PART III: RESEARCH METHODOLOGY
6. Introduction: Complex problem, complex analysis
6.7 The coding process
Coding was used to reduce the data. Following Charmaz (2006:9), I adapted some of the practices of grounded theory by applying thematic analysis using in vivo and a priori codes. I commenced with open and ‘in vivo’ coding, followed by axial and later selective coding. From these, I selected two linked core categories – one related to the macro context of inequality and transition, and the other micro and related to what happens in the black box of restorative justice processing. These themes are discussed in chapter seven and eight. Other key codes were gathered around these core categories. These linked codes and categories from the ‘ground’ up, made the story concealed in the data explicit.
I used deductive and inductive reasoning sequentially and later combined the codes that were surfaced through these ‘top-down’ and ‘bottom up’ methods. This was done to complete the picture and to locate the study in the literature consulted. The questionnaire was based on the conceptual framework which was derived from empirical observations in the course of my work as a practitioner, and from multi-disciplinary literature. Data collection was bounded by the three interlocked concepts in the conceptual framework - peacebuilding, restorative justice and social justice.
6.7.1 Open coding
I initially changed the word documents (of the transcripts) to text files for use in the Weft QDA software, an open source qualitative data analysis software package. The Weft method, while useful, proved too cumbersome and I switched to La Pelle’s (2004) method of analysis. After doing a line by line reading of each interview, I coded key words, ideas and sentences and highlighted relevant paragraphs throughout the text. I tried to transcribe and code on the same day that I conducted interviews or as soon as possible on the days thereafter. The four interviews that I did not transcribe and code immediately were ones that did not add anything new. I later transcribed and coded two and listened to recordings of the final two that I did not transcribe, to deepen and thicken the analysis. This was done two weeks after the interviews were completed.
Instead of coding subsequent interviews with codes from previous interviews, I first coded each interview on its own to ‘form general impressions and intuitions’ (Dey,
175 1993:66). I did this to remind myself of each individual, their body language, how they came across, and what I thought and felt during the interview. This generated pages of codes, some of which were duplicates that were removed.
The reflective review process as suggested by Dey (1993:67) helped me to generate an interim analysis which in turn helped me to focus questions for subsequent interviews. In the two week period, before transcribing the last interviews, I ‘pawed’ (Ryan & Bernard 2003:88) the coded interviews on a daily basis. I sorted the codes in various
configurations and wrote insights and memories about what body language, choice of words and silences at particular moments could have signified on two flipchart boards and in my note book which was always at hand. I used different coloured markers to write relationships, connections, singularities and similarities on the flipcharts as a thought or question occurred to me. When I compared each victim and offender’s pattern of experience throughout the restorative justice processing period, I started to discern potentially serious ethical issues in one of the cases. I wrote to the organization seeking to clarify if my analysis was correct. I received no response and turned to my supervisor for advice. This issue is discussed in chapter seven, section 7.5.3 (a).
6.7.2 Axial coding
To prepare for axial coding, I highlighted open codes with similar characteristics in different colours. By opening two Microsoft word windows at a time, I could drag codes from the main list across to the other window and created documents for different coloured codes. These were then ‘pawed’ and pored over to see what name emerged for the theme. I kept a copy of the main colour coded list.
I then printed out all the individually coloured open and in vivo codes after enlarging the font. The codes were cut up and arranged on a long table. Duplicate codes were discarded and the rest were sorted into piles. These piles were put into folders and filed because no clear themes were emerging at that point.
Instead of returning to the Weft QDA software - which I found did not enable me to view the data from different angles - I decided to continue using Microsoft word as proposed by La Pelle (2004). The triangulated analytical process I was following was sufficient to lead me to key themes. Initial open, in vivo and axial coding of data led to a fine grained
176 analysis. I am aware that case study research is not regarded highly unless it is in depth and thick. My son assisted me by double checking my numbers and colour coding of word frequencies and linguistic connectors (Ryan & Bernard, 2003:91) using Microsoft word. I focused only on words and phrases within the data that were relevant to the research question. The totals of word frequencies and linguistic connectors for each interview were typed onto the transcript and a key was developed for the colours assigned to each word. At my direction, he also created a document where all responses to the different questions for each sub-unit were cut and pasted together in word documents.
These manipulations of the data are usually done using Qualitative Data Analysis software. I found the software too limiting and incapable of manipulating the data in the ways that I wanted it to. During the following week I created some distance from the data and attended to other aspects of the research while performing Bernard’s (2000:445) tongue in cheek but effective ‘interocular percussion test – which is where you wait for patterns to hit you between the eyes’. These insights ‘hit’ me over a period of time as I used various data displays to arrange the data so that I could see the patterns, themes and ideas that constitute the restorative justice/peacebuilding/social justice ‘quilt’ produced by the bricolage. The final insight ‘hit’ me after completion of the conclusion chapter, when I realised that the manifestations of denial that formed the patterns of denial referred to in chapter eight, were embedded in a culture of denial in South Africa. This denial is rooted in the master narrative of the ‘miracle/rainbow nation’ which results in many South Africans closing their eyes to anything that does not ‘fit’ the miracle/rainbow metaphor. As a result white beneficiaries of apartheid and market democracy; and black beneficiaries of market democracy are unable or unwilling to accept that trans-historical and growing inequality belies the notion of a ‘miracle’ for all those who live in the land.
I revised the thesis according to this insight about the culture-pattern-manifestations of denial which resonates with a combination of Galtung’s notion of ‘cultural-structural- direct’ violence (1996:2), Cohen’s ‘states of denial’ (2001:25), Zerubavel’s ‘conspiracy of silence’ (2010:32) and Liem’s ‘multi-layered model of the silencing of historical memories’ (2007:172).
6.7.3 Selective Coding
During the selective coding stage, the links between themes became obvious as well as the similarities and differences within and between themes. At this point I checked to see
177 if the data was beginning to answer the overall research question and sub questions, by using the a proiri concepts in the title and conceptual framework, as well as the research questions as organizing categories. I selected the core themes and the sub-themes to explain what the data was saying. By juxtaposing the sub-themes, the paradoxes, dilemmas and contradictions were laid bare as exemplified by the interplay findings in chapter eight.