• No results found

3 Methodological Approach and Research Strategy

3.3 Methods of Data Analysis

For data analysis, I relied on triangulation, combining insights from literature review, document analysis, and interview evaluation. Data interpretation was conducted through content structuring qualitative content analysis (inhaltlich strukturierende qualitative Inhaltsanalyse) following Kuckartz (2005), while also relying on structuring techniques described by Mayring (2015). Data analysis was supported by the software program MAXQDA 12.

Content structuring qualitative analysis is a common strategy used to evaluate qualitative data, particularly interviews. Like all forms of content analysis, it seeks to reduce data complexity. It is based on an assessment of topic frequency, identifying similarities, differences, and the relationship between topics using a category and coding system. Codes are understood as a unit that captures the essential content of an interview sequence and are as such preliminary elements of structuring. Clustered codes form categories based on identified patterns in the data (Saldaña 2012) (Figure 3). The aim is to develop analytical categories that link content to theory (Kuckartz 2005, 43–44).

Therefore, categories need to be identified both deductively, based on the theoretical model, and inductively, from the empirical material. Certain aspects, such as the local perception of lithium

THEME

CATEGORY

SUB-CATEGORY

CODE

governance, could only be assessed inductively through open coding. To support the analysis, interviews are clustered in groups of interview partners (members of the local, regional, or national government; academics and analysts; private sector employees; etc.). Moreover, different criteria are used to distinguish individual interviewees, for example, gender, political affiliation, or geography. This supports a balanced analysis down the road, which requires a consideration of different groups.

Prior to fieldwork, a theoretical model was developed from the existing literature which serves as a basis for the structured analysis (see Chapter 2.5 and Appendix I). It considers the different resource governance dimensions with relevance for local actor grievances. These dimensions serve as thematic categories for data evaluation and were used to develop the directory of interview questions and thus structured data gathering. In the model, resource governance was differentiated along four relevant components: project planning, stakeholder participation, revenue management, and cost management. Ownership is another important factor in this publicly organized extraction initiative. The model also delineates grievances along these governance dimensions, distinguishing between socioeconomic grievances, livelihood grievances, and political grievances. These grievances, it is hypothesized, also interact with factors beyond the resource project, linking to local experiences, values, beliefs, and perceptions.

In order to better understand the link between lithium governance and the development of grievances that can incite mobilization, four sequential steps were taken to evaluate the data.55

Categories were developed and assessed, a coding manual was created, data was coded, and finally the material was evaluated.

55 This structure was inspired by a process used by Hopf, Rieker and Schmidt (1995) in a study on authoritarianism

Development and Assessment of Categories

The evaluation began in the interview phase. Initial observations on relevant categories and sub- categories beyond the theoretically defined governance dimensions were annotated in the field diary. Structured evaluation then started with intensive text work, in which about 15 percent of interview material56 was read carefully, guided by two questions:

• Which issues are mentioned by the participants relating to the governance of lithium? • How do these issues shape the perception and interpretation of lithium and the lithium

project; e. g. how do governance dimensions relate to positive or critical observations about the resource or the project?

The subsequently developed categories / sub-categories highlighted different aspects such as: • Stakeholder involvement and individual interests vis-à-vis the lithium project • Individual knowledge of the lithium project

• Assessment of costs and benefits of lithium exploitation

• Evaluation of lithium governance dimensions and government actuation

This first evaluation helped to further develop the coding manual, linking theoretically based categories on governance to response categories of the interview partners. Thereby, a focus was placed on indications of grievances and the identification of expectations towards the program, including local evaluations and expressions of values and beliefs in the salar communities linked to e. g. a mining-based development model. After distinguishing the main categories, the next code layer was established in which base categories were further specified and differentiated.

I coded based on the techniques described by Kuckartz (2005) and Saldaña (2012). New codes were added for unforeseen topics, developing additional sub-categories that had not been deduced from theory. If it was not clear how to describe a certain idea as a code, I used a paradigmatic quote as a provisional code or relied on paraphrasing techniques. Out of these indicators, more specific codes were developed at a later stage. Decisive quotes were also highlighted in the interview transcripts to be used as examples in the analysis. I also added memos to interesting or unclear text passages

for further reflection in the evaluation phase. Text segments not relevant for the research were left uncoded and did not enter the evaluation.

Development of a Coding Manual

Based on data assessment, a coding manual was developed in which all categories, their sub- categories, and codes were defined and typical examples for each were given (see Appendix IV).

Coding of Material

In a second cycle, the complete material, including the material already revised in the first coding cycle, was analyzed and categorized based on the coding manual. The interview material was thereby matched with the categories and sub-categories. Already assigned categories and codes were in some cases changed and adapted. An example of a coded text segment is given in Appendix V.

Again, particularly interesting quotes, expressions and ideas were highlighted and commented on in memos. For each interview, a short summary was written. Bullet points summarized the interview partner’s position towards the lithium project and important ideas of the interview. As interviews had be clustered in groups of interview partners based on certain characteristics such as origin, gender, and political affiliation, I could check based on these summaries whether certain ideas were more pronounced in specific regions or individual interview partner groups. At the same time, I could identify new groups based on similar ideas expressed by different interview partners.

Evaluation of the Material

After the categorization was finalized, MAXQDA was a helpful tool to gain an overview of category and code distribution by interview partner groups as a basis for an interpretative assessment of the research question. The structured summaries of the interviews were worked through again to gain an understanding of different groups-based interpretations of the lithium program and its governance.

The main actor groups and central categories of evaluation served as a structure for the interpretive part of this thesis. The analysis was carried out per chapter, combining relevant primary data,

secondary literature, and the interview material (by re-reading and re-checking notes and memos connected to the relevant categories for the chapter). Selected quotes from the interviews were included in the thesis, providing space for local opinions, and opening academic reasoning and categorization up to scrutiny.

While the outlined evaluation strategy allowed for a structured, controlled, and detailed evaluation of the material, it also embodies certain shortcomings. Methodological literature on content analysis, for example, recommends interview evaluation as group work with at least two researchers coding the material. Kuckartz (2005, 83), however, also underlines that this is an expensive and time consuming strategy which cannot always be followed, particular in PhD research. For the outlined reasons, I also coded alone and have consequently been dedicated to transparency in my data analysis process.

As recommended by Gerring (2004, 344), the secondary case on the Argentinean lithium program was analyzed predominately through a literature review, extended by three interviews with Argentinean academics. Consequently, the analysis process had less diversity in the data and was therefore less objective than the micro analysis of the contemporary lithium case, which is based on triangulation of a larger data set from different sources and the possibility of cross-checks. The Argentinean case can hence support an assessment of the validity of the study’s findings beyond the Bolivian lithium program, but it is also limited in its scope, providing initial indications that would need to be assessed in more detail in further research.

The data analysis process is further specified in the Appendix, which includes the interview directory, the coding manual of applied categories, sub-categories, and codes as well as an excerpt from a coded interview as an example of the data evaluation process.