Case study research, either single or multiple, has been increasingly used in a number of disciplines in the last few years (Creswell, 2007; Yin, 2014). However, at the same time, there has been an increasing number of critics as well. The critics argue that case study research is problematic. The arguments include that case studies lack methodological rigour (Gibbert, Ruigrok, & Wicki, 2008), are ambiguous in data collection methods (Verschuren, 2003), and have limited generalizability (Hillebrand, Kok, & Biemans, 2001; Kennedy, 1979). Despite the criticism, some experts argue that case studies offer a better understanding of specific phenomena that is often neglected by ordinary research designs that focus more on the generalisation (Creswell, 2007, 2014; Stake, 1995; Yin, 2014).
This essay reviews a collective case study design by Ho, Woodley, Cottrell, and Valentine (2014) in Vietnam, entitled “A multilevel analytical framework for more effective governance in human-natural systems: A case study of marine protected areas in Vietnam”. The purpose of the essay is to critically highlight the case study strengths and weaknesses based on a range of issues. It addresses research approach and research design, theoretical framework, data collection methods, validity, data analysis and findings, and generalisation. The essay found that the case study has some minor weaknesses, but overall it is methodologically vigorous. It is acknowledged that, as this essay only relies on the final report of the case study, this essay itself has limitations in assessing validity and reliability of the research.
Ho et al. (2014) found that there have been numerous studies about “institution”, “institution analyses”, and “governance processes” for natural resource management. However, there have been limited studies trying “to integrate the components of institutions and governance” into a “comprehensive framework” to study the complexity and particularity of “human-nature systems” in a Marine Protected Area (MPA) context (2014, p.12). They defined the human-nature system as dynamic symbiotic relationships between people and nature in complex institutional structures and governance processes. While the research questions are: “1) How problems occurred because of improperly formed institutions affect the governance among state-actor groups and subsequently non-state actors; and 2) How problems occurred with non-state actors affect the governance undertaken themselves and the state actors?”
The framework for the phenomenon reflects Ho’s et at. (2014) worldviews of existence, how it is built and how to be understood. The ontological question is “what is the form and nature of reality and, therefore, what is there that can be known about it” (Guba & Lincoln, 1994, p. 108). The answer is that the “human-nature system” is very complex, involving various institutions and governance processes, but the meanings are chaotic, evolved, hidden, and little is known (Andrade, 2009; Guba & Lincoln, 1994; Stake, 1995). While the epistemological question is “what is the nature of the relationship between the knower or would-be knower and what can be known” should consistent with an ontological view (Guba & Lincoln, 1994, p. 108). The answer is that the researchers had limited knowledge of “human nature system”, including their knowledge of what has been known, while what have been known varying, changing and often conflicting among the stakeholders. (Creswell, 2007, 2014; O'Leary, 2014; Trent, Artiles, & Englert, 1998). Therefore, the researchers have to work closely with the stakeholders, make interpretation and become ‘insiders’, but the negative consequence is the researchers can make bias based on their own values (Creswell, 2007). This ontological and epistemological assumption justified the worldviews of social constructivism in this case study research (Miles & Huberman, 1994) which “emphasize that the world is constructed by human being as they interact and engage in interpretation” (O'Leary, 2014, p. 293), and “if so little is known... then the naturalistic approach with its more qualitative methods might be used ”(Westbrook, 1991, p. 242).
In a more specific argument for assessing the rationale for using a research design, some expert arguments that case study is used for particular phenomena and settings is in accordance with the nature of Ho’s et al (2014) study including the research questions (Berg, 2009; Creswell, 2014). The uniqueness and complexity of the phenomena are the first two reasons to use case study research (Anderson, Crabtree, Steele, & McDaniel, 2005; O'Leary, 2014). While the research questions should begin with ‘how’ and ‘why’, the temporal of the phenomenon should contemporary, and the setting should natural or without manipulation or control (Yin, 2014). However, Yin (1994) argues that case studies with ‘what’ question are used to explore new phenomena. Understanding context and related factors are the keys to asking and answering the questions (Hakim, 2000; Ragin, 1997). These requirements are to produce a specific time and place report that is deep, detailed and life (Miles, Huberman, & Saldaña, 2014).
The framework Ho et al. (2014, p.11) built has four interdependent components: (i) organisational structure; (ii) informal institutions; (iii) institutional performance; and (iv) governance. From these components, they identified 20 principles and groups of themes which serve as “operational definitions and measure of empirical phenomena” (Modell, 2005, p. 237). The framework reflects the theoretical underpinnings and serves as a map to see causal logic among the components. The benefit of a clear and detailed framework is to “structure data collection enough to provide comparability across cases without overly constraining the search for site-specific explanations” (Greene & David, 1984, p. 75). It helps the researchers to select the cases and variables, draw their relationships, collect and analysis data, find rival theories and make a generalization (Miles et al., 2014; Yin, 2014). Therefore, the framework leads the study to investigate “what it claims to investigate” (Gibbert et al., 2008, p. 1466; Greene & David, 1984).
Ho et al. (2014) found from secondary data that MPAs in Vietnam are managed under three different jurisdictions, and different management approaches, involving National Departments (ND), Local Communities (LC), and Private Sectors (PS). Therefore they put the MPAs into three categories. O'Leary (2014) and Creswell (2014) argue that case selection takes into account the research problems and research questions, and consideration of the characteristics of a case that has rich information, accessible, and strong theoretical relevance. Further Yin (2014) adds that case selection needs to consider “represent confirmatory cases (i.e., to replicate identical phenomenon), contrasting cases (e.g., a success vs a failure; top-down vs participatory approaches), or theoretically diverse cases (e.g., location, level, size). This could be the argument of Ho et al. (2014) to select three case studies. However, they did not provide information on the total number of MPAs. Greene and David (1984) argue that the number of selected cases should represent the number of cases in each category. The benefits are that researchers can do replications and extension to produce a better understanding of theory (Eisenhardt, 1991; O'Leary, 2014) and improve the ability to theorise about a broader context (Stake, 1994). Replication refers to collective support for a proposition from a number of individual cases; meanwhile, extension refers to the elaboration of a theory (Eisenhardt, 1989).
The requirement to do analytical generalisation is that it needs at least four to ten cases to undertake a vigorous cross-case analysis (Eisenhardt 1989). It will be difficult to build theory if there are less than four cases, but it will be hard to manage if more than ten cases. It is debatable as Hakim (2000), and Johnson and Christensen (2004) argue that two or three is appropriate, depends on the research problem. This can justify Ho’s et al (2014) choice of the number of cases. Gerring (2004) argues that the size of sample shapes the trade-off between comparability and representativeness. Single-unit case studies allow comparability, but they lack representativeness. On the contrary, multiple case studies excel in representativeness but they lack comparability. Therefore Yin (2014) suggests that in order to do comparative (cross-case) analysis, the researchers can undertake a number of case studies (different levels, programs, departments etc.) or embedded case studies in a context, instead of undertaking multiple case studies in a number of contexts.
From the three case studies, Ho et al (2014) chose 83 participants, represented National Departments (N), Provincial Agencies (P), Local Governments (L) and Local Communities (C). The absent information is that Ho et al. (2014) did not elaborate on how they selected the participants, the representatives of each group of stakeholders and whether in numerous stages or once-off selection. In selecting samples, case studies purposely select the participants based on theories rather than using random sampling. The number of respondents does not matter. It can be a single person, an organisation, a unit or a community. It focuses more on finding individuals that provide rich information about the uniqueness of the phenomenon and on answering research questions (Stake, 1995). However, within-case sampling should ensure a balanced representation of each unit/case so that cross-case analyses can be undertaken (Greene & David, 1984).
In order to select the right participants, phased selection can offer better choices. However, Ho et al. (2014) did not elaborate on this issue. Miles and Huberman (1984) argue that good data collection in qualitative approach is iterative, not a linear process. At the beginning, the selection of respondents might not satisfy the needs of studies. As Charmaz (2006, p. 102) argues, “theoretical sampling involves starting with data and then examining these ideas through further empirical enquiry”. Therefore, researchers need to have alternative participants, keep evaluating their representations, and seek information from other previous studies (Miles & Huberman, 1994). The iterative process means that the focus should be on the sufficient theoretical meanings for researchers to build and work on constructs that emerge during data collection and analyses (Straub, Boudreau, & Gefen, 2004). Therefore the duration of data collection continues until the researchers find no theoretical insights any longer (Ford, 2010).
Another unclear information related to a unit of analysis. Ho et al. (2014) did not provide explicit information on the selection of unit analysis, but in data collection and analysis, they treated the groups of stakeholders (N, P, L and C) as units of analysis or embedded case studies (Yin, 2014). It means that there were twelve units of analysis or four units of analysis per case. Previously they selected three MPAs as the cases. Therefore, they could have undertaken a number of data analyses at different levels, and produce rich data and strong findings. As Baxter and Jack (2008) argue that researchers can undertake data analysis within the subunits (within case), between the subunits (between case), and across the subunits (cross case).
The strength of the case study research is in using triangulation to collect, analysis and confirm data (Gibbert et al., 2008). Ho et al. (2014) undertook data collection through a variety of methods and sources. The first stage, they used semi-structured and open-ended questionnaires before holding focus-group discussions with the participants. Stakes (2009, p.64) argues that from all various sources of the information, “interview is the main road to multiple realities”. It enables the participants to express their views freely, covers many aspects of an issue, does not comply with certain structure, and often leads to unexpected discoveries (Creswell, Hanson, Plano, & Morales, 2007). The second stage, they undertook personal observations and held workshops and conferences. The benefits of triangulation are to enrich collected data, avoid personal and methodological biases, and provide the researchers with a better understanding and various perspectives of the phenomena (Decrop, 1999). These data collection methods provided Ho et al. (2014) with holistic data and understanding of the phenomenon (Verschuren, 2003). The agreement of data from the two stages, including allowing participants to review and provide feedback on the draft of the case, is the most critical approach to establishing construct validity (Beverland & Lindgreen, 2010), but it cannot be assumed that the research has achieved construct validity (Andrade, 2009). Construct validity has been also linked to achieving “theoretical sufficiency”, instead of data saturation, which refers to “build up and work upon constructs which emerge from problem under investigation” (Andrade, 2009, p. 48). Another approach is to prolong involvement for building trust with participants while identifying bias that originated from anomalies brought by both the participants and the researchers (Flyvbjerg, 2006).
In addition to the framework and triangulation methods, Ho et al. (2014) enhanced internal validity for data analysis by identifying themes during interviews using identified institutional problems (Gibbert et al., 2008). The themes were grouped according to the 20 principles they had identified. These principles can serve as propositions to guide data collection and discussion (Baxter & Jack, 2008, p. 552). Then they held some sessions with the participants for further explanation building, including contrasting opinions of identified causes and reasons for the outcomes and consequences of governance. Together with the participants, they identified, analysed and summarised barriers and factors that shape human nature system. This data collection procedure is in accordance with Yin’s identified tactics for internal validity (2014). Some of the tactics are matching similar patterns of inputs (themes) from all the groups with the patterns that have been proposed in theories and asking for justification of answers provided by groups, including opposing opinions from other groups (Gibbert et al., 2008). It means that data analysis is undertaken together with data collection. Stake (1995) refers to it as direct interpretation; collected data are reviewed, tested and analysed over and over until the researchers find the realistic answers to the research questions.
From data analysis, Ho et al. (2014) identified eight thematic findings. The findings were compared with patterns (themes) that had been identified in the framework of the research; it means that they had both predicted identical findings (literal replication) and anticipated different findings (theoretical replication). However, they did not explain if the findings from each unit/case were converted into a set of propositions to facilitate replications of the three MPAs, nor did they link the kind of replications to certain phenomena and cases (Greene & David, 1984). A finding is robust if it holds in a case, also holds in a comparable case, but not in contrasting cases (Miles et al., 2014), and if a finding does not hold in comparable cases, it demands either theory change or refutation (Lloyd-Jones, 2003). They had demonstrated that these findings came from the units of analysis (N, P, L and C), but they did refer these units to their original cases (ND, LC & PS) to identify their comparability and contractibility. As the result, it is unknown if the three approaches of MPAs produce different or similar outputs of a human-nature system. Therefore it is unknown to what extent and under what management approach the framework works as expected (literal replication) or does not work for predictable reasons (theoretical replication) (Yin, 2009, p.18).
Ho et al. (2014) undertook generalisation by referring the thematic findings to theories, and not to other cases as the contexts of the three cases are different (Creswell, 2007). This aligns also with the argument of Stake (1995) that the focus of case study is on the analytical generalisation of the particularities, instead of statistical generalisation. They identified the similarities and explain the reasons. By doing so, they determined the level of confidence of the findings (Eisenhardt, 1989). The findings were correct and similar, there were some differences but none was in conflict with theories, which led them to adjust the framework. Therefore the findings strengthen internal validity, increase the cover of generalisation and support their analytical framework (Eisenhardt, 1989). The missing thing is that Ho et al. (2014) did not identify and explain the theoretical perspectives on the differences in findings. Onwuegbuzie and Leech (2007, p. 245) argue that reporting cases that do not fit the framework “allow readers to evaluate them and to draw their own conclusions”. However, Yin (1994) argues that exploratory and descriptive cases do not look for negative cases or alternative explanation.
Based on the generalisation, Ho et al. (2014, p.16) identified and classified factors that shape (restrain) governance of the MPAs in Vietnam to answer the research questions. The factors were put into three groups, instead of four as they identified earlier in the framework. The “institutional performance” in the initial framework was taken out; it contrasts with the findings. The new factors related to “(i) formal institutions, (ii) political behaviours and organisational structure and (iii) social capital and socio-economic conditions”. In the second stage of primary data collection, they identified specific factors that shape each group of stakeholders in order to answer research questions. However, they left a number of unanswered questions such as which case(s) posed contradictive findings to the framework, what factors in each case differed from the initial framework, and what revision they did to the framework before being tested with the other cases (Johnston, Leach, & Liu, 1999).
To sum up, from ontological and epistemological perspectives, the researchers have demonstrated a strong argument for using case study design. The researchers established a clear framework for the case study, employed a number of triangulation techniques to portray and confirm the findings of the phenomenon. Therefore they have strengthened construct validity, internal validity and external validity. The essay found some issues related to missing information on sampling methods, population, proposition, and replication, but this essay treats them as the result of limited available space for the researchers to provide the information in their report.
Overall, the researchers have tried to a build an analytical framework for the assessment of institutional and governance arrangement of MPAs from a human-nature system perspective. They made some claims about three interdependent issues that shape relationships of a number of stakeholders and policy outcomes. Their final recommendation is about the establishment of a body to manage governance processes in MPAs. Therefore they suggested further research to understand governance processes in order to avoid unnecessary problems as the result of establishing a bridging body.
Most importantly, this research has refuted some critiques on the weakness of case study research design. The research has demonstrated that case study research vigorously uncovers particular and complex issues that are overlooked by other research designs that tend to focus more on drawing statistical generalisation. Case studies have a vigorous methodology, employ numerous data collection methods and analysis, and therefore have strong validity and reliability.
Reference:
Comments
Post a Comment