Data Sources and Collection

From the reviewed literature, we derived a unified understanding of circular economy in the agricultural sector and a range of proposed determinants. This has formed the foundation of

our research and guided our data collection. Our data consists of primary and secondary sources. According Saunders et al. (2009), primary sources is classified as first occurrence of a piece of work whereas secondary sources is classified as subsequent publications of primary literature. Our primary sources consist of six interviews with smallholder horticulture farmers in Kenya and four key informant interviews (Appendix 1-10) as well as observations.

The secondary sources consist of journals, reports, textbooks, news articles, academic articles and indexes, which have been used to substantiate the findings from our primary sources.

As we are doing an exploratory study, we want to find out what is happening and seek new insights. Therefore, our main method of collecting our primary data has been through semi-structured interviews. These are a non-standardised form of interview that follows some key questions and themes central to the study. However, the questions might vary from interview to interview. We deemed semi-structured interviews the best tool for data collection as it assured that our key topics were covered while the interviewees also had a chance to develop their own ideas and arguments. Nevertheless, before going into the field, we conducted some informal and unstructured interviews in order to get advice on doing research in the Kenyan context as well as to establish contacts we could use during our fieldwork. As these interviews were used mostly as background knowledge and to get started with our research process, they have not been included in our findings. In addition to our interviews, we have also collected primary data through observations. We have done participant observations, which is a research strategy in which the researcher immerses in the research setting with the objective of sharing in people’s lives while attempting to learn their symbolic world.

Participant observations are qualitative and works with the meaning of people’s actions (Saunders et al., 2009). We have used the observations as it was important for us to observe people in their social setting and understand what is going on. Especially observing the farmers’ farms, fields and how they arranged their production were very insightful observations as it shows working conditions that might affect drivers and barriers and that we would not have realised or asked in to without physically being present on the farms.

3.4.1 Sampling Method

The population from which our sample is drawn consists of the horticulture smallholder farmers in Kenya and experts working within the field. As it has not been possible to

interview the entire population, we have selected a non-probability sample for our interviews. Non-probability sampling is of more qualitative character and may be used when there is only access to a limited number of the entire population. Even though no statistical conclusions can be drawn, you still may be able to generalise from non-probability samples just not on statistical grounds (Saunders et al., 2009). Our key informants were chosen by purposive sampling as these were selected to be cases, or informants, that would best enable us to answer our research question. This form of sampling is used when you want to select cases that are particularly informative and may also be used by research adopting a grounded theory strategy like in this paper. Nonetheless, purposive sampling does not allow for generalisation about the total population but helps to provide an understanding of the topic and context (Saunders et al., 2009). Due to time and cost constraints, it has not been possible to interview farmers from all counties in Kenya. Nevertheless, with the assistance of prior contacts, we ended up having access to farmers in three counties. Operating in the horticulture sector with smallholder farmers entailed various barriers hindering accessibility.

This meant that our research participants were primarily selected by convenience sampling.

We are aware of the fact that the counties in Kenya differ in various ways in regard to regulations, crops grown and climate. This might entail different barriers and drivers in terms of implementing circular economy practices. Therefore, our findings might not be applicable to all counties.

3.4.2 Interview Process

In order to ensure completeness in our data collection as well as to increase the reliability of our findings, we constructed an interview guide (Appendix 11) for our interviews with the smallholder farmers as recommended by Yin (2003). Throughout the data collection process, we adjusted the interview guide several times as we discovered new areas and points that needed to be included and elaborated on. The interview guide included an introduction to our research and us in which we made it clear that the interviewees were anonymous and could stop the interview at any time. This is very important, as the informants usually will feel more relaxed in the situation and will speak more openly. Moreover, we explained our research purpose and the principles of circular economy in the agricultural sector to the interviewees.

The main part of the interview guide contained our key themes related to agricultural and circular economy practices and the drivers and barriers that might affect the implementation

of circular economy provided by existing literature. Further, the interview guide served the purpose of ensuring that our interviews were conducted in a uniform fashion. As we wanted different things out of our interviews with the key informants and the smallholder farmers, the process of these interviews differed. In relation to the key informants, we did not have a uniform interview guide that we utilised for all four interviews. As the key informants had different backgrounds and expertise areas, we would prepare a set of questions or areas that we wanted covered in each interview to make sure we would get the needed information.

Moreover, conducting the interviews semi-structured allowed the informants to shed light on and cover some topics that we might otherwise not have discovered in our research.

Throughout the interviews with the smallholder farmers, we tried to formulate questions in a neutral way and to avoid leading questions in order to assure validity and comparability between the interviews. Due to the technicality of the topic, we tried to simplify the concept and made sure to frame the questions in an easily understandable way. We realised that the understanding and talkativeness of each informant varied a lot and hence questions in some of the interviews were leading, which we deemed necessary in order to grasp and understand the informant as complete as possible. After each interview, we offered our contact information so they could contact us in case of any further questions. From the informants that agreed, we recorded the interviews in order to assure that no points were missed. For the ones that did not want the interview recorded, extensive notes were taken throughout the interviews. However, we know that this puts the data at risk of interviewer bias, as the notes reflect our interpretations of the information provided by the informants.

3.4.3 Data Processing and Coding

To ensure that we got the most out of our data and that it was available for a comprehensive analysis, we transcribed the interviews shortly after they were conducted (Appendix 1-10).

This meant that our memories were still present and it decreases the risk of missing important points. In order to analyse our data systematically, we coded our farmer interviews (Appendix 12). We divided our findings into seven main determinants that we considered the most important for the smallholder farmers’ adoption of circular economy from the interviews. Having identified the seven main determinants, we were able to more precisely categorise our findings in terms of the main determinants’ impact on circular economy

adoption. Coding our data, we had in mind our unified understanding of circular economy practices in the agricultural sector. Our secondary data has been used to support or oppose the findings from our primary data.