Coding the interviews: Labelling the narrative luggage

After the transcribing process, I started coding the content of the conversations. I approached the coding process as a way to organize my data and generate a foundation for working with interpretative analysis (Coffey & Atkinson, 1996). I applied coding as a tool to identify in which passages in the interviews the leaders were talking about something that seemed to be a core theme that caught my interest in relation to the research question. What were they talking about, when they talked about their experiences as leaders? I thought of this work as having a more abstracted meta-conversation with the data after having had the actual conversations with the leaders: ‘How can I understand the use of this particular word or this specific sentence? How does this particular utterance possibly relate to how the leader is making sense out of his/her experiences as a leader?’.

In this manner, I applied coding as a way to generate organising principles for the processing of my data, relating to the codes I generated as ‘tools to think with’ (1996, p. 32). Aligned with what Van Nes et al suggest (2010), I chose to do the coding in the original language (Norwegian) to secure as much quality and accuracy as possible in terms of validity, meaning wise. Hence, I coded the interviews to get an overview of main themes occurring in the data, thinking of the different segments of texts as pieces of ‘luggage’ that I labelled with thematic tags in order to be able to identify each segment and retrieve it among the others (Wengraf, 2006). As my research question and the interview questions were abductive and open-ended, I did not test the interviewees answers to my questions against a defined hypothesis or particular theory. Because of the semi-structured format, I found that many themes would not occur linearly but be found in various and different passages in the interviews.

The coding process was done in four steps. I did the coding of the interviews manually, where each transcribed interview was coded in one separate word document. After I had transcribed the interviews, I inserted the transcribed text in a two columned table. I structured the interview by speaker’s turns and marking my turns in blue and the leader’s turns in black. Similar to the samples of coding offered by Coffey & Atkinson (1996), in the left column I put specific content (what was said), and then wrote verbatim ‘in vivo’ codes (Saldaña, 2013) in the form of key themes and words occurring in the stories, in the right column. Verbatim coding aims at identifying words and expressions used by the interviewees themselves, that can be found in the actual raw data (2013). For this project, the first step of verbatim coding was important for how I was able to identify the emic categories that I used for the further analysis and discussion, particularly concerning the categories of trust and self-trust. Next to the second column, I started to write down my ‘meta conversation’ with immediate associations, ideas for interpretation, and annotations to the emic categories, themes and key words. In this way I was able to map emic categories and themes that had something in common across the twenty interviews, and could identify similarities and differences in the kind of luggage that each interview contained, while at the same time starting to work on the etic codes.

While some qualitative researchers will claim that coding your data is not the same as analysing it (Coffey & Atkinson, 1996), others argue that coding of data is a deep engagement with the data’s meanings and therefore should be thought of as a vital part of the analyzing (Miles, Huberman & Saldaña, 2018). For this project, the process of coding was vital for how I was able to make sense of the large amount of interview data and generate insights that related to my research question. However, coding is always an important procedure to prepare for further data analysis, because coding is an analytic approach to break down data in manageable parts to identify further relevant questions about the data. How data is coded is vital for how one can do the analysis of it. For this project, I aimed at generating codes at a general level to be able to work with data as chunks of conceptually related content. In accordance with Seidel and Kellc (1995), I applied coding at this level to (a) noticing relevant phenomena, (b) collecting examples of those phenomena, and (c) analysing those phenomena in order to find commonalities,

differences, patterns, and structures (p. 55-56). As Coffey and Atkinson stressed, coding of data as conceptually related chunks of content can be thought of as data complication (1996, p. 29) because it can function as a means to develop and reinvent ideas about what the data signify, facilitating for new perspectives and insights.

I carried out the coding work in four steps or cycles. After I had done a first cycle (Saldaña, 2013) of verbatim coding of each interview, I did a second cycle of coding as a within-case analysis (Creswell, 2007) to look for correspondence and connections between main categories within each interview. As a third step, I then coded each interview axially (Strauss & Corbin, 1990) in order to identify if there were subcategories that could be linked to any of the main categories within each interview (Charmaz, 2010). I looked for connections, patterns, idiosyncrasies, mysteries, overlaps, redundancies and other characteristics in the leaders’ language. For the axial coding I sketched a matrix showing all the interviewees and main categories mapped against the most frequent themes and key words what I thought of as subcategories and relevant attributes that appeared interesting and of relevance in a leadership identity context. Aligned with how Schreier described the process of coding qualitative data and how it is concerned with describing meaning in context, for the axial coding I tried to go beyond the specifics of any particular passage as I worked with identifying codes of relevance, and aimed at levelling the abstraction of the meaning of the passage. I did this in order to identify categories that applied ‘to a number of concrete, slightly different passages’ (p. 170). I then used the matrix overview to perform the fourth step of coding, doing what Creswell (2007) described as a cross-case analysis, mapping all the interviews with each other, looking for corresponding categories and phenomena across the twenty stories.

How I worked with the coding of data in these four steps to generate main categories first and then looked for subcategories corresponds to what I understand as a strategy of subsumption (Schreier, 2014), working systematically to examining one passage after another, going through the content step by step and mapping main categories with subcategories, subsuming overlapping subcategories or creating new subcategories (2014, p. 8).

The result of the four steps of coding made it visually very clear to me that the specific word

‘trust’ stood out as an emic category, occurring repeatedly in the interviews in form of the Norwegian word ‘tillit‘ ¹. The word “self-trust” also emerged as an empirical category, occurring in the form of the word ‘selvtillit‘² and in phrases referring to having trust or faith in one self. I recognised trust as a predominating term occurring as an empirical category in all the interviews.

From having a first impression that trust was something that several of the leaders talked about, after the coding of the data, I could point to the actual occurrences of the word ‘trust‘ and conclude that trust was in fact a vital theme in the leaders’ stories. I also recognised that self-trust transpired as a central theme, often in the form of the word self-trust (selvtillit), but also in other forms that

1 In Norwegian language, ‘tillit’ means trust

2 I Norwegian language, ‘selvtillit’ means to have trust, faith, in one self. The Norwegian word

‘selvtillit’ is what I understand as the concept of self-trust in a Lehrian term (see discussion chapter).

in the Norwegian language equal the meaning of the term self-trust. I then started to look for other words co-existing with the words trust and self-trust in sequences, and found some commonalities regarding the leaders’ choice of words and expressions to describe their leadership experiences.

The point I would like to stress here is that I did not censor variety in the leaders’ vocabulary to manufacture simplicity in the place of complexity, or lump together empirical diversity to produce an unambiguous concept of trust. On the contrary, in the leader’s stories, the concept of trust and self-trust stand out as core themes in how they describe their understanding of their leadership project. They all have slightly different ways of depicting how trust and self-trust become important and defining phenomena in their lifeworld as leaders, but they all relate to trust and self-trust as core elements in their leadership identity, and their descriptions have much in common when it comes to depict what I understand as the essence of how they understand their leadership identity. Hence, I experienced that the coding process led to an empirically generated and saturated set of data which I could proceed with for further analysis.

Therefore, I argue that trust and self-trust in this case are primarily emic categories. This is an important thing to underline in order to prepare the reader for the analysis where trust and self-trust as concepts occur in many variations, but still with what I understand to be in a very homogenous meaning. The homogeneity concerning trust as it emerged in the data material as I transcribed and coded the interviews is precisely the reason why I became curious about the concept in the first place, and why I started to note how trust and expressions for self-trust so often occurred in the same context. Regarding self-trust, the concept is also an emic occurrence in my interview data, but I also understand self-trust as a concept that I developed further as an etic category. While it is possible to claim that the word and concept of “trust” (‘tillit’) emerged as a purely empirical category, I understand self-trust as an emic category because the term is used by the leaders themselves in the interviews. However, I also understand self-trust as an etic category, the latter resulting from my interpretation of how the word self-trust (‘selvtillit’) or utterances referring to the act of trusting one self, was used in the leaders’ narratives and how the use related to the specific context the word was used in. Hence, the coding process for this project can be understood partially as a process to identify emic categories emerging from the interviews, but also partially as a qualitative content analysis in order to ‘go beyond the specifics of any particular passage’ (Schreier, 2014, p. 170).

I will be returning to the performative aspect of trust and self-trust in the discussion chapter and also dwell more in detail with the possible explanations for how trust can be found to be such a vital concept in the leaders’ narratives, relating back to Discourses-in-practice and discursive

practices. However, at this point I find it worthwhile to highlight that the occurrence of trust and self-trust in the leader’s stories is not something I deliberately forced in order to present a certain image or interpretation that I wanted to make. Trust and self-trust emerged as empirical phenomena in the leaders’ stories. Having said this, I am well aware that qualitative data can never speak for themselves. Language is neither transparent nor value-neutral (Eisenhart, 2008). Its’

meaning is constantly decoded through interpretation, which is again context dependent. This illustrates how reflexivity in regard to the interpretation of narratives is the qualitative researcher’s best tool.

I therefore went over each interview several times looking for potential phenomena in the leaders' stories that could shed light on how they understood their leadership identity. For the translation and analysis, I did a first selection of sequences and passages from each interview for the various categories and themes I had identified as a result of the coding process. After I had done this, I proceeded with translation and analysis.