3. METHODOLOGY
3.7 Analysing data
Corbin and Strauss’ (2008) methodological guidelines were followed to provide a clear and systematic process for analysing the collected data. This approach aligned with the epistemological assumptions of the study, placing priority on the phenomena under study, and seeing “both data and analysis as created from the shared experiences of the researcher and participants, and the researcher’s relationship with participants” (Charmaz, 2003, p.313). In the analysis, the aim was to explore senior managers’ implicit meanings and experiential views. Based on the extant literature on qualitative research in accounting and MC, a framework was developed to assist in analysing the data.15 Figure 3.2 summarises this framework.
Figure 3.2: Framework to analyse the collected data
Source: Researcher based on literature review
The framework included six guidelines used as a checklist while collecting and analysing the data. The first was an iterative process of data collection and analysis. This iteration was not a repetitive mechanical task, but a reflexive process to catalyse insights and develop meaning. The second guideline related to coding the
15 Twenty-four accounting papers that based their findings on an inductive approach to data
analysis were reviewed (Gurd, 2008). Analysis of these papers is beyond the scope this thesis but is available on request.
data, which was a process through which concepts were identified in the data. The main idea was to identify concepts, break them down into categories and sub- categories, and then transform them into findings.
The third guideline was theoretical sampling, which was a process of selecting events or interviews, making comparisons between responses, and gathering data to guide the development of higher-level concepts (Parker and Roffey, 1997). Next came the writing of memos and diagrams. The objectives of these mechanisms were to reflect on the data, discover gaps in earlier interviews, make explicit comparisons, and generate new questions. Charmaz (2003) affirms that, through such memos, the researcher begins to analyse and write early in the research process, and thus avoids being overwhelmed by piles of unanalysed data. The data collection and analysis should continue until further analysis neither adds to the existing categories nor creates a new category. At that point, the study reaches theoretical saturation, which is the point where new sources of data no longer offer additional information (Goddard, 2004). Qualitative case-study research should be understood as the development of inductive findings; thus, the research work implied the need to assess whether the previous guidelines resulted in a contribution.
Again, as a checklist to assist with the data analysis, the framework took account of Suddaby’s (2006) warning of potentially serious misconceptions in inductive qualitative research. He claims that some authors use labels such as “grounded theory” as an excuse to ignore the literature or to do research without a clear methodology. Case-study research is not easy; it requires discipline, flexibility and rigour (Carmona and Ezzamel, 2005; Suddaby, 2006). Additional problems identified by Suddaby (2006) are the fact that researchers may stop collecting data too early, present raw data (qualitative research requires a higher level of abstraction than the data itself), or focus on how to verify emerging codes rather than on how to understand the nature of the phenomenon being studied. The objective of case-study research on MC is to understand the development of accounting and control in its natural context. According to Carmona and Ezzamel (2005, p.150):
Case-study research is a unique way of understanding organizational micro-processes, in their most minute detail. As the researcher delves deeper and deeper into such micro-processes, the researcher is likely to encounter many similarities reported by other researchers in other organizations.
Scholars must be aware that creativity is an essential ingredient of case-study research (Strauss and Corbin, 1998; Yin, 2009). Researchers should also endeavour to be transparent and offer sufficient detail in the presented data to create a sense of verisimilitude (Suddaby, 2006). Furthermore, the logic of inductive research demands that successive questions should be asked about participants’ experiences to probe the researcher’s theoretical insights; hence, the study must show the researcher’s reflexivity (Charmaz, 2003).
All interviews were conducted in Spanish, audio-recorded, and then transcribed verbatim. Triangulation is a strength of case-study research because it uses evidence from different sources to corroborate the same fact or finding (Rowley, 2002). Thus, as mentioned in the previous section, data were retrieved from a number of different sources to form a single body of data. This was an iterative process of collecting and analysing the data. A case-study database of the gathered evidence was developed as the data were collected and analysed. This database included case notes, archival documents collected during the research, interview transcripts and analysis of the evidence.
In reality, data collection and analysis were not distinct stages. The analysis began as soon as the first interview was in progress, trying to make sense of the stories that were being told in order to ask the right questions. Research site notes and transcripts are a research activity that involves interpretation, including repeated listening to recordings and noticing new data features (Silverman, 2010).
Over the course of the data analysis procedure, both manual and electronic processes were employed. However, the analysis process varied between the different stages of the research. During the exploratory stage, analysis was based on the initial propositions of the research.16 The idea was to seek evidence to corroborate the initial propositions and then record relevant evidence and make a judgement on whether the positions were substantiated. In stage two, a qualitative software package (NVivo) was used to code the data and to facilitate theoretical sampling. At this point, the research aimed to analyse relationships between MC and the assumptions about human nature behind it. Hence, the first step of the coding was to identify incidents in the interviews that showed such assumptions by
16 In the initial research proposal, Simons’ (1995b) LOC was used to theorise that, based on
their assumptions about human nature, some leaders rely on boundary systems and heavily stress diagnostic systems, while managers with different beliefs about human nature emphasise interactive systems and use value systems to inspire and guide organisational search and discovery. These were the initial research propositions used to seek evidence in the exploratory stage.
senior managers. Review and analysis of the interviews revealed that managers’ assumptions at Compartamos could be understood by tracking and coding managers’ perceptions of people, the organisation in which they worked, the work their subordinates performed, and the context in which their organisation was embedded. These four categories were the first to emerge from a line-by-line analysis of the interviews. The coding process is illustrated by examples presented in Table 3.4.
Table 3.4: Example of Compartamos Bank data coding Comparing incidents
From… Identifying conceptual categories To…
“…we are a great place to work not because we get paid the same salary as in an investment bank, but because the employees feel highly satisfied. I can assure you that 97% of my subordinates would say that they love what they do.”
Perception of the organisation in which they work
“Compartamos’ mission is to help our clients accomplish their objectives, to work with self-fulfilled people who achieve their goals. It is also to reach into every Mexican home, trying to do good and creating a micro-business opportunity for each one of our customers.”
Perception of the organisation in which they work
“Our balanced scorecard was developed to measure the way Compartamos generates social and human value, not just economic value.”
Perception of the organisation in which they work
“We understand that a person has many facets in his or her life. You cannot conceive of an employee only from the professional aspect because you would be a little short-sighted. We conceive of people from the physical, intellectual, social, spiritual, and
professional aspects.”
Perception of people
“… I would say that 99% of my team have the same principles. Because I think the values I inherited from my parents are the ideal ones. That is what I would like to see in everyone. I think we all live in the same way.”
Perception of people
“Here, at Compartamos, it is not necessary to push people to do their job; people accept the commitment and go for it.”
Perception of people
Source: Researcher’s own analysis
Line-by-line analysis of the interviews and operationalisation of the LOC within the interview questions facilitated the coding relating to MCS. From Compartamos’ interviews, the following codes were created: performance evaluation, strategic goals, individual goals, tracking and monitoring employees’ performance, value system, rewards and punishments, and scorecards. Three additional codes helped provide an understanding of interviewees’ opinions of Compartamos Bank: history of the organisation, Compartamos’ growth, and business methodology.
Identification of relationships between the codes relied heavily on visual display, for instance mapping categories and their propositional statements. NVivo software facilitated this process, allowing categories (nodes) to be visually displayed, and to be arranged and re-arranged in forms that reflected changing lines of thought (horizontally, hierarchically and chronologically). It also allowed the production of diagrams to demonstrate this thought progression, and as categories were electronically generated, expanded, merged and collapsed, the movement of each data segment was logged in the database. NVivo allowed the researcher to interrogate the data to identify patterns and trends, and to query the unstructured data from the interviews.
Some of the analysed data were then translated into English in order to reflect on it, discover gaps in the interviews, make explicit comparisons, and generate new questions. In line with Charmaz (2003), memos were also a valuable tool for analysing and writing early in the research process, and thus avoid being overwhelmed by piles of unanalysed data. These memos were written in formats that looked like rough versions of the case. Discussion of and feedback on the preliminary versions helped the researcher return to the research site and generate new data.
The coding process continued to seek relationships between comments, subject matter, categories and secondary data. Sub-categories were developed to classify different perceptions, for example “perception of the organisation as a generator of assets to society”, “perception of people as honest and committed individuals”, and “perception of people as maximisers of resources without restrictions”. Furthermore, variations in phenomena were explored by comparing apparently related patterns of dimensional locations, and relationships were identified between sub-categories, categories and phenomena.
At this stage, in response to senior managers’ assumptions about human nature in Compartamos Bank and their perceptions of the organisation and the context in which the bank was embedded, several new codes were identified, including self- realisation, human dignity, economic value, human value and social value. The coding relating to MCS generated three new codes: FISEP17 scorecard, Code of
Ethics, and the “We Expect More from You” programme. New data were continuously integrated through coding from questions about the senior managers’ values, their perceptions of human behaviour, and how these two factors influenced their exercise of MC.
As previously mentioned, an interview guideline was developed with the intention of operationalising Simons’ (1995b) LOC framework. Thus, through the codes, it was possible to link the MC literature with Compartamos’ actual MC mechanisms. At this stage, Laughlin’s (1995) middle-range thinking approach was useful to ensure a constant interplay between the data and the literature. As explained by Broadbent and Laughlin (2009), the conceptual “skeleton” provides alternatives that are likely to be presented in any empirical situation, and a language with which to discuss
17 FISEP is a Spanish acronym that stands for Físico (physical or health), Intelectual
(intellectual), Social-familiar (social-family), Espiritual (spiritual) and Profesional
(professional). See Section 5.5 in Chapter 5 for an analysis of the concept and how it is used as an MCS in Compartamos.
them. Compartamos’ empirical data provided the “flesh” to make the “skeleton” of Simons’ LOC meaningful, and a resource to reshape the conceptual framework should it fail to express the empirical situation adequately, in a “skeletal” sense. Thus, middle-range theory helped provide a language for structuring the empirical description (Broadbent and Laughlin, 2009).
In order to gain additional feedback, the memos were organised and a next preliminary version of the rough analysed data was presented at two conferences (Casanueva, 2014a, 2014b). After reviewing new literature recommended at the conferences, the data analysis and coding process continued. This part of the analysis culminated in the formation of propositions based on relationships that showed a category’s causal conditions, phenomena, context, intervening conditions, actions and consequences. As suggested by Norris (2002), to identify these relationships, the data were questioned and compared, and four analytical steps were performed almost simultaneously: 1) continually searching for dimensions of variation within issue categories and sub-categories and in secondary data (e.g. internal financial reports, strategic maps, code of ethics and annual reports); 2) exploring variations in phenomena by comparing dimensions for apparently related patterns of dimensional locations; 3) speculating about relationships between sub- categories, their categories and the phenomenon; and 4) verifying these propositions against actual data. Corbin and Strauss (2008) explain that it is this constant interplay between proposing and checking that leads to the emergence of contributions.
The data analysis, the constant interplay with the literature, and feedback on the memos brought about a new line of study. While trying to understand the organisation’s MCS, mainly in terms of the value system and the FISEP scorecard, a new research approach emerged from the data: Compartamos’ senior managers were aiming not only to influence people to behave in ways that lead to the attainment of organisational goals (the “what”), but also to lead managers to act in accordance with the organisation’s purpose and principles (the “how”). Hence, the focus of the thesis turned to exploring why and how Compartamos Bank had attempted to influence the “how”. Thus, the research entered its third stage.
An alternative approach to data analysis was adopted in stage three. Here, based on Marginson (2009), a descriptive framework was developed to organise the case- study (Rowley, 2002). This framework included sections reflecting the themes in the case-study and a collection of evidence within those relevant themes. The data in these categories were then analysed and compared in order to produce a
description of the case-study that could be corroborated from multiple sources of evidence. Again, Laughlin’s (1995) middle-range thinking approach was useful in using empirical detail to provide the “flesh” which made the “skeleton” meaningful, as well as providing a resource to reshape the conceptual framework of Marginson’s (2009) value systems (Broadbent and Laughlin, 2009).
In segmenting the data at this stage, a matrix was developed so that, during the analytical process, incidents could be readily traced back to their original context in the interview transcripts. A set of first-round provisional categories was then generated, to which the segmented data were coded. These categories took two forms which might best be described as “interviewee-driven” and “researcher- driven”. The first were derived from the researcher’s familiarity with the interviewees’ context and language, while the second were derived from theoretical interest in the phenomenon under study (the “how”). Hammersley and Atkinson (1983, p.153) clarify the importance of interviewee-driven categories in the data analysis process:
The actual words people use can be of considerable analytic importance as the “situated vocabularies” employed provide valuable information about the way in which members of a particular culture organize their perceptions of the world, and so engage in the social construction of reality.
Although the interviewee-driven categories were simply general descriptions of concepts (e.g. “Mística guardians”, “Integration meetings”, “Serviazgo leadership programme”, “the person at the centre”), and the categories changed during the analytical process, generating these two forms of category provided a framework for moving back and forth between “natural” and “theoretical” discourses (Hammersley and Atkinson, 1983).
Segmenting and placing the body of data in the matrix allowed examination and analysis of incidents according to the developed framework. During the analytical process, the research questions were constantly borne in mind, together with assessment of whether the data provided information to answer those questions. The next task was to carry out analysis that aimed to go beyond the identification and description of broad themes to exploration of coherent meanings embedded in the data, and then to draw up a theoretical statement that would express the collective meaning of the analysed data. The data collected in the first two stages were then re-read, analysed and compared with the identified meanings. Throughout the analytical process, some categories were quickly justified, while
others collapsed or needed to be re-defined. Also, new categories emerged and other themes were eliminated as irrelevant to the research focus.
Throughout this iterative process, identified meanings changed as the researcher developed and refined theoretical insights into the phenomenon under study. As the process drew to a close, corroborated meanings, themes and insights constituted the roughly formed outcomes of the study. In order to present evidence of these insights, selected quotes were chosen from the full body of data. Some of these themes and insights stood alone, explaining aspects of the phenomenon under study, while other findings were interrelated.