1. Introduction
Grounded theory is a qualitative research methodology used to develop theory. We describe the core tenets of grounded theory, revealing how to collect and analyse data applying its fundamental tenets as introduced by its original proponents, Glaser & Strauss (1967), but reflecting the subsequent analytical approach of Corbin & Strauss (2015). We explore divergences and debates on how to apply grounded theory in practice.
Grounded theory is a suitable research methodology to develop theory for three reasons. First, grounded theory has an established reputation for the study of human behaviour and for making knowledge claims about how individuals interpret reality (Suddaby, 2006). Second, grounded theory’s central aim is theory building, rather than theory testing. It is a suitable design when a theory does not fully explain a process (Creswell, 2007; Goulding, 2005; Thornberg & Dunne, 2019). Grounded theory facilitates recording and interpreting individuals’ subjective experiences. Through the methodological process of theoretical sampling and constant comparison, it enables abstraction of individuals’ subjective experience into theoretical statements (Fendt & Sachs, 2008). Third, grounded-theory methodology has established guidelines for conducting research and interpreting data, particularly Corbin and Strauss’s (2015) systematic approach.
Our paper contributes to the prior literature in the following three ways. First, since Elharidy et al. (2008), Gurd (2008), von Alberti-Alhtaybat & Al-Htaybat (2010) and Sutton et al. (2011), a gap has emerged in the up-to-date literature examining grounded theory in accounting research. Second, we consider the essential features of grounded theory in depth, as a valuable resource, especially for novice interpretive researchers considering adopting this method. Third, we summarise prior accounting studies using grounded theory. We hope this sensitises accounting researchers to the potential of using grounded theory as a method in accounting research.
Section 2 describes grounded theory. In Section 3, we discuss the approaches to grounded theory by the main proponents of this methodology, identifying their primary differences. In Section 4, we review grounded theory in prior research, including accounting studies using grounded theory. Section 5 discusses quality characteristics in the context of interpretative research. We conclude the paper in Section 6.
2. Description of Grounded Theory
This section reviews the origins of grounded theory, its core tenets and the divergent approaches in the prior literature.
2.1 Origins of Grounded Theory
Glaser & Strauss (1967) originally devised grounded-theory methodology. Their approach was largely a protest against (a) a methodological climate in which qualitative research was considered preliminary to the ‘real’ methodologies of quantitative research (Goulding, 2006) and (b) the positivism permeating most social research (Suddaby, 2006). Glaser & Strauss (1967) were also motivated by a desire to dismiss the myth that all good theories had been discovered and that research should focus on testing theories through quantitative empirical approaches. Glaser came from a tradition of rigorous, positivistic quantitative research learned at Columbia University. He sought to apply this training to qualitative research (Charmaz, 2000). Strauss studied at the University of Chicago with its tradition of symbolic interactionism and qualitative approaches of inquiry, such as observation and intensive interviewing: “Hence, Strauss brought the pragmatist philosophical study of process, action, and meaning into empirical enquiry through grounded theory” (Charmaz, 2000, p. 512). The two researchers devised the methodology while researching the experiences of chronically ill patients. The crux of Glaser & Strauss’ (1967) grounded theory is that the adequacy of the theory developed depends on the research process used to derive it. The theory derives concepts from the data and develops them by collecting, coding and analysing data concurrently. This approach ensures that the theory produced fits the phenomenon under investigation (Glaser & Strauss, 1967). The approach contrasts with the more traditional logical-deductive approaches, which use existing theories to generate hypotheses, and then test them empirically.
The original proponents, Glaser & Strauss (1967) define grounded theory as the discovery of theory from data. Corbin & Strauss (2008) describe grounded theory as denoting theoretical constructs derived from qualitative analysis. Both definitions reflect the same fundamental methodological principle: theoretical interpretation of a phenomenon generated from data using core methodological guidelines. Grounded-theory researchers do not commence with a theory. Theory evolves during the research process and is produced from the continuous interplay between data analysis, data collection and resulting theory (Corbin & Strauss, 2015; Glaser, 1978; Glaser & Strauss, 1967). The emerging theory leads to further data collection and analysis, further developing the theoretical constructs. Grounded-theory research seeks to make sense of the data collected to determine its meaning and significance (Parker & Roffey, 1997).
2.2 Core Tenets of Grounded-theory Methodology
The original proponents of grounded theory, Glaser and Strauss, diverged on the application of the methodology. Several variants have emerged. Notwithstanding this divergence, the approach uses fundamental elements regardless of the variant of grounded theory adopted: (i) coding, (ii) development of concepts/categories, (iii) constant comparison of data, (iv) theoretical sampling, (v) theoretical saturation, (vi) theoretical integration and (vii) use of memos to reflect researchers’ analytical thought processes. Sutton et al. (2011, p. 62) include a useful glossary of grounded theory terminology. Figure 1 reflects the application of grounded theory, including its core tenets. We describe the grounded-theory approach in seven stages.
Stage 1: Research Problem
Stage 1 in Figure 1 involves identifying the research problem. Glaser & Strauss (1967) advocate starting with a broad substantive area. Others advocate identifying a specific research problem and research question (Corbin & Strauss, 2015; Suddaby, 2006). This might involve a preliminary review of the literature and/or drawing on professional experience. Once researchers identify the research problem, they select a research methodology.
Stage 2: Field Research 1
Stage 2 in Figure 1 involves entering the field and the simultaneous collection and analysis of data, such as interview or other types of data. This stage should begin with a general target population. With interview data, researchers conduct and transcribe the first interview and then analyse the transcript line-by-line (Charmaz, 2006; Corbin & Strauss, 2015). The analytical process in grounded theory involves the use of coding strategies (known as open coding). Researchers analyse data for meaning and disaggregate them into units of meaning, labelled (coded) to generate concepts (Goulding, 2006). Concepts are the foundation for Corbin & Strauss’s (2008, p. 51) analytic method: “the categories for which data are sought and in which data are grouped; they usually become the chief means for establishing relations between data; and they are the anchor points in interpretation of finding … The use of concepts provides a way of grouping/organizing the data that a researcher is working with”. Codes denote participants’ words or incidents as concepts derived from observation. Researchers use memos throughout the analytical process (Corbin & Strauss, 2015; Glaser & Strauss, 1967). Memos comprise written records of analysis that depict relationships between analytical concepts. As such, when researchers identify codes in the data, they record their thought processes around identification in memo format. Memo writing begins with the first analysis of data and continues throughout the analytical process. Memos might include short quotes of data as a reminder of what generated a concept or idea. When it comes to writing up, researchers use much of this memo writing to illustrate the concepts. Researchers regularly update memos on individual concepts as the analysis progresses, thereby evolving into memos of greater depth and complexity (Corbin & Strauss, 2015).
Concepts identified and coded, say in the first interview, require further investigation. Researchers conduct a second interview to develop the concepts identified in the first interview. This second interview might provide additional insights into these concepts but may also yield new concepts that researchers code and explore in subsequent interviews. The principle of gathering data based on evolving concepts by alternating data collection with analysis is known as theoretical sampling (Corbin & Strauss, 2015). The objective is to develop the varying properties and dimensions of a concept (Stage 3 in Figure 1). Unlike conventional methods of sampling, researchers sample concepts in the data, not people. Researchers identify concepts and further questions for exploration through the analysis. These concepts drive the sampling process, i.e., the next round of data collection. Interviewees provide the data that elaborate on these concepts. Developing the concepts in terms of their properties and dimensions involves the constant comparison of data. As researchers collect data, they constantly compare new data to prior data for similarities and differences. This increases concept generality and explanatory power (Glaser & Strauss, 1967, p. 24).
Stage 3: Conceptual Development
In Stage 3 of Figure 1, researchers systematically develop concepts in terms of their properties and dimensions. Properties are characteristics that define and describe concepts. At the same time, researchers validate interpretations by comparing them against incoming data (Corbin & Strauss, 2015; Glaser & Strauss, 1967). Validation does not imply testing hypotheses but refers to researchers assessing interpretations both with participants and against emerging data as the research progresses. This circular process of data collection and constant comparison continues until the research reaches the point of theoretical saturation; that is, the point in the research when all the concepts are well defined and explained (Corbin & Strauss, 2015).
Corbin & Strauss (2008) offer techniques for analysing data for concepts. Table 1 summarises the analytical tools in the coding/analysis process, describing each tool and its benefits. The first technique involves the use of questioning. When analysing the data line-by-line, researchers ask questions of the data: Who, what, when, where, how and with what consequences. As concepts become more developed, researchers might question whether there is a relationship between one category and another. The second technique involves making comparisons. The process of constant comparison involves comparing each incident in the data with other incidents for similarities and differences. Incidents are then placed under the same or different codes. In subsequent interviews, incidents labelled under the same code are compared for similarities and differences (within-code comparison) to uncover the different properties and dimensions of the concept. The third technique involves drawing on personal experience. This is used when the researcher has life experiences similar to those of the participants and can use this experience as a comparative case to stimulate thinking about various properties and dimensions of concepts. The fourth technique is the flip-flop technique, which involves looking at the opposite or extreme range of a concept to bring out its significant properties and dimensions. The fifth technique looks at the language used by interview participants. On occasion, the language used can be so expressive it can translate as a code. This is called an in-vivo code, indicating the term comes from the data. Finally, researchers can look for words that indicate time (e.g., when, after, since, before, in case, if). Such words can denote a change or a shift in perception, thoughts, events or interpretations of events.
Delineating the context under which something happens is as important as identifying the right concept (Corbin & Strauss, 2008). The approach to analysing data for context is similar to the approach to analysing for concepts, in that researchers continue to question and make comparisons. Where researchers identify a concept about context, they can employ additional strategies to expand upon these contextual concepts. Corbin & Strauss (2008) provide two tools for analysing context: the paradigm and the conditional/consequential matrix. Corbin & Strauss’s (2015) conditional/consequential matrix helps researchers make connections between macro and micro conditions that influence the phenomenon under investigation. Researchers seldom use this matrix in grounded-theory work (Creswell, 2007). Table 2 summarises the paradigm approach, describing the approach and its benefits. The paradigm comprises questions applied to data to draw out contextual factors and to identify relationships between context (i.e., structural conditions) and process (response to events) (Corbin & Strauss, 2008). It suggests looking for keywords that signal a line of action or an explanation for something, then following that thought through in the data. The basic components of the paradigm are: (a) conditions (participants reveal the circumstances or conditions that lead them to take a particular course of action), (b) interactions and emotions (responses made by individuals or groups to situations and events) and (c) consequences (consequences answer questions about what happened as a result of those actions/interactions or emotional responses). Researchers use the paradigm approach to understand the circumstances that surround events, thereby enriching the data (Corbin & Strauss, 2015).
Analysing data for process is a critical step in theory building. The contents of the dataset, and researchers’ interpretation of these, determine how process is conceptualised or described. In analysing data for process, researchers try to capture how participants react to certain events or situations and how these reactions vary over time or under different structural conditions. Corbin & Strauss (2008) provide questions for analysing data for process, included in Table 3, along with their benefits.
Stage 4: Category Development
Concepts vary in levels of abstraction. There are basic-level concepts and higher-level concepts that Corbin & Strauss (2015) call categories. Categories have wider explanatory power than concepts. Researchers initially cluster concepts into descriptive categories. As the research progresses, researchers re-evaluate the concepts for their interrelationships (Stage 4 in Figure 1). Corbin & Strauss (2015) call the process of cross-cutting or relating concepts to each other axial coding. Through a series of analytical steps, researchers gradually aggregate the concepts into higher-order categories, including one underlying central or core category. These higher-order categories and the core category suggest an emergent theory. At this point, researchers may conduct a second round of field research to further validate or elaborate on the categories developed (Goulding, 2006).
Stage 5: Theoretical Integration
With grounded theory, it is vital to lift the analysis to a more abstract level, beyond description, to theory development (Corbin & Strauss, 2015). Theoretical integration involves linking categories around a central or core category and refining the resulting theoretical formulation. Categories pull together all the identified concepts into a theoretical framework. Researchers may decide to present a preliminary theoretical framework to a group of interview participants (Stage 5 in Figure 1) and/or colleagues for feedback. At this stage, researchers reflect on the framework, identify any gaps in the theory and refine as required.
Stage 4 describes the development of theoretical categories. We describe the methodological process of integrating these categories (as presented in Table 4) in this section. “Integrating means choosing a core category, then retelling the story around that core category using the other categories and concepts derived during the research” (Corbin & Strauss, 2015, p. 107). Researchers require analytical tools to lift the analysis beyond description towards theory development. Corbin & Strauss (2015) present several techniques designed to help researchers achieve theoretical integration. Researchers use these techniques in Stage 5 of Figure 1. Table 4 summarises these techniques. The first technique involves writing the storyline. Corbin & Strauss (2015) suggest researchers begin writing, in a few descriptive sentences, about what seems to be emerging from the data. In doing so, a story or description begins to emerge. The second technique involves moving from the descriptive story (or sentences) to the theoretical explanation. Once researchers identify a core category, they tell the story around this core category using the other categories (and related concepts) derived from the research. The third technique Corbin & Strauss (2008, 2015) advocate is integrative diagrams. Researchers extensively use integrative diagrams throughout the research process, especially during the theoretical integration process. Constructing diagrams enables researchers to distance themselves from the data, forcing researchers to work with concepts at the category level rather than at the level of detail contained in the numerous memos. It also forces researchers to think carefully about the logic of relationships. Diagrams focus on those categories that have reached the status of major categories. The fourth technique involves returning to the academic literature to reflect on new theory by reference to prior theory in the literature. The fifth technique involves reviewing and sorting through memos, the running logs of analytic thinking (Corbin & Strauss, 2015). Reviewing the memos reminds researchers of the thought process involved in identifying concepts and categories.
Stage 6: Contextualise in Literature
Stage 6 in Figure 1 involves contextualising the theoretical framework within the existing literature. Doing so assists in identifying the similarities and differences between the constructed theoretical framework and prior theoretical frameworks. It also serves to highlight the contribution (theoretical significance) of the study.
Stage 7: Present Core Category and Theory
The final stage involves presenting the core category and the theoretical framework in its final form.
Grounded-theory research assumes that attitudes, beliefs, norms and processes within the social world under investigation are capable of being observed and that “it is possible to generate knowledge about and evidence for them” (Mason, 2002, p. 17). Researchers can construct concepts and theories out of stories framed by research participants. Research participants try to explain and make sense of their experiences/lives, both to researchers and themselves. Out of these multiple constructions, researchers and participants together create knowledge or understanding.
Grounded theory building favours data collection methods that gather rich data directly from individuals experiencing the phenomenon (Shah & Corley, 2006). Critically, the purpose of grounded theory is not to make truth statements about reality but to make statements about how social actors/interview participants interpret reality. The purpose of grounded theory is to elicit fresh understandings about patterned relationships between social actors and how these relationships and interactions actively construct reality (Glaser & Strauss, 1967; Suddaby, 2006). Researchers place themselves in the context where the phenomenon is occurring and develop interpretations of the phenomenon based on personal experiences, as well as the experiences of those living it.
2.3 Illustrating the Application of Grounded Theory
Constructivist grounded theory captures the interplay between the form and content of data (Charmaz, 2017). Corbin & Strauss (2015) present analytical techniques (the mental strategies researchers use when coding) to use to make sense of qualitative data. They acknowledge that interpretation cannot be formulised and argue that their approach aims to teach researchers how to think more self-consciously and systematically about data. Critically, the Corbin-and-Strauss (2015) approach provides a sense of structure, process and analysis to the research, while allowing researchers to be flexible in the application of their analytical techniques. Research design should not apply grounded-theory methodology in an orthodox or fundamentalist form (Corbin & Strauss, 2015; Fendt & Sachs, 2008; Locke, 2001; Suddaby, 2006). In Figure 2, we illustrate the application of grounded theory in Stages 2 and 3 of the research from Cullen & Brennan’s (2017) approach.
3. Divergent Approaches to Grounded Theory
How the two original proponents of grounded theory, Barney Glaser and Anselm Strauss, conceptualise and operationalise the methodology has diverged. Table 5 summarises the aspects of grounded theory on which this divergence centres (also compared by Sutton et al., 2011). Strauss & Corbin (1990) provoked accusations from Glaser (1992a, 1992b) of distortion and infidelity to the central objectives of parsimony and theoretical emergence (Goulding, 2006). Glaser (1992a) documents his main objections to Strauss and Corbin’s work by reproducing much of their text with changes to reflect his perspective (see Melia, 1996 for a detailed discussion of Glaser’s rebuttal). This dispute led to the emergence of two dominant approaches, Glaserian and Straussian grounded theory. Critically, both approaches continue to adopt the core tenets of grounded theory, suggesting more similarities than differences.
Denzin & Lincoln (2000) identify four interpretive paradigms structuring qualitative research: (i) positivist and postpositivist; (ii) constructivist-interpretive; (iii) critical and (iv) feminist-post-structural. Corbin and Strauss’s version of grounded-theory research reflects an interpretive, and specifically constructivist, paradigm on building knowledge. “The constructivist paradigm assumes a relativist ontology (there are multiple realities), a subjective epistemology (knower and respondent cocreate understandings), and a naturalistic (in the natural world) set of methodological procedures. Findings are usually presented in terms of the criteria of grounded theory or pattern theories.” Denzin & Lincoln (2000, p. 21). The constructivist grounded theory is a contemporary version of Glaser & Strauss’s (1967) original statement (Charmaz, 2017). Within a constructivist-interpretive paradigm, individuals seek understanding of the world in which they live and work. The goal of research is to rely as much as possible on participants’ views and redirect qualitative research beyond positivism (Charmaz, 2000, 2017). Rather than starting with a theory, constructivist researchers generate a theory or pattern of meaning (Creswell, 2007). Constructivist researchers often address the “processes” of interaction among individuals. They focus on the specific contexts in which people live and work to understand the historical and cultural settings of the participants: “a constructivist grounded theory fosters the development of qualitative traditions through the study of experience from the standpoint of those who live it” (Charmaz, 2000, p. 522). Constructivist grounded theory has retained its pragmatist foundation through Anslem Strauss and is a direct methodological descendent of the pragmatist tradition (Charmaz, 2017).
Researchers use prior literature (theoretical and empirical) (i) to generate the study’s research problem, (ii) to make comparisons, (iii) to enhance sensitivity to subtle nuances in data, (iv) to provide questions for initial interviews, (v) to stimulate the analytical process and (vi) to confirm findings. Where concepts emerge from the data with properties similar to concepts identified in the literature, researchers examine both concepts for similarities and differences (Corbin & Strauss, 2015). Grounded-theory researchers diverge on how and when to engage with the extant literature. As noted in Table 5, Glaser (2013) remains committed to delaying the literature review until the end of the analysis. Thornberg & Dunne (2019) acknowledge the importance of the literature debate and Glaser’s role in it, arguing that the debate has highlighted potential risks associated with researchers’ unquestioning acquiescence to dominant theoretical frameworks. They argue that early engagement with the existing literature offers benefits which far outweigh the drawbacks. While Corbin & Strauss (2015) do not advocate entering the field with an entire list of concepts, they acknowledge that certain concepts identified in the literature may emerge from the data, thus demonstrating their significance. If this happens, they suggest that researchers ask themselves whether the concepts were truly derived from the data or imposed by researchers on the data due to their familiarity. Where there is a glaring discrepancy between the research findings and the findings in the prior literature, the research findings require further investigation through theoretical sampling. In this way, researchers use the prior literature to stimulate questions during the analysis process.
4. Prior Literature on the Application of Grounded Theory
To avoid the risk of (a) becoming overly consumed by the methodological debates on grounded theory and (b) ignoring the pitfalls in its application, we explore literature and opinion on the application of grounded-theory methodology in prior management research (Fendt & Sachs, 2008; Goulding, 2005, 2006; Locke, 2001; Shah & Corley, 2006; Suddaby, 2006). Suddaby (2006) warns against methodological slurring. In research using a grounded-theory approach, there should be consistency between the research problem, the research questions and the methods used to answer these questions (Suddaby, 2006). Several academics have reflected on the importance of methodological disclosure and demonstrating the process surrounding the use of grounded-theory methodology (Fendt & Sachs, 2008; Seale, 1999; Shah & Corley, 2006; Suddaby, 2006). For example, Suddaby (2006) points to the poor presentation of research methodology in papers purporting to adopt a grounded-theory approach. He argues that grounded-theory methodology should be transparent enough to demonstrate that researchers followed the core analytic tenets (i.e., theoretical sampling, constant comparison, theoretical saturation) in generating the research. Readers can then assess how researchers used the data to generate key conceptual categories. He suggests that researchers make apparent to readers the process of data analysis, including coding techniques and category creation, in the methodology section. Also, the research should provide illustrative examples of coding techniques and the evolution of conceptual categories in a table or appendix. Suddaby (2006) further refers to papers that begin with an interesting question, are written well and follow a well-constructed methodology but present incomplete data and/or obvious findings. He suggests that this is the result of one, or a combination, of three errors in the application of grounded theory: (i) confusion between grounded theory and phenomenology; (ii) a failure to ‘lift’ data to a conceptual level due to incomplete analysis of the data; (iii) or the absence of sufficient data.
4.1 Grounded Theory in Prior Accounting Research
Elharidy et al. (2008) and Gurd (2008) review grounded theory applied in an accounting context. von Alberti-Alhtaybat & Al-Htaybat (2010) describe their experiences applying grounded theory in practice. Covaleski & Dirsmith (1983) justify adopting grounded theory as their object is generating theory. Gibbins et al. (1990) highlight the benefit of grounded theory in its ability to describe the experiences of decision makers. Elharidy et al. (2008) highlight the benefit of developing theory grounded in everyday practices. In Table 6, we summarise a selection of papers using grounded theory in prior accounting research. While grounded theory features in prior accounting research, it is not common but is increasing. A challenge to publishing this kind of research is the positivist hegemony in accounting’s “mainstream” journals. The method continues to encounter scepticism due to its perceived lack of rigour and credible findings. Gibbins et al. (1990) is a ground-breaking grounded-theory study, in that the paper is published in the highly positivist Journal of Accounting Research. Many would find Gibbins et al.'s (1990) application of grounded theory quite positivist in style. Table 6 shows that researchers have applied grounded theory to a wide range of topics in accounting. Such studies appear in a wide range of journals and are especially favoured in Accounting, Auditing & Accountability Journal. Grounded-theory studies rely primarily but not solely on in-depth interview methods. Researchers adopt a range of forms of grounded theory, from Glaser & Strauss’s (1967) pure form to later more structured and prescriptive forms. Most studies provide the theory generated in the form of diagrammatic theoretical frameworks.
5. Research Quality: Criteria for Evaluation
Researchers can experience difficulty in publishing qualitative research. This is because inexperienced researchers do not understand the unique quality characteristics applying in qualitative research compared with quantitative research. With grounded theory, terms such as credibility, applicability, transferability, dependability, and confirmability replace the usual positivist criteria of internal and external validity, reliability, generalisability and objectivity. When considering research quality, the main proponents of grounded theory do not favour terms such as “validity” and “reliability” (Corbin & Strauss, 2015; Glaser & Strauss, 1967), preferring instead to use the term “credibility” (Corbin & Strauss, 2015; Glaser & Strauss, 1967). They argue that researchers cannot apply the same criteria across qualitative methodologies. For Corbin & Strauss (2015), quality of findings and validity of findings are not synonymous. They too are uncomfortable using the terms “validity” and “reliability” when discussing qualitative research. They prefer the term “credibility” over the term “truth”. For Corbin & Strauss (2008, p. 302), “‘credibility’ indicates that findings are trustworthy and believable in that they reflect participants’, researchers’, and readers’ experiences with a phenomenon but at the same time the explanation is only one of many possible ‘plausible’ interpretations possible from the data”. Corbin & Strauss (2008, p. 302) describe quality qualitative research as that which “resonates with readers’ and participants’ life experiences…..that blends conceptualisation with sufficient descriptive detail to allow the reader to reach his or her own conclusions about the data and to judge the credibility of researchers’ data and analysis…that stimulates discussion and further research on a topic”. Drawing on research methodology literature, Corbin & Strauss (2015) list general criteria for evaluating the quality of research findings.
6. Conclusion
We outline the approach to methodological analysis and disclosure for the ‘non-sequential’ (Suddaby, 2006) steps followed using a grounded-theory methodology. Our outline reflects the core tenets of the original grounded-theory methodology (Glaser & Strauss, 1967), the systematic analytical approach of Corbin & Strauss (2015) and the recommendations of experienced researchers around the application of the methodology in research. The grounded-theory perspective, as conceived by Glaser & Strauss (1967), is the most widely used qualitative approach in the social sciences (Denzin & Lincoln, 2000). Grounded theory is an established methodology. The research design we describe in this paper reflects not only the core tenets of the original grounded theory, but also current thinking on the application of the methodology in management research. As such, it represents an improved understanding of grounded theory in management research.