of this entire book. most social science textbooks have failed to consider the them from the other research strategies that have been discussed. The case study inquiry the wode of the Chicago school of sociology. Nachmias & Nachmias. 1.

yes. gator may draw according to a given situation. the case study strategy should not be the Watergate scandal. categories (U.. case studies have a distinctive place in evaluation research (see Variations Within Case Studies Cronbach et al. Whatever the application. some investigators distinguish between quantitative a single. case studies can illustrate certain topics within an evaluation. Though some fields. 1986. The most important is to explain the causal links in real-life Yes. by two reporters from The Washington Post (see confused with "qualitative research" (see Schwartz & Jacobs. indicating the situations in which doing As a further note.. Van Maanen. 1994).j the strategy but may be considered variations within case study research and one favors qualitative or quantitative research. Sechrest. the counterargument can still be posed-that regardless of whether I! t. 1987. And.~ tl 14 15 CASE STUDY RESEARCH INTRODUCTION t1 . however. evidence. Strauss BOX 3). A second tenns as the comparative case metlwd as a distinctive fonn of multiple-case application is to describe an intervention and the real-life context in which it studies. 1979. as the but on the basis of wholly different philosophical beliefs (e. 1993. 1980. 1990).and multiple-case studies. Fifth. 1979. In fact. research strategies in social science. Dabbs. quantitative intervention being evaluated has no clear. Likewise. case study research can include both single. Guba & Lincoln. interventions that are too complex for the surveyor experimental strategies. Actually. 1986). the evidence. ethnographic research does This chapter has introduced the importance of the case study as a research not always produce case studies (for example. an empirical topic by following a set of prespecified procedures. 1986). historical research can include enonnous amounts of quantitative And. . 1983. case studies can include. 1988. one constant theme examples. evidence. 1990. Stake. have In evaluation langnage. 1980. 16). 4). Smith & Heshusius.there is a strong and essential R also provide answers to common questions. 1975). the case study strategy may be used to explore those situations in which the And. 1990). Instead. certain journalistic efforts can qualify as case studies. Patton. The case study. General as a Research Strategy Accounting Office.g. Some qualitative research follows ethnographic methods and seeks to satisfy two conditions: (a) the use of close-up. In addition. the contrast between quantitative and qualitative evidence case study may be a "meta-evaluation"-a study of an evaluation study does not distinguish the various research strategies. 1989. like other research strategies. & Corbin. Lijphart. as analogous (N. 1982). yes. single set of outcomes. for instance. 1981. Smith. U. 1991. and even be limited to. case studies need not always include direct. and not quantitative.S.or multiple-case study may be preferred. to doing a research and qualitative research-not on the basis of the type of evidence survey. . General Accounting Office. There are at least five different applications. the explanations would link program implementation tried to delineate sharply between these two approaches (and have used such with program effects (U. detailed observation of the natural SUMMARY world by the investigator and (b) the attempt to avoid prior commitment to any theoretical model (Jacob. Lincoln & Guba. 1982. Fourth. Jacobs. and multiple-case studies are in reality but two variants of case study designs again in a descriptive mode---<:ven from a journalistic perspective. Stake. Lincoln. common ground between the two (Yin. chap. The basic Lincoln. These case studies can be based on any mix of quantitative and qUalitative evidence.. see the brief ethnographies in strategy. & Faulkner. However. Some situations may have no clearly preferred strategy. nor are case studies limited to these two conditions. 1990. (see Chapter 2 for more). 1990. 1991. Although some believe that these philosophical beliefs are irrecon. Van Maanen et al.S. 1970). 1991. Van Maanen. is a way of investigating G. Guba & strengths and weaknesses of the various strategies may overlap.s. procedures will largely dominate the remainder of this book. Note that. yes. General Accounting Office. some experiments (such as studies of psychophysical perceptions) is that program sponsors--rather than research investigators alone-may and some survey questions (such as those seeking categorical rather than have the prominent role in defining the evaluation questions and relevant data numerical responses) rely on qUalitative. is to consider all the strategies in a pluralistic fashion-as These distinctions have produced a sharp debate within the field of evaluation part of a repertoire for doing social science research from which the investi- research. George. see Agranoff & Radin. one of the best written and most interesting case studies is about As a related but important note. detailed observations The chapter also has attempted to distinguish the case study from alternative as a source of evidence. Third. p. finally. Yin. occurred. such as political science and public administration. single. approach.~ " I Certain other features of the case study strategy are not critical for defining cilable. 1989.

i. that you had successfully conducted your A Journalistic Case Study case study). I. and the book·s appeal lies in its piecing together of fact after school course). each piece adding up curiously and then potently to an explanation for the different than research case studies. or experimental (but not The "case. and (c) exploratory case studies. the object must research. Examining case studies used/or teaching purposes: Obtain a copy of a case questions: How did the coverup occur.g. Identify the specific ways in which this type of "teaching" case is fact. to a three types of case studies used for research (but not teaching) purposes: (a) explana- best-seller." Further. if any. descriptive. The book is dramatic and sus- penseful. and serendipitously represents a 3. even though case studies have traditionally been considered to be "soft" not to be . orgam- EXERCISES zations. coverup. a survey. to do such research in the first place. Robert Stake (1994) has yet another approach for defining case studies. Describe the ways in which the findings of this study are Nixon administration more generally. the significance of your findings? Would you have advanced some major theory? Would you have discovered some- Although public memory of President Richard M. to a colleague. Does the teaching case cite primary documents. Now assume that you were actually able to answer these BOX 3 questions with sufficient evidence (i. Locate a re. What aspects of the question. historical. Defining a case study question. Does it advance some major theory? Has it discovered something rare? complex set of events that occurred in the aftermath of the burglary. He considers them hard. management survey. economic analysis). or an experiment (but not a case study) in order to answer this question. we must all work hard to overcome the problems of doing case stody research. If the case involves a significant public event and an appealing 5. people. perhaps you should consider receded. as in All the President~ Men. or exploratory? Why? Finally. Defining different types ofcase studies usedfor research purposes. "significant" case study questions. Identifying "significant" qu£stions in other research strategies. Develop a question that would be the rationale for a case study you ntight conduct. Every study of entities qualifying as objects (e. the "softer" a research strategy. Bernstein and Woodward continually confront the reader with two "how" and "why" 4. (b) descriptive case studies.) fascinating account of the Watergate scandal. Define the explanation. Compare the situations in which these different types of case studies would be most applicable. relies on solid journalistic methods. Diifjning. Name a topic you thiok is worthy of making the subject of a case study. or even the case study) methods. Bernstein and Woodward's All the President~ Men (1974) remains a redefining the major questions of your case. This definition is too broad. Case study research is remarkably 1."8 melhodological choice but a choice of object to be studied. now imagine that you could only do a history. by skill or dispo- sition. tory or causal case studies. Rather.e. regardless of the methodology used (e. NOTE including the recognition that some of us were not meant. the ingredients may add up.g.. could not be answered through these other research strategies? What would be the distinctive advantage of doing a case study to answer this question? 2. or display data? Does the teaching case have a conclusion? What Establishing the how and why of a complex human situation is a classic appears to be the main objective of the teaching case? example of the use of case studies.g. common design for case studies. However. and then name a case study you would like to conduct. and countries) would then be a case study. and why did it occur? Neither is study designed for teaching purposes (e. Identify the three major questions your case .. contain evidence. the harder it is to do. be a "functioning specific" (suclus aperson or classroom) but nota generality (such as a policy~.. Paradoxically. the case is the "coverup. is not the Watergate burglary itself.." in this book. the chapter has discussed some of the major criticisms of case study research and has suggested that these criticisms are misdirected.16 CASE STUDY RESEARCH INTRODUCTION 17 study would try to answer. Instead of doing a case study." a significant. a case in a textbook used in a business answered easily. Would it be explanatory. search study based solely on the use of survey. How would you justify. Nixon's resignation has thing rare? (If you are unimpressed by your answers. whether done by journalists or social scientists. psychological experiment.

1966). as in ~eslgnmg any other type of research investigation. In identifying the research strategy for your research project. the design is the logical sequence that ~e deve~opment of this research design is a difficult part of doing case connects the empirical data to a study's initial research questions and. each of plan that thes~ cho~ce. a comprehensive "catalog" of re. ~our major types of designs are relevant. p. In the most elementary sense. about these questions.and multiple-case studies. and there is some set of conclusions (answers) ~onsIder:'tions as the assIgnment of subjects to different "groups. As a vanous response measures (see Cochran & Cox. There are no textbooks for getting from here to there. Fisher. if not explicit. One pitfall to be avoided. Colloquially. Chapter I has shown. (b) internal validity (for explanatory or design still may be considered flawed. Between "here" and ''there'' may be found a number tion of dIfferent stimuli or experimental conditions. gUIshes between holistic and embedded designs. Although these designs GENERAL APPROACH TO wiIl need to be continually modified and improved in the future. postlest-only design as a quasi-experimental of any deSIgn: (~) construct validity. It is a logical madel of proof that allows the researcher to draw I L . such as conclusIOns to be drawn) to the initial questions of a study. however. 6). The next task is to design your case study. 1986. UnlIke other research strategies. 96). In fact. analyzing. mately. Nor have there emerged any common designs-for example. study was but one type of quasi-experimental design (the one-shot postlest- F~r case studies. including the collection and analysis of relevant data. In a laboratory experiment. I" . (c) external validity. 1957. ' design. The book. and interpreting 18 observations. to its conclusions. strategy that has its own research designs. research design. IS based on the unit or units of analysis to be covered-and distin. If not explicit. mc s t udy has an Imp IClt. chap. ~e second parr. 1935. is to consider case study designs to be A research design is the logic that links the data to be collected (and the a subset or variant of the research designs used for other strategies. the case study is a separate research I~vestigator should deal with these four aspects of quality control is summa. when you should select the case study strategy. identification with the one-group post-test-only design" (Cook & CampbeU. which summarize the various research designs for quasi-experimental situations. scholars incorrectly thought that the case .t parr. and (d) reliability.:. The first parr o~ cate~ories consists of single-case and multiple-case lowing statement appearing in a revision on quasi-experimental designs: deslgn~. research IS needed. Every emp' . The c~ study investigator also must maximize four aspects of the quality In other words. case studIes only). or the identification of of major steps. case study research designs have not been codified. or research design Every type of empirical research has an implicit. For the longest time. nzed ID Chapter 2 but also is a major theme throughout the remainder of the Unfortunately. How the recognized as something different. covering such design questions to be answered.earch designs for case studies has yet to be developed.teg~es. the one-shot. SImIlarly. ulti- studies. al experiments. Sidowski." the selec. following a 2 x 2 only design). where here may be defined as the initial set of ?t lIke those the biological and psychological sciences. For this purpose. cited in summary definition. but the case study has now been ~ausal. 1979. there are not even textbooks like the well-known volumes guides the investigator in the process of collecting. 1957. whICh can occur in combination with either of the "Certainly the case study as normally practiced should not be demeaned by fi. as opposed to other Definition of Research Designs stra. DESIGNING CASE STUDIES 19 2 by Campbell and Stanley (1966) or by Cook and Campbell (1979). following chapter therefore expands on the new methodological ground broken by the first edition of this book and describes a basic set of research designs for doing single. with the fol- ma~. a research design is an action plan s. ''panel'' stud- Designing Case Studies ies-such as those now recognized in doing survey research (see Kidder & Judd. This misperception has finally been corrected. a plan. another textbook has described a research design as a Cochran ~ Cox. in their DESIGNING CASE STUDIES present form they will nevertheless help you to design more rigorous and methodologically sound case studies.s reflects an important logical connection to the issues being studIed.

Philliber. what data to collect. "case" may be an individual. Although the substance of your questions will vary. of such roles as juvenile delinquents or derelict men." "where. This ftrst component has already been described in Chap- and the individual is the primary unit of analysis.o~ld still be needed to help identify the relevant information about thIS mdlvldua/ or vant research strategy to be used. assume that your research. You also can ~agme case studies of clinical patients. on the topic of intero~g~iza­ research. Jennifer Platt (1992a. be appropriate for "how" and "why" questions. Copenhagen Business School. This is the condition-which exists in ex~ments. 21 20 CASE STUDY RESEARCH DESIGNING CASE STUDIES inferences concerning causal relations among the variables under investigation. m the ClasSiC case study. This third component is related to the fundamental. "cases" might be included in a multiple-case study. a 3. lem of defIning what the "case" is-a problem th~ has plaguc:<l many mves- 2. Chapter I sug- gested that the form of the question-in terms of "who. Nevertheless. Your research ganizations collaborate because they derive mutu~ be~efIts.:. . dealing with at least four problems: what questions to study.. ProPOSI~O~S . however. Denmark. you might thl~ that o~­ example. not in your work plan. Borum. for example." Every exploration. complete your study by examining only one organization. Schwab.pes of leaders. each proposition di- rects attention to something that should be examined within the scope of the Another way of thinking about a research design is as a "blueprinf' of study.ome propoSitions deals with a logical problem and not a logistical problem. 77-78. and several SUC? individuals ?r "how. these ''how'' and "why" questions do not POl~t to does not address the initial research questions. tigators at the outset of case studies. also ~glDs to Such questions can be answered only if you collect information directly from tell you where to look for relevant evidence (to defme and ascertam the ex- the other organizations and not merely from the one you started with. pp." relevant individual would be collected. Copen. The case study strategy is most likely to . suppose you want to study a single organization. a research design what you should study. however. have to do with the organization's relationships with tion. some studies may have a legitimate reason for not havmg accurate conclusions about interorganizational partnerships. should still have some purpose. Can you imagine how you would ask for support from Queen Isabella to do your For case studies. 1991. capturing what you are really mterested Note that a research design is much more than a work plan. in addition to reflecting an important theoretical ISSue (that oth~r m- other organizations-their competitive or collaborative nature. In each situation. study questions. so your initial task is to clar- The research design also defines the domain of generalizahility. you cannot draw At the same time. The main in answering. a manu- F. began with the question: How and why do organizations data are relevant. fIve components of a research design are especially important: exploratory study? Unit of anlliysis.. ThIS propO~I­ questions. emphasis added) Study propositions. For instance. personal communication. the design for an explora~ory study ~hould rn: state a purpose. As for the second component. "exploration. facturer and a retail store collaborating to sell certain computer products)? hagen." "what." and "whY"-provides an important clue regarding the most rele. the logic linking the data to the propositions. if any. an individual person is the case bemg studIed.prob- 1. the criteria for interpreting the findings. Instead of stating propositions. whether ify precisely the nature of your study questions in this regard. 1992. the obtained interpretations can he generalized to a larger population or to different situations. . Only if you are forced to state s. 1992b) has. a study's questions. no~ h?w 4. and the early case studies in the Chicago school of SOCiology were bfe ~ISto~es 5. and how to analyze the results (see collaborate with one another to provide joint services (for example. as well as the criteria by which an exploration will Judged Components of Research Designs successful. led you to the case study as the appropriate strategy i~ the purpose of the design is to help to avoid the situation in which the evidence fIrst place. of exemplary students. its unites) of analysis. that is. The outcome could have been and the other research strategies alike-in which a tOPIC IS the subject of avoided if you had developed an appropriate research design in the fIrst place. For instance. (Nachmias & Nachmias. or of ~n ty. its propositions. This is a flaw in any propositions. your research design. Consider the analogy in BOX 4 for exploratory case studies. In this sense. what tiona! partnerships. Information about each ter I. centives for collaboration do not exist or are unimportant). As a simple will you move in the right direction. 1980). These "how" and "why" questions. For instance. If you tent of specifIc benefIts to each organization). & Sarnsloss. s~rveys.

the book also exploratory case study.:. Beware of these types over another. his exploration began with some rationale and direction conceptualization of the computer and ends when the engineering team relin- even if his initial assumptions might later have been proved wrong (Wilford: quishes control of the machine to Data General's marketing staff. For example. The more a study contains specific propoSItions. investigators have confused case studIes of ne~ghborhoods ~Ith tion of the program. see BOX 5a). How a general As a general guide. . differen! definition. Without such propositions. you want to study the role of the United States in the same phenomena Street Comer Society (Whyte." which is impossible to do.tion. ~ other the case) is related to the way the initial research questions have been defined. Selection of the appropnate umt about programs. rfyour questions do not lead to the favonng of one umt of analysIs these ~IDgle case~ In s?ciology and political science.': For example. including the importance of 9. it does not need ~ollect "everything. The unit of neighborhoods when in fact they are case studies of s~all groups (note of analysis for your' case study might he a country's economy. about the implementation process.t the dynamics of a small group---the engineering team? The answer IS cnlIcallfwe want to understand how the case study relates to a broader body of knowl- edge-that is. an economic policy. . Any case study of such a program would therefore have case studies of small groups (for another example. & Sjoberg (1991) contains some classic examples of questions. is about the development of a new his "exploration" of the New World. The book begins with the initial tered them. the. Is the case study about the minicomputer. He also had This easy-to-read book describes how Data General s engmeenng team some (mistaken) criteria for recognizing the Indies when he actuallyencoun. phenomena can be quite different than how a small group copes WIth these Suppose. the definition of the unit of analysis (and therefore of area such as a neighborhood copes with racial transition. and he had some rationale for directly with one produced by Digital Equipment Corpo. M?st actors.:owing of the relevant data. Peter Drucker (1986) has written a provocative essay about 1 in Chapter 1 of this book) and Tally:' Comer (Liebow. Each unit of analysis would call for a slIghtly ~Ifferent defined than a single individual. have often been mistaken for beIDg case studIes "capital movements" independent of the flow of goods and services. illustrates a fundamental problem in doing case studies-that of defining the unit 0/ analysis. or .is it. even though . upgrading. for instance. nor does it. 1992). going westward (Why not south? Why not south and then east?). also a best-seller. 19431l955-a1so see BOX world economy. an industry in that in neither book is the neighborhood geography descnbed. frequently. research design and data collection strategy. with a small group in organizational studies. confusIDg an IDnovauon to confront these conditions in delineating the unit of analysis. the proposi. an investigator might he tempted to dynamics topiC. tio. 1967~so seeB~X fundamental changes in the world economy. he had to have some reasons for asking minicomputer produced by Data General Corporation.ase. a case study of a specific program may reveal (a) Sometimes the unit of analysis may have heen defined one way. In short. provide an answer. the "case" also can be some event or entity that is less well between two countries."c. 23 22 CASE STUDY RESEARCH DESIGNING CASE STUDIES BOX 4 BOX Sa "Exploration" as an Analogy for an What Is the Unit of Analysis? Exploratory Case Study The Soul of a New Machine (1981) was a Pulitzer prize-winning book by When Christopher Columbus went to Queen Isabella to ask for support for Tracy Kidder. Such topics already represent a vast m~. This same degree of rationale and direction should underlie even an·· The book is an excellent example of a case study. even vanations ID program definition. Because the book is not an acadenuc study.ns In studying these individuals might involve the influence of early chIldhood or the role of peer relationships. whether to generalize to a technol~gy topi~ or to a group individuals. and (b) program components that existed prior to the fonnal designa. The book. Orurn. this chapter). for example. and about organizational of analysis results from your accurately specifying the primary research change: Feagin. the more it will stay within feasible limits. However. intended to compete for three ships (Why not one? Why not five?). or the trade or capi~ flow Of course. the world marketplace. your questions are probably either too vague or too numerous-- of topIcs-none IS easIly defined in tenns of the heginning or end points of and you may have trouble conducting your case study. ~bou. to. depending upon the perspective of different though the ~enomenon heing studied caI1s for a. Case studies have been done about decisions. invented and developed the new computer.

the begmDlng and end of the case.t of analysis become important. In this ~~YSIS. conditions in psychological experiments (which is the way that hypotheses and data are connected in psychology). each case study and unit of analysis either should be similar to those previously studied Tw~ of the cases appear similar but in fact have different main units of by others or should deviate in clear. s~natlDg manner-in the field of world trade. If the two potential patterns are outsIde It (the context for the case study) Similarly I'fth . on the level of inquiry. regarding the impact of the new speed limit law). but without labeling it as such . I' . . whereby several pieces of information from the same case may be related to some theoretical proposition.tIy competitive. Most Busmess Battles Shapzng America's Future (1989) .geographlc area. the persons to be included within the u the What Campbell did was describe two potential patterns and then show that ml~edI~te tOPIC of the case study) must be distinguished from tho. and criteria for interpreting the findings. :tive r umt-Samsung's development of the microwave Oven as an illus. ?nce. manner. . about the Korean finn Sarnsung.J~~: defi:~ though the entire study consists of only a singlecase(the state of Connecticut). TIy to explain to that person what questions tion of the fatality rate. One final point needs to be made about defining the case and the unit of ~ Magaziner and ~ark Patinkin's book The Silent War: lnsidethe Global analysis.y elps the reader understand a real-life situation of intema: this reason. researchers will want to compare their findings with previous research. ':fi' " ecaselsa utservlces m a spec~ c. the previous literature therefore also can become a guide for defining nO~~eIO?ment is part of the context. In a related article on one type of the small gro ups I'IVed'm a small area WIth . the criteriafor interpreting a study'sfindings. Linking data to propositions can be done any number of ways. ~a11 ~oup. the pattern- whose ~IStriCt boundaries do not coincide with the area Finall ~ aim es matching technique is a way of relating the data to the propositions.of ~al~sis-is actually an illustrative example of how national P o Cles ect IOreIgn lDvestrnents. .: IC s . is a case stud of the cnti~al poliCIes that make the finn competitive. clear neighbo-'-ood . If the unit of an~~~i:7. ) pattern-a time-series pattern--Campbell (1969) illustrated this approach BOX 5b h HI Imp IcatlOns . but none These two cases show how the definition of the main and embedded . over a number of years before and after the legal y~U are trymg to answer an~ why you have chosen a specific case or grou change.. casestlas ~dwa~ ~ answen~g those questions.' owever. : of traffic fatalities. f?r mslance. even any top~c ~at might be chosen. as well as the definition of the contextual events surrounding t~ruts to": llDIts. All of these types of questions need to be This article also illustrates the problems in dealing with the fifth compo- nent. search. presents a good example of how units of anal be defined m a more di . showed unsystematic fluctuation rather than any marked reduction. the key definitions should not be idiosyncratic. decisions need to be made about public servic proposition. to . 24 CASE STUDY RESEARCH DESIGNING CASE STUDIES 25 BOXSb considered and answered to define the unit of analysis and thereby to deter- A Clearer Choice Among Units of Analysis mine the limits of the data collection and analysis. pertaining to the role of the available research literature.::'::~~ studies. and mcorrec y I enttfymg the UOlt of analysis. considered rival propositions (an "effects" proposition and a "no effects" . Rather. one good practice is to discuss the state law limiting the speed to 55 miles per hour. Understanding Ko~ eco. .To reduce the confusion. However. bo .owfo(~e the data matched one better than the other. further examina- potential c~ WIth a colleague. and the case study also contains an the case and unit of analysis. specific time boundaries ~ n. These components represent the data analysis steps in case study re- emli~dedaftiurut. The Apple computer factory ex.1). has become as precisely defined as the assignment of subjects and trealInent of :uuuyses.?. This may help you to avoi~ A simple eyeball test was all that was needed to show that the actual pattern ? looked unsystematic rather than following a downtrend (see Figure 2. One promising approach for case studies is the idea of "pattern-matching" described by Donald Campbell (1975). . The main unit of analysis is likely at the level being addressed by the main study questions. depends. Campbell first showed how the annual number of traffic . operationally defined ways. Campbell's data matched I 1 "-. about the development of an Apple computer Linking data to propositions. Most mves~gators wIll encounter this type of confusion in defining the fatalities in Connecticut had seemed to decline after the passage of a new umt of analYSIS . OIY Smgapore. ~~ case. for Each ase tud h presents mne case studies uDna economic competition. and a research design should lay the foundations for this analysis.. the general definition of the case has been established oth I ' fi thus Campbell concluded that the speed limit had had no effect on the number cations m the uni. yses can In his article.Is 10 fact a case study of Singapore's critical li' th The fourth and fifth components have been the least well developed in case make the cO. ' fact ~nod~ct The ~th~r case.

. 27. the relevant field contacts YEARS depend upon an understanding---<>r theory---<>f what is being studied. the one pattern much better than they matched the other. the complete research design should not only indicate what data are not sufficient for successful implementation. 1985. Using the same case.. Currently. students . least two rival Propositions. The statement presents the nutshell of a theory of MIS implementation- One hopes that the different patterns are sufficiently contrasting that (as in that is. (Markus. 210 221 II "- IJ . a ''no effects" pattern wrongly think that by using the case study method. and (c) Its UOlts of analysIs. An Example of Pattern-Matching Theory development.. 1982) and "grounded theory" (Strauss & Corbin. and not just overlay the new MIS on the number-the number of fatalities for that year-for which one could not old organizational structure.. theoretical propositions at the outset of an inquiry. 1969) . As a result.d by (a) a study's questions.. . 26 CASE STUDY RESEARCH DESIGNING CASE STUDIES 27 a prior. prior to the conduct of any actual obse. and they have been encouraged zoo . 1983) to be col~ected:--as indicate. But how close does a simplest ingredient of a theory is a statement such as the following: match have. there IS no precise way of setting the criteria for interpreting these types of findings. A research design should include five components. (Markus. The design also should tell you what is to This second statement presents the nutshell of a rival theory-that is. 'JYpically.m. Van Maanen et al.J to make their ''field contacts" as quickly as possible.1. Using a case study on the implementation of a new management information system (MIS) as an example (Markus. that be done after the data have been collected-as indicated by (d) the logic MIS implementation fails because of the resistance to change on the part of . because each data point in the pattern was a single zation was able to re-structure itself. design phase is essential... Van Maanen. theory development as part of the SOURCE: COSMOS Corporation. methods such as ethnography (Lincoln & Guba. 1988. Among other considerations. 1986. to. these related methods deliberately avoid specifying any :. 1983). Nor would a statistical test The case study will show why implemeniation only succeeded when the organi- have been possible. is one point of difference between case studies and related (Campbell. Although the current state of the art does not provide detailed guidance on the last The case study will also show why the simple replacement of key persons was two.. J. Figure 2. 1983) ~alculate ~ variance and could not conduct any statistical test.Uons. be so as to be considered a match? Note that Campbell did not do any statistical test to make the comparison. 1990).. hi: The Role of Theory in Design Work Covering these preceding five components of research designs will effec- vJI" 1234117 12'4'87 tively force you to begin constructing a preliminary theory related to your YEARS YEARS topic of study. For case studies. propositions I linking the data to the propositions and (e) the criteria for interpreting the • Uno eft'ecb" pattern findings . No gnidance could be I 'Sl'S2 '51'14 'SS'56 '17 'S8 'SD more misleading. they can proceed quickly into the data collection phase of their work. (b) its propositions.•. whether the ensuing case study'S purpOSe is to de- velop or to test theory.. This role of theory deVelopment. that organizational restructuring is needed to make MIS implementa- Campbell's case) the findings can be interpreted in terms of comparing at tion work. data collection. an additional ingredient might be the following statement: Summary.

and the ruling class exhibited immoral. lying the answers. as noted earher financial markets and the real economy of goods and services. He points to the ''uncoupling'' between the primary products (raw materials) economy and the industrial economy. there were bitter class to on such issues as (a) the purpose of the descriptive effort. Using a Metaphor to You can see that. . and yet others explaining the exploration. revolutions is the topic of Crane Brinton's famous historical study-:-The ~t­ 'This theory should by no means be considered with the formality of grand ()T1IY of a RevolutWn (1938). government machmery was meffi- cient. 1984). others focusing on specific industries. n?~ that the MIS example For yet other topics.Ideas WIth col- realistic range of topics that might be considered a "complete" description of leagues or teachers. Drucker claims that the world economy has changed significantly from the past. In general. or inept behavior (or to the collection of any case study data is an essential step in doing case studies. In this sense. rather than relying solely on this "factors" approac~ to description. and that thiS IS but one type of theory available literature wiII provide no conceptual framework or hypotheses of that can be the subject of study. Russia. to overcome the barriers to the- In other situations. you should be aware of the full range of theones theoretical base-and research design-for your study. However. as the author's purpose is not. To test these with the illustrative case in BOX 4. and asking yourself challengmg quesl10ns about what what is to be studied. you should try tu prepare for your case study by ?oing BOX 6. the five components of the needed research design. a similar uncoupling between theoretical statements. and revolutions as to determine whether they followed smular courses. and also BOX 1 for another example). this requires theoretical propositions. Other types of theories for you to consider note. even an exploratory case study should be propositions might require different studies. Then. as these two initial ingredients are elaborated. why you are proposing to do the study. ~b) th~ pu~ose of uncouplings. and (c) the criteria by which the exploration wIIl be Judged the plight of specific countries. rising to a critical poi~t and work for designing a specific case study. discussing your ~PIC and . and (c) the likely topic(s) that will be the essence of the you are studying. existing works may provide a rich theoretical frame. s~ much to explrun the Rather. Economy" (1986) is an exceptional source of theories and hypotheses. France-all underwent similar courses of events during their major political the complete research design embodies a "theory" of what is being studied. to describe the ebb and flow of events m the economic development. descriptive manner. propositions. and your concern should focus such things as reviewing the literature related to w~at yo~ would . The author 1989). Nevertheless. nor are you being asked to be a masterful theoretician. the complete research design will The "cross-case" analysis reveals major similarities: All societies were on provide surprisingly strong guidance in determining what data to collect and the upgrade.e lD a theory in social science. theory development prior als deserted from positions of leadership. including the rationales under. adeptly uses the cyclic pattern of fever and chills. For instance. For some topics. the inte~lectu­ the strategies for analyzing the data For this reason. will help you go a long way toward developing the needed As a further reminder. and what you hope description. economically. Whether four countries-the American colonies. Each different study would likely call for a successful. and the uncoupling between characteristic of being an "exploratory" study. and ings-that is. from a fever as a way of describing the pattern of events over tIme. BOX 6 ment for implementation to succeed. and that the replacement of such people is the only require. the simple goal is to have a sufficient blueprint for your study. to learn as a result of the study. and any new empirical study is likely to assume ~e low labor costs and manufacturing production. Illustrative types of theories. (b) the full but study (also see Cooper. and criteria for interpreting the find. Drucker's theoretical framework would provide gUidance for designing these studies and even for collecting relevant data. allthree). L . the stated Develop Deseriptive Theory ideas will increasingly cover the questions. some focusing on the different preceded by statements about (a) what is to be explored. England. that might be relevant to your study. dissolute. different unit of analysis. theory development takes time and can be difficult (Eisenhardt. and the iIIustrates MIS ''implementation'' theory. Such a knowledge base does not lend itself to the development of good include the foIlowing: !. logic connecting data to propositions. the existing knowledge base may be poor. If you are interested in international followed by a false tranquility. Peter Drucker's ''The Changed World four revolutions. units of analysis. for instance. Tracing and analyzing these events IS don. Good answers to these questions. the author also develops the metaphor of a h~ body suffermg However. 29 28 CASE STUDY RESEARCH DESIGNING CASE STUDIES individual people. the appropriate theory may be a descriptive theory (see ory development.

should be only facilitate the data collection phase of the ensuing case study. for instance. such as federal. analysts have frequently confused the theory of the program formulas for determining the confidence with which generalizations can be (e. state. 1987). The ap- considered like multiple experiments (or multiple surveys)..2.1 I- work teams.:". Theory development does not individual case srudies are to be selected as a laboratory investigator selects the topic of a new experiment. 1 1 making theory (Carroll & Johnson. theories of individual development. personality. Decision- LEVEL ONE ~. 1992). cOgnitive the • . international behavior. the analysts unfortunately recommend managerial steps (e. organiza. Rather... JI r------- ." in the case study results will occur.2. or social groups.. This role of theory has been characterized which a previously developed theory is used as a template with which to throughout this book as "analytic generalization" and has been contrasted compare the empirical results of the case study.. individual perception. I rIvol th• . an inference is made about a population Analytic generalization can be used whether your case study involves one (or universe) on the basis of empirical data collected about a sample. SOURCE: COSMOS corporation.. this is the most common way of generalizmg ers want to know the desired substantive steps (e. supervisory-employee coordination. I behavIOr. and interpersonal networks • Organizational theories-for example. can involve individu- als.g. the logic of replication and the distinction between statistical f L . thIS Let us take the more commonly recognized way of generalizing-"statis- tical generalization"-flfst. a common topic of Figure 2. This mismatch can be avoided by giving closer A fatal flaw in doing case studies is to conceive of statistical generalization attention to the substantive theory. ••• rt--t----.. 31 30 CASE STUDY RESEARCH DESIGNING CASE STUDIES • Indivi~ua1 theories-for example. hire a good project director).g. Lavrakas. This is or several cases. Generalizing from case study to theory. rival theory. replication may be claImed.g. learning and disability. although it is the less relevant one fordoing case type of generalization is shown as a Level1Wo Inference in Figure 2. theories of family functioning. studies. informal groups.1 This method of generalizing studies.g. organizations. 1987). Fowler. Making Inferences: Two Levels case studies is the evaluation of publicly supported programs. describe a newly effective when doing surveys (e. are shown with another way of generalizing results. may be considered yet more potent if two or more cases suppo~ the same theory but do not support an equally plausible. Further. how to make education more effective) with the theory of program made. 1987). This is because cases are not "sampling units" and should not be chosen for this reason. cultural institutions.-----------I j L--.• SURVEY ! CASE S'l'UDY Elr1'E1UMENT Harrison. Graphically. depending mostly upon the size and internal variation within the implementation (e. and inter- 11mp~t I I=:'u:'!. theories of urban development."! I LEVEL personal interactions TWO • Group theories-for example. how to install an effective program). theories of bureaucracies. Multiple cases. the development of a theory of how a program is supposed to work is essential to the design of the evaluation but has been commonly underemphasized in the past (Bickman. Moreover. In this situation. I •I tional structure and functions.g. and it is an integral curriculum). According is commonly recognized because research investigators have ready access to to Bickrnan..2. (though not the only) part of generalizing from experiments. known as "statistical generaliza- to support the same theory." Understanding the distinction between these two types of generali- zation may be your most important challenge in doing case studies.. universe and sample. or local programs. Under these propriately developed theory also is the level at which the generalization of circumstances. and interorganizatioual partnerships • Societal theories-for example. excellence in organizational performance (e. the method of generalization is "analytic generalization. 1988. As another example. Wberepolicymak... in this sense.. as the method of generalizing the results of the case. The empmcal results tion. In statistical generalization.g. technological developmen~ and marketplace C~=C8 I -==y I I~ I functions Other examples cut across some of these illustrative types. Iftwo or more cas. which shall be later referenced as single-case or multiple-case shown as a Level One ['lference in Figure 2.

. ...feh'nB' -. nt ..... not only is an immense aid in defining the Figure 2.... whereby certam :on- I Accounting Office... you should aim for Level7Wo Inferences when doing .w.-l.... have been commonly used to establish the quality of any empirical social research.. you also can judge the quality of any given design according to certain logical tests.. or exploratory. Rather than resisting such a requirement. Case Study Tactics for Four Design Tests appropriate research design and data collection but also becomes the main SOURCE: COSMOS Corporation. .... vehicle for generalizing the results of the case study. The main point at this juncture is that you should try to aim toward analytic generalization in doing case studies. in fact requires the develop- ·-valJdltF . and not for des<:"ptive . Summary. 32 CASE STUDY RESEARCH DESIGNING CASE STUDIES 33 and analytic generalization will be covered in greater detail in the discussion of multiple-case study designs..u...... the four tests also are relevant to case study research. Because case studies are one form of such • External validity: establishing the domain to which a study's findings can be generalized empirical research. of . IS that the several tactics to be used in dealing with these tests should be applIed throughout the i......... cIaaIa. case studies.. Figure 2.26-29): Because a research design is supposed to represent a logical set of state... as distinguished from spunous relationships Four tests. Credibility.3. Interaal.. .. deScriptive. a<eport in terms of Figure 2.. with the same results tactics for dealing with these four tests when doing case studi~s. Because of the imPOrtance of this distinction between the two ways of generalizing.. 1986... -aaecaee.. -.. aae malUpJe aoarcN . and data dependability (U..In_.) explicit attention.-.. In other words... the tests I. do pa...... oapJoaalloa-lndJ..S.... ... The use of theory.. aae repUcatI. a good case study investigator should make the effort to develop this theoretical framework.... CRITERIA FOR JUDGING have been summarized in numerous textbooks (see Kidder & Judd. For case studies..3 lists the four widely used tests and the recommended case study tactics as f well as a cross-reference to the phase of research when the tactic is to be This list is much more complex than the standard "validity" and "reliabil- used.. Because the four tests are common to all social science methods... """""". ditions are shown to lead to other conditions.... I • Construct validity: establishing correct operational measures for the concepts ments.. ~d e:wh item deserves book. -- data ColleeUOIl ment of a theoretical framework for the case study that is to be conducted. data aaaI. f L .earc:h III wIdcIt taetlc OCC1ll'll uta eoUeelioa. .... you will find repeated examples and discussion throughout the -. THE QUALITY OF RESEARCH DESIGNS pp. in doing case studies. and you should avoid thinking in such confusing terms as "the sample of cases" or the "small sample size of cases..t.I.f • Reliability: demonstrating that the operations of a study-such as the data Therefore. phaae of re. or exploratory studies): establishing a causal relationship........t_ _.. draft caM ._oa data eolleetloa respondent in a surveyor a single subject in an experiment.. data coJlectioa whether the study is to be explanatory...... -aaaIJoIo remainder of this chapter as well as in Chapter 5.. Iogl.2. 1990). an Important revelallon.. Concepts that have been offered for these tests include trust- worthiness.w.. This subsection has suggested that a complete research design... no matter rellabUflJr -.Dty ot_ .-..." as if a single case study were like a single .. an important innovation of this book is the identification of several collection procedures can be repeated. covering the five components described earlier....ta . . conf1flDability.udlty _ .. .. General I being studied • Internal validity (for explanatory or causal studies only. (Each tactic is later described in detail in the relevant chapter of this ity" notions to which most students hav~ been exposed. lop-- ... however..

have changed their character. 1\vo related analytic tactics. dealing with spurious effects.3 shows for doing case studies. three tactics are available to increase construct validity. threat to internal validity. People who have been critical of case studies often point to the fact that a case study This second test has been given the greatest attention in experimental and investigator fails to develop a sufficiently operational set of measures and quasi-experimental research (see Campbell & Stanley. As one set of suggestions. based on interview and documentary evidence collected as part I. alrea. 1966. t to deal with the overall problem of making inferences and therefore the specific problem of internal validity. L . In the simplest example. This is especially true in doing case studies. If the investigator incorrectly concludes that there IS a causal cannot tell whether the rec. Numerous "threats" to validity have been identifIed. ." areader to event y. also relevant during data collection (Chapter 4). tern of urban services. In this is to have the draft case study report reviewed by key informants (a procedure sense. r " 34 CASE STUDY RESEARCH DESIGNING CASE STUDIES 35 subsequent conduct of the case study. The second step now demands that you I tify. described further in Chapter 6). As Figure 2. In a manner encouraging convergent lines of inquiry. studies.dy touched to be the standard measure used in the FBI Uniform Crime Reports) as your r· upon but to be described further in Chapter 5. operational events that constitute "change. a case study borhoods. for case study research. Cook & C~p­ that "subjective" judgments are used to collect the data. and not just at the middle-income residents in "gentrifying" neigh. the specific tactics for achieving this result are difficult to ~den­ study the rise in neighborhood crime. also are described in Chapter 5. analysis."s it appear to 2.:. maInly as studying "neighborhood change"-a common case stody topic. The third tactic The third test deals with the problem of knowing whether a study's findings are generalizable beyond the immediate case study. concerns have arisen over how certain urban neighborhoods cover this topic. or Second. the research design has failed to deal with some investigator's impressions only.:s ":fId their consequences. internal validity is a concern only for causal (or explanatory) case type~ of chang. suppose you satisfy the fIrst step by stating that you plan to I However.3 shows that the analytic tactic of pattern-matching. Demonstrate that the selected measures of these changes do indeed reflect the be airtight? A research design that has anticipated these questions has begun specific types of change that have been selected. changes in the pat. Me an example such bell. For example. given that large validity. only two points need to be made here. "design work" actually continues beyond the initial design plans. extended to the broader problem of making inferences. and this tactic is relevant External Validity during data collection (see Chapter 4). Perhaps this is not a valid measure. Any number of case studies have examined the First. may be the turnover from low. without any prior speci. A second tactic is to establish a chain of evidence. or experi- racial turnover. Thus to cover two steps: an investigator will "infer" that a particular event resulted from some earlier occurrence. in which an investigator is trying to determine whether event x led fIcation ofthe slgmfIcant. ments). Is the inference correct? Have all the rival explanations and original objectives of the study) and possibilities been considered? Is the evid~n~e convergent? D. However. housing deterioration and abandonment. To meet the test of construct validity. because so many textbooks already Over the years. which are not concerned with making causal statements. However. the concern over internal validity. an investigator must be sure involves an inference every time an event cannot be directly observed. surveys. Select the specific types of changes that are to be studied (in relation to the of the case study. shifts in a neighborhood's economic institutions. is one way of addreSSIng Internal measure of crime. Basically. . The fIrst is the use of multiple sources ofevidence. explanation-building and time-series proportions of crimes are not reported to the police.orded changes in a case study genuinely reflect relationship between x and y without knowing that some third factor-z- critical events in a neighborhood or whether they happen to be based on an may actually have caused y. also justify why you might be using police-reported crime (which happens Figure 2. 1979). Note that this logic is inapplicable to descriptive or Neighborhood change can indeed cover a wide variety of phenomena: exploratory studies (whether the studies are case studies. Construct Validity Internal Validity This fIrst test is especially problematic in case study research.

( h· h is another way of dealing with reliability). ' mixed uses the need for small blocks. .0 could repeat the procedures and amve . not on "replicating" the results of one case by doing another Case study.!~::ntation Most people are probably already familiar with this final test. ~the~ an also will have identified those types of neighborhoods within which gentri.':p::''':~fI. If a study had focused on "gentrifica- tion" (see Auger. analogou~to For example."oaching the reliability problem is ~o make as many . the procedure for selecting a neighborhood for study De!U. the investigator is striving to generalize a particular set of results to some IS TheYproblem lies in the very notion of gene:aHzing. an auditor is also. Yet no s~t of cases. 36 CASE STUDY RESEARCH DESIGNING CASE STUDIES 37 if a study of neighborhOOd change focused on one neighborhood. ·caI1·nquiries were made in other locales to . In principle. one or ana er th till tands as a number of similar neighborhoods.3. theJacob's U1 ng book created heated controversy in the planning profession. This replication logic is the Same that underlies the use of experiments (and allows scientists to generalize from one experiment to another) and. steps as operational as pOSSIble and to conducttiresearch alwa s looking over your shoulder. examme replication has been made. in which a "sample" (if selected correctly) readily generalizes to a larger universe. . .~:~::an~ther tactic. the results might be accepted for a much larger th facet Of her rich and provocative ideas. I tead an analyst should try to generalize findmgs to 'theory. r r doing case studies is therefore to conduct the research so that an au Ito Without such documentation. broader theory (see BOX 7). theories about changes in all of these . even though further replications have not became the vehicle for examining other cases. been perfonned. the chapter tOPICS.:ver I of neighborhood parks. sense. case s~dy re- w IC procedures have been poorIy documented' making .. . (Note that the emphasis is on doing the same case over again. performing a redl s are followed. I ~ w . a selen the . are the results applicable to another neighborhood? The external validity problem BOX 7 has been amajor barrier in doing Case studies. 1979). In accoun ng as If someoneone and bookkeeping . tion. The book IS based mostl~on periences from New York City. 18(Note ) at .:!"ufe of Great American Cities (1961). the development as described by an earlier investigator and conducted the same Case study all Over again. theslummIng of need an UDSIUmnu·ng. urban planning suchastheroleofsldew • ero e neighborhoods would be the target to which the results could later be gener- ~orpnmdary I~ 10 . the theory of neighborhood change that led to a Case study ns way. if a later investigator followed exactly the same procedures :~~~~~~:d~:i1~~~ ~'. ti t does not attempt to select "representative expenmen . Once such Partial result. ex . with experiments) rely on analytical generalization. A theory must be tested b "Jdi of a theory of urban planrung.external reviewers ReliabiUty :7::' I search 3 will suspicious of the reliability of the case S~~y·t~ . will be discussed further in this chapter in the section on multiple-case desigus.' were is . likel to deal satisfactorily with the complaint. goal of reliability is to minimize the errors and biases in a study. Critics typically state that single How Case Studies cases offer a poor basis for generalizing. However. an~ the eory s s significant contribution to the field of urban planDlng. a where the theory has specified that the Same results should occur. to othe~ case studIes. such critics are implicitly Can Be Generalized to Theory contrasting the situation to survey research. • these issues in fact represent The generalization is not automatic. in the frrst place is the same theory that will help to identify the other cases the s~en s roach is well illustrated by Jane Jacobs in her ~amous book. As. . no matter how large. you could not even repeat your own work IL . as shown in Figure 2. Agood g\lideline roducethesameresultslfthesameproce ure d. whereas Case studies (as "representative" case or set of cases. In the past. the aggregate . Thus analysts fall into the trap of try10g to select a survey research relies on statistical generalization. The objective is to be sure that. In thIS r ability check and must be able to One prerequisite for allOWing this other investigator to repeat an earlier Case study is the need to document the procedures follOwed in the earlier Case. at the same results .. and the processes alized. new empm. This is because from ~~e case to another. . The to which the results are generalizable.. fication Was occurring. al aYys aware that any calculations must be capable of bemg generalizes from experimental results . through replications of the findings in a second or even a third neighborhood. However. Ii. This analogy to samples A mmon complaint about case studies is that it is difficul~ to generalize and universes is inco"ect When dealing with case studies. to theory. 10 essence. Yi Ie broader theorelicallSsues reflecting the single expenences of New °alks" . In analytical generaliza. Her theory. however.) The I : of ~:g:~::r:~a:r":. the later investigator should arrive at the same findings and conclusions.

and (d) multiple-case (embedded) designs. based on a 2 x 2 matrix (see Figure 2. The single case can ones. 1978). This syndrome appears to be due to some phYSIcal Injury to the whether some alternative set of explanations might be more relevant. studies is between single. This has commonly been the situation in clinical critical case in testing a well-fonnulated theory (again. on whether a single-case study or multiple cases are going to be used to address the research questions. or (in some cases) . like Graham Allison's comparison of three theories of bureaucratic establish any common patterns (Yin. A primary distinction in designing case SOURCE: COSMOS Corporation. For Instance.4). Basic Types of Designs for Case Studies Rationale for single-case designs. (b) single-case (embedded) designs. or compositional phases of the research and are therefore described in greater detail in subsequent chapters of this book.and multiple-case designs. there also can be a unitary or multiple units of embedded analysis. for the case study strategy. This means the need for a decision.and multiple-case studies reflect different design situations and that. One rationale for a single case is when it represents the extreme or unique case. Some of the tactics occur during the data collection. data analysis. though not all of the tactics occur at the formal stage of designing a case study. Yet the syndrome occurs so rarely that scientists ha~e been unable to manner. recall that a single-case study is analogous to a single experiment. and tional innovation. such patients are unable to recognIze loved lr\Ie. (c) multi- ple-case (holistic) designs. BOX 2). single-case study is an appropriate research design whenever a new person the single case can represent a significant contribution to knowledge and with this syndrome-known as prnsopagnosia--is encountered. in which a specific injury or ~isorder . farruliar faces. What Are the Potential Single-Case Designs? Figure 2. or extend the theory. The case t L . 1970. within these two types. The theory has specified a clear set of propositions single case is worth doculI)enting and analyzIng. in the field of organiza- First. in an entire field. the four types of designs are (a) TYPE 2 TYPE 4 (multiple units single-case (holistic) designs. prior to any data collection. note the analogy to psychology.may be so rare th~t :my the critical experiment). (See BOX 8 for another example. meeting all of the conditions for testing the theory. various tactics are designs designs available to deal with these tests. Four tests may be considered relevant in judging the quality of a research design. 39 38 CASE STUDY RESEARCH DESIGNING CASE STUDIES single-case multiple-case Summary. In such CIrcumstances.) many of the same conditions that justify a single experiment also justify a A second rationale for a single case is one in which the case represents an single-case study. The of analysis) rationale for these four types of designs is as follows. Four types of designs will be discussed. The matrix assumes that single. In this brain. holistic TYPE I TYPE 3 (single unit CASE STUDY DESIGNS of analysis) These general characteristics of research designs serve as a background for conSidering the specific designs for case studies.theu: own Image then be used to determine whether a theory's propositions are correct or in a mirror. To confirm. friends. In designing and doing case studies. pictures of famous people. challenge. the functioning and the Cuban missile crisis (described in Chapter I. there may exist a single case. Given visual cues alone. theory-building.4. Thus. one ~areclll~~al as well as the circumstances within which the propositions are belieVed to be syndrome is the inability of certain clinical patients to recogruze. Such a study can even help to refocus future investigations The single-case study is an appropriate design under several circumstances.

appeared to account for the outcomes. be pro- case study. such barriers had been prominently cited as the major reason that (\%7). Prior to the study. a potential vulnerability of the slOgle-case design phenomenon previously inaccessible to scientific investigation. rather than barriers. shOwed that. / even though a case study might be about a single ~ubhc ~ro~am. Tally's Comer (see BOX 9). The book is about a single group of black ~ l~ failed but that the failure could not be attributed to any barriers. Single-case designs therefore require careful mv~tigation of ~e latter-day example is Elliot Liebow's famous case study of unemployed I potential case to minimize the chances of misrepresentanon ~d to ~ax~­ blacks. to scientific observation. used such a design by focusing on a single school in their book. ~ not to commit oneself to the single case until these major concerns have been day lives. bu~ one ~t . However the units are selected. study would document the person's abilities and disabilities to detemUne the exploratory devices or such as the conduct of a pilot case that is the fIrst of a precise nature of the face recognition deficit but also to ascertain whether multiple-case study. These three rationales serve as the major reasons for conducting a single- ! (and possibly even some quantitative analyses of ~arge num. Liebow had the opportunity to meet the .bers of . could not be claimed to suffer from ''barriers to innovation. This occurs when. to unemployment and failure.L . Gross et al. the embedded UDits :uso ~Ight. at- case). There are other situations in which the single-case study may be cess" units~uch as meetings. the resuI'nng d ' eslgn i .had watershed in innovation theory. In an organizational study. For lOstance. such as is that a case may later tum out not to be the case it w~ thou~t ~o be at the Whyte's Street Comer SOciety. an innovation also inner-city neighborhood. the book. cities for a long penod of time. even though the problems ! Holistic versus embedded case studies. the anal- such conditions justify the use of a single-case study on the grounds of its ysis might include outcomes from individual projects within the pro~ revelatory nature. Neal Another rationale for selecting a single--case rather than a multiple--case Gross et al. their coping behavior. The book provides !"sigh~ into a subculture that In this manner. because few social scientists had previously had the opportunity to investigate these problems. previously described in Chapter I. formed a significant case study. A third rationale for a single case study is the revelatory case. 1985). since the study. This situation Whatever the rationale for doing single cases (and there ~ay be more ~an exists when an investigator has an opportunity to observe and analyze a the three mentioned here). innovations failed. roles. By befriending these me~. In either sltuatlon.S." In the prevailing Such was the situation in Elliot Uebow's sociological classic. A outset. the single-case study related disorders exist. and 10 particular therr sensItivIty processes. ~ese i bedded units can be selected through sampling or cluster techDlques conducted as a prelude to further study. DC. represents a has prevailed in many U.~roJ­ ects). in this school. design is that the investigator has access to a situation previously ~accessible Implementing Organizational Innovations (1971). such as the use of case studies as ! em (McClintock. The single case showed how mvestigations identification of barriers. 40 CASE STUDY RESEARCH DESIGNING CASE STUDIES 41 . though limited to a single case. the a~thor was. the literature has been much more of such topics could be done. within a single . analysts had focused on the been only obscurely understood. The same case study may involve were common across the country (as distinguished from the rare or unique I more than one unit of analysis. Tally's Comer theories. BOX 8 The Single Case Study as the Critical Case BOX 9 The Revelatory Case as a Single Case One rationale for selecting a single-case rather than a multiple-case design is that the single case represents the critical test of a significant theory. unCOVer some prevalent phenomenon previously inaccessible to scientists. stimulating much further research and eventually dominated by studies of the implementation When other investigators have similar types of opportunities and can tention also is given to a subunit or subunits (see BO. BOX I. His observations of and insights into the problems ofunemployment I covered. and to learn about their every.I mize the access needed to collect the case study evidence. the development of policy actions. living in a poor. cannot be regarded as a complete study. A farr warmng IS men in one neighborhood in Washington. or locations.on Its own. However. in these latte~ instances.X 10). Implementation about their lifestyles. The case study is therefore worth conducting because The school was selected because it had a prior history of innovation and the descriptive information alone will be revelatory.

g ] ] •• .!:::' ~~ ~ .!:l e a ..g~ a ._. <:: ~ ~ ~.• '§ . with a new research design. ~l -'i! e0 a·· ~ o > . however. ~ ] e e2 . ~'.e-• . a holistic design would have been used (see Figure 2. the largest '"• ~§:S'i B ~...1::'0 j ..~ ~= · 0 ~ ~~ '> u..u ·n . The holistic design is advantageous when no logical subunits can be ~ • '" ~8~ '5 3 j:Q 't. i:i oi study.0 0 0 focusing a case study inquiry... When a global approach ~ . f 8 : __ .- Sl ~ ~ o ~. o • . Because of this problem. do change.~ ~ ~C -Ii• :.~ ~= -Ii 's . but as the case study proceeds.. E 0 j:Q .~ ~ . e ~. At each level of analysis.: -1l ~ c~ '6 'if Il.. if the relevant research questions really . and James Coleman. iii ~6l! g 1:1 .g"" ~ 'a '"~ -c1. the smallest unit was the individual member.• 0> Both variations of single-case studies have different strengths and weak. >! . Single-Case Design . Type I).. L2!:i . .~ .~ .. oS ..-" ~ :l go .9 C 0. 1983).:~.~ -a.~g.g• allows an investigator to avoid examining any specific phenomenon in op- I~ 'f 21·s 'a a ~ 8 !l. unbeknownst to the researcher.~ ~ ::I 0 B :5 'a ..~ * N ~ .l! • ~S! -'So '" .. Co :S } ~ ~ .• B .0. Il.. and several intermediary units also were important. '" •• I!i::i .g~ . .• " " . H> B e~ ~ ~> ii0 analysis.~ . nesses.! 8 measures or data. ~ ~ :5''. '!ype 2).2 t ""0 § ~~ > l" '-" 8:§?!l"l . Another typical problem with the holistic design is that the entire case study may be conducted at an abstract level. ~ !:I. With such SUbunits. In ~ 1 8 If '" ~~~ contrast..:j ~ . '.~ ·• >-~ ~'il j . Bateman. ~ 0 > u 0 '. '= " U" . .4. ranging from historical to survey ~~ .g ·8 erational detail.. A further problem with the holistic design is that the entire nature of the ~'a~ c:l'!<I ~. an embedded design can serve as an important device for •~ E! Sd ~ ... ~ . a holistic nature.. .i:!0 8.~ academicians--Seymour Martin Lipset.s>. if the case study examined only the global nature of a program or of an organization..... Although some people have claimed J. such unsus- ~~ ~~ .. '" e. 42 CASE STUDY RESEARCH ..5u =: t-. Martin Trow. begins to address different questions..l ~.5 U ~ ..§•" pected slippage needs to be avoided.. 'E . & Moore.:..7 ~ ~ criticism of case studies is based on this type of shift-in which the original .~ 1: ~ .~ ~ • 'E 'Ii 'fio. lacking any clear i .Ii would be called an embedded case study design (see Figure 2.~ e.2 lii .. :l 6~~ l1 43 L .. ~ i':E B ~t g .S s.:] ~ ~ '"•• One way to increase the sensitivity to such slippage is tu have a set of subunits.4.. g :s 6~ o 'Cd '" " > and involves several units of analysis (see the follOwing table).... in fact. The initial study questions may have reflected one orientation.. t:3 e c c ~ E '". :5 ~..~ ~g:. identified and when the relevant theory underlying the case study is itself of e 1:a• 1 ~ >. . as 'I::.. 'S e !l. . different data collection techniques were used.!:! 0 -5 ." 0 BOX 10 An Embedded. - g Q 1:1::.9. "0 '..... ~ 'So ~ ~ s:: " ~.:l ·n :Ii such flexibility to be a strength of the case study approach. >. .8 ~ a ~ ~ £. ~ . Potential problems arise.. 001 ·5 'I.." 8 ~ ~ research design is no longer appropriate for the research questions being asked (see Yin. and the evidence ~ s::: ] ~a E 1:I§·85- e ~ ~ ..@ s::: 'l:I .~ H il .~ ~ -! 6 ~ ~ :f'~ b .5 .ll "" . you should simply start over again.9 s ~ case study may shift... The case study is about the inside politics of the International'!ypographical Union "." ~ i! .. during the course of o: ..!f 0 § . The main unit •• ~ ~ was the organization as a whole. Union Democracy (1956) is a highly regarded case study by three eminent 'E~ !l. a different orientation may emerge. ...§~~ v .

so of inquirY. and the study as a whole would have used a multiple-case design. if the data focus only on individual employees. The evidence from multiple cases is often embedded units of analysis. the (or to the mUltiple subjects within an experiment}-that is. and if the larger. and the overall study is therefore regarded as justifiable under certain conditions-where the case represents a critical test being more robust (Herriott & Firestone. also has some pitfalls. not sampling logic. however. or is . holistic aspects of the case begin to be ignored. Moreover. is made-to ensure that the case in fact is relevant to the issues and questions Therefore. the critical case.versus single-case designs. However. The choice is considered one of research design. the ration- of existing theory. At the same time. both anthropology and political science have developed one acteristics as a subunit of analysis. lightly." Similarly. is repeating certain critical experiments. The subunits can consider multiple experiments-that is. and Multiple-case designs have distinct advantages and disadvantages in com- two types have been described: those using holistic designs and those using parison with single-case designs. to follow. Replication. and the revelatory case are all likely to A major step in designing and conducting a single case is defining the unit involve only single cases. sampling logiCS. enhancing the is far different from a mistaken analogy in the past. what have been considered "comparative" (or multiple-case) studies (see a study of organizational climate may involve individual employees as a Eckstein. Single cases are a common design for doing case studies. however. the single-case design is eminently considered more compelling. The replica- What Are the Potential Multiple-Case Designs? tion logic is analogous to that used in multiple experiments (see Her~e~ & Barlow. but it should not come as a surprise to the revealed by the different rationales underlying the replication as opposed to investigator. or new technology) in which syndrome. same methodological framework-and no broad distinction is made be- Whathas happened is that the original phenomenon of interest (organizational tween the so-called classic (that is. single-) case study and multiple-case climate) has become the context and not the target of study. A program evaluation that includes project char- For example. Overall. replication is independent innovations OCCur at different sites. Here. which incorrectly con- insights into the single case. is limited toa few cases due to the l expense or difficulty in performing a surgical preparation in animals. becomes a project study if set of rationales for doing single-case studies and a second set for doing no investigating is done at the larger unit-that is. the decision to undertake multiple-case studies cannot be taken of interest. Thus each site might be the said to have taken place. for multiple-case studies. 1976). the larger unit of analysis. A major one occurs Multiple. This often add significant opportunities for extensive analysis. teacher aides. The methodological differences between these two views are shift might in fact be justifiable. 1979). chmcal The same study may contain more than a single case. unusual or rare case. If similar results are obtained from all three cases. by definition. This pling" logic. Every case should serve a specific purpose within the overall scope Within the single case may still be incorporated subunits of analyses. A common example is a study of school inno- thereby producing evidence that the three cases did indeed involve the same vations (such as open classrooms. An operational definition is needed and some ple-case study can require extensive resources and time beyond the means of precaution must be taken--before a total commitment to the whole case study a single student or independent research investigator. 44 CASE STUDY RESEARCH DESIGNING CASE STUDIES 45 An embedded design. or where the ale for single-case designs usually cannot be satisfied by multiple cases. 1975. a major insight is to consider multiple cases as one would that a more complex-or embedded-<lesign is developed. if too much attention is given to these sidered multiple cases to be similar to the multiple respondents in a survey subunits. the choice between single. From the perspective of this book. However. multiple-case studies when the case study focuses only on the subunit level and fails to return to have been considered a different "methodology" than single-case studies. syndrome in psychology or medical science. the conduct of a multi- of analysis (or the case itself). subunit of study. studies. the appropriate research design I the study has to use a mUltiple-case design. In some fields. with both being in- cluded under the case study strategy. This replication logic is the same whether one subject of an individual case study. where the case is a rare or unique event. Summary. The case serves a revelatory purpose. a "sam- case study itself will have shifted its orientation and changed its nature. in frequency in recent years. and such designs have increased is one in which the same results are predicted for each of the three cases.and multiple-case designs remains within the the study will in fact become an employment and not an organizational study. Thus. if one has access only to three cases of a rare. to follow a ''replication'' logic. the "program. When this occurs. George. 1983). for instance.

and the analysis must follow cross-experiment rather than Multiple-Case Studies within-experiment design and logic. According to the sampling logic. Furthermore. universe or pool of potential respondents and then a statistical pr~edure for three or four cases might be selected in which both types of applications are selecting the specific subset of respondents to be survey~ . three or or pool. the United Sta~s was considered to have no counterpart support structures. again similar to the role played in cross-experiment nine to twelve cases. so that data To take another example. ThIs logIC IS present. or impractical to survey the entire universe or pool. these six to ten cases. The sampling logic demands an operational enumeration of the entire but not either alone. surveys. whether applied to experiments or to c~e studIes. One industry was much more dependent aggregate.46 CASE STUDY RESEARCH DESIGNING CASE STUDIES 47 limited by the rarity of occurrence of a clinical syndrome. Three or four additional cases might be selected in which only administrative applications are present. In each of these situations. The ability to conduct six or ten case planning and support by government agencies. is similar to the way scientists deal with contradictory experimental findings. and not just academic. a few cases advocates: the role of the U. theoretical framework. Finally. shows how the defense department supported the development of these indus- tions must be revised and retested with another set of cases. (See BOX 12 for another example of a multiple-case empirical cases do not work as predicted. a number of respondents (or subJects) are policy-oriented theory. Each case must be carefully selected so that it either (a) predicts similar results (a The international marketplace of the 1970s and 1980s was marked by Japan·s literal replication) or (b) produces contrasting results but for predictable prominence. but for ~ifferent tion) as well as the conditions when it is not likely to be found (a theoretical reasons than the administrative-only cases (another theorelIcal replicatIOn). with the same prediction of little increase in use. Again. If this entire pattern of results across these multiple cases is indeed found. is analogous to excellent case study (1990) points to a counterexample. that theories can be practical. (two cases) whose cases and conclusions are tied together by a practical. To pursue this proposition in amul tiple-case study design. and support of R&D. arranged effectively within a multiple-case design... The framework needs to state the conditions under which a particular phenomenon is likely to be found (a literal replica.S. to six) might be designed to pursue two different patterns of theoretical Hooks provides quantitative data on two cases-the aeronautIcs mdustry replications. with inferential statistics used to establish the confidence intervals four other cases would be selected in which only instructional applications for which this representation is actually accurate. in the aggregate. to determine whether. would have provided compelling support for the initial set of upon government than the other. one might consider the initial proposition that from &"Smaller number of persons are assumed to represent the data that might an increase in microcomputer use in school districts will occur when such have been collected from the entire pool. are present. Gregory Hooks s studies. assumed to ''represenf' a larger pool of respondents (or subjects).) . just as with experimental science. defense department in implementing anindustrial (two or three) would be literal replications. modification must be made to the replication design. this logic tries through financial support. . If the cases are in some way contradictory. However. The theoretical framework later becomes the vehicle for gener. ensuring demand. . the initial proposi. In contrast. the alizing to new cases. This replication logic. If all the cases turn out as predicted. . whereas a few other cases (four planning policy within defense-related industries. frequently ignored by the ability to conduct six to ten experiments on related topics. too. An important step in all of these replication procedures is the development of a rich. The study in BOX 11 contains an excellent example of a multiple-case study must be distinguished from the sampling logic commonly used In.. an individual case or subject is considered akin to a single BOX 11 experiment. The resulting da~ from the sample that is actually surveyed are assumed to reflect the enl!re UnIverse little increase in use (predicting a theoretical replication). would provide substantial support for designs. in fact. in both cases. with the prediction being f. theory. and a Policy-Oriented Theory The logic underlying the use of multiple-case studies is the same. microcomputer use did increase over applicable whenever an investigator is interested in de~~nIng the pre~a­ a period of time (the investigation would be predicting a literal replication in I lence or frequency of a particular phenomenon and when It IS too expensIve these three or four cases). replication). if some of the the initial proposition. Much of its strength was attributable to the role of centralized reasons (a theoretical replication). technologies are used for both administrative and instructional applications. in the and the microelectronics industry. Hooks's evidence propositions. but from the field of urban studies. Remember.

. ' .. reviewed r------------------~1 . In !urn. in which convergent evidence is sought regarding Ihe facts United States..N:....... Yet a third group . concluding that failure was therefore not necessarily inherent in the academic enteJprise... 48 CASE STUDY RESEARCH DESIGNING CASE STUDIES 49 . and you are interested in sJudyingIhe behavior ofIhe incumbent and conclusions for Ihe case.1 . DESlGiV PREPARE.. many other topics. hMQ.dd. following Ihe sampling logic. (This figure is derived from research on Ihe case sJudy meIhod... COUECl'... case sJudies should not generally be used to assess Ihe incidence of replication logic is followed.. 1 1 1 1--"] Szanton starts with eight case studies... . you would probably not have Ihe resources to Across cases.. A Multiple-Case.TZE 4 _. Figure 2.. leading to :.. . Replication Design Acornmon problem oftM 1960s and 1970. if a ~ampling logic had to be applied to all types of research..... Within each of the four groups of case studies. 1---I-~_ftPCWl the reader of a general phenomenon. Any sampling logic simply would be misplaced under a summary report. of Ihe Republic. . For each individual case....rn ~ coaPct J... if Ihe First.... . _report JCODl"'. .. The eight cases are sufficient ''replications'' to convince c . • c~ . . yielding a large number of potentially relevant 2. Peter Szanton's book...... ZDd .buDwld. .Iop -. Across the four groups..CO:::. he has illustrated theo. Ihe report should indicate Ihe extent of Ihe replication logic and l .... however. Ihe sIudy is eminently feasible... was how to get good advice to city gov~ments. BoIh Ihe faiIhful to Ihe complexity of reality..S. Bateman. phenomena. you would still have too many variables in relation to Ihe 42 data points available)...... This potent case study design can and should be applied to SOURCE: COSMOS Corporation.i.. conduct a fuIl study of all 42 incumbents (and even if you did..ahIbqr .. Szanton has illustrated the principle of literal replication..1-.. I1 1 The book is an excellent example of a multiple-case......J... ing Ihe sJudy must consist of Iheory development and Ihen shows Ihat case . 1 1 1 of cases shows how university groups have successfully helped business... BOXU __=====. to be at all be Ihe information needing replication by oIher individual cases. ....::.. Moreover...._.. Third. showing how different university groups 1 _ .5. this would require an impOSSibly large number of cases.t I. Ihe report should indicate how such circumstances... many selection and Ihe definition of specific measures are important steps in Ihe Important tOPICS could not be empirically investigated.. must incorporate dozens if not hundreds individual cases and Ihe multiple-case results can and should be the focus of of relevant variables. in which nonuniversity groups also failed.. 1983. and sectors other than city government A final set of three cases shows that those few groups able to help city government were concerned with implementation and not just with the production of new ideas. --~ I-i----. t ••• I the expenences of numerous attempts by university and research groups to collaborate with city officials. the major conclusion that city governments may have peculiar needs in receiv- ing advice... . design and data collection process.. .. badhId.. 1 1 1 1 engineering firms. Case Study Method retical replication... Each individual case sIudy consists of a ing problem: Your investigation deals wiIh Ihe role of Ihe presidency of Ihe "whole" sJudy. replication design. type of sJudy just could not be done... a case sJudy would have to cover boIh Ihe phenomenon The replication approach to multiple-case sJudies is illustrated in Figure of interest and its context.. .. . as Ihere have been only 42 presidents since Ihe beginning and why a particular proposition was demonstrated (or not demonstrated)........ .... ·AI..:. This Any application of this sampling logic to case studies would be misplaced.CUJD=~E. Second. Not Wen Advised (l98!)... .. all failed to help cities..TZE . see Yin. such as in Ihe follow...-. each case's conclusions are Ihen considered to from some leadership perspective.. DEFINE . . variables...- 1 studies.) The figure indicates that Ihe initial step in design- too large to allow any statistical consideration of Ihe relevant variables... Szanton then provides five more case I=--J -..... . The leadership perspective.. & Moore.

may be either. a further question you will encounter The difference between these two designs depends upon the type of phe- has to do with the number of cases deemed necessary or sufficient for your nomenon being studied. one would still want a minimum of to be either holistic or embedded. The fact that a design calls cases-if any-were predicted to have contrasting results.01" is not derived from any replications). For example. three to replicate a direct result (literal tical studies is the selection of the criterion for establishing levels of signifi. The individual cases. centers be included as case studies.4. These data may be highly quantitative.05" or "p < . judgmental choice. six. you should think is concerned with the delivery of services by different community mental of this decision as a reflection of the number of case replications-both literal health centers (facilities) (see Larsen. Each center may rightfully be and theoretical-that you would like to have in your study. ences or if you want a high degree of certainty. the results of each survey will your multiple-case results (as with the higher criterion for establishing sta. the important consideration is related to your sense of the complexity of the realm of external validity. and the data will the rival theories are grossly different and the issue at hand does not demand be used along with archival information to interpret the success and operations an excessive degree of certainty.g.. the survey data are pooled across centers. When an embedded design is used. for each individual center. For instance. including the use of surveys within each case. I L . the topic of a case study. formula but is a matter of discretionary. you may want to articulate these relevant conditions more coming more prevalent. should serve in a manner similar to mUltiple experiments. Figure 2. the selection of In all nine centers. within a multiple-case two or three literal replications). and an investigator must choose each case carefully. 1982). want to include at least some number of cases that varied along ethnic or The replication design does not necessarily mean that each case study needs racial lines (and within each type of case. not a sam- discussing external validity (see the section titled "External Validity"). asmaller vidual case study may in fact include the collection and analysis of highly number of theoretical replications is needed. the typical the conduct of a survey at each case study site. the theoretical framework may dictate that nine such For the number of literal replications. the survey data will be part of the findings tistical significance. with single cases: Each individual case may still be holistic or embedded. whereas other Multiple-case studies: Holistic or embedded. at the given center. For example. However. logic. each indi- thought to produce much variation in the phenomenon being studied. Type 3) or of multiple embedded cases (see Figure 2. for multiple-case studies does not eliminate the variation identified earlier ~gain. Majchrzak. in the neighborhood example used previously in Any use of multiple-case designs should follow a replication. because a sampling logic should not be used. TYpe 4). This section has dealt with situations in which the same inves- you are uncertain whether external conditions will produce different case tigation may call for mUltiple-case studies. not be pooled across centers. replication) and six others to deal with contrasting conditions (theoretical Cance. A study of gentrification would therefore explicitly at the outset of the investigation. or case. a multiple-case study may consist of multiple holistic cases (see colleagues before proceeding with any case study design. a multiple-case study design is no longer being used. When Summary. you may press for five. However. with similar results 1984) is that ethnically and racially different neighborhoods do not usually (a literal replication) or contrasting results (a theoretical replication) predicted follow similar courses of change. likely to be using a survey rather than case study design. if your rivals have subtle differ. When using a multiple-case design. in contrast. an appropriate analogy from statis. In This IS a difficult step to perceive and is worth extensive discussion with other words.4. In an embedded design. but they are more expensive and time-consuming explicitly at the outset of your study and identify a larger number of cases to to conduct. the greater certainty lies with the larger number of cases).5 depicts a very different logic than that of sampling design. If. These types of designs are be- study results. Much as the choice of "p < . a study may even call for study. Instead. F~gure 2. Rather. an embedded design is used because surveys of the the number of replications depends upon the certainty you want to have about centers' clients will be conducted. be included. 50 CASE STUDY RESEARCH DESIGNING CASE STUDIES 51 why certain cases were predicted to have certain results. quantitative data. The cases common coneem from the standpoint of policy research (e. supposing a study criteria regarding sample size also are irrelevant. For the number of theoretical replications. when external conditions are not study design. and the investigation is or more replications. you may want to settle for two or three literal replications when focusing on the attitudes and behavior of individual clients. In contrast. However. a pling.

:uonales for using a single-case study design. Chapter 5' similar data found.. Identify some basic questions to be answered by your case study.or multiple--case deSIgn. For all three types or res 1 ~ todies this might include searching changing the purpose or objectives of the study to suit the case(s) that were level and also caninvolve inferences-forexamp e. but still illegitimate change. hniq be described for surveys or experiments-for example. are cbanged. What is the main unit of analysis in this book? What alternatives did you consider. The former situation is much like changing experiments when it is for attems among converging types of evidence. ~gure . then n~e tho~ for usmg a As an example. or~ S furthe . In the event of a multiple-case design. The point is that the flexibility of case study designs is in selecting cases different from those 2 2 focuses only on the fonnal research design process. but only under stringent circumstances. thelatter is a more subtle collection tee ues can . an investigator has every right to 5 Defining the criteriafor judging the quality ofresearch designs. pilot case studies may reveal· inadequacies in the initial multiple-case design.and multiple-case studies. the investigator can correctly be accused of exercising a NOTE bias in conducting the research and interpreting the findings. p. . Defining the unit of flI/Lllysisfora case study. SIt 0 S C or does it have embedded units of analysis? A final reminder is that a case study design is not something completed only at the outset of a study. Describe the research design of this case study. How did it justify the relevant evidence to be sought. h li ti on the evidence? Is the design a single.I an appropriate and desirable use of pilot studies. and (~) reliability. data collection t<chniques might be dePIcted as a third actiVlties. t be answered? What methods were used to draw conclusions. the selection of cases may have to design? be modified because of new information about the cases. (b) . This is . What are the advantages 0 ea all. not on data collecti?n initially identified (with appropriate documentation of this shift) but not in l. EXERCISES I. Give examples of each type of deSIgn. Select a topic for a case study you would like to do. If these. At the same time. Examine or read the case study The Soul of a New Machine. . (c) external validity. Nam~ the initial stages of a study. The design can be altered and revised after the 4 E tablishing the rationale for single. Defin~ ~e four conclude that the initial design was faulty and to modify the design. Select one of the case studies desctibed in the BOXES of this book. 3. criteria for judging the quality of research designs: (a) construct validIty. (Also see Chapter 3 for more internal validity. an investigator must be careful not to shift. Does the naming of these questions clarify the boundaries of your case. Give an example of on pilot case studies. ' design for surveys or stimulus presentation strategies for expenments. and why did you select the unit that you did? Carry out the same exercise for some other case study of your choosing. 53 52 CASE STUDY RESEARCH DESIGNING CASE STUDIES . after some early data collection and analysis. what ~~es di described in the BOXES of this book or from other case studIes f &type~ was thought to be a revelatory or unique case may not turn out to be so after with whi& you are acquainted. euch. the theoretical concerns or objectives. rather than the cases them- selves. with regard to the relevant length of time for which evidence is to be collected? The relevant organization or geographic area? The type of evidence that should be collected? The priorities for doing analysis? 2. as descnbed rID. given the basic research . unknowingly. based How Case Study DesIgns Can Be Kept Flexible questions o · .. Defining a case study research design.) each type of criterion in a case study you lDlght want to do. Defining the boundaries of a case study. ? I . . eIther from ~e design or may help to articulate it In the event of a single-case design. questioruwre obvious that an experimental procedure is infeasible. In other words.