3. Method and Data
3.1. Qualitative methods
3.1.2. Qualitative data analysis
Data analysis was iterative and undertaken in an inductive manner, but was informed by key concepts in the literature 133, 134. Each interview transcript, set of observational notes, and archival document was read several times, generating and coding themes according to both issues identified in the literature, and features of the data that emerged inductively. Analysis was conducted with the assistance of N-Vivo 8.
Our data analysis involved three stages. Across all stages, two members of the authorial team who had conducted the fieldwork undertook initial coding of the data. In advance of the analysis, we assembled all of the documents, interview transcripts and field notes for each of the cases into a single data file. This enabled us to share data across the research team. In order to understand the complexity of each project 135, we coded the data on a within case basis 133. We began by analysing the data collected in stage one, involving all nine CLAHRCs, with a focus on the initial founding conditions of the nine CLAHRCs and their models for closing the second translation gap. We then analysed the four in depth cases where we focused on examining how the CLAHRCs were introduced over time.
Across both stages of analysis, we began with a fine-grained reading of the data 136. After
inductively creating a list of first order codes from the case evidence, we consolidated all of our codes across the nine and four cases respectively, and progressed with axial coding, structuring the data into second-order concepts and more general aggregate dimensions 137. In doing so, we engaged in deductive reasoning whereby we linked our inductive codes with existing concepts and frameworks 138. While we accept that our accounts are one of many potential interpretations 138, we worked in two ways to ensure that we did not retrospectively fit the data to service our theorising 139. First, we triangulated between data types. Second, we triangulated across analysts with one member remaining independent of the data who was able to challenge and interrogate the coders’ knowledge and interpretations 140. Our coding
50
was based on interview and archival data, and we only included codes in our work where we had at least four archival and/or interview sources for each of our cases where relevant. The observational data was employed to contextualise and corroborate the interview and archival coding.
In order to avoid identification of individuals or CLAHRC cases, we reveal the social position of the interview respondent only. To define their specific role in a CLAHRC (e.g. Director etc.) would pinpoint the respondent and organisation in question. The process through which we classified actors by their social positions is outlined in Chapter four. Our approach aligns with our research protocol following ethics approvals, and a need for particular sensitivity given ongoing CLAHRC refinancing.
Consistent with the preceding point, in presenting our qualitative analysis we do not draw explicitly on our observational data as much of it contained information that was either: (i) not intended for public consumption, i.e. may be a commentary of others; (ii) may be highly personal views that were communicated in confidence; (iii) or revealing the details of meetings etc. may risk revealing the identity of respondents. As such, we draw exclusively on data from formal interviews in the text, but employ the observational data as a means of informing our analysis.
We began by examining the founding conditions for each of the nine CLAHRCs in terms of the local institutional conditions, the social positions of the main actors involved, and the nature of the process through which the CLAHRC bid was developed, the analysis for which is presented in Chapter four. Founding conditions constitute a key concept in institutional entrepreneurship theory, which are used to explain variation of institutional entrepreneurship work and outcomes.
The next stage was to take the data for the nine CLAHRCs and examine how the founding conditions shaped the institutional work undertaken by the CLAHRC actors. We identified distinct ‘theoretical categories’ of institutional work that was undertaken, across the nine cases initially, and then in the four in depth cases following the later stage developments. All of the different forms of institutional work relate to the process of enacting institutional
51
entrepreneurship for institutional change. The themes were developed inductively through first order coding, which then progressed to axial coding, working back and forward between the data and theory, as outlined above. First, this work was related to the way in which actors interpreted the CLAHRC mission and responded accordingly, as shaped by the founding conditions. We termed this work the ‘envisioning’, which we defined as the process through which actors developed an ‘embryonic’ vision of change, based on the interplay between themselves and the context in which they are situated 27. This work is presented in Chapter
five. Second, was the work that actors undertook in signing up stakeholders of the CLAHRC, which we term ‘engaging’ and present in Chapter six. Third, was the work around the sustaining of the CLAHRCs, which we term ‘embedding’ and present in Chapter seven. Finally, there was a distinct area of institutional work that involved the rethinking of initial decisions and learning from the process of establishing the CLAHRCs, which we term ‘reflecting’, and is presented in Chapter nine. All of the different themes are expanded on in the introduction sections to the relevant chapters. Across these different areas of institutional work, we were able to identify distinct differences in the nature of the work undertaken by actors, as shaped by the institutional antecedents of the local institutional context and the social position of the actors involved.
The final stage of our analysis involved re-examining our data to understand the temporal sequencing of the theoretical categories of institutional work and the nature of relationships between them. As is common in process-based studies, it is difficult to establish whether or not a process is entirely linear 141, however, we present the second order codes in the temporal sequence in which they generally emerged in the case histories. Employing both inductive and deductive reasoning, and travelling back and forth between data and theory, we aimed to develop an understanding of how an actor’s social position shaped the envisioning process 126, 133, 136. Our emerging theoretical arguments were based on the interplay between
theory and data, drawing on both within and cross-case analysis 142. Based on our analysis of
the temporal sequencing of the different forms of institutional work to promote KT we developed a process-based model of institutional entrepreneurship as presented in Figure 1.
52
Figure 1: A process model of institutional entrepreneurship
In addition, and based on the work of the different IEs, we identified five different archetype models that were employed by actors to try and close the second translation gap. In Chapter ten we explore those models and the issues arising from each. In doing so, we emphasise that the models do not represent any one CLAHRC, but rather the elements of the different models may be exhibited in any one CLAHRC.