Chapter 3: Research Design, Methods and Study Site
3.3 Research Methods
3.3.10 Fieldwork Observations
In all three phases of the research, the overt observation (where everyone knows they are being observed) method was used to investigate cross-cutting issues that were particularly of relevance to the components of this thesis (Taylor-Powell and Steele, 1996; Gray, 2009). Various aspects of the EPWS program intervention, and household livelihoods were observed to allow the validation of responses from household interviews, key informant interviews and focus group discussion and thus contributed significantly to the triangulation process. Photographs and field work notes were taken throughout the research period. Probing questions such as ‘how’, ‘why’, ‘when’, ‘how much/how many’ were asked in a non-intrusive manner to tease out details on issues of particular research interest.
3.3.11 Field Experience
Permission to conduct research in Kibungo Juu ward was granted by the Ward Executive Office of the Kibungo Juu ward, District Administrative Secretary of Morogoro rural district, and Regional Administrative Secretary of Morogoro region with institutional support from the University of Dar es Salaam. Formal introductory meetings were held in each sub-village on entry with leaders and community members through the facilitation of the Village Executive Office in each village. In these meeting, the role of the researcher was explained in the village and people were told that a research student was conducting research on environmental and livelihood issues and that the purpose of the research was to learn from them about these issues.
Throughout the research, the researcher maintained vigilance of the research process. Temporary research assistants with a degree qualification were employed to assist during the household questionnaire interviews. These research assistants were selected after examining their potential to act objectively within interview and group situations. Prior to implementing the surveys widely and following the pilot in Kibungo village, all field assistants were trained and practiced posing the questionnaires to one another. The researcher rotated among field assistants as they performed interviews to ensure that the assistants asked questions in a consistent manner. Each evening the researcher examined questionnaires to ensure that there were no inconsistencies, and in the event of discrepancies, the interviewee was followed the following day for clarification.
In each village, one local research assistant was hired to direct the research team in the village and sub-villages, and to the respondents’ homes. This was important because of their considerable local knowledge and to assist in introducing the interviewers to households, as they might be less intimidating for respondents. The approach to the respondent's house was usually made by the local research assistant. It was explained to the respondent that the interviewer was a visiting researcher, and would like to talk to them, if possible. Interviews were then either conducted inside or outside the home in the absence of the local research assistant. The interviewer started by explaining the purpose of the research and that the researcher was a student studying these sorts of projects and was interested to hear their experience.
Kiswahili is the national language in Tanzania and was spoken at the study sites. As such, there was no language problem because Kiswahili is the common language for the researcher and research assistants as well as the respondents.
Numerous forms of bias can enter into data collection. The researcher’s main concern was the identification of the research with CARE/WWF or the government, which could impose biases in the responses. To avoid this, a car was hired and CARE/WWF or government transport was avoided. Also, research assistants/interviewers were asked by the researcher to be as clear as possible to avoid being associated with the project. As such, respondents were helped to understand the purpose of the research in order to gain trust and limit biases. However, some individuals would inevitably look upon the interviews with suspicion. Yet this suspicion cannot be solely attributed to the researcher’s relationship with CARE/WWF because suspicion of outsiders is common in the area.
In addition, respondents sometimes asked questions at the end of the interview showing that the distinction between the interviewer and project management was not very clear in their minds. In this scenario, people asked the interviewer to change things about the project. However, the interviewers were trained to reiterate that they were not involved in the project management and, although the findings would be available to CARE/WWF, the researcher could not guarantee anything would change. This was done purposely to ensure the researchers’ presence would not raise unrealistic expectations.
In the second phase of the research, people were much more interested in the researcher and the research work and they became more comfortable talking due to the teams’ engagement with people in some local activities and as they got to know the research team. For example, the team inspired students at the Kibungo Secondary School by discussing with them the importance of education and how to handle challenges and also worked with the community in making bricks for the local secondary school and in road clearing.
Another concern was about the researchers’ relationship with CARE/WWF staff. However, this did not pose any conflict of interest in the researchers’ critical evaluation of their intervention. They made clear to the researcher that they wanted the brutal truth and they were open in describing both the successes and failures that they perceived in the intervention. As such, the researcher does not believe that bias was introduced in the analyses and conclusions due to a desire to impress CARE/WWF.
Throughout the research process, strict ethical procedures approved by the University of Leeds were followed. Firstly, to address the risk of disclosure, all data were stored securely and anonymised in presentation. It was particularly important that the researcher never reported to CARE/WWF any information directly linked to any individual and that was emphasised to respondents. The research assistants were also briefed on these ethical codes. For seeking informed consent, the interviewers explained to respondents the purpose of the research and what the researcher would do with the information they gave. After
this, they were given a choice as to whether or not to be involved. Due to literacy issues and sensitivities over using signatures, verbal consent was sought.
Another consideration relates to reward for research participation. The researcher felt the presence of poverty and the fact that people gave up time to participate warranted something more than gratitude. However, this was considered inappropriate because people would want to be interviewed just to receive a gift. After a discussion with the research assistants, the team decided to participate in some local activities such as inspiring students at the Kibungo Secondary School by discussing with them the benefits of education and how to handle challenges and responding to questions asked by students. Also, the team worked with the community in making bricks for the local secondary school and in road clearing.
3.4 Data Analysis
3.4.1 Quantitative Data
Data entry templates were designed and adapted as interviews were carried out. The quantitative data were analysed with Microsoft Excel and IBM SPSS Statistics for Windows for basic, descriptive statistics and R for propensity score matching.
The analysis of the household data reflects the experimental study design by comparing the experimental (EPWS program participants) and the control group (Non-EPWS program participants). A propensity score matching technique which involves the prediction of the probability of the EPWS program participation was performed for chapters five and six (Abadie and Imbens, 2005). The nearest
neighbour matching with replacement variant was performed to match the treatment group with the control group. Out of 117 members of the control group, 67 were matched to 116 members of treatment group. The data set was first disaggregated and compared for the two groups (EPWS program participants and non-EPWS program participants). Subsequently, to investigate the equity question, the data was further disaggregated by the four asset wealth groups and by gender. The gender analysis used sex of the household head (male = 151, female = 81) as grouping variable before matching and (male = 178, female = 54) as grouping variable after matching.
Both parametric and non-parametric tests were used. Parametric tests were performed for household level variables that passed assumptions of parametric test (Kolmogorov-Smirnov test), while non-parametric test was performed for the household level variables that could not satisfy the assumptions of parametric test.
The data gathered through the household survey was, where appropriate, triangulated with results from the qualitative data gathered through personal observation, key informant interviews and focus group discussions.
3.4.2 Qualitative Data
To complement the use of multiple methods, data analysis followed a grounded theory approach (Strauss and Corbin, 1998; Heath and Cowley, 2004). Themes, concepts and ideas based on data collected during the first phase of fieldwork in
Kibungo village (Phase 1) were taken forward during further data collection phases in the same village and in Nyingwa, Lanzi and Dimilo villages (Phase 2 and 3). This approach helped to maintain confidence in research outcomes through the constant comparison across types of evidence (Bailey et al., 1999). Additionally, this iterative process ensured focussed and relevant research development. Data from focus group discussions, key informant interviews and observations were manually coded and grouped on similar themes (Neuendorf, 2002). The themes allowed similarities and differences between data to be easily identified and relevant quotes to be easily extracted. Data were constantly revisited throughout this process and new connections between data were formulated.