«Deborah Duveskog Faculty of Natural Resources and Agricultural Sciences Department of Urban and Rural Development Uppsala Doctoral Thesis Swedish ...»
Survey indicators for the impact questionnaires were defined through stakeholder workshops carried out in each country in 2004/05. The workshops were 3-days and included about 20 persons; FFS member farmers, farmers from FFS networks- producer organisations, field extension staff, and project coordinators. During the workshops participants, through interactive facilitation, developed indicators for expected outcomes of FFS education, based on the local perspective. Indicators related both to short term aspects such as access to services, agency, organisational skills etc. and more longterm outcomes such as improved livelihoods and well–being. Even though there were some variations of indicators in the three countries, the outcome of the events were surprisingly similar, supporting the idea of the possibility for standardising survey content across the three countries. The process applied in these seminar events built on principles of co-operative inquiry used to enable groups of people to gain better understanding of their everyday experiences and develop new and creative ways of making changes (Heron 1971; Heron 1996: Reason and Bradbury 2001).
Individual in-depth interviews For more in-depth understanding of how empowerment and transformation played out in individuals’ lives (Merriam, 2002), individual interviews were carried out with sample FFS members and graduates. The specific aims of the interviews were to understand (a) respondents’ perceptions of their experiences of FFS in terms of both instrumental and personal gains, and (b) changes induced at the personal level (skills and world views) and in respect of relationships at the household/community level following FFS participation.
Individuals in Kakamega district, Kenya, were purposely sampled, with assistance from local FFS network leaders, to (a) represent typical FFS graduates, (b) be informative examples of personal changes resulting from involvement in FFS, and (c) to ensure gender balance among respondents.
Twenty individuals were interviewed, half of which were graduates of FFS from about year 2000, while the other half were made up of current FFS members or more recent graduates. The in-depth interviews followed an interview guide developed to ensure that certain questions were covered. In association with the in-depth interviews ten key informant interviews with FFS facilitators and FFS network officials were also carried out with. The approach permitted flexibility to explore and probe topics of interest to each respondent (Patton 1990). Translators were used during the interviews that were audiorecorded for transcription supplemented by handwritten notes. Interviews transcripts were analysed, using a constant comparative approach, party using NVIVO-QSR (version 8) in order to identify its essential elements and for coding.
Figure 11. One of the respondents for the individual interviews with his wife and outside their home.
(Photo by D. Duveskog) Focus group discussions Focus group discussions among people from similar background or experiences, brought together to discuss a specific topic was used to gain more in-depth understanding of certain issues. The structure of these focus group discussions was kept to a minimum, allow feelings and characterizations to emerge from the participants themselves (Dawson et al. 1993). Focus group discussion were used to generate data in terms of background information, opinions, ideas, perceptions, and beliefs and experiences on aspects and factors that influence opinions, behaviours and motivation among farmers. Focus groups were also used at the design stage of the household survey and other quantitative tools to frame indicators and questions, and used to shed light on quantitative data collected. Focus groups were used mainly among groups of FFS farmers, FFS networks and among village/ward committees and other local institutions. Data was recorded either by written notes or recorded and transcribed.
Figure 12. A group interview undertaken in Kakamega, Kenya. (Photo by D. Duveskog)
Participatory visualisation tools Various visual and interactive facilitation tools, building on PRA practice (Chambers 1994) were used where appropriate and in particular in connection with focus group discussions and explorative workshops. A range of tools was used during the preparation phase of the research, such as; mapping, evaluation wheel, flow diagrams, network diagrams, change tool etc. Such tools help stakeholders improved practice in programme and intervention context (Chambers 1993; Guijt and Braden 1999) and were thereby a valuable means to induce change and action among respondent participants and contribute to a spirit of action research.
Key informant Interviews Key informant interviews were used to capture data related to institutional and policy issues and opportunity structures and to understanding local contexts and situations. Semi-structured or fully open-ended interviews (Patton 1990) were held with selected key informants such as local and national extension managers, extension workers, government officials and village leaders. Most of these interviews were recorded and transcribed. During field visits, meetings and training events, consultative meetings and project visits informal talks also occurred, and were documented through field notes.
Secondary data sources Secondary data were also reviewed and analysed, including policy documents, extension management guidelines and procedures at the local and national levels. The aim was to gain an understanding of how institutional issues influence service provision to farmers and constraints and facilitating factors in responding to farmers’ demands by the government and other extension actors.
Further, background materials about the bio-physical, socio-economic and cultural contexts in the various study sites were analysed so as to gain a better understanding of the local situation.
Personal diary In the context of viewing my research as an action research inquiry within my workspace I kept a diary for descriptive accounts of my everyday research and work experiences. Observations and reflections in everyday events such as visits to farmer groups, discussions and meetings with extension staff, participation in national policy processes and meeting etc. was recorded and reflected on as part of the research process.
Direct observations Direct observations played an important role throughout the research period in contextualising findings and understanding contexts (Patton 1990) as well as to understand relationships and interactions among individuals and sub-groups.
Observations were made during regular FFS group meetings attended, villages visited, stakeholder and community events attended. In particular, observations were useful in understanding gender dynamics by observing the interactions between men and women during FFS group sessions and events.
Action research within the workspace This research was party carried out as an action research process within my regular workspace. For much of the study period I was acting as researcher while at same time program advisor for the development intervention under study. In a way I was researching my own practice and studying aspects and phenomenon that my actions and me were an important component of. While these dual roles provided challenges in terms ensuring true objectivity of the research it also provided a range of opportunities for ensuring direct impact of the research in informing practice. My aim was to generate concrete and practical knowledge to enable those responsible for making policy, managing programs and delivering services to make more informed judgements about their activities, thereby make services more appropriate and effective for the people they serve. This perspective is fully consistent with the motives and objectives of participatory action research (Stringer 1999; Reason 1994) where apart from producing knowledge and action useful to the community, it also empower people to construct and use their own knowledge. This implies less emphasis on uncover generalizable truths and more focus on the emphasis on the realities of individuals/groups in local contexts (Stringer 1999). This action research perspective allowed me to be a researcher while at the same time act as a change agent and assume benefits related to my role as ‘insider’. Research that operates at a distance from the everyday lives of practitioners, and largely fails to penetrate the experienced reality of their day-to-day work (Stringer 1999). Reason (1994) points out the fact that we can only understand our world as whole if we are part of it, as soon as we stand outside we divide and separate. In action research there is no functional distinction between the researcher and the researched. They are all defined as participants, and have equal footing in determining which questions to be asked, information to be analysed, and conclusions to be made (Stringer 1999). To undermine the possible drawbacks of problems in objectivity I deliberately teamed up with research colleagues that did not have an involvement with the FFS program under study for data analysis and this provided a continuous check on my research findings and conclusions, to ensure that my own biases did not undermine the research. Reason (1994) argues that true objectivity does not exist and that the observer is always inseparable from that which is observed.
Instead he refers to the term ‘critical subjectivity’, arguing that the validity of our encounters with experience rests on the high quality, critical, self-aware, discriminating and informed joint judgments of the research actors and subjects.
The below table gives an overview of the use of the various methods explained above in the different papers.
8.3 Data analysis 8.3.1 Analysis of interviews Interview data was analysed in an inductive manner, where themes were generated based on emerging similarities of expression in the data material. A constant comparative approach where the various groups of respondents were treated the same. All individual interviews and most group interviews were recorded and, by a third party, transcribed and translated into English where needed. The software NVIVO-QSR (version 8) was used to separate data from the transcript and identify essential elements. Many of these segments later provided quotations in the write up of research findings, where pseudonyms were used in order to protect the anonymity of the respondents. Each transcript was systematically reviewed and responses coded (Miles and Huberman 1994) based on which common themes were identified and sub grouped thematically.
Analysis continued until there was a consensus on interpretation and each category was ‘saturated’, that is, further analysis appeared to yield no new information (Lincoln and Guba 1985). Many of the sub-headings in the findings section of this summary as well as in the published papers represent themes that emerged through this analysis process.
8.3.2 Statistical analysis of survey data A variety of survey tools were applied for this research and data was accordingly analysed in a variety of ways with statistics computed by use of the SPSS software. Questionnaire sections relating to the expressions of wellbeing among respondents were analysed according to the established wellbeing ranking methodology where a poverty index is based on a field tested set of poverty indicators that form the basis for a household poverty index computed (Ravnborg et al. 2004; and Friis-Hansen 2005). For analysis empowerment-related variables were separated into two groups: 1) selfperceptions and attitudes among farmers towards their power and agency in life, such as the power to influence their lives and community, trust and gender relations; and 2) actual expressions of agency in their daily lives, such as productive assets, knowhow, access to services and the ability to plan. The questions in the survey relating to self-perceptions and attitudes were captured by a three-point summative scale (Likert 1932): 1 = agree, 2 = neither agree nor disagree, 3 = disagree. An example of the attitudinal statements used was, ‘I feel I can make this village a better place to live in!’. Questions related to expressions of agency in everyday life were mainly included in the questionnaire as binary items (yes/no questions) such as ‘Are you a member of a savings or credit organization?’ All statistical analysis was carried out using SPSS software. A factor analysis was carried out in order to aggregate empowerment factors. The eleven questionnaire items with a summative scale on perceptions and attitudes were subjected to principal component analysis (for each country data set) to segment the variables into fewer factors of self-perception of agency.
Correlations between the input variables were first checked and all variables were shown to correlate with at least a few others. Spearman and Pearson correlations were checked to confirm that the limited three-point ordinal data set was applicable for factor analysis. Models for four, five, six and seven factors were computed (factor rotation method Equimax with Kaiser Normalization). Correlations and factor-loading coefficients were used to understand the nature and structure of the four factors. Finally the factors were labelled and saved as individual variables subjected to the cross-tabulation in the same way as the binary dataset items.
Levels of significance were tested on both the emerged factors and categorical variable items by cross-tabulation and Pearson Chi-square tests.
The binary data items, generated from the qualitative interview work, and emerging factors were analyzed through cross-tabulation to identify dependences between various variables, such as between the empowerment variables and poverty level categories. In comparing the categorical variables and testing of significant differences between groups, the Pearson chi-squared (χ2) test was used.
Participants whose responses were incomplete were excluded automatically by SPSS in the data analysis. More detailed description of the survey data analysis is presented in the published papers.