In order to answer my main research question I devised a set of subsidiary questions.
The core subsidiary questions are listed in Table 3.1, matched to their data collection methods. I also needed to investigate the social, economic and cultural context in order to facilitate the analysis. Further research questions concerning the study context are listed in Appendix 2.
Overall research question: How important are climate threats (and by analogy, climate change) to human security in a rural African context of multiple risks, shocks and stresses and what do the socio-cultural aspects of current threats and responses imply for climate adaptive capacity?
Subsidiary research questions
What information do I need? How can I obtain the information? Table 3.1: Core research questions matched to methods
My research methods traced a definite trajectory during the course of the fieldwork, in that although I began by relying on specifically selected and designed research instruments during the individual interviews, after the pilot interview stage I adopted a more open, less structured and more flexible approach to collecting and analysing
data. Also, as our fieldwork proceeded, the semi-structured interviews began to be complemented by more natural encounters with villagers, usually with people who had already been interviewed. Talking to them informally during these chance meetings turned out to be an excellent source of data and a considerable proportion of my evidence is derived from such fortuitous conversations. Throughout the fieldwork I used observation.
Table 3.2 sets out the distinct phases of my fieldwork and the methods I used during each phase. I then go into some detail about the different phases of the work and the individual methods.
Table 3.2: Fieldwork methods in each phase
Reconnaissance visit
I first visited the Mount Elgon region in October-November 2008, when I spent three weeks identifying a suitable location. Before leaving the UK I read relevant literature on Uganda and in particular Bugisu, the south-eastern region of the country that is inhabited by the Gisu people. During the visit I had several meetings with local government officials and elected officers in Bududa and Manafwa Districts and had a chance to see the topography of the area, seasonal agricultural activities and University in Kampala and had worked as a research assistant for various British and European academics. People with this level of qualification and experience are few and far between in Bugisu, so I quickly arranged to hire him as my senior assistant
for the duration of the fieldwork. He made a very significant contribution to the study, for instance undertaking a follow-up visit in 2010 as I explain below, and showed professionalism as well as a genuine interest.
After travelling around both districts and talking to a large number of people, I decided to choose Bubufu as my location. I have already explained that it represents a paradigmatic case of communities in Africa that are poor, dependent on natural resources and exposed to multiple risks, shocks and stresses. It is also fairly typical in other ways. It is neither particularly remote nor close to a major town, and it is not as badly exposed to landslides as some villages in the area, where I thought people might be completely preoccupied with that particular source of risk. My choice was confirmed by the welcoming attitude of local government officials and the village chairmen.
Preliminary fieldwork
I returned to Uganda in January 2009. This phase included recruiting and training junior research assistants to carry out the household enumeration survey and translate for me during my own interviews with women in the village. Local government officials spread the word that temporary work was available for local people with a specific educational standard. Several people applied by letter, and the final selection of three extra assistants was made by interview. Two ‘fixers’ from the village were also recruited by word of mouth; it was their job to contact villagers chosen for interview and arrange a convenient time and place with them.
After a two half-days of training in a local venue we began the household enumeration survey, initially with Mr. Balungira and me accompanying the other assistants. For more information on training the fieldwork team see Appendix 3.
Semi-structured interview instruments and topics were then piloted with ten individuals. Several key informant interviews were also carried out at this time with local government officials and village elders, and a participatory wealth-ranking exercise was conducted. Appendix 3 contains details of the methods used in the preliminary and main fieldwork phases.
Main fieldwork
The remaining semi-structured interviews and key informant interviews, focus group discussions and participatory rural analysis (PRA) discussions were all carried out between the start of March and the end of June 2009. Appendix 4 contains the semi-structured interview guide. Interviews were nondirectional in that they did not steer people towards talking about climatic threats despite my particular interest in this topic.
The semi-structured interviews varied greatly in the amount and quality of evidence they produced. The relatively large number of interviewees (58 individuals, some of whom were interviewed twice) was vital to achieve data saturation, the point beyond which no new topics or themes emerge (Bryman, 2008:462). They also provided a matrix of data against which the validity of insights from the most productive, detailed and interesting interviews could be appraised during the analysis (see below).
Follow-up visit
In 2010, after I had returned to the UK, I commissioned Mr. Balungira to return to Bubufu to conduct some follow-up interviews, instructing him by email on the topics to pursue. Although a follow-up visit had been planned for some time, a landslide that killed over 300 people a few miles away at the beginning of March injected urgency; Mr. Balungira made his visit to Bubufu a few days later. The aims were, first, to find out how the village had been affected by the heavy unseasonal rains that, according to the media, were causing havoc in the Mount Elgon area at the time.
More specifically, I also wanted to find out what had happened to some of the individuals we had interviewed in 2009. Had the worries they articulated in 2009 been realised in the intervening months? If so, how had they dealt with them? This visit produced very useful data. As well as revisiting individuals who had been interviewed already, Mr. Balungira also spoke to people whose houses were at risk from landslides but who had not been in our original sample.
Analysing transcripts and notes
Once all the recordings were transcribed and the notes typed up, back in the UK I used NVIVO8 to organise the data and code it according to emerging themes. As social differentiation is an important theme running through my study, I looked for patterns and variations of perspective associated with my sampling categories;
gender, wealth rank, household headship and life-stage. A common criticism of qualitative research is that it allows ‘verificationism’, the tendency of researchers to look for evidence that supports their preconceived ideas and to ignore any evidence that undermines them. As I developed interpretations I tried to avoid this trap by deliberately looking for counterexamples in the data.
Sampling
I treated two Local Council 1 areas (LC1s) as a single sampling frame because neither was large enough to serve as a sampling frame in itself. Until a few years ago they had comprised a single administrative unit. Administrative boundaries in this area are frequently changed in response to rapid population growth and the need to legitimise claims to government resources; two LC1s are eligible for more government funding than a single LC1. Most villagers I asked did not know where the boundary between the two LC1s ran, and shortly after my fieldwork they were further sub-divided anyway, so I am confident that conflating the two LC1s made no difference to the evidence I gathered. I defined a household as a group of people who ate together. Before they marry, young men commonly live in separate huts in their father’s compound but eat with their parents; I counted such young men as members of their father’s household.
My sampling strategy was purposive and ‘theoretical’, in the sense that it was driven by my hypotheses about what criteria would be likely to differentiate perspectives and experiences within this small community (Gomm, 2008). First, I wanted to interview roughly equal numbers of men and women in order to avoid gender blindness. I also needed to talk to men and women in a range of structural positions, defined as ‘the perceived set of possible interactions with other actors and the perceived likely outcomes of these interactions’ (Homer-Dixon, 1999:137). The ethnographic, African risk perception and gender and development literatures
suggest that gender, wealth rank, household headship and life-stage/marital status would all be significant.
Each household in the village was assigned to one of three wealth ranks through a participatory wealth-ranking exercise (Appendix 3). Working in three villages near my own study location, Dolan (2004:647) found that 20 per cent of all sample households were female-headed, so clearly this was an important category to include in my sample. Female household heads have to be distinguished between de facto and de jure female household heads. I decided to interview female household heads from households assigned to all three wealth ranks: however, in practice all the village widows fell into the lowest wealth rank, which reflects the custom of stripping a wife of land and household assets after her husband dies, and only one of the de facto female household heads belonged to a household in the highest wealth rank.
I also included in the sample young unmarried men and women aged 18 or over:
these young people are poised to pass into adulthood and start their own households so they might be expected to have specific perspectives. Previously-married men now living alone were also included in a separate sample segment. According to the ethnographic literature (La Fontaine, 1959; Heald, 1998), Gisu inheritance traditions have the effect of creating impoverished men at both ends of adult life, and Gisu men living alone, whether widowers, divorced or never married, often experience specific vulnerabilities. It was important, then, to interview a sample of the men who fell into this group and take into account the reasons that they were living alone.
Based on these criteria, I created 15 sample segments, trying to get roughly equal numbers of interviewees in each segment (Appendix 5). The sampling was done by populating the segments with named individuals based on information from the household survey and selecting interviewees from those names. It was not entirely random, because I tried to interview spouses and children over 18 from the same households with a view to highlighting intrahousehold differences. In the event, that was seldom possible and the 58 people interviewed individually were drawn from 40 separate households. As well as being interviewed in 2009, a good number of these people were also interviewed in 2010 depending on their availability at the time of Mr. Balungira’s follow-up visit.
Appendix 6 lists the people who were interviewed individually about threats and responses, using the pseudonyms I assigned to them to protect their anonymity. It also shows where interviewees belonged to the same household. In the event, in some categories we interviewed more people than we had originally intended, for reasons I explain below.
Participants of the group discussions were segregated by gender because I anticipated that it would be a key differentiating factor in threat perceptions and responses. One group consisted mainly of married women and the other of married men. In both cases there was a mix of all three wealth ranks to try to access differences in perspective along wealth lines; see Appendix 3 for more information about the participants.