Chapter 3 Methodology
3.5 Data analysis and theorisation
3.5.1 Open coding and first-order concepts
Following those established models of qualitative data analysis, the discursive- based data were coded into terms or short sentences that were primarily ‘informant- based’ (Gioia et al, 2012; Langley and Abdallah, 2011), and the ethnography data were coded faithful to ‘descriptive facts’ (Van Maanen, 1979). This process was difficult, in that the codes created during this analysis should neither be too extensive to manage, nor too generic and vague so that emerged meanings are prematurely enveloped into a narrow theoretical framework. For this issue, Gioia and colleagues (2012) helpfully introduce the conception of ‘concept’, which, in their term, is a less ‘well-ridden wheel’, but an inductive tool. It explains social phenomenon at a higher level than describing, and allows creative, innovative, and data-driven meaning creation. Thus, concepts could be potentially developed into well-defined constructs and core dimension of new theories.
But as a fresh qualitative researcher, I still did not find it easy to pick up ‘concepts’ from the messy and somewhat unfamiliar (not similar to any empirical study I had reviewed by the time) data. Langley and Abdallah (2011) state that researchers doing a single case study might be confronted by the ‘trap of having nothing but a boring sequential narrative to tell’. On the contrary, my study involves too many interesting issues and a large variety of ways of coding them, which paradoxically rendered my data as flat as that is described by Langley and Abdallah.
A second challenging aspect of this study is to distinguish the generic socialisation path that unfolded over time from the identity construction process that was experienced by newcomers at individual level. Although based on the philosophical underpinning of this study ---- social constructionism ---- such separation does not ontologically exist, it is not unacceptable to slightly differentiate between the socialisation processes and experiences of individual newcomers, as the former were infused with discourses, institutions and structures that were, most of the time, deep-levelled, slowly transformed, semi-immune to everyday acts (Berger and Luckmann, 1967; Sewell, 1992; Watson, 2008; Ybema et al., 2009), and the latter were
more idiosyncratic and fluid. These are analogous to what Van Maanen (1979: 541) tells about ‘the studied scene and the member interpretations of what stands behind’. As discussed in earlier section, the separation would make it possible to view how individual newcomers made sense and engaged with social and cultural context to construct their identities.
However, at points it appeared rather ambiguous in the data. For instance, training faculty’s strict discipline and lack of trust on student officers could be seen as either general properties of the academy or the subjective perception of some of the studied newcomers. In a word, it was a delicate analysis process that demanded reflexivity to define the ‘boundary’ in empirical data in order to both keep faithful to the data and construct a neat theoretical framework.
After repeated data reviewing and some preliminary coding practices, some interesting, unexpected and inspiring observations emerged, which suggested directions of thematic interests. These observations provided some provisional answers to the two research questions and constructively pushed analysis to move forward. To wit, when asked what influenced them most or attached most attention from them at a given phase, recruits spontaneously mentioned different aspects or elements in that context; and given aspects that drew vital attention from some newcomers might be barely noticed by some others. One rudimentary observation in this regard is that a few newcomers particularly pointed out the stress gendered from the fact of ‘being assessed by’ and ‘wanting to impress’ knowledgeable tutors and colleagues in role-play exercises, whereas, some others, albeit equally unsure about how to perform the role-play tasks, conceived it a chance to learn from each other and expressed little sense of inferiority. Another example is recruits’ reflections on PDU1, with some focused on how they were treated better than at the academy while others on how much progression they had made in doing various kinds of jobs. These salient differences in their interpretation of and reflection on the shared experiences indicated the divergence in their propensity to define their situated position. Different individuals were preoccupied with different dimension of the social context, to which they directed most efforts to engage with.
This data-driven analysing process was only undertaken to sharpen eyes so that the first-order codes could be more meaningful than, otherwise, gestalt and random. It
helped to solve the first challenge mentioned above of ‘getting lost’ in data, for where informants’ attention, emotion, and effort was mostly attached and what they perceived, believed and cared about constituted crucial comparative elements (Langley and Abdallah, 2011) to derive deeper and reflective understanding of the data. It also facilitated addressing the second challenge, in that the idiosyncratic interpretations of informants rendered subjective accounts more visible and easily distinguishable from the commonly recognised ‘essential realities’. For example, following this rationale, the feeling of nerve at role-play exercises was regarded as a naturally invoked emotion that generally existed in such a situation, but how informants accounted for this feeling and let it shape their self-perceptions and behaviours reflected their different ways of sensemaking and identity producing. To wit, some recruits considered the nerve as a result of the position of ‘being assessed’ and let this recognition reinforce their sense of vulnerability, whereas, some saw it as normal for every rookie and motivated themselves to get involved in more practices alike. Nevertheless, this thematic dimension was carefully applied to prevent it from imposing constraints on or overruling open coding.
Moreover, the data collected showed overt discontinuities at the frontiers between the four waves outlined earlier, each bearing its distinctive contextual characteristics, namely where newcomers were, who they interacted with, and what they were preoccupied with. Given this, data were chronologically broken into 4 units of analysis at the open-coding stage. Drawing on Langley’s (1999) review of ‘temporal bracketing strategy’ for theorising in organisation studies, the method of analytical division could be utilised to generate comparison and integration of multi-dimensional causality of a process. Inspired by that template, the four chronological episodes were comparatively evaluated as alternation between police academy, PDU sector and operational policing domain in shaping the identity of newcomers. Additionally, this breaking-to-integrating pattern of analysis allowed both inter-personal comparison as well as examination on intra-personal trajectories.
Through these practices, at the first round of open coding, 130 first-order concepts were coded. After getting more familiar with the data and the practice of open coding, the researcher continued with some minor aggregation and selection based on two criteria ---- a given issue’s significance (e.g. issues peculiar to a small number of people were eliminated or merged into other or new concepts), and relevance (e.g.
focused on issues relevant to inter-personal and intra-personal comparisons). This activity was not dissimilar to Pratt and colleagues’ (Pratt et al., 2006) practice of ‘abandonment and revision of provisional categories’ based on data-category fit.
Via this process, the number of first-order concepts was brought down to 54. 20 codes were extracted from the set of individual identity matters, which are shown as the headings without bullet points in the rectangular boxes in the first row in Figure 3.1 to Figure 3.4. These codes denote the general aspects that had attached individuals’ attention and stimulated their reactions in identity construction in the four distinctive contexts. Additionally, 34 codes were drawn out from the set of contextual descriptions, which are shown as the headings in the rectangular boxes in the first row in Figure 3.5 to Figure 3.8. These codes describe the key influences and elements of the social contexts on individuals’ identity construction.