As with other research approaches, the first step in designing a case study is to clearly define your research question. Because many case studies are exploratory, the theoretical and empirical literature may provide only a sketchy foundation for your study (Eisenhardt, 1989). Nevertheless, a thorough literature review will be your first step. Other critical steps include identifying your unit of analysis, selecting the case or cases that will be the focus of your study, and planning your data collection procedures. Each of these steps is discussed here.
Identifying the Unit of Analysis
The unit of analysis “is the major entity that you are analyzing in your study” (Trochim, 2006). Most studies focus on individuals as the unit of analysis, but other possibilities include aggregate entities like groups or organizations; projects; or events such as decisions made or information seeking episodes. A primary defining charac- teristic of a case study is that it focuses on a single instance of the unit of analysis. For example, a case study may focus on a single person, a single organization, or a single event. Within the focus on the single case, the study is likely to take multiple perspectives by gathering data based on multiple units of analysis, then aggregating it to understand the case that is the focus of the study. For instance, Tan et al. (2005) used a
54 APPLICATIONS OF SOCIAL RESEARCH METHODS
case study approach to examine how knowledge had been managed in one organization: the National I.T. Literacy Program (NITLP) in Singapore. The unit of analysis is one team within one organization: the implementation team within the NITLP. To fully un- derstand knowledge management within the team, the researchers examined interactions among team members, interactions between the team and other parts of the organization, and the perspectives of individual members of the team. This approach exemplifies the multifaceted character of case studies.
Selecting a Case
Instead of randomly selecting a case from a population of cases, you will strategically select a case (or several cases) based on your theoretical purposes and the relevance of a case to those purposes (Eisenhardt, 1989; Glaser & Strauss, 1967). Stake (2000) advises us to choose the case from which we can learn the most. If the case is chosen because of its theoretical dimensions, case selection is called theoretical sampling, and the goal is to choose a case or cases that are likely to replicate or extend an emergent theory. Both purposive sampling and theoretical sampling are quite different from statistical
sampling, which focuses on selecting study participants that are representative of a
population of interest.
Some studies will focus on a single case, while others may compare two or more cases. A single case study provides in-depth investigation of and rich detail about phenomena. Yin (2003) identified five possible reasons for selecting a particular case: (1) it is a representative or typical case that captures the circumstances and conditions of an everyday or commonplace situation; (2) it is a critical case that is essential for testing a well-formulated theory; (3) it is an extreme or unique case that represents a rare phenomenon that needs to be documented and analyzed; (4) it is a revelatory case that illuminates previously inaccessible knowledge; or (5) it is a longitudinal case that can be repeatedly studied at several different points in time. For example, Kari’s (2006) study of the relationship between information searching and personal development employs a single case study method. Only one person was selected to represent a revelatory case, so that the investigator was able to empirically observe the whole searching experience of the study participant. In addition, it was a typical case, based on sex, age, location, education, employment, and Internet experience.
Multiple-case studies (also called comparative case studies) are basically a combina-
tion of two or more single case studies. This approach contributes to cross-case analysis and the extension of theory to additional individuals or settings. There are two logics supporting a multiple-case study design: literal replication and theoretical replication.
Literal replication selects cases that are very similar to each other, and the researcher
expects the same outcomes from each. Theoretical replication selects cases that differ from each other in theoretically important ways, and the researcher expects either to expand the scope of the underlying theory or to rule out the theory’s applicability in certain settings. The second example presented in this chapter exemplifies this approach: Hughes (1998, 1999) conducted a series of four case studies, then compared the results across the cases.
Collecting Data
Because a case study is intended to generate rich data concerning a particular case, it is almost inevitable that multiple methods of data collection will be used. These methods
Case Studies 55
might include analysis of existing documentation and archival records, interviews, direct observation, participant observation, and examination of physical artifacts (Yin, 2003). In addition, quantitative methods, such as questionnaires, may be incorporated into the research design. Among these methods, direct observation is most frequently used and, because direct observation is a major research tool for field studies, some investigators (e.g., Fidel, 1984; McTavish & Loether, 1999) equate case studies with field studies. Yin (2003) noted that even though most case studies use direct observation, data collection methods for case studies are not limited to direct observation.
The results from the different data collection methods are combined through triangu-
lation. This approach to data integration was developed by Denzin (1978) for use in socio-
logical studies. Stake (2000) defines it as “a process of using multiple perceptions to clar- ify meaning, verifying the repeatability of an observation or interpretation” (p. 443). Four types of triangulation were identified by Denzin (1978): (1) data triangulation (combin- ing data from different sources); (2) investigator triangulation (combining data collected by multiple researchers); (3) methodological triangulation (combining data collected via different methods); and (4) theory triangulation (combining data collected from multiple theoretical perspectives). This fourth type, though possible, is quite rare in ILS and other social science research. The others have all been used in ILS studies and, in fact, could all be included within a single case study. For example, imagine that you and a couple of colleagues are studying the metadata generation practices in a particular organization. You will need to integrate the data collected by each researcher, requiring investigator triangulation. In terms of data triangulation, you may have data from several sources: cat- aloging staff, executive staff, policy and procedures manuals, and Web server transaction logs. In terms of methodological triangulation, you may have used direct observation of the workplace, interviews (with staff), content analysis (with policy manuals), and trans- action log analysis (with the Web server logs). In each case, the triangulation process will require the comparison of findings from each investigator/source/method, cross- checking carefully to ensure that the study findings are valid conclusions drawn from the data.