• No results found

PART 1: FARMING SYSTEM ANALYSIS OF THE OKAVANGO RIVER BASIN (ORB)

1.5 Empirical data gathering and analysis

1.5.5 Focus group interviews

Focus group interviews represent the third method of empirical data gathering of this study. They were carried out within each study site and had two main goals: First, to validate the information obtained in semi-structured farmer interviews and second, to gather both qualitative and quantitative data on mean labour-use within farm-households and on inputs into smallholder production.

The idea behind complementing the qualitative semi-structured interviews and the quantitative household survey with focus groups interviews is data triangulation (Walliman 2005, Maher et al. 2015). Each of the methods described has its own strengths and limitations, yet by combining them and cross-checking their results, more reliable information may be gained. Following Wolff et al. (1993), this study uses focus groups in data triangulation to:

i) illustrate and confirm conclusions from survey analysis, and to

ii) determine new explanatory categories by examining a topic from different dimensions, thus arriving at a better understanding than would be possible using either of the triangulated approaches alone.

To increase the degree of triangulation success, all three methods were designed concurrently and implemented to provide asymmetrical but independent observations of the study sites (see Wolff et al. 1993). This strengthens a researcher’s ability to draw conclusions as well as the confidence in the conclusions themselves (ibid., 133). Furthermore, the choice of smallholders included in both the focus group- and semi-structured interviews was based on the household data base created via the quantitative survey.

In both Mashare and Seronga, separate focus groups were conducted with smallholders that were purposively selected from both the poorer, non-ox-owning cluster and the wealthier, ox- owning cluster. In Cusseque, the survey sample was not grouped into two clusters but regarded as one set or cluster of households (which follow comparable farming practices and that rely on a comparable resource base). For each cluster, two focus groups were planned: one concerning traditional smallholder agriculture and agricultural labour needs and a second one dealing with non-agricultural household tasks and livelihood activities (gathering data on mean annual labour-needs for household chores, natural resource use and animal husbandry). Both focus groups were supposed to be conducted with the same participants. In Mashare, an additional focus group focused on conservation agriculture (CA) and in Seronga, a complementary focus group focused on animal husbandry. In the end, a total of ten focus groups was conducted, five in Mashare (two on arable agriculture, two on non-agricultural tasks, one on CA), three in Seronga (two on arable agriculture, one on animal husbandry) and two in Cusseque (one on arable agriculture and one on non-agricultural household activities). However, in Mashare, the focus group dealing with non-agricultural household activities of the wealthier cluster needed to be aborted, because participants lacked enthusiasm to lead a lively discussion and the validity of the answers began to appear questionable. As there was no time to conduct the focus group a second time, labour-data on mean non-agricultural household activities in Mashare are based solely on the information provided by the poorer cluster. This had implications for Seronga as well. Here, another researcher from the TFO project (Eigner 2012) had already collected mean annual labour-data on household activities. As the non-agricultural labour data in Mashare were used for both clusters of households, the

same approach was chosen for Seronga. To increase data fit, Eigner’s (2012) assumed mean household size (and resulting labour needs) were adapted to each cluster’s mean household size18. In Seronga, the time saved by skipping these two focus groups was re-invested into a focus group on animal husbandry and additional semi-structured interviews.

For the choice of focus group participants, random stratified sampling was chosen. This means that the site- and cluster-specific samples were differentiated by gender of the household head, its education (not higher than successful completion of primary school vs. attending secondary school or higher) and its age (using the age median of the respective study site to divide the samples into an older and younger group). This allowed for creating a total of 8 sub-samples for each site and cluster, from each of which one household head was randomly selected to participate in both focus group (i.e. the agricultural and the non- agricultural). In order to ensure that each focus group participant could freely express his opinion, it was ensured that no person of official acknowledged authority, e.g. priests or traditional authorities, were included in the interviews. However, potential non-official power relationships could not be controlled for in the choice of participants.

All focus groups were carried out using a uniform set of guidelines, as well as a local moderator. This moderator was an employed member of the TFO project who was originating from the study sites’ wider area and who understood the local cultural context. This study’s researcher was responsible for introducing the project and specific tasks to the participants and for documenting the process, while the moderator was responsible for ensuring a smooth discussion and a general understanding of the tasks required from the participants.

In general, a relatively high level of control was maintained over the discussion and new tasks and topics were regularly introduced by both the moderator and the researcher. They kept the conversation focused, but also provided check-up questions or interjected in the discussion to make sure the participants elaborated on all topics of interest and did not forget any important topic in the course of the discussion. Special attention was paid to ensure that all participants could clearly voice their opinion and that no idea or objection was discarded by more dominant participants. To ensure a smooth flow of the discussion, these interventions happened rarely and only if a participant was clearly ignored in the discussion.

A central task of the focus groups was to provide quantitative data on mean annual labour- needs per household (for non-cropping activities) or per area (for crop production). This was achieved by creating two main outputs per focus group, namely: i) a seasonal calendar of all tasks carried out within a typical households (as defined jointly by the focus group participants) throughout the year, based on a brainstorming exercise and, for non-cropping activities, on ii) a table indicating which of the tasks identified in step one is carried out in which months, for how many days each month and for how many hours on that day by how many household members, separated into adults and children. These products represented the condensed output or results of the focus group and no transcript on what was said by which

18 Eigner (2012) assessed the labour needs of a household of ten members, consisting of three adult males, three

adult females and four children. Acording to the household survey in Seronga, the mean household size of all crop-producing households is 3.85. The actual household size is therefore 38.5 % of the value used by Eigner (2012). Therefore, the labour needs of all non-crop-production related activities reported by Eigner (2012) were multiplied by 0.385 and only then used for this analysis.

participant was created (also due to the different languages). All subsequent analysis focused on these products.

For creating this output for the crop-production related activities, this was achieved by leading the focus group participants to a field of 20m*25m (=500m²). Here, they were asked to estimate (individually), which agricultural tasks they would carry out on that field, how many days each task would last and how many hours per day how many household-members (of which gender) would have to work to finish this task. After 20 minutes, the focus group commenced within a nearby building. It entered into a discussion phase where participants were asked to discuss and jointly agree on which tasks needed to be carried out and how much time was needed for each task (considering an average work speed of a healthy adult, while always being aware that some smallholders might work quicker than others). The final output of this exercise was a value on labour hours needed per task upon which all participants could agree. A similar discussion took place on the next day for the non-crop production related tasks.

The validation of semi-structured interview results took place via probing into the main agricultural tasks mentioned by the focus group participants and on how commonly they were applied. The degree to which the moderator scrutinized a task depended largely on the degree to how quickly and unanimous the participants agreed on this task’s timing and mean labour- needs. Those tasks mentioned in the semi-structured interviews, but not during focus group, were introduced by the moderator and their importance discussed within the group. If the group assumed that this task was relatively common to the study site, it was added to the seasonal calendar. If it was unheard of or basically non-existent, it was not added to the calendar (such as irrigation in Mashare and Seronga, which was carried out by only a handful of farmers in both study sites).

In Mashare and Seronga, where households were differentiated into two clusters, the mean value of the labour-needs stated by the two clusters was used for all subsequent analysis. This means that for all tasks which both clusters were considered to be experts on (such as weeding, planting, harvesting), the mean was used for gaining a credible approximation of what could be the specific tasks actual labour-need. In cases where one of the clusters was assumed to be an expert (such as the poorer cluster and soil preparation by hoe or the wealthier cluster and manure & fertilizer application), only the value reported by the “expert”-cluster was used for subsequent analysis. In the end, the approximated labour needs were validated by comparing them with similar smallholder farming systems of the tropics. The results of this comparison can be found in the site-specific analyses (see tables 1.14, 1.21 and 1.30 in chapters 1.6.1 – 1.6.3).

In general, the labour-data generated in the focus groups closely resembled those reported in literature for similar farming systems (see site specific analyses). This confirms that trustworthy results on agricultural labour-needs per hectare can be obtained by having homogenous farmers’ groups discuss and compromise on a set of plausible labour-values for an exemplary, clearly demarcated field.

It is likely that basing the labour-values on individual farmer interviews would have resulted in less accurate results, because these values would have relied on farmer recall only. In

general, the longer ago an activity was carried out, recall becomes more blurred (Spencer 1991). In order to avoid these significant memory lapses in individual interviews, they have to be carried out via frequent visits and surveys (ibid.). However, I had not sufficient time and resources to carry out regular visits. Therefore, focus group discussions provided the best available means to collect reliable data - with the one exception of conservation agriculture; this issue and corrected labour data will be presented at the end of the chapter. Both the consideration of study site and the choice of conducting separate focus groups for the identified clusters introduced a simple level of analytical control, which might be similarly accomplished in a quantitative analysis through the use of statistical methods (see also Wolff et al. 1993).