Researching fair trade empowerment using mixed methods
5.4 Selecting samples and methods
Applying a postcolonial lens to un-mask imagined or constructed categories of producer and community and the relations embedded within them reveals multiple actors involved in the Kasinthula scheme, the cane production process, the Chikwawa district, and in trade relations beyond the local scale. Avoiding assumptions of others reveals the fact that these actors have various degrees of financial, social, and political control and influence. In that context some have greater degrees of financial and political voice than others. Highlighting potential research participants, this section explains the sampling strategies adopted and selection of mixed methods to obtain credible evidence.
5.4.1 Research Credibility
The empirical focus of this research is at the local level with cognition of relevant factors at other scales of analysis. So far in this and the previous chapter I have introduced the study area, the local production network, communities of interest (Figures 5.1 and 5.2), and defined fair trade empowerment indicators. The final technical elements of the impact assessment process are to identify research participants and methods of data collection.
As stated earlier, I employed a mixed methods approach to this research. Building an argument that this approach enhances credibility of evidence and levels of
understanding, Baxter and Eyles (1997) define credibility as the connection between lived experiences and concepts recreated through interpretation. To enhance
credibility ‘researchers focus on respondent selection procedure, interview practices,
and strategies for analysis’ (Baxter and Eyles, 1997: 513). These elements are explored through this chapter.
5.4.2 Sampling research participants
In the vein of a grounded approach to data collection and analysis, Narin et al. (2005:
222) argue that ‘the goal of research is to generate meaning from data’. Therefore, the quality of material collected is crucial if the research is to be credible and of a high standard. Referring to the earlier discussion of triangulation, source triangulation requires the presentation of quotations from several different participants to cross-check themes and arguments from evidence (Baxter and Eyles, 1997). Recognising constraints of time and financial resources, it is important to use a good cross section of participants and multiple sources to increase the epistemological utility of the material.
Research participant selection procedures in a grounded theory approach begin as a common sense process of talking to informants most likely to provide early
information (Draper, 2004). Therefore, selection is likely to be based on established contacts and from networks of those contacts to identify participants. From identified populations a sample frame of research participants can be selected. The sample frame is an up-to-date full list of potential participants from which samples are taken.
In qualitative research the sample sizes are usually relatively small. Ritchie and Lewis (2003) identify three reasons for this. First, there will be a saturation point where no new evidence is obtained, and, therefore, a point of diminishing returns is reached where increasing sample size will not generate new evidence. Second, in qualitative research there is no requirement to ensure statistical validity; instead the selection is subjective to the researchers’ objectives such as eliciting perceptions, experiences, and attitudes of participants. Third, qualitative studies yield rich material and, therefore, it is necessary to keep sample sizes smaller to do the process justice. In support, Saunders et al. (2003) suggest that the smaller number of cases generated from a sampling process versus a census provides more time to design, pilot, and collect more in-depth material.
There are a number of options regarding sampling techniques; the following represent the options I adopted. First, in a simple random process, discussed by Bryman (2008), I ensured each unit of the sample frame had an equal probability of inclusion in the final sample. Second, before randomising, I stratified the sample frame by categories such as age or gender to make sure that some participants from each unit of analysis identified in the sample frame had representation in the random sample. In support, Roche (1999) advocates the use of stratified random sampling for this reason of overcoming potential issues of exclusion or discrimination. Finally, to select
particularly informative cases I took a purposive sample. In this method participants are sampled ‘in a strategic way so that those sampled are relevant to the research questions being posed’ (Bryman, 2008: 415).
The final point to be discussed regarding sampling relates to the question of when to stop. Referring to ideas within grounded theory, Glaser and Strauss (1967) suggest a criterion of theoretical sampling. Material is gathered until a point of theoretical saturation is reached, a point where the researcher has heard a range of stories and nothing new is being learnt (Crang and Cook, 2007), related to the point of
diminishing returns above. With no statistical procedure for sampling, the ‘final sample size is always a matter of judgement’ (Saunders et al., 2003: 155). As a general rule of thumb, Ritchie and Lewis (2003) suggest a maximum sample size of fifty, but ultimately the final sample size will be determined by the theoretical saturation point. To begin my sampling procedure of KCGL I referred to Figures 5.1 and 5.2 to identify the range of sites from where I could find research participants.
Table 5.5 identifies potential participants from the KCGL scheme and the Chikwawa district.
Producer community groups Participants
KCGL KCGL management team
KCGL Phase 1 and 2 committee members KCGL Fairtrade committee members KCGL farm managers
KCGL farmers – male and female
KCGL employees – identified permanent and seasonal
Workers committee members SVCGT and KCGL board Members of Illovo and KCGL
Illovo Illovo Nchalo management team
Illovo national management team
TA governors TA chief
Village headmen / women Local government members District Commissioner
Distract Executive Committee Local villages Non-Kasinthula villagers Table 5.5 Potential participants
Focussing on non-management members of KCGL first, to organise a list of potential participants I requested lists of both farmers and employees. The HR manager of KCGL was able to provide Excel spreadsheet lists of farmers and permanent employees.103 To ensure that I secured representation from as wide a range of groups from within KCGL as possible, I stratified the lists by phase (I and II, Figure 5.1), gender, and job description for the employees. By stratifying the sample frame I provided space to hear perspectives from representatives of all groups in the
Kasinthula production community in subsequent interview processes. This was an important process for this research that is attentive to the different contexts in which different participants are embedded and the fact that in different circumstances people will experience fair trade in multiple ways.
To randomly sample participants from that stratified sample field I used a computer-based randomiser program to generate a random sample.104 To begin the sampling process I assigned a number to each farmer and employee as a sequential four digit number. From the generated random lists I started with the first name and worked down the list sequentially to select potential participants.105 As I only had a short space of time to conduct these interviews I needed to be flexible in who I was able to
103 The issue of accessing seasonal employees will be covered in section 5.6.2.
104 Online randomiser at http://www.random.org/lists/
105 Lists of all people interviewed over the course of both field visits can be found in Appendix B.
interview. Therefore the first samples were based on convenience, selected on the criterion of who could be available at relatively short notice. I discussed my lists of potential participants with the management team at KCGL who were able to contact the selected individuals. If a first choice was not available then we went for the next name on the random list. The first sample I created was purposely small to conduct pilot interviews that I discuss in section 5.6.
5.4.3 Selecting mixed methods
After selecting and sampling participants, there are a range of methods (or tools) available to collect qualitative research material. Principally the methods are based upon some form of interview or discussion with either individuals or groups of participants. Such methods have been used in postcolonial research, critical
development geography, and fair trade impact studies (for instance see Bacon, 2004;
Dolan, 2008; McEwan and Bek, 2009; Paul, 2005; Utting, 2009). Generally the objective is to capture non-monetary impacts (Paul, 2005) by asking what, how, and why questions rather than how much or how many. However, as already stated in this chapter, as part of my mixed methods approach I argue there is a case that some quantitative elements can complement qualitative material collection. I return to this point again later in the chapter.
In discussing qualitative research methods, Bryman (2008) and Ritchie and Lewis (2003) note the flexibility of qualitative methods and highlight three major types of method. First, biographical methods capture life stories or narratives of the lived experiences of participants. Second, individual interviews capture personal
perspectives. Third, focus group discussions (FGDs) capture multiple, intersubjective accounts from a group on a particular issue or phenomenon.
As part of a triangulation process to enhance research credibility, I decided to use interviews and FGDs. In a mixed method approach, group discussions may be
conducted before or after one-to-one interviews. There is no right or wrong approach;
it depends upon the circumstances of the research process. However, Kanbur (2005) argues for one-to-one interviews first to elicit key themes and issues and then follow up in purposive group discussions to expand on those identified themes and issues. As is presented in subsequent sections, I adopted this approach. With time and resource
constraints I decided not to also conduct biographical methods as I wanted to obtain perspectives from as many participants as possible and I was satisfied that I obtained sufficient evidence of lived experiences through interviews and group discussions.
Before I discuss the theory and my application of the principal methods employed, the next section introduces research challenges related to interpretations and