Case studies in humanitarian logistics research

Among research methods used in logistics and supply chain management (SCM) research, the case study has been frequently considered as one of the most powerful (Ellram, 1996; Voss et al., 2002; Seuring, 2008) because of its usefulness in assessing “real world” examples (McCutcheon and Meredith, 1993). Yin (2009, p. 18), defines it as “an empirical inquiry that investigates a contemporary phenomenon in depth and within its real-life context, especially when the boundaries between the phenomenon and the context are not clearly evident, and copes with the technically distinctive situation in which there will be many more variables of interest than data points, and as one result relies on multiple sources of evidence, with data needing to converge in a triangulating fashion, and as another result benefits from the prior development of theoretical propositions to guide data collection and analysis.” Since its early days, humanitarian logistics (HL) research has benefitted from the method, providing understanding to this “absolutely fascinating”, but complex subject (Van Wassenhove, 2003).

Studies of HL state of the art (Natarajarathinam et al., 2009; Kunz and Reiner, 2012; Leiras et al., 2014; Chiappetta Jabbour et al., 2017) have confirmed the need for more empirical evidence pointed out by previous authors (cf. Van Wassenhove, 2006). While literature reviews on the use of case studies in discipline-related fields identified an increasing trend toward using more qualitative case studies (Dubois and Araujo, 2007; Barratt et al., 2011), evidence shows that this trend is already a fact on HL and SCM literature, as qualitative methods are preferred over quantitative or mixed methods (Chiappetta Jabbour et al., 2017). However, HL literature is lacking a comprehensive study on how qualitative methods, particularly case studies, are used in this specific context. Addressing this issue, we conduct a theory-driven content analysis of published HL case studies in logistics and SCM journals, following Patton (2002). Results show ambiguous use of the term “case study,” insufficient information about the research process, and deficient use of theoretical frameworks, contributing to the critics and misconceptions so common for case study research (Ellram, 1996; Flyvbjerg, 2006). We call for more rigor when carrying out and documenting case research and, based on the results from the content analysis, propose a framework for designing case studies in HL. The paper is a first attempt to investigate case study research on HL, seeking to provide guidance and better understanding for both academics and practitioners. The next section overviews existing frameworks for case study research on logistics literature. Section 3 describes the methodology used for the content analysis and Section 4 presents results followed by a discussion and a framework for crafting case study research in Section 5. The paper concludes with some final remarks and avenues for further research.

Case study research has been frequently considered as one of the most powerful research methods in operations management (Voss et al., 2002). It is used to investigate a specific phenomenon, allows to study operations in their natural setting, and is useful in early phases of research where there are no prior hypothesis or previous work for guidance (Sachan and Datta, 2005). Although a diverse number of case study frameworks from logistics, SCM or operations management literature exist to conduct rigorous research (see Table I), evidence shows that case study-based articles do not seem to ensure rigor (Seuring, 2008; da Mota Pedrosa et al., 2012), which contributes to the existing criticism of this research approach. Historically, case study research in logistics and SCM was not largely used compared to other methods (see Dunn et al., 1994; Mentzer and Kahn, 1995; Näslund, 2002; Frankel et al., 2005; Sachan and Datta, 2005; Guinipero et al., 2008). However, a recent study by Hilmole (2018) shows a different picture.

In the study, the author analyzes 1,699 “supply chain” publications between 1995 and 2015 from 214 journals that state the use of case studies. Evidence shows that since 2009, the number of published case studies exploded, mostly due to the shock waves of the economic crisis that lead to an increase on environmental and sustainability issues on the agendas of governments and corporations. As a result, “case studies have achieved visible status in supply chain research works, and case study research is increasingly applied as a research method” (p. 97). Not surprisingly, HL and SCM case study-based publications are not significant enough to be highlighted in Hilmola’s study, despite the fact that these also followed the burst (Yu et al., 2015). Nevertheless, it does not change the fact that qualitative research, and in particular case studies, is the most common method on HL and SCM (Chiappetta Jabbour et al., 2017).

Data collection can be a time consuming, emotionally draining and sometimes physically dangerous process (Neuman, 2007). This can be particularly true in HL and SCM research, as the characteristics of the context (cf. Holguín-Veras et al., 2012) impede access to data (Jahre and Fabbe-Costes, 2015; Jahre et al., 2012; Peretti et al., 2015) and researchers have to find ways to overcome these issues. However, regardless of the characteristics of the context, methodological soundness and rigor should be at the heart of any research endeavor (Halldórsson and Aastrup, 2003). Thus, investigating how case studies are used in HL research can help us to understanding how established frameworks are adapted to this particular context, what are the gaps regarding rigor (Näslund, 2008; Seuring, 2008), and how can these be addressed.

Following the recommendations of Seuring et al. (2005) and Seuring and Gold (2012), we conducted a content analysis based literature review. Consisting of a systematic, rule-governed and theory-driven analysis of fixed communication (Mayring, 2008), we follow four main steps: material collection, descriptive analysis, category selection and material evaluation. This process model for content analysis has previously been used for conducting literature reviews in HL (e.g. Kunz and Reiner, 2012; Leiras et al., 2014; Vaillancourt, 2016), as well as other SCM fields and contains both quantitative and qualitative content analysis techniques. This paper presents a qualitative analysis, the purpose being to study “the appropriateness of the procedures used relative to a chosen context,” rather than the “explicitness and objectivity of scientific data processing” (Krippendorff, 1980, p. 87).

Relevant articles were identified following a process similar to that of Vega and Roussat (2015). Seeking to avoid a restrictive sampling bias (Fabbe-Costes et al., 2009) and ensuring appropriate journal and publication quality (Wilding and Wagner, 2014), logistics/SCM international refereed journals were selected using EbscoHost, Emerald Insight, ScienceDirect and Wiley to identify articles published by 2017. Only those written in English and including “case study” and “humanitarian operations” or “humanitarian logistics” in all text were selected to capture articles that do not include the term “case study” in title or abstract, but only in the main body.

The broad scope of search terms and type of content identified many articles not relevant for our study, such as “case study” included only in the reference list. Therefore, we followed a process similar to that of Abidi et al. (2014), screening through all the identified papers before selecting. Inclusion criteria identified articles conducting or reporting on single or multiple case studies, using case study only or combining it with other methods. We excluded articles reporting literature reviews including the keywords, journal editorials, introduction to issues, book reviews, doctoral dissertations or articles on other themes such as sustainability reporting, defense logistics or human trafficking.

The process explained above resulted in a final sample of 69 articles out of a total of 215 identified, which were then classified following the categories explained below. Instead of examining articles that fully match the proposed set of categories, our aim is to reveal gaps in HL literature based on articles that, for different reasons, fail to reach a clear definition of the case study in relation to the proposed categories.

Guba (1978 in; Patton, 1980) states that categories result from patterns that revealed through convergence (fit) and recurring regularities in the data. As a set, categories should ensure internal and external plausibility (consistent and holistic), be reasonably inclusive (absence of unassignable cases), reproducible and credible to the sources. Table II shows categories used in previous HL literature reviews, as well as the approach used for category selection. Other than research method, some commonalities can be found on the different sets of categories, including phase of the disaster management cycle and type of disaster.

Based on this overview and seeking to meet Guba’s (1978) criteria for category selection, we developed a set of categories following a theory-driven deductive approach and combined the results from Table I with methodology literature to select categories. Table III lists the resulting set. As they are subject to comprehensive description and discussion in the referenced works, we referred to these for more detailed explanations.

Mayring (2014) describes three qualitative techniques for evaluating the material; summary, explication and structuring. Considering our purpose and the process chosen to build categories, the structuring technique appears the most relevant. Called deductive category assignment, the technique’s purpose is to extract a certain structure from the material, and requires creation of the category system based on theory, other studies or previous research before coding the text.

Seuring and Gold (2012) state that although literature reviews are preparatory work for further empirical research, they should be considered as “fully-fledged research methods” (p. 551). Thus, measures such as transparency, objectivity, replicability, validity and reliability should be used to ensure research quality. Considering Halldórsson and Aastrup’s (2003) suggestion to make sure that quality criteria fit with the method we use the criteria proposed by Seuring and Gold (2012) inspired from Avenier’s (2010) work; explicitness and ostinato rigore, to ensure research quality.

In accordance with Avenier (2010), the previous sections ensure explicitness by a thorough description of the research steps to expose linkages between our decisions, information gathered and inferences drawn. Further, the following section presents in detail the information gathered and differentiates it from the inferences drawn by the researchers, allowing readers to build autonomous assessments of the knowledge generation process thus ensuring trustworthiness. Regarding ostinato rigore, Avenier (2010, p. 15) defines it as “an obstinate quest for becoming still more rigorous in the way researchers collect information, read and reread academic literature and field documents, and draw inferences.” Triangulation, negative cases (as a mean to prevent researchers from settling too quickly on a particular interpretation) and member checks are some of the techniques used to ensure rigor. Triangulation was difficult to perform in this research, but rigor was vindicated through constant member checks of the different choices and decisions throughout the research process between the researchers, constantly challenging initial impressions and interpretations.

The following section presents results from the sample analysis based on the categories above. Addressing key findings, it provides the basis for discussion in Section 5.

The first analysis of the sample relates to the overall configuration and rationale for the use of case study method, as introduced in Table I. Results show that almost half of the sample (46.4 percent, n=32) does not describe any methodology regarding the case study or, if mentioned, the methodology section is limited to practical aspects of the research process. There could be at least two reasons for this. First, many articles are based on operations management where case studies are rarely seen as a research methodology in itself, but rather used to illustrate (e.g. Gatignon et al., 2010), examine the appropriateness or applicability of a methodology (e.g. Besiou et al., 2011; Kretschmer et al., 2014), or test a model (e.g. Kilci et al., 2015; Alem et al., 2016). Second, some authors actually use the research method but do not present it as such. While structured as a case study, the methodology is termed as qualitative research (e.g. Sheppard et al., 2013) or semi-structured interviews (e.g. Santos et al., 2016). This can create problems because it contributes to the critics and misconceptions about case studies by presenting something as a case study, which does not qualify as such, or failing to present research as a case study when it in fact is. A clear definition of what is case study research could reduce this problem in both instances.

The other half (53.6 percent, n=37), on the other hand, clearly state and argument for the use of case studies research in a well-structured way, presenting basic information about the research process. However, a theoretical framework for case study rationale, data collection and analysis and research quality is often partially missing. With one exception (Goffnett et al., 2013), papers employing case studies refer to different frameworks when arguing the use of this method and its design. To a lesser extent, articles use frameworks for data collection, i.e. data collection techniques, sampling, etc. (33.3 percent, n=23), data analysis (26.1 percent, n=18) or research quality (26.1 percent, n=18). Nonetheless, we found good examples (10.1 percent, n=7) of the use of frameworks, i.e. the methodology itself (see Table AI). Although not many, these articles show rigor and allow assessing research quality, something that should be found in case study research at large.

Very few articles state a specific objective for the case study making it hard to classify its purpose. Nevertheless, the formulated research question makes it possible to infer for what purpose the case study was conducted. The largest portion of the sample (37.7 percent, n=26) adopts an exploratory approach, based on questions such as “what are the priorities and challenges of health commodities management systems” (Ibegbunam and McGill, 2012), “how purchasing power and purchasing strategies are related” (Pazirandeh and Norrman, 2014), or “how can supply chain performance be measured” (Haavisto and Goentzel, 2015). The second largest portion (30.4 percent, n=21) follows a descriptive logic, with cases aiming to “describe and analyze a successful training exercise in detail” (Gralla et al., 2015), “summarize the current state of reverse logistics within the HL sector” (Peretti et al., 2015), or “describe and analyze the forecasting and order planning process” (van der Laan et al., 2016). A smaller portion (18.8 percent, n=13) can be defined as explanatory, seeking to understand, for example, “the impact of cooperative purchasing on purchasing power” (Pazirandeh and Herlin, 2014), or investigate “the relationship between the disaster management process and supply chain resilience” (Scholten et al., 2014), while the smallest portion (2.9 percent, n=2) could be described as predictive. For a small amount of articles (10.1 percent, n=7), such classification was impossible, as the purpose of using the case study is so unclear that it could be put in any of the subcategories above. This underlines the need for clear definitions of and rigor in conducting case studies in HL research.

Organizations appear as the preferred unit of analysis (62.3 percent, n=43). In particular, results show that international organizations are studied as individual units using single or multiple case designs. Case studies of specific countries (37.7 percent, n=26) or operations (36.2 percent, n=25) constitute the second most popular unit, followed by networks or clusters 18.8 percent, n=13). While single units represent almost half of the sample (46.4 percent, n=32), the remaining articles focus on a combination of units (see Table IV). This shows perhaps one of the complexities of conducting case studies in HL research. The humanitarian context is characterized by being a project-based setting similar to construction or events, implying that studies of temporary projects (i.e. operations) must be conducted as well as or combined with more permanent aspects (i.e. preparedness) (Jahre et al., 2009) in organizations, networks, communities or countries. Not surprisingly, many of the studies are conducted with major international organizations and very few focus on small/medium or local organizations. Rationale for unit selection is missing in most of the articles. A detailed explanation on how the unit was selected could ensure research reliability and reduce bias.

Concerning data collection, most studies are based on semi-structured or open-ended interviews (79.7 percent, n=55), which do not come as a surprise as this technique is one of the most used in case study research (Yin, 2009). However, as pointed out by Eisenhardt (1989a) case studies often rely on different sources of data. Such is the case of the articles studied in this research. With some exceptions (31.9 percent, n=22), case studies in HL ensure data triangulation from interviews, official documentation, academic and/or practitioner literature, press and direct observation. Moreover, presentation from practitioners has also been used as data sources (e.g. Kovacs and Spens, 2009; Day et al., 2012), as means to understand their perspective on a specific topic being discussed in a controlled environment, i.e. a meeting or a workshop. Nevertheless, the high percentage of cases using a single source of data can be representative of the difficulty in getting access to timely and reliable data. The question of quality regarding data triangulation must therefore be addressed, perhaps by including diverse research methods.

The last category relates to the disaster management cycle phases and geographical location. First, almost a quarter of the articles (23.2 percent, n=16) do not explicitly refer to the stage/phase focused in the study (e.g. Balcik et al., 2010; Sandwell, 2011; Makepeace et al., 2017). It could consist of more than one, or none. Similar to the findings on research design, lack of basic information regarding case boundaries lowers research quality. Another interesting result is the almost non-existent case studies of the recovery phase. Rietjens et al.’s (2014) study of reconstruction and Noori and Weber’s (2016) study on rehabilitation are the only two identified (2.9 percent). Not surprisingly, the majority of case studies are on preparedness (42 percent, n=29), followed by response, i.e. operations (24.6 percent, n=17). Preparedness topics include pre-positioning (e.g. Roh et al., 2015: Jahre et al., 2016), network optimization (Klibi and Martel, 2012; Alem et al., 2016), and vehicle fleet management (Pedraza Martinez and Van Wassenhove, 2013; Pedraza Martinez et al., 2011; Besiou et al., 2014; Kunz et al., 2015). On the other hand, coordination is the most studied issue in emergency response (cf. Akhtar et al., 2012; Sodhi and Tang, 2014; Ergun et al., 2013). During an emergency, it is difficult to capture data and reliability can be lost when asking interviewees to reflect on past events, which may explain the difference between the number of cases conducted on each of these two phases. Nevertheless, overcoming this issue by exploiting documentation related to emergency response, like situation reports, and conduct content analysis, should be possible.

When reviewing the results from the analysis, two interesting findings deserve attention. First, many articles use the term “case study” when actually such method has not been used. This can be due to what Gerring (2004, p. 352) defines as “definitional penumbra” where authors describe their research as a case study because the research method is qualitative; is ethnographic, clinical, participant-observation, or otherwise “in the field”; is characterized by process-tracing; investigates the properties of a single case; or investigates a single phenomenon, instance or example. We find some of this reasoning in articles in the sample. Others refer more to the organization and presentation of data in order to understand a context, which is more related to a case study for teaching (Ellram, 1996). This misuse could lead to generalized misconceptions of what a case study is. Defining the case is thus paramount.

Second, many of the articles do not refer to methodological frameworks when presenting the techniques used for data collection and analysis. This leads to two well-known misconceptions of case study research, i.e. “case studies do not use a rigorous design methodology” (Ellram, 1996) and “the method maintains a bias toward verification” due to lack of rigor (Flyvbjerg, 2006). While many justify the use of different techniques for case selection, sampling, data collection and analysis, and validity (cf. Table AI), it is unfortunately not the case for most studies in HL research. Presenting how the case study was conducted using methodological theory should be at the core of the research endeavor.

These two issues – from a large logistics research perspective – have also been pointed out by Gammelgaard (2017), for whom “there seems to be a lack of knowledge on what a qualitative case study is and how to conduct it” (p. 910). Despite the efforts of scholars calling for more rigor regarding case study research, particularly in logistics and SCM (Seuring, 2008; Näslund, 2008, da Mota Pedrosa et al., 2012), HL research seem to have fallen into the same path. Our results support the evidence from previous studies on purchasing and supply management literature (Dubois and Araujo, 2007), and operations management literature (Barratt et al., 2011) regarding the lack of rigor when using case studies.

In order to overcome these issues, and based on the results from this study, we propose a framework that can help crafting case study research in HL, in the same line of Gammelgaard’s (2017) guidelines on the form of “check questions.” The proposed framework was inspired by Watson’s (1994) “What, Why and How Framework for Crafting Research” (see Figure 1). Additional material that could help explain some of the gaps found in the analysis, such as the difficulty in classifying cases (Mayring, 2014), were included in our analysis.

Patton (2002) suggests that qualitative research design must be open and flexible to allow exploration of whatever the phenomenon offers for study because the design emerges also while data are collected. Hence, the proposed framework suggests that researchers can design the study by starting in any of the boxes as long as they cover all four. Iterations to adapt the why, what and how to each other ensure consistency and rigor.

The top left addresses the issue of defining the case study’s purpose. As discussed in Section 3, most HL case studies explore or describe a phenomenon, with a few used for explanation or prediction purposes. While our analysis classified the studies in these four subcategories, HL research can use case studies for other purposes as well. Siggelkow (2007) suggests three different ways. First, cases can motivate a research question. Arguing for a study using a real-life example, i.e. case, can be much more appealing than a pure theoretical motivation. Second, cases can be used as inspiration for new ideas. In the case of limited theoretical knowledge about a phenomenon, an inductive research strategy that lets theory emerge from the data can be a valuable starting point. Third, cases can be used to illustrate. Operationalizing constructs and illustration of causal relationships is a key advantage of case research compared to other empirical work. Accordingly, HL researchers can motivate the use of case studies in a number of ways. What is important is to be explicit about it.

The top right concerns contextualization of the research method, and addresses the gaps identified in two categories during the analysis concerned with identifying the study’s focus. Similar to other contexts, HL focuses on various units of analysis. Patton (2002) categorizes these as people focused (e.g. individuals, groups), structure focused (e.g. organizations, projects), geography focused (e.g. countries, regions), activity focused (e.g. crises, events) and time based (e.g. particular days, weeks or months, seasons). The author states that the key issue when selecting units of analysis is to decide what it is you want to be able to say something about at the end of the study, i.e. “what data collected during what period describing what activities will most likely illuminate the inquiry?” (p. 229). For the analysis, the categories organization, network, country and operation were defined as units in HL research, referring to structure. Projects, programs and units in organizations, should also be included and defined as units of analysis. Moreover, we used three phases in the analysis to categorize cases, but found that some studies were difficult to classify because they did not explicitly state anything regarding this. A reason for this is that in HL, case studies can investigate phenomena within a single phase, across phases or between phases (e.g. how preparedness affects response) and thus, the definition becomes difficult. Because they have profound implications for the extent to which (and to what) the results found can be generalized, unit of analysis should be explicitly described and argued. The type of disaster and its geographical location (and with this the vulnerability and coping capacity as well as probability of hazards, see e.g. INFORM, 2016) should also be included, as both criteria have big implications for preparedness, response and recovery and thus, for the studied phenomenon.

The bottom right relates to data source and analysis, and addresses the gap found in the analysis regarding lack of proper theoretical background for case selection (sampling), data collection and data analysis techniques. Most HL research use organizations as the unit of analysis and with a few exceptions, the case selection appear as spontaneous and do not follow any sort of rationale. Although HL as an emerging and rather unique field, should not mean that research cannot rely on strong methodological basis. For instance, a deliberate choice of case relates to a purposeful or judgement sampling (Patton, 2002), and thus different techniques can be used to bound the collection of data and define from whom data is collected? Some include extreme case sampling, critical case sampling, typical case sampling, theoretical sampling (Patton, 2002), snowball sampling, criterion sampling, opportunistic sampling, and convenience sampling (cf. Miles and Huberman, 1994). Further, during the analysis a number of good examples of theoretical background for research design were found (cf. Table AI). Not surprisingly, most of these studies share a similar, very explicit and detailed chain of evidence of the case study process, which contributes to rigor and quality assessment of the research (Seuring, 2008). By asking what techniques can be used to data collection and analysis the researcher has to evaluate the different possibilities and make a structured decision on the type of technique and type level of analysis that match the study’s purpose. For instance, the lack of access to sites and data may require more retrospective studies (Voss et al., 2002) in this context compared to others. Qualitative content analysis of historical data (e.g. situation reports), and experience feedback from trainings or workshops provide a viable alternative. Limitations from such a choice should be included when discussing research quality.

Finally, the bottom left addresses an aspect in relation to data collection and analysis, a feature that results significant when crafting case study research in HL. It concerns the extent to which cases are built and discussed in relation to theory. It is the opportunity to consider how existing frameworks contribute to achieve the study’s purpose and how the case outcomes contribute to the development of existing frameworks. During recent years, HL research has strived to develop a consistent theoretical framework, partially by “borrowing” theories, as shown by Tabaklar et al. (2015). These include primarily operational research- and SCM-related theories, such as systems theory, stakeholder theory and resource-based view. As stated by Kovacs and Spens (2011, p. 37), a “myriad of logistics concepts have been applied to the field,” constituting a conceptual basis that should be used for selection of interview questions and the definition of dimensions/structure for analysis. However, this endeavor falls short when aiming to contribute to the field, because research has struggled to develop an understanding for the context and operational constraints of HL (Van Wassenhove, 2010). Theory development constitutes a basis for future research. Therefore, we suggest discussing case studies in HL in relation to frameworks or theories. An iterative process is thus completed allowing enough flexibility to approach the phenomenon.

The purpose of this research was to explore the case study research in HL. For this, we conducted a content-based literature review based on selected academic logistics and SCM journals. The results from the analysis allowed us to propose a framework for crafting case study research in HL, attempting to contextualize this method by pointing out what is relevant and particular for this field. For instance, regarding purpose, HL research goes beyond the objectives of case studies and includes the idea of “use” of case studies, bringing together both OM and Logistics/SCM perspectives of the method. When it comes to boundaries of a case study, research in HL entails a broader scope regarding where phenomena is to be found in relation to specific treats of the humanitarian context like disaster phases. This determines the focus of the study, i.e. the layer from which data is gathered, that for HL can go from an organization to a program or an operation, and that can include geographical locations or even seasons. As for how data are collected and analyzed, the unique features of HL encourage using innovative techniques to overcome lack of access, but calls for a strong methodological and theoretical background to nurture the current streams used in HL research and contribute to the development of HL theory. Some limitations and avenues for further research can be identified. First, the review was limited to SCM and logistics journals. Further research could include other journals within fields such as disaster management, public management and not-for-profit management as well as transport geography. Second, in order to classify the articles, our deductive approach to category selection proved to be somewhat limiting in the analysis of a very heterogeneous sample. A replication adopting an inductive or more abductive approach could help determining new categories and confirming those used for this research. Finally, the article concludes with a framework for crafting case study research in HL. The authors encourage further refining and testing of the framework that constitutes a first attempt of contextualization of case studies in HL.

In terms of practical implications, the study provides humanitarian organizations and other actors with more understanding on how to call for and evaluate case study research. Well- documented, trustworthy and rigorously crafted case studies can provide organizations with powerful communication tools to demonstrate usefulness of new concepts, help arguing in discussing with donors and other counterparts, for training and internal and external communication.

The author would like to thanks Professor Marianne Jahre for her contribution on previous versions of this paper, as well as the two anonymous reviewers for their comments and suggestions.

Table AI