Extended qualitative content analysis: researching the United Nations and other international institutions

Abstract

Purpose

This paper aims to introduce the extended qualitative content analysis (EQCA) method to integrate data-reducing and data-complicating research steps when conducting qualitative research on the United Nations and other international institutions.

Design/methodology/approach

EQCA supplements the method of qualitative content analysis, which enables researchers to deal with large amounts of data, with two elements from grounded theory, which allow detailed analysis and interpretation of codes and sub-codes. The elements in question are axial coding and theoretical sampling.

Findings

EQCA provides a method to generate middle-range theories by combining theoretical and empirical analysis to address and theorize the complex interactions between actors, structures and norms in international institutions. The value added by the proposed method is demonstrated with a case study of a United Nations intergovernmental working group in the issue area of business and human rights.

Originality/value

Based on the concepts of interpretation and social causality, this paper contributes to the body of qualitative research that transcends the dichotomy between positivist and post-positivist approaches in the disciplines of international relations and international political theory.

Citation

Mende, J. (2022), "Extended qualitative content analysis: researching the United Nations and other international institutions", Qualitative Research Journal, Vol. 22 No. 3, pp. 340-353. https://doi.org/10.1108/QRJ-11-2021-0127

1. Introduction

When conducting qualitative research, scholars of international relations (IR) and international political theory (IPT) usually have to choose between data-reducing and data-complicating methods. The former methodology allows researchers to systematically deal with large amounts of data, but may have limited applicability in terms of addressing and theorizing the complexities at hand. The latter provides rich resources for building inductive theories that help the researcher “move beyond a model of science that views simplicity, coherence and reduction as primary goods. We need models of science able to incorporate the chaotic complexity of the international system” (Wight, 2009, p. 294). However, data-complicating approaches require intensive time and resource commitments, and are difficult to apply to vast databases such as those maintained by international institutions.

The divide between the two methodologies is partly due to the separation of qualitative research methods in IR and IPT into positivist and post-positivist approaches. Advanced methodological discussions in the social sciences and more recent discussions in IR and IPT bridge these two approaches. Building on these, this paper suggests combining the advantages of each (positivist and post-positivist as well as data-reducing and data-complicating) into a method it calls extended qualitative content analysis (EQCA). EQCA supplements qualitative content analysis (QCA) [1] with elements of grounded theory methodology for two purposes: (1) to facilitate dealing with large amounts of data and (2) to allow in-depth analyses that enable theory building beyond the falsification or verification of initial hypotheses. This new method can be applied to written data such as documents and transcribed interviews. EQCA thus complements other approaches that integrate data-reducing and data-complicating methods, with a particular focus on (but not limited to) international institutions.

The paper proceeds by outlining the ostensible gap between positivist and post-positivist approaches, which is particularly visible in the IR and IPT literature (Section 2), and how it can be bridged with the concepts of interpretation and social causality (Section 3). Section 4 introduces the EQCA approach, extending the coding frame from QCA with axial coding and theoretical sampling from grounded theory. Section 5 discusses the detailed research steps required to conduct EQCA, exemplified by a study on a United Nations (UN) forum, though the method is applicable to other international institutions and contexts as well.

2. The methodological divide in qualitative research

A dichotomy in the IR and IPT literature divides qualitative research into two strands. This divide is visible in other social sciences as well, as Gläser and Laudel point out:

One strand in this discussion is concerned with the question how causal arguments can be made with qualitative data. This strand […] just assumes that the data are there, i.e. can be produced in the form necessary for theoretical analysis. […] A second strand of the methodological discussion is focused on the ways in which qualitative data (texts and pictures) can and should be analyzed but is rather vague about what such an analysis is supposed to achieve. […] Theory building does not occur […] and generally appears to play a minor role (Gläser and Laudel, 2013; §6f., also cf. Stake, 2010; Goertz and Mahoney, 2012).

The first of the two strands in IR and IPT is referred to as positivist, explanatory or empiricist research (Burchill et al., 2005, p. 3; Klotz and Lynch, 2007, p. 11; Lamont, 2015, p. 17ff.). It strives to produce facts and explanations by identifying factors and causalities that are given in the real world. Its criteria for reliability and validity resemble those found in quantitative research. In their famous book on the matter, King, Keohane and Verba emphasize that “‘qualitative’ research is [not] fundamentally different from ‘quantitative’ research, except in style” (King et al., 1994, p. 5).

The second strand is known as post-positivist or interpretive research (Burchill et al., 2005, p. 3; Lamont, 2015, p. 15; Klotz and Lynch, 2007, p. 11). It seeks to understand how the real world and its meanings have been constructed in the first place by investigating identities, norms and ideas in discourse, metaphors, pictures, narratives and practices, for example.

3. Bridging the divide: interpretation and social causality

In response to the gap resulting from the divide between the two strands (Hollis and Smith, 1994, p. 244; Lamont, 2015, p. 17), a growing body of international studies research rejects this dichotomy between positivist and post-positivist approaches [2]. Instead, it seeks to develop ways to combine the two by pointing out that both methodologies make use of interpretation, and both can contribute to explanations: “Certainly the terms are not interchangeable, but in practice there is considerable overlap. Those who say they explain behavior also interpret meaning, and those who focus on understanding language also explain action to some degree” (Klotz and Lynch, 2007, p. 15, emphasis added; also, cf. Klotz, 2009, p. 1; Wight, 2009, p. 29f.).

This paper contributes to this body of literature by integrating both approaches on the basis of two aspects emphasized in qualitative research scholarship: interpretation and social causality.

First, while the interpretation of data is usually connected to interpretive (post-positivist) approaches, positivist approaches are interpretive, too. “There is no such thing as purely descriptive, a-theoretical analysis, since all description involves selection and interpretation of meaning according to implicit, informal theories-in-use” (Spencer et al., 2003, p. 201, also, cf. Schreier, 2012, p. 2; Ginger, 2006). Even artifacts and documents that appear to simply be given do not deliver objective facts. Rather, they are “social products” that reflect the interests, positions and values of their authors or their institutional context (Hammersley and Atkinson, 2007, p. 130). They contain both intended and unintended values and meanings (Hitchcock and Hughes, 1995, p. 231, cf. Saldaña, 2012, p. 54ff; Bazeley, 2014, p. 337f.). Accordingly, “official documents are a site of claims to power, legitimacy, and reality” (Lindlof and Taylor, 2011, p. 232). The “institutional settings in which they are constructed, interpreted, and used” affect their character, scope and content (Miller, 1997, p. 78). The analysis of data is both context-dependent and open to multiple interpretations.

However, there is a habit of omitting the interpretive dimension of positivist research. International studies tend to treat the data collection process as a method in itself (e.g. by referring to “conducting interviews”). Some refer to “document analysis” as a general frame, rather than a method of interpretation; others only state the type of computer-assisted software they use. The pivotal process of how data are interpreted is only described with a reference to “coding” – without specifying the ontological, epistemological or methodological choices that were made in deciding between different types and procedures of coding and decoding, or how the results were treated.

One reason for these omissions is that interpretation “is an ill-structured activity for which no algorithm can be provided. At the same time, the widespread reluctance to define intermediary steps and their outputs makes it often difficult to assess the contribution of a specific method” (Gläser and Laudel, 2013, p. §2). However, it is disclosing, rather than neglecting, the interpretation process that enables the transparent documentation and reflection of qualitative research, and that secures a study's validity, intersubjectivity and reliability (cf. Strübing et al., 2018; King and Horrocks, 2010, p. 160ff; Barbour, 2014; Burns, 1989).

The second concept that bridges the divide between positivist and post-positivist research is social causality. It provides the basis for the claim that post-positivist approaches can be explanatory, too (Bevir, 2006; Soss, 2006). This is because “human behavior is intentional and meaningful” (Bazeley, 2014, p. 329). Action, agency and actors' intentions interact with comprehensible – and thus explainable – norms and ideas, and with institutional designs and logics. Causes are understood as social relations, rather than mechanical or deterministic deductions. Social causality can, therefore, hardly be reduced to a one-dimensional conclusion that “if a, then b”; nor is it prone to be replicated in experiment-like settings. Rather, social causality can be revealed in relations and patterns, in motivations and reasons, in effects and implications. It entails multi-causal relations between a and b. As Miles explains:

We consider qualitative analysis to be a very powerful method for assessing causality. […] Qualitative analysis, with its close-up look, can identify mechanisms, going beyond sheer association. It is unrelentingly local, and deals well with the complex network of events and processes in a situation. It can sort out the temporal dimension, showing clearly what preceded what, either through direct observation or retrospection. It is well equipped to cycle back and forth between variables and processes – showing that “stories” are not capricious, but include underlying variables, and that variables are not disembodied, but have connections over time (Miles and Huberman, 1994, p. 147; emphases deleted, similarly Bazeley, 2014, p. 327; Spencer et al., 2003, p. 205).

Thus, social causality extends the notion of explanation that grounds in positivist hypothesis testing by also incorporating interpretive research that analyzes inter-subjective, social and normative meanings and relations.

Drawing on the concepts of interpretation and social causality, this paper introduces EQCA as a method that uses both the interpretive and explanatory power of qualitative research.

4. Extended qualitative content analysis

EQCA is designed to integrate positivist and post-positivist approaches to both systematically reduce the data and add to their complexity as a basis for theory building. It thus extends the data-reducing qualitative content analysis (following Schreier, 2012) with two data-complicating elements from grounded theory, namely, axial coding and theoretical sampling [3]. While the former makes the data more manageable, the latter generates empirical and theoretical heuristics and reflection (Coffey and Atkinson, 2013, p. 30) to address and theorize complexities.

Mixing methods in this way must be undertaken with care, since not all methods are compatible (Tesch, 1995, p. 115). This paper suggests that QCA is compatible with elements from grounded theory for two reasons. First, QCA's flexibility and openness allow it to be used as a methodological toolbox that can be modified, extended and combined with other approaches (Schreier, 2014, p. 22ff.) [4]. The second reason is that grounded theory and QCA share a number of elements, including the plurality of data sources, the pivotal role of coding, [5] the application of several coding cycles and the development of categories and subcategories from the empirical data.

At the same time, QCA and the grounded theory approach have different epistemological assumptions and practical implications. QCA is designed to deal with a large amount of data in a systematic way. It reduces the data and structures them coherently in a coding frame. The “aim of QCA is to systematically describe the meaning of your material” (Schreier, 2012, p. 3) by classifying the data, identifying variables and developing categories and subcategories.

By contrast, grounded theory aims to complicate the data by exploring their complexity. The methodology of grounded theory [6] entails a plural set of “systematic, yet flexible guidelines for collecting and analyzing qualitative data to construct theories ‘grounded’ in the data themselves” (Charmaz, 2006, p. 2).

While QCA can be conducted either inductively or deductively (Schreier, 2014; Steigleder, 2008; Mayring, 2000), the grounded theory approach is an inductive approach characterized by its openness to all kinds of new data, even during the coding and analysis process. Hence, grounded theory contains elements of uncertainty and unpredictability in terms of resources, time planning and theoretical output. When applied properly, [7] it is very resource intensive. This paper therefore takes QCA's coding frame as a data-reducing starting point and incorporates the theory-building power of grounded theory by integrating two of its elements, axial coding and theoretical sampling, resulting in the EQCA framework.

4.1 The coding frame

At the heart of QCA lies the construction of a consistent coding frame that includes selected categories and their subcategories. This frame describes the meaning of the categories and identifies their timing, frequency and co-occurrence. The approach even facilitates capturing “latent meaning, meaning that is not immediately obvious” (Schreier, 2012, p. 15; cf. Kracauer, 1952). Yet, coding is prone to mistakes, projections, bias, self-reinforcing presumptions and unreflected presuppositions. As Schreier explains, “one reason why just reading through your material is not enough is that we invariably perceive the world selectively” (Schreier, 2012, p. 128). Accordingly, codes are not simply objectively performed reductions of data. Rather, they add value and interpretive meaning to the data: “a code is a researcher-generated construct that symbolizes and thus attributes interpreted meaning to each individual datum for later purposes of pattern detection, categorization, theory building, and other analytic processes” (Saldaña, 2012, p. 4).

Main categories and subcategories form the structure of the coding frame. The main categories specify relevant aspects of the data that may already be included in the research question (Schreier, 2012, p. 59). They are further differentiated into subcategories that capture their different dimensions.

In accordance with QCA, EQCA constructs a coding frame and codes the data in several cycles. It may initiate coding deductively using comprehensive main categories developed in previous theoretical work. This step reduces and conceptually sorts the data into “meaningful categories” (Coffey and Atkinson, 2013, p. 36), thereby focusing the research when there is a vast amount of data available (Schreier, 2012, p. 128).

In the subsequent coding cycles that use elements of grounded theory, the data are further scrutinized for dimensions, patterns, causes, explanations and effects, in a departure from the QCA frame.

4.2 Axial coding

Axial coding means to “code intensively and concertedly around single categories. By doing this, the analyst begins to build up a dense texture of relationships around the ‘axis’ of the category being focused upon” (Strauss, 1987, p. 64). This type of coding works with the previously identified concepts to interpret and explain them. It involves searching for relationships between and within subcategories, such as causes, actions and interactions – and inactions, if possible – and for norms and rules, deviance, effects, outcomes and consequences (Strauss and Corbin, 1998, p. 131ff.). It emphasizes the process of a phenomenon (asking “who” and “how”), its structure (“why”) and the inextricable link between the two (Strauss and Corbin, 1998, p. 127). Axial coding not only identifies causal relationships; it also helps elucidate patterns, connections and structures between or within subcategories.

In the context of the grounded theory approach, axial coding is undertaken after several cycles of open coding, which generate descriptive categories by identifying and decoding phenomena in the data (Strauss and Corbin, 1998, p. 101ff.). Hence, axial coding works with decoded data. This allows axial coding to be integrated into the EQCA frame as a follow-up to the coding frame. At the same time, as Tesch points out:

While the coding in descriptive/interpretive research proceeds according to topics (for example according to the question: does this segment of text represent an instance of the topic […]), the code in theory-building research usually has to give an indication of the content of the segment as well (Tesch, 1995, p. 125).

Axial coding, therefore, extends the coding frame and sometimes redefines its initial subcategories.

4.3 Theoretical sampling

Theoretical sampling involves the subsequent search for, coding and analysis of further data. It aims to multiply and differentiate subcategories, including their relationships, negative cases and possible gaps, to “maximise opportunities to discover variations among concepts and to densify categories” (Strauss and Corbin, 1998, p. 201). Theoretical sampling differs from other sampling methods, in that it is not focused on varying the persons, groups or cases in the sample; rather, it seeks to include a rich variation of concepts and their dimensions, aspects, conditions and effects. This may involve integrating new data (such as further documents or interviews), but new dimensions can also be found in the original data when a new perspective is applied. A central reason for this step is that the data may contain gaps, particularly in research involving official documents and strategic or diplomatic discourse.

While a purposeful sample is selected at the outset of the study for a predetermined purpose, theoretical sampling progressively and systematically tailors data collection to serve the emergent theory. Theoretical sampling is thus always purpose-driven; the sample is selected for the purpose of explicating and refining the emerging theory (Breckenridge and Jones, 2009, n.p.).

Theoretical sampling is applied until saturation is reached, i.e. “when no new properties, dimensions, conditions, actions/interactions, or consequences are seen in the data” (Strauss and Corbin, 1998, p. 136). The theoretical sampling step is undertaken after several cycles of coding, once the draft categories and subcategories have been established, which allows it to be integrated with the EQCA's extended coding frame.

In EQCA, the initial coding cycles develop the coding frame, and axial coding extends this frame by further interpreting a chosen range of phenomena. At the point of theoretical sampling, the analysis switches from refining the coding frame to establishing a theoretical framework, since an entire coding frame cannot easily be re-assembled (Schreier, 2014, p. 128). Hence, new insights from theoretical sampling inform the theoretical framework rather than the coding frame itself. In sum, EQCA can capture phenomena not only by asking “what,” but also “how” and “why,” thereby enabling one to refine existing theories and develop new ones.

5. Extended qualitative content analysis in the United Nations

This section explains how to conduct EQCA step by step. Section 5.1 describes how the UN is a particularly relevant research context for IR and IPT scholars, but EQCA can be applied to other contexts and international institutions as well. Section 5.2 identifies eight research steps, which proceed from raw data to a theory in a manner that strives to be valid, transparent, reliable and comprehensible. In practice, the process is reiterative and moves back and forth between the steps.

To illustrate the research steps, this section uses examples from a case study that was conducted in the research project “Business Actors beyond Public and Private: Authority, Legitimacy and Responsibility in the United Nations Human Rights Regime” (BAPP) [8], which analyzes the deliberations about business responsibilities for human rights in a UN forum with the mandate to draft a binding treaty on business and human rights (known as the Treaty Process).

5.1 The United Nations context

International institutions consist of norms, programs, rules, networks, processes, structures and actors. They are marked by “their principled and shared understandings of desirable and acceptable forms of social behavior […], a strong element of intersubjectivity” (Kratochwil and Ruggie, 1986, p. 764, also Johnston, 2001, p. 492). While international regimes combine issue-specific norms, rules and principles with normative and behavioral expectations, international organizations are formal, intentional actors with clear rules of membership and decision-making procedures that transcend issue areas (Rittberger and Zangl, 2010, p. 7).

The UN, one of the most encompassing international institutions, is an organization that hosts various international regimes simultaneously. It entails organs, fora, programs, networks, principles and norms, as well as several kinds of formal and informal participation mechanisms of states and non-state actors. To account for the fact that its agents are reflexive actors and its structures are institutionalized processes, this paper draws on the perspective that agents and structures are interdependent and mutually constitutive. Agents constitute and reproduce structures, and are in turn shaped (i.e. enabled and limited) by these structures, mediated by intersubjective processes, ideas and norms (Mende, 2016, p. 49ff.). This basic conceptual understanding informs constructivist and neo-institutionalist approaches to international institutions (Wight, 2009, p. 296; Adler, 1997; Wendt, 1987). They emphasize “the role of ideas in constituting political action, the power of persuasion in political debate, the centrality of deliberation for democratic legitimation” (Schmidt, 2010, p. 2). Actors are “not only able to think, say, and act but also to think about their thoughts, reflect upon their actions, state their intentions, alter their actions” (Schmidt, 2010, p. 17). Their actions and interactions are embedded within institutions, in which normative and behavioral expectations converge through processes of socialization, persuasion and social influence (Johnston, 2001, p. 492). Institutions not only provide a ground for shared understandings and shared knowledge; they also make “acting a particular way public and observable” (Johnston, 2001, p. 502).

These factors are what make deliberations (and other practices) in international institutions conducive to qualitative research with the aim of revealing and interpreting social causalities. Integrating data-reducing and data-complicating research will not capture all aspects of the relationships among agents, ideas and structures in international institutions. But, EQCA integrates its various elements into a theoretical framework that links structures and agency and incorporates norms and ideas.

5.2 Steps of extended qualitative content analysis

EQCA involves eight steps of research (Figure 1). This section describes each step in turn.

• Step 1: Categorization of documents

International institutions produce a large amount of data documenting their processes, mechanisms, discourses and results, which is gathered in databases [9]. Since the UN is a multilateral intergovernmental organization, only states are official members and have the right to vote. Yet, non-state actors also play a pivotal role in various UN fora, where they may participate in deliberations and have a significant influence on decision-making processes. Therefore, a decision must be made about which documents to include in the analysis. This decision requires knowledge of the type of deliberation and the decision-making process used in the case of interest. Attention may also be given to documents that have been filed outside the forum at hand, or to actors that boycott or are excluded from the forum.

Once the documents about a certain conflict, from a certain forum and/or a certain time period, are collected, they are sorted and categorized. Each document is classified with variables [10] such as the forum it is relevant to, whether it was submitted directly to this forum or filed externally to it, the time/year it was submitted and/or discussed, the type of actor who submitted it and the stance it displays on the matter of conflict of interest.

In the BAPP project, the deliberations take place in an Open-Ended Intergovernmental Working Group (OEIWG) that documents all written statements and submissions by all stakeholders on its webpage [11]. In this case study, the documents were supplemented with statements from outside the forum to include categorical rejections of the OEIWG as well as more extended explanations of comments by stakeholders, resulting in a corpus of approximately 1,000 documents. These documents were classified with variables including basic factors such as the year of submission, factors that sometimes required more detailed background research such as the type of actor (state, non-governmental organization (NGO), individual expert, UN forum, other international organization, union, business actor) and, finally, the document's normative evaluation of the Treaty Process (positive, negative, neutral/mixed), which was more difficult to assess and may change over time. This dataset, including its variables, has been published (Mende, 2020) to make the analysis transparent.

• Step 2: Coding frame

The initial coding cycles are applied to construct a coding frame. QCA suggests the development of a coding frame based on a representative sample of the data (Schreier, 2012, p. 58ff.). This frame may be either deductively derived from theoretical assumptions or inductively identified based on the data. In each scenario (and in approaches that employ a mix of both), the coding frame consists of categories and subcategories, though their number may vary (Schreier, 2012, p. 65ff.). The coding frame is tested and further refined until it differentiates between the categories and subcategories. After saturation, the framework is applied to the entire data set.

The BAPP project included several case studies, which allowed the main categories to be deductively coded. One of the case studies traced how, why and with what intent the statements addressed the regulation of supply chains. A first deductive coding cycle coded all segments that addressed supply chains (including synonyms). Further coding cycles established a second tier of subcategories, capturing the normative position for or against integrating supply chain regulation into a future treaty (including neutral or mixed positions). A third tier reveals the dimensions of these positions, including the differentiation between single elements of supply chain regulation and the reasons for each position.

• Step 3: Axial coding

After the data have been reduced in Step 2, the axial coding step adds complexity. This step seeks to identify dimensions, causes, effects, consequences and implications in the coded segments – in a conceptual and explanatory (rather than descriptive) way (Strauss and Corbin, 1998, p. 124f.). It is conducted by asking the how, when, where, who and why of (selected) concepts or phenomena. “Answering these questions helps us to contextualize a phenomenon, that is, to locate it within a conditional structure […], to relate structure with process” (Strauss and Corbin, 1998, p. 127, emphasis added). This step detects interactions among agents, ideas and structures by examining both the categories established in the coding frame and the variables that capture institutional and structural positioning within the UN. The results from this step extend the coding frame in the form of analytical text passages and memos.

The BAPP case study identified common denominators and points of reference among the normatively very different positions on supply chain regulation, including the question of who bears responsibility for violating human rights in supply chains.

• Step 4: The extended coding frame

The extended coding frame registers how axial coding augments the initial coding frame – not by reassembling it, but by supplementing parts of it with further insights. It addresses relationships between the variables from the classification process (Step 1), the categories and subcategories of the coding frame (Step 2) and the causal and explanatory relationships detected during axial coding (Step 3). This step is concerned with the identification of relationships, key themes and/or patterns that permit the construction of typologies and matrices. Strategies for identifying said relationships, key themes and/or patterns include ordering the data according to the classificatory variables, looking for deviations or extreme/negative cases, making comparisons within and across cases, looking for differences or similarities, making comparisons according to context and time, rearranging data, looking for correlations by testing combinations of categories and variables, and evaluating the outcomes (Bazeley, 2014, p. 254ff.). All of these strategies rely heavily on working with analytical memos, and going back and forth several times (Bazeley, 2014, p. 279).

In the BAPP project, connections were drawn, for example, between an actor's evaluation of the Treaty Process and their positioning toward supply chain regulation. This connection was additionally triangulated with the type of actor in question, showing that supply chain regulation was primarily embraced in documents stemming from unions, and mostly rejected in statements by business actors. Furthermore, the analysis demonstrated that companies mainly address states as the bearers of responsibility, while NGOs target companies as the actors responsible for supply chain regulation.

• Step 5: Gaps and open questions

In a step closely related to (and in practice, often addressed simultaneously with) Steps 3 and 4, the extended coding frame is examined for questions and gaps, as well as possibilities for further differentiation or refinement and blank spaces motivated by political, strategic or other concerns. UN documents may contain strategic or political language that deviates from the discussions and ideas that initiated them. Some blank spaces or double meanings might be interpreted based on previous theoretical knowledge or observations made in the field; others may be latent and only discovered by coding and scrutinizing the data.

In the BAPP project, the triangulation between types of actors and their respective justification revealed how common references to shared values or meanings were nonetheless often marked by diverging interpretations (also, cf. Mende, 2021). For example, both sides jointly acknowledged the complexity of supply chains – only to then argue conversely either for or against supply chain regulation. This made further understanding of how supply chains can be regulated (and by whom) necessary.

• Step 6: Theoretical sampling

In the theoretical sampling step, further data are gathered to address and explore the remaining gaps or open questions. In the UN, these may include shadow reports or other contributions from NGOs that accompany or evaluate a certain process or forum but might not officially be part of it. Additional data can also be gathered through interviews with relevant actors and stakeholders. Interviews have the strong advantage of allowing tailored questions that focus on the previously identified issues and blank spaces. They also allow one to address topics and perspectives that do not find their way into documents for political or strategic reasons, but that nonetheless guide behavior and decisions. Conferences or regular meetings of the UN forum of interest potentially allow the researcher to combine interviews with field observations. However, they pose their own challenges, which include obtaining access to the meeting (not all UN fora are open to the public), finding the time and room for interviews during a busy conference and the missing voices of stakeholders that do not attend. The choice of the data is oriented toward theoretical saturation, as is the choice of questions and guidelines for the interviews. Ideally, the data collection process is guided both by the gaps identified in Step 5 and insights gained from sifting through the data while they are being gathered. This process allows new relevant issues and perspectives to emerge.

The data analysis in the BAPP project was accompanied by the participation in OEIWG meetings and, in addition, in related UN fora, to capture the perspectives of relevant actors that were not part of the Treaty Process. Additional 20 interviews were conducted, guided by the variables from Step 1, which covered each type of actor and each position on the Treaty Process. Furthermore, the existing dataset was re-assessed for new categories from Step 3, e.g. the references to complexity.

• Step 7: Coding of new data

The new data are classified and coded according to Steps 1 to 3; however, in this penultimate step, axial coding is emphasized over the coding frame. Proceeding from the established coding frame, this step explores how to integrate the new findings into the extended coding frame. It might be possible to integrate some results in this way; others will fall outside the frame and must be captured in accompanying text passages or memos. In any case, at some point, the analysis will provide the basis for a theoretical framework.

In the BAPP project, two main categories stood out: the category of complexity and the responsible entity (state or company). While the latter could be integrated into the extended coding frame, the meanings and applications of complexity were explored beyond supply chain discussions and thus left the coding frame behind, opening up a new and unexpected avenue of research.

• Step 8: Theoretical framework

Here, the findings from Step 7 are incorporated into the theoretical framework, thereby refining and differentiating, extending, validating, correcting and changing it. This step produces a theoretical model with explanatory power that can be accommodated with previous theoretical assumptions and/or serves as the basis for the development of a middle-range theory.

The two main categories in the BAPP project were assessed using two different theoretical frameworks. The question of which actor is made responsible for supply chain regulation was integrated with the information regarding positioning, reasoning and elements of supply chain regulation found in the data and subsequently connected with other discussions in the field of business and human rights. This analysis was able to demonstrate that even in a context that features highly contrasting positions, a complementary model of human rights responsibility that includes both states and companies can present a common basis for further discussions. The category of complexity, however, was shown to be applicable beyond supply chain regulation. The same pattern of using complexity emerged for a number of other issue areas as well: complexity was acknowledged, but with different consequences regarding the assessment of a treaty as impossible or necessary, thereby either legitimating or delegitimating the Treaty Process. Complexity was also identified as a performative practice that demonstrates solutions to the issue of said complexity by dealing with the problems at hand, without necessarily stating whether the Treaty Process can live up to such expectations. Therefore, the analysis of complexity inductively provided a possible explanation for why documents that were coded as neutral or unclear toward the Treaty Process still offered productive solutions to strengthening it.

6. Conclusion

The paper proposes the EQCA method to conduct qualitative research on international institutions. It focuses on the UN as a hub of global politics that encompasses actors and structures as well as norms, ideas and law. EQCA is a simultaneously data-reducing and data-complicating approach that facilitates the construction of middle-range theories. It extends QCA with steps from the grounded theory approach, thus advancing research that transcends the dichotomy between positivist and post-positivist perspectives. The EQCA method may start with theoretical assumptions, but these assumptions or theoretical constructs are not only verified or falsified through empirical testing. Rather, the empirical analysis provides theoretical constructs with dimensions, relations and explanations that enrich – and may even transform – the initial assumptions. Theory building in this sense is not restricted to inductive approaches that construct theories based on the available material. Nor is it about testing hypotheses. Rather, it generates theoretical models that are based on interpretation and explanation, as well as theoretical and empirical analysis. These are conducive to studies in the field of international politics and beyond.

Figures

Figure 1

Steps of EQCA

Notes

1. While the abbreviation QCA can refer to either qualitative content analysis or qualitative comparative analysis, the paper only refers to the former.

2. While it does not necessarily refer to the terms outlined above, this body of literature includes, for example, Aradau and Huysmans (2013), Yanow and Schwartz-Shea (2006), Jackson (2011), Mearsheimer and Walt (2013), Finnemore (1996) and much earlier Dessler (1989), Carlsnaes (1992), Ragin (1987).

3. Gläser and Laudel (2013) similarly aim to combine exploratory and explanatory qualitative analysis, but set other foci and combine elements of QCA and grounded theory differently.

4. However, there is controversy regarding which variants of QCA allow further explanatory research steps. Cf. Schreier (2012, p. 4).

5. Coding is understood as a general tool to assess qualitative data. It is not a priori inherent to a certain method. Cf. Saldaña (2012).

6. This paper uses the Straussian version of grounded theory in Strauss and Corbin (1998). For different versions of grounded theory that diverge from its initial formulation in Glaser and Strauss (1967), see Morse and Niehaus (2009, p. 95).

7. For criticisms of the inflationary but inadequate references to the methodology of grounded theory, see Suddaby (2006), Benoliel (2016), Hardy and Bryman (2009).

8. Funded by the Deutsche Forschungsgemeinschaft (DFG, German Research Foundation), project number 398306144.

9. The UN provides several publicly accessible data banks in its Official Document System at https://documents.un.org/prod/ods.nsf/home.xsp) and http://unbisnet.un.org/. Statistical data are available at http://data.un.org/. Digital recordings of sessions in Geneva are provided at https://conf.unog.ch/digitalrecordings/. An overview of UN databases is provided at http://www.un.org/en/databases/index.html. Additionally, UN fora publish their work, meetings, documents and results on their own websites, and NGOs provide supplementary material. Additional data include treaties, conventions, policy papers, policy debates, etc.

10. Qualitative data analysis software supports the analysis of variables and codes, but it does not replace the researcher's substantial engagement with and interpretation of the data.

11. https://www.ohchr.org/EN/HRBodies/HRC/WGTransCorp/Pages/IGWGOnTNC.aspx.