Data Collection and Analysis

According to the replication logic involved in case study methodology, data collection and analysis occurred in tandem. Initial descriptive frameworks were used to guide the data collection during the execution of the case studies, and interview questions were adapted based on the findings in each subsequent round of data collection. This allowed for adaptation and refinement of the interview protocol which would not have been possible if all data were collected prior to conducting any analysis.

The primary data source consisted of interviews conducted with informants at each MNE. Upon conducting each interview, access to other suitable contacts was requested if the participants were willing to provide it. The interview approach used was partially based on ethnographic interview technique as this allowed interviewees to explain practice transfer and adaptation in their own words (Spradley, 1979). Where the researcher defines these terms on behalf of the informant, the credibility of the findings is greatly reduced (Guba and Lincoln, 1989). Hence, each interview began with questions related to the actual experience of managers in establishing transferring innovation practices within the MNE network, asking the interviewee for specific examples of the type of activities they themselves mention in the interview. The questions were then adapted using language consistent with that of the interviewee. This also allowed for the identification of other informants who were personally involved in the transfer process and management of other subsidiaries. The semi-structured interview protocol used is presented in Appendix 3. Prior to conducting interviews with each of the informants,

numerous exchanges occurred via email. In most cases, the respondent requested and was provided with sample questions and descriptions of the interview protocol. These informants then sought clarity on any of the questions which they had regarding the nature of the research and how it might apply to their own personal experience. These exchanges occurred over a period of many weeks, and sometimes months. When the informant felt they understood the nature of the research and what would be asked of them, they proceeded to schedule the interview. These email exchanges improved the efficiency with which the interviews were conducted, and often were necessary to gain their consent.

During each interview, the participant was asked to explain the initial expectations for transferring innovation practices, and what outcomes were obtained. The participant was also asked about the various challenges that were faced in transferring practices, in terms of resistance from subsidiary employees if extant, learning challenges, difficulties in obtaining shared understanding of the purpose of the practice, and so forth.

This stage of the interviewing was semi-structured and many probing questions were used to uncover the linkages between challenges, benefits and the national and organizational institutional elements which may influence these processes.

Table 6 provides a summary of innovation practice transfers discussed during the interviews, according to the sender unit, the recipient, and a basic characterization of the practice, organized according to each MNE in the sample. Note that many of the innovation practices discussed are idiosyncratic to the MNE, and hence do not

necessarily have a specific name. Furthermore, in developing capabilities within subsidiaries with innovation mandates, innovation practices are often transferred as a

„suite‟, as opposed to individually. Hence, descriptions of each practice are more or less specific, depending upon the nature of the innovation practices discussed by each interviewee. In addition, the original source and intended recipient of each practice transfer is not known, as the process for diffusion appears to be more organic rather than centrally controlled.

Table 6: Innovation practice transfer discussed across four MNE cases.

Practice(s)

N/A N/A Virtual (online) identification of individual thought leaders to be

Table 6 continued...

Additional data pertaining to the industry and firm in question were gathered from secondary sources such as annual reports and company websites, as required, to improve understanding of contextual factors. These data gave additional insights into the general and business environments in which the MNE operates. Furthermore, this secondary analysis helped to discern the phenomenon from the context, as managers sometimes over-emphasize idiosyncratic events in their discussions which, while they do represent an account of actual events, are not representative of their most typical experiences with the phenomenon in question.

All interview transcripts were stored and analyzed with the assistance of NVivo^© version 9 software. Analysis of the data began with open coding of each transcript, keeping in mind both the research questions and the descriptive frameworks presented in Chapter 4. While some prefer a more grounded approach, a more structured approach was opted for, as particular research questions had already been specified, and the phenomenon was deemed too complex to be addressed efficiently through a purely inductive approach (Miles and Huberman, 1994). The coding was not completed in a single, linear fashion but instead as new codes were established prior transcripts were recoded accordingly. In addition to open coding, annotations and memos were used to store and build upon initial thoughts. As coding proceeded, themes representing a particular variable, construct or important aspect of the phenomenon were organized into collections or „tree nodes‟ (Bazeley, 2007).

As the coding scheme for the data became increasingly complex and saturated, comparisons were made across cases on different nodes. At one point, a summary table of the basic qualities of each node, or variable, across each case was produced. The overall purpose of this analytical technique was to reduce the data into more manageable and theoretically relevant sets. It also permitted the formation of linkages between different nodes, which would eventually represent relationships between variables. Brief case summaries of each MNE are provided in Appendix 4. Through this process, structure was imposed on the interview data, in order to derive theory (Richards, 2005), which is presented in the following chapter.