“THEY SAY WEALTH IS IN THE SOIL”: LOCAL KNOWLEDGE AND AGRICULTURAL EXPERIMENTATION AMONG SMALLHOLDER FARMERS IN CENTRAL MALAWI By Michele T. Hockett A THESIS Submitted to Michigan State University in partial fulfillment of the requirements for the degree of Community, Agriculture, Recreation and Resource Studies – Master of Science 2014 ABSTRACT “THEY SAY WEALTH IS IN THE SOIL”: LOCAL KNOWLEDGE AND AGRICULTURAL EXPERIMENTATION AMONG SMALLHOLDER FARMERS IN CENTRAL MALAWI By Michele T. Hockett For smallholders in central Malawi, farm management is complex and dynamic. Farmers’ seasonal decisions are determined by a range of factors including resource availability, environmental changes, and farmer priorities. Moreover, management decisions are influenced by a combination of local knowledge and expert recommendations. Although local knowledge is developed over centuries of experimentation within volatile agroecological systems, smallholder experimentation processes are not well documented in literature and are underutilized in agricultural development projects. This study aimed to examine the decision-making processes of experimenting farmers and explore the drivers of on-farm experimentation. A mixed methods design incorporated field observations, survey data, and in-depth interviews, where quantitative and qualitative threads had multiple points of interface. This study found that Malawian farmers across a range of socioeconomic characteristics are inclined to experiment. While experimental methods differ between farmers, there are commonalities in the drivers of experimentation, including climate change, income generation, and improving household nutrition. Farmers’ current practices should be taken into account in the development and implementation of agricultural intervention projects so that such projects might work effectively with smallholders to improve Malawian farming systems. ACKNOWLEDGEMENTS Thank you to USAID and Africa RISING for providing the financial support that made this research possible. Many thanks to Dr. Robert Richardson, Dr. Sieglinde Snapp, Dr. Regis Chikowo, Dr. John Kerr, Dr. Anne Ferguson, and Dr. Maureen McDonough of Michigan State University for your academic support throughout this project. Thank you to Alex Smith, Issac Jambo, Emmanuel Jambo, Kondwani Khonje , Miriam Mhango, Elian Majamanda, the AEDOs, AEDCs, lead farmers, and village headpersons for your invaluable insight and support in the field. Thank you to Brendan Cooper for emotional support and creative ideas. Special thanks to Dziwani Kambauwa for being my voice and sounding board. Most importantly, immense gratitude and appreciation to the farmers of central Malawi who gave their time and shared their stories with us, and to whom this project congratulates for their tenacity, curiosity, and innovativeness. iii TABLE OF CONTENTS LIST OF TABLES ....................................................................................................................... vi LIST OF FIGURES ................................................................................................................... viii KEY TO ABBREVIATIONS ..................................................................................................... ix 1.0 INTRODUCTION................................................................................................................... 1 1.1 Purpose of Study ..................................................................................................................... 3 1.2 Research Questions ................................................................................................................. 4 1.3 Basis of Study .......................................................................................................................... 5 2.0 REVIEW OF RELATED LITERATURE............................................................................ 8 2.1 Theoretical Framework ........................................................................................................ 14 3.0 METHODS ............................................................................................................................ 18 3.1 Overall Approach.................................................................................................................. 18 3.2 Study Areas............................................................................................................................ 19 3.3 Phase One: Household Survey ............................................................................................. 22 3.3.1 Target Population ............................................................................................................. 22 3.3.2 Data Collection: Survey Instrument and Field Research Team ....................................... 23 3.3.3 Data Collection: Sampling Framework ............................................................................ 25 3.3.4 Survey Data: Inspection, Entry, and Analysis ................................................................. 28 3.4 Phase Two: In-depth Interviews .......................................................................................... 32 3.4.1 Target Population and Sampling Framework................................................................... 32 3.4.2 Data Collection: Interview Structure and Content ........................................................... 36 3.4.3 Qualitative Data: Translation, Transcription, and Analysis ............................................. 37 3.5 Respondent Characteristics ................................................................................................. 39 3.5.1 Household Survey Respondents ....................................................................................... 39 3.5.2 In-depth Interview Respondents....................................................................................... 44 4.0 RESULTS .............................................................................................................................. 47 4.1 “Experimentation” versus “Everyday Practice” ............................................................... 47 4.2 Types of Experiments ........................................................................................................... 49 4.3 Innovative Farmers at the Household Level ...................................................................... 55 4.4 Innovative Farmers Within the Household ........................................................................ 59 iv 4.5 Farmer Motivations and Methodologies ............................................................................. 62 4.5.1 Initiating an Experiment ................................................................................................... 62 4.5.2 Designing an Experiment ................................................................................................. 69 4.5.3 Repeating, Adjusting, or Abandoning an Experiment ..................................................... 72 4.6 Drivers of On-Farm Experiments ....................................................................................... 79 4.7 Summary of Results .............................................................................................................. 82 5.0 DISCUSSION AND RECOMMENDATIONS FOR FURTHER RESEARCH ............. 86 5.1 Farmers Who Experiment ................................................................................................... 86 5.2 Experimental Crops, Varieties, and Techniques ............................................................... 92 5.3 Managing an Experiment from Start to Finish .................................................................. 93 5.4 Smallholder Experiments and the Theory of Reflection-in-Action .................................. 95 5.5 Implications for Development .............................................................................................. 96 5.6 Limitations and Suggestions for Future Research ............................................................. 98 5.7 Conclusions .......................................................................................................................... 101 APPENDICES ........................................................................................................................... 104 APPENDIX A: Household Survey Instrument ...................................................................... 105 APPENDIX B: In-depth Interview Question Guide .............................................................. 124 APPENDIX C: Qualitative Coding Audit Trail..................................................................... 130 WORKS CITED........................................................................................................................ 134 v LIST OF TABLES Table 1. EPA Agronomic Information…….……………………………………………………..20 Table 2. Survey Sample Groups……………….………………………………………………...23 Table 3. Experimentation Classifications by Sample Group……………………………………32 Table 4. Characteristics of Household Survey Population (by EPA)…...………………………40 Table 5. Education of Household Head (by EPA)…...………………………………………….40 Table 6. Characteristics of Household Survey Population (by Sample Group)……...…………41 Table 7. Education of Household Head (by Sample Group)……………………………………41 Table 8. Coefficients for TLU Conversion………………………………………………………43 Table 9. Asset Values used for Wealth Index…………………………………………………....44 Table 10. Socioeconomic Characteristics of In-Depth Interview Population………….……….45 Table 11. Examples of Farmers’ Experiments from Interviews…………………………………54 Table 12. Gender of Household Head by Experimentation Groups…………………………….56 Table 13. Grouping (Dependent) Variable Means by Independent Variable Groups…………..56 Table 14. Gender and Experimental Crops, Varieties, and Techniques………………………...60 Table 15. Respondent’s Gender by Gender of Experimenter……………………………………62 Table 16. Farmer Motivations for Experimentation……...……………………………………...64 Table 17. Farmers’ Use of a “Control” in Experiments………………………………………...70 Table 18. Farmers’ Separation of Crops………………………………………………………..71 vi Table 19. Ideas of “Success” and “Failure”……………………………………………………73 Table 20. Description of Predictor Variables Used in Logistic Regression Analysis….………..80 Table 21. Logistic Regression Analysis of Experimentation…………………………………….81 Table 22. Qualitative Coding Audit Trail………………………………………………………131 vii LIST OF FIGURES Figure 1. Experimentation to Adoption Process……………………………………………......15 Figure 2. Revised Theory of Reflection-in-Action……………………………………………….17 Figure 3. Map of Study Areas…………………………………………………………………...22 Figure 4. Total Frequency (%) Distribution of Experiments……………………………………50 Figure 5. Frequency (%) Distrubtion of Experiments (by Sample Group)……………………...51 Figure 6. Frequency (%) Distribution of Experiments (by Experimenter Classification)……….52 Figure 7. Frequency (%) Distribution of New Crops (by Experimenter Classification)………...52 Figure 8. Frequency (%) Distribution of New Crops (by Experimenter Classification)………...53 viii KEY TO ABBREVIATIONS AEDC Agricultural Extension Development Coordinator AEDO Agricultural Extension Development Officer Africa RISING Africa Research in Sustainable Intensification for the Next Generation ANOVA Analysis of Variance EPA Extension Planning Area (a geographical boundary around a specific region; similar to a county in the American system) GPS Global Positioning System MSU IRB Michigan State University Institutional Review Board TLU Tropical Livestock Unit USAID United States Agency for International Development ix 1.0 Introduction Over the past several decades smallholder agricultural production in Malawi has stagnated due, in part, to rapid population growth and its associated effects on the country’s arable land (Schulz et al., 2003; Snapp et al., 2010). Population pressure has forced farmers to utilize smaller pieces of land and reduce or eliminate fallow practices, which has led to low soil fertility and decreased crop production. These agronomic changes have especially affected maize, the country’s staple food crop, which is grown continuously and often without crop rotations (Bezner Kerr, 2005; Gilbert, 2004). Approximately 90% of Malawi’s rural population are smallholder, subsistence farmers (producing enough food to survive and only enough revenue for immediate needs) who rely on maize-mixed, rainfed fields that are smaller than two hectares (IFAD, 2010). As many families in Malawi are dependent on maize, reduced crop production is made manifest in high rates of malnutrition, where an estimated 47% of children under five years old suffer from growth stunting (UNICEF, 2006-2010). Although the situation for Malawian farm families seems dire, many farmers are actively experimenting with crop diversification, local weed/disease control methods, food storage techniques, and improved crop varieties as a means to increase food security and generate income at the household level. Until the mid 1970’s, local knowledge and farmer experimentation practices were largely unacknowledged by the international research and development communities (Scoones and Thompson, 1994b; Sumberg and Okali, 1997). Thanks to a major paradigm shift over the past several decades, however, a faction of international development researchers have recognized the importance of incorporating local knowledge and utilizing farmer participation in the design and implementation of agricultural intervention programs (Chambers et al., 1989). 1 Agricultural intervention programs are usually designed and implemented by development practitioners, researchers, extension agents, and other experts, and the primary goal for many interventions is to improve agricultural production in developing countries like Malawi. Intervention projects typically introduce new crops, varieties, or techniques in rural communities and encourage farmers to try improved farming methods. While many of these projects actively encourage independent farmer experimentation, farmers are often trying new things in conjunction with these projects instead of experimenting independently (although project participants frequently adapt or adjust project recommendations to suit their own objectives). Often, the technology or crop brought in by an intervention is not readily adopted by participants because it is contextually inappropriate. To create appropriate, desirable techniques that will pique the interest of farmers, interventions should incorporate local knowledge and farmers’ current practice into their projects. In order to do this, however, practitioners must first understand what kinds of experiments farmers are conducting on their own, independent of interventions. Past research has illustrated the multitude of resourceful and effective ways local people independently manage the natural resources on which their livelihoods depend (Chambers et al., 1989; Scoones and Thompson, 1994a; Warren et al., 1995). Local knowledge is built upon centuries of experimentation with and adaptation to changing agro-ecological conditions. For example, in Malawi (and throughout much of sub-Saharan Africa) the practice of intercropping grain legumes with maize—a technique which has been widely promoted by extensionists, nonprofit projects, and research institutions in recent years—has been implemented by local farmers for generations. Although more and more development specialists are beginning to recognize the importance of incorporating local knowledge in agricultural interventions, the literature 2 surrounding farmer experimentation in developing countries (that is, experimentation independent of outside interventions) is relatively scant. There is still much to be learned, therefore, about local agricultural practices and experimentation processes, the influences that shape experimentation, and the ways in which agricultural development projects can be informed by farmer-led experiments. Agricultural researchers, development practitioners, and most importantly Malawian smallholder farmers would greatly benefit from the integration of local practices, experiments, and priorities into agricultural development projects. In the words of one farmer who was encouraged by an intervention project to conduct agricultural experiments as a way to find solutions to her farming challenges, “They say wealth is in the soil”. Through the successful partnership of expert and local knowledge, smallholders will be better equipped to find solutions to their agricultural dilemmas and to maximize on the wealth all around them. 1.1 Purpose of Study For the purposes of this study, experimentation was defined as any instance where a farmer attempted to plant an unfamiliar crop or variety, or to implement an unfamiliar technique for the very first time. Note the difference in time horizons between a farmer’s adoption of a specific crop, variety, or agricultural technique and their experimentation with something new: where adoption is the repeated and unchanging use of a specific crop, variety, or agricultural technique over the long term, experimentation is the initial trial of a new or unfamiliar plant or technique—the introduction of a new element into a smallholder’s farming system—and is iterative and constantly evolving from season to season. This study was particularly focused on independent experimentation, where farmers tried new things without the guidance of an expert such as an agricultural intervention project or an extension officer. This type of independent experimentation has been termed “folk experimentation” by Bentley (2006). Due to the inherent 3 negative connotations commonly associated with the word folk, however, this study instead employs the terms local or smallholder experimentation to differentiate these processes from formal or professional experimentation. The aim of this study was to understand the decisionmaking processes of those farmers who were experimenting independently of intervention projects, draw distinctions between methods used across smallholder experiments, and explore the motivations (e.g. attitudes) and drivers (e.g. physical resources) of independent experimentation. 1.2 Research Questions This project focused on the experimentation processes and decision-making of Malawian smallholder farmers (especially, but not exclusively, women) who have tried unfamiliar crops, varieties, and/or techniques in the past two agricultural seasons. In this context, unfamiliar crops, varieties, and techniques are defined as those that a farmer has never previously tried. Experimentation with legume crops was chosen as the primary focus of this study as legumes are common in central Malawi (although legume production is less prominent than sole maize production in this region; see Snapp et al., 2002a), and because legume crops are of critical importance to a multitude of ongoing research and development projects that aim to improve nutrition, soil fertility, and agricultural production (Bezner Kerr and Chirwa, 2004; Gilbert, 2004; Schulz et al., 2003). Additionally, leguminous food crops such as groundnuts are typically secondary only to cereal crops (such as maize and millet) in both cultural importance and prevalence in much of Malawi (IFAD, 2010). The importance of legume crops to Malawian farmers will be further discussed in Section 2.0. The following research questions relating to farmer experimentation with legume crops were investigated: 4 RQ1: What are the characteristics of farmers who experiment with unfamiliar legume crops, varieties, and farming techniques related to legume production (e.g. gender, socioeconomic status, etc.)? RQ2: What motivates these farmers to experiment with unfamiliar legume crops, varieties, and farming techniques related to legume production? RQ3: How are these farmers managing their experimental crops, varieties, and techniques? 1.3 Basis of Study Past studies have found that different members of the household often have different agricultural and household roles. For example, land preparation tasks (e.g. building/shifting planting ridges) are often undertaken by men, whereas planting, managing (e.g. weeding, pest prevention, etc.), and harvesting tasks are primarily the responsibility of women (Bezner Kerr and Chirwa, 2004; Ferguson, 1992; 1994). Additionally, Central Malawi is predominantly comprised of members of the Chewa tribe. Chewa culture is matrilineal, meaning that land is passed down through the woman’s side of Chewa families. Regarding RQ1, the aforementioned studies and cultural norms led to the theory that female farmers would be more likely to experiment with unfamiliar legume crops and varieties than would male farmers, given that legumes are traditionally planted and managed by women, and a family’s land is culturally held by matriarchs and their male relatives. As the literature surrounding the determinants of experimentation is relatively scant, this study’s predictions about the wealth and assets of experimenting farmers were informed by adoption and farm management literature. Although adoption of technologies and farm management practices are not synonymous with experimentation, these ideas are interrelated (as 5 will be discussed in Section 2.0) and studies of adoption and management can shed light on the socioeconomic factors associated with on-farm decision making, including the decision to experiment with something new. Previous studies relating to adoption have found that farmers will be more likely to adopt an agricultural technology if they perceive that the benefits of a new technology will exceed its costs (Pannell, 1999; Asfaw et al., 2012). Similarly, a recent study of food security and innovation found that less food secure households made fewer management changes (i.e. experimental trials) on their farms than did households that were more food secure (Kristjanson et al., 2012). Based on these studies and the relationship between adoption, management, and experimentation, it was posited that farmers with fewer assets and/or less physical capital (i.e. the most vulnerable households in the population) would be less likely to conduct on-farm experiments because experimentation requires resources (e.g. extra land, new seed, etc.) and has unknown (and potentially undesirable) outcomes. There are countless factors that influence a farmer’s decision-making processes, and this is especially true for farmers who are trying new crops, varieties, or techniques on their farms. All farmers are situated in different physical and socioeconomic contexts (Pannell, 1999), and thus all farmers have different motivations for experimenting and employ different strategies to manage their experiments. To best understand the drivers of on-farm experimentation relating to legume production (RQ2), in-depth interviews were conducted and questions focused on the decision-making processes of farmers who experimented with unfamiliar crops, varieties, and techniques in the past two growing seasons. This research question was explored by speaking with farmers, and not to them (Box, 1989), about their life circumstances, farming systems, and agricultural experiments, and the farmer insights gained through these in-depth interviews subsequently added context and clarity to the survey results. 6 Regarding the management strategies employed by farmers from the beginning of an experiment to the following season (RQ3), previous studies have found that during the first attempt, a farmer will typically plant a new or unfamiliar crop in a small quantity, or on a small plot of land (Sumberg and Okali, 1997). In the second season, the farmer will scale out the same experiment only if she is satisfied with its performance during the previous season (Rhoades and Bebbington, 1995). If, however, the farmer is dissatisfied with the outcome of the experiment in the first season, she will change her management strategy during subsequent seasons (e.g. attempt a different spacing arrangement within the row, use a different field or area, implement at a different time in the season, etc.) (Schön, 1983). Regarding farmers’ use of a comparison plot or “control”, past studies have found mixed results, where some farmers consciously use a control, some farmers use a “historical control” (comparing experiments to their prior understanding of the farm system), and yet other farmers do not actively use any kind of control (Sumberg and Okali, 1997). The exploration of this research question attempted to validate the aforementioned literature through exploratory interviews with innovative farmers, where questions focused on the management of experimental crops, varieties, and/or techniques. This thesis employed a rural household survey and in-depth interviews to characterize smallholder farms, examine experimentation processes, and explore decision-making at the household and individual levels. The following section will ground this study with existing literature, and subsequently the theoretical framework, methods, and results will be presented. Finally, the implications for future research, extension, and development projects will be discussed. 7 2.0 Review of Related Literature Case studies from the existing experimentation literature provide insight into the innovative capacities of local farmers and suggest that farmers are essentially constant experimenters because they are continually adapting to the dynamic conditions (e.g. economic, climatic, etc.) on which their lives and livelihoods depend (Chambers et al., 1989; Scoones and Thompson, 1994b; Warren et al., 1995). Local people have a long history of agricultural experimentation, from potato storage techniques in the Peruvian Andes (Rhoades, 1989) to cocoyam intercropping in northern Ghana (Millar, 1994), to legume varietal selection in Malawi (Ferguson, 1992; 1994). Farmer-led, or “folk” experimentation (Bentley, 2006), however, differs from formal research experiments in both structure and purpose. According to Bentley (2006), “folk experiments do not have to be scientific…[because farmers] may be knowledgeable and creative but not strictly scientific” (p. 459). When farmers experiment, they “have very specific goals in mind and the results of [their] experiments must be practical” (Rhoades, 1989, p. 9). In other words, farmers must continually experiment and adapt in order to sustain themselves and their families. Past literature has compared local knowledge of farm systems to a musical performance (Richards, 1989); while researchers attempt to conduct experiments (“play an opus”) under precise and controlled conditions, local farmers must experiment in dynamic and unstable environments. In the same way that musicians must deal with bad acoustics, stage fright, and many uncontrollable factors, farmers must adapt to changing climates, variable markets, limited resources, and a host of other challenges for which researchers in controlled settings may not need to compensate. Thus, while local experiments often disregard the precision of the scientific 8 method, (Bentley, 2006), they are necessary, carefully planned, and can result in life-changing innovations. Smallholder experiments differ from formal scientific experiments not only in structure and method, but also in measures of external validity (Misiko and Tittonell, 2011). While replicability and generalizability are necessary measures that validate an experiment in formal science, these measures are often neither necessary nor possible in local experiments. Farmers may not have the luxury of replication due to ever-changing and restrictive factors such as weather and resource availability. Additionally, local experiments are not constructed to be applied or generalized across a wide range of contexts, but rather they are crafted to fit the conditions of one specific farm system. Where formal experiments can be controlled and repeated, local experiments are the “real practice” (Misiko and Tittonell, 2011, p. 1137) and “can only occur ‘in time’, where they are embedded in particular agroecological and sociocultural contexts” (Scoones and Thompson, 1994b, p. 20). Formal scientists and smallholder farmers also use different criteria to judge the “success” of an experiment. For scientists and researchers in the formal sector, a successful experiment may be that in which one hypothesis is rejected in favor of another hypothesis (Schön, 1983), or where the relationship between two variables yields a high level of statistical significance. Smallholders, however, may deem an experiment successful if it can help them adapt to their circumstances and make it through to the next season, if it can survive or thrive over the long-term, if people like the outcome, or if it “leads to the discovery of something there” (Balée, 1994; Misiko and Tittonell, 2011; Rhoades and Bebbington, 1995; Scoones and Thompson, 1994b; Schön, 1983, p. 145). These epistemological differences between professionals and smallholders should be taken into account to best understand the wide range of 9 criteria that distinguish a successful experiment from a failure. Success, it seems, lies in the eyes of the innovator. Smallholders conduct many different types of experiments and are driven by a variety of goals. Bentley (2006) claims that local experiments are “motivated by changes in the environment and the economy, and seek to resolve labor and capital constraints” (p. 451). Other studies have classified local experiments into different types, where some farmers experiment out of curiosity, in what has been termed an exploratory experiment (Schön, 1983 as cited in Stolzenback, 1994) or a curiosity experiment (Millar, 1994; Rhoades and Bebbington, 1995). Other farmers innovate to produce a positive change in their farming system, often in response to conditions that are out of their control (e.g. climate change and variability). This type of experiment has been called a move-testing experiment (Schön, 1983) or a problem-solving experiment (Millar, 1994; Rhoades and Bebbington, 1995). Additionally, Millar (1994) argues that the most frequent kinds of experiments conducted by farmers are “adaptive”, whereby a farmer starts with a new technology or technique (e.g. learned from a relative, an old tradition, an extension agent, etc.) and reinvents it to suit his or her specific context. Adaptive experiments can also occur when a farmer takes a familiar technique and applies it to a new environment, in the case of migration, for example (Rhoades and Bebbington, 1995). Bentley (2006) states that the best local experiments are adaptive, where farmers do not simply replicate an idea or technique, “but combine new ideas creatively with local knowledge” (p. 452). From the outside, it is relatively clear to distinguish between multiple types of experiments in local agricultural systems, but many smallholder farmers do not readily label their adaptations and on-farm practices as experiments at all. The difficulty, then, lies in how to 10 measure and categorize what agricultural researchers see as “local, on-farm experiments” when those experiments seem like everyday life to the innovators themselves. To circumvent this epistemological difference, Misiko and Tittonell (2011) used the language of farmer “tryouts” rather than “experiments” in their 2003-2007 study of research partnerships between local farmers and agricultural scientists. For smallholder farmers who do not have access to extra land, seed, time, and other precious resources, there is no practice round or preparation period before beginning an experiment—“tryouts are usually the real practice… [and] when a technology seems practical to smallholders, they try it out under their household’s social and farm-level ecological conditions” (p. 1137). For this study, Misiko and Tittonell’s language of “trying out an unfamiliar crop, variety, or technique for the first time ever” was used to convey questions about legume experimentation to farmers who were surveyed and interviewed. This study focused on experimentation with legume crops and varieties and the associated management techniques of leguminous crops for many reasons. Past research has shown that the traditional Malawian practice of intercropping grain legumes in maize systems can increase soil nitrogen (N) levels and enhance soil fertility, leading to greater maize yields among other benefits (Gilbert, 2004; Snapp et al., 2010; Snapp et al., 2002b). In fact, farmers are more likely to use legumes as intercrops if the legume will provide multiple benefits beyond enhancing soil fertility, alone (Snapp et al., 2002b). Common legume crops used by Malawian farmers as intercrops include pigeon pea, cowpea, common bean, soybean (soya), and groundnut. These crops can be thought of as secondary crops, as the primary crop for Malawian subsistence farmers is almost always a cereal such as maize or sorghum (IFAD, 2010). Depending on the legume crop(s) used, maize-legume intercrop systems have the potential to 11 yield multiple benefits such as: N fixation in soil; reduced pest and disease pressures; improved soil organic matter and water infiltration (through the incorporation of crop residues); income generation (in the case of groundnut, which can be sold as a cash crop); low seed costs; use as fodder for livestock; minimal labor requirements; late maturity (available when other food sources are not); vast increases in calories and protein in diet (as compared to a maize-dependent diet); potential for use in porridge for young children; and secondary use as medicine for earaches and diarrhea in children (Bezner Kerr and Chirwa, 2004; Gilbert, 2004; Snapp and Silim, 2002). One of the most highly valued benefits of leguminous crops is their high protein and caloric content—a trait that is especially prized by female farmers who are responsible for feeding children. A farmer will often choose what to use as an intercrop based on a given crop’s associated benefits. Just as all farmers have different motivations for experimenting (or not), all farmers have different priorities and goals for their farms (Pannell, 1999). These differences in priorities are most striking, however, between males and females. Past research has found that while both men and women farmers prefer high yielding crops (Schulz et al. 2003), women tend to value crops as a food source (protein-rich legumes, in particular), while men tend to value crops for their potential to generate income (Bezner Kerr, 2008; Bezner Kerr et al., 2007; Snapp and Silim, 2002). Women’s preference for food crops over cash crops may be the result of food scarcity worries. While many smallholder farmers are resource-poor, female-headed households, in particular, have inadequate access to credit, labor, and agricultural inputs (e.g. fertilizer, seed, etc.) compared to male-headed households (Bezner Kerr, 2005; Snapp et al., 2002a). This lack of resources decreases the comparative food security of female-headed households. Although female-headed households may be less food secure than male-headed households, women are 12 key participants in the Malawian agricultural system. Ethnographic studies in sub-Saharan Africa have revealed that women are intimately involved with the agricultural tasks of weeding, land use, harvesting, and seed selection (Bezner Kerr, 2005; Sharland, 1995). Women farmers’ familiarity with these important steps in the agricultural process suggests that “they therefore have more intimate and personal knowledge of the crops themselves and they are the ones who are involved in the key stages of production” (Sharland, 1995, p. 387). Women are not only responsible for seasonal seed selection, planting, and crop management (Bezner Kerr, 2005; Ferguson and Mkandawire, 1993), but they are also the repositories of detailed crop knowledge—from seed storage to plant growth to post-harvest usage. Qualitative studies have found that when women are the primary decision-makers regarding seasonal seed selection, they will often plant a wide variety of different crops to meet different farm and household needs (Ferguson and Mkandawire, 1993; Ferguson 1992; 1994). By intercropping many different food plants including leguminous crops, women are able to diversify their families’ diets (e.g. adding protein through legume consumption), satisfy secondary household needs (e.g. fuel for cooking, livestock fodder, medicinal needs, etc.), ensure year-round food availability (through intentional combinations of early and late maturation crops), and enhance the resilience of their farms through biodiversity (bolstering pest and disease resistance, drought and flood tolerance, etc.). Where women have decision-making power over what to plant they will often grow leguminous crops to bolster their families’ protein intake (Ferguson and Mkandawire, 1993; Ferguson 1992), but where men have decision-making power over what to plant they will often grow cash crops to generate income (Bezner Kerr, 2005; Bezner Kerr and Chirwa, 2004). 13 Sometimes the decision-making power over what crops are planted rests with the head of household (Bezner Kerr and Chirwa, 2004), and sometimes the division of labor is such that seasonal seed selection decisions are made by the woman of the house (even if she is not the household head) (Ferguson and Mkandawire, 1993). Therefore, in both the surveys and the interviews, explicit questions were asked regarding who was responsible for planting experimental legume crops and how that decision was made, because it cannot be assumed that the head was the only “experimenter” in any given household. 2.1 Theoretical Framework Under the theory of adaptive rationality, Nitsch (1990) contends that farmers manage their complex farm systems through “a continuous interaction among visions, experiences, and experimentation” (p. 69). Similarly, Malawian farmers have a vision for how they would like their farm to develop, a lifetime of agricultural experience, and an array of new experiments, adaptations, and problem solving efforts that they employ to merge their vision with their experienced reality. Schön’s theory of Reflection-in-Action (1983) asserts that practitioners who are confronted with uncertain, unstable, complex, or unique situations (in this case, local farmers who manage dynamic, resource-constrained farm systems) will reflect on the complexities of the situation, take inventory of their own knowledge and prior experience, and then conduct an experiment so as to better understand a phenomena and/or to create a change in a situation. During these experiments, the innovator (farmer) is in constant dialogue with her environment; a local farmer assesses how the farm “back-talks” during an experiment (p. 164) and engages in a sort of conversation with the soil and crops. After an experiment, a farmer may proceed in one of several ways, depending on her perception of the experiment’s success. In the case of a dissatisfactory experiment, the farmer 14 may critique her method or theory, make adjustments, and attempt the experiment again. Alternatively, she may terminate the experiment. In the case of a satisfactory or successful experiment (an “innovation”), a farmer may choose to scale-out the innovation and/or repeat it in subsequent seasons. The repeated use of an innovation (which resulted from an experiment) corresponds with the definition of “individual (farm-level) adoption” made by Feder et al. (1985), where a new technology is used “in long-run equilibrium [and] when the farmer has full information about the new technology and its potential” (p. 256). Furthermore, Schultz (1975) states that experimenting with new technologies will lead a farmer towards equilibrium, where adoption of an innovation is possible. Thus experimentation is the first step on the adoption spectrum, where experimentation leads to the development of an innovation, and the long-term use of an innovation with repeated successful outcomes will lead to the adoption of that innovation. Such innovations may be adapted or adjusted by farmers in future experiments, making the process truly iterative (Nitsch, 1990). This relationship is depicted in Figure 1. Figure 1. Experimentation to Adoption Process Experimentation with an new idea Successful idea becomes an innovation New idea Adaptation and adjustment of innovation Long-term adoption of an innovation 15 While experimentation and adoption are closely related, they are not interchangeable terms. These clarifications are provided not only to distinguish between experimentation with an unfamiliar technology and adoption of an innovation, but to assert that the primary focus of this study is on-farm experimentation and the decision-making processes involved therein. While Schön’s Reflection-in-Action theory details many drivers of experimentation (e.g. practitioners’ past knowledge and experiences), it neglects to account for the influence of gender and other socioeconomic factors on the decision-making processes involved with experimentation. This omission of gender from Reflection-in-Action necessitates a revision if the theory is to be used in explaining agricultural experimentation (see Figure 2), as culturally prescribed productive roles place women at the forefront of agricultural processes such as legume seed selection and crop production in Malawi (Bezner Kerr and Chirwa, 2004; Ferguson, 1992; 1994). Likewise, other characteristics that are relevant to agricultural experimentation, in particular, should be included in Schön’s theory (e.g. socioeconomic status and access to information). Given the characteristics that were explored through RQ1, close attention was therefore given to the decision-making processes and experimental capacities of female farmers in Malawi, and the influence of resource availability on experimentation processes. 16 Figure 2. Revised Theory of Reflection-in-Action The influence of gender in on-farm experimentation and decision making must be examined and accounted for both in theory and in practice. By understanding the drivers of experimentation and decision-making from the perspective of Malawian smallholders (across a range of socioeconomic characteristics), development practitioners and extensionists can work effectively with farmers to support and enhance their existing capacities and farm management techniques. 17 3.0 Methods 3.1 Overall Approach This study used a mixed methods framework that utilized field observations, household surveys, and in-depth interviews. The quantitative and qualitative strands of this study had several points of interface: at the design level, during data collection, and during the interpretation of results. An adaptation of the explanatory sequential design described by Creswell and Clark (2011) was used, where quantitative data were collected during the first phase of field work using a survey instrument, and initial analysis of these data informed both the case selection and the development of the interview questionnaire for the second, qualitative phase of field work. The qualitative data were then collected through in-depth interview sessions which were conducted using an interpreter and audio recorder. Immediately following the interview sessions, the recorded conversations were translated into English and transcribed. After completing both phases of the field work, quantitative survey data were analyzed comprehensively using the statistical packages SPSS and R. Results from these statistical tests informed the analysis of the qualitative interview transcripts, which were thematically coded using NVivo 10 qualitative software. The quantitative data helped to identify a sample and relevant questions for the qualitative interviews and also gave insight into possible emergent themes that were explored during qualitative analysis. Likewise, the qualitative data helped to contextualize and explain the statistical findings gleaned from the survey data by bringing farmers’ voices and personal experiences to light. Rather than using quantitative or qualitative methods alone, the integration of methods used in this study yielded a more thorough, richer understanding of the drivers 18 behind on-farm legume experimentation, the experimental methods used by smallholders, and the characteristics of innovative farmers who conduct on-farm experiments. 3.2 Study Areas Field work was conducted in two phases over a ten week period from late-May to lateJuly, 2013. Data were collected from 22 village clusters within five Extension Planning Areas (EPAs), namely Linthipe, Golomoti, Mtakataka, Kandeu, and Nsipe, across two districts of Central Malawi, namely Dedza and Ntcheu. Before entering each study area, the field research team followed politically and culturally appropriate protocol and met with the Agricultural Extension Development Coordinator (AEDC) and at least one Agricultural Extension Development Officer (AEDO) for each EPA, and explained the nature of the research and the project timeline, and expressed gratitude to them for their assistance with the project. Subsequently, the team met with the village headman (or headwoman) and at least one lead farmer in each village cluster to explain the project. As a result, the AEDCs, AEDOs, and village authorities generously provided comprehensive lists of local area households, which were used for random sampling purposes. Study sites were determined by agro-ecological zone, market access, and the presence or absence of interventions by the agricultural research project Africa RISING (Research in Sustainable Intensification for the Next Generation). Africa RISING is a USAID funded agricultural research initiative that conducts research for development projects in six countries across sub-Saharan Africa, including Malawi. Africa RISING promotes several “best-bet” legume crops (e.g. peanuts (groundnuts), soy bean (soya), cowpea, and pigeonpea), and several novel growing techniques (intercropping two legumes in the first season (doubled-up legumes) and following with a maize crop in the second season). These crops and techniques aim to provide multiple benefits to farmers such as increasing soil 19 fertility, enhancing maize production, generating income, and improving household nutrition. The project works through the government of Malawi’s agricultural extension staff and encourages farmer experimentation through a participatory approach called the mother-baby trial design (Snapp et al., 2002a). Farmers work on researcher-managed, multi-treatment demonstration plots (“mother trials”) and replicate their favorite treatments from the mother trial on their own plots (“baby trials”). During the field study (2012-2013 growing season), Africa RISING was concentrated in eight sites across the Dedza and Ntcheu districts of Central Malawi, with two mother trials in each EPA: Linthipe (Mkuwazi and Mbidzi); Golomoti (Msamala and Kalumo); Kandeu (Dauka and Katsese); and Nsipe (Amosi and Nzililongwe). During this period, the project worked with 450 smallholder farmers in the aforementioned villages. Below, Table 1 describes agronomic information about each EPA that hosted a mother trial. Table 1. EPA Agronomic Information Characteristic Elevation (meters above sea level) Linthipe 1238 Golomoti 555 Kandeu 904 Nsipe 868 1005.5 890.6 799.7 810.8 Dominant Soil 1 Loamy clay Loamy sand Sandy clay, loam Dominant Soil 2 Clay Clay loam Sandy clay, loam Loam Distance from small market (km) 5 1 2 9 Distance from large market (town) (km) 40 40 35 20 Maize, tobacco, groundnut Maize, cotton, groundnut Maize, tobacco Maize, tobacco, groundnut 4623 (Mposa) 2232 (Golomoti Centre) 2362 (Kampanje) 1758 (Mpamazi) Annual rainfall (mm) Major crops Number of farming families in the project sections 20 Sandy clay Participation in Africa RISING was voluntary. After enrolling, farmers participated in the preparation and management of the mother trial plot, and seed for the baby trial replications was provided to participating farmers as both an incentive and a means to participate in the project. This study worked in partnership with Africa RISING, and the study’s sample population was based on farmer’s participation (or lack thereof) in Africa RISING. The study’s sample population was therefore divided into three groups:  Intervention households: at least one household member was actively participating in Africa RISING  Local control households: no household members were participating in Africa RISING; households were located within the same village as an Africa RISING mother site  Distant control households: household members had no prior exposure to Africa RISING; households were located in separate villages from Africa RISING sites, but had similar agroecological conditions and market access as the nearest Africa RISING site. Intervention, local, and distant control households were chosen from within the same EPAs in Linthipe, Kandeu, and Nsipe. In Golomoti EPA, however, intervention and local control households were chosen within the same EPA, but distant control households were chosen from an adjacent EPA with a similar agroecology (Mtakataka). The map in Figure 3 depicts the Africa RISING sites (where intervention and local control households were located) along with the distant control sites. 21 Figure 3. Map of Study Areas 3.3 Phase One: Household Survey 3.3.1 Target Population The target sample size for the first phase of field work was 320 smallholder farmers (160 intervention households, 80 local control-group households, and 80 distant control-group households), distributed evenly amongst at least three villages per EPA: Linthipe; Golomoti/Mtakataka; Kandeu; and Nsipe. Household surveys were administered to a sample of 324 farmer participants (97 males and 227 females) with 163 in Dedza district (approximately 2.61% of the population of Dedza) and 161 in Ntcheu district (approximately 3.41% of the population of Ntcheu). Population information is according to the 2008 Malawi Population and 22 Housing Census: Spatial Distribution and Urbanization Report (National Statistics Office of Malawi, 2008). Of the total sample, 162 participants were members of intervention households, 81 were from local control households, and 81 were from distant control households, thus the target sample size was slightly exceeded. Below, Table 2 depicts the sample population according to EPA and sample group. The survey instrument, sampling framework, and respondent characteristics will be discussed in the following sections. Table 2. Survey Sample District: EPA Intervention HHs Local Control HHs Distant Control HHs Total HHs Dedza: Linthipe 42 20 21 83 Dedza: Golomoti 40 20 0 60 Dedza: Mtakataka 0 0 20 20 Ntcheu: Kandeu 40 21 20 81 Ntcheu: Nsipe 40 20 20 80 3.3.2 Data Collection: Survey Instrument and Field Research Team The household survey was conducted over a three-week period, from May 23 to June 13, 2013. An 18-page survey instrument was used to collect household and farm level data, and question topics included: socioeconomic and demographic characteristics; respondents’ memories of the climatic history of their EPA; cropping systems and land use; food security and agricultural production; on-farm experimentation; and participation in Africa RISING and other agricultural intervention programs (Appendix 1). The initial survey instrument was designed with contributions from Michigan State University faculty and Malawian agricultural experts. Survey enumerators: (1) were fluent in both English and Chichewa, (2) had completed 23 Bachelor’s degrees, (3) had prior experience administrating survey questionnaires to local farmers, (4) were not members of the traditional authority, and (4) underwent training and certification according to MSU Institutional Review Board requirements prior to beginning field work. The survey was refined, translated, and pretested in partnership from the enumerators prior to beginning fieldwork. Every enumerator was shadowed at least two times throughout the survey process, by one of two research supervisors. This was to accomplish several goals: to ensure that all enumerators were asking the survey questions in a consistent manner; to act as a reference in cases where a respondent gave an unorthodox answer to a question; to assess the enumerators’ performance (which later helped to inform who was hired on as an interpreter for the in-depth interviews in the second phase of field work); and to take detailed field notes. Survey sessions were conducted in Chichewa, and surveys were only conducted with household members over the age of 18 who verbally agreed to participate after listening to an MSU IRB-approved informed consent script. With the exception of one village (where survey interviews were conducted in the center of the village), all survey interviews were conducted at the home of the respondent, and the location of the home was recorded using a GPS device. An average interview lasted approximately 45 minutes for local or distant control households. Intervention household questionnaires contained one subset of questions relating to respondents’ participation in Africa RISING that was omitted from control household questionnaires and therefore the average interview time for an intervention household was 1 hour 15 minutes. At the end of each survey session and upon the agreement of the respondent, enumerators took digital photos of respondents and their families. These photos were then printed and given to survey 24 participants at the end of the field work period as a gift for participating in the survey process. Respondents were informed of this gift only after they had completed the survey questionnaire. 3.3.3 Data Collection: Sampling Framework Intervention households were selected to participate in the survey process using a stratified random sampling method (Vaske, 2008). First, a list of all participating farmers in Africa RISING, divided by EPA (Linthipe, Golomoti, Kandeu, and Nsipe) was obtained. Before randomly selecting a sample from this list, however, participant names were crosschecked with a list of farmers who had recently participated in a separate agricultural household survey led by Wageningen University (Netherlands) in the previous month. Any farmers who had participated in the Wageningen study were eliminated from the sampling pool in order to avoid respondent fatigue or over-surveying of any particular household. Names were then randomly selected from within each EPA using a random number generator. Forty households were randomly selected from within each EPA as a primary sample, with 20-25 additional households selected as alternates in the event that primary households were unavailable. In each EPA, primary and alternate households were selected with an approximately equal distribution between the two Africa RISING mother trial site village clusters. In the field, key informants from each village cluster helped the team to determine the approximate location of the target households, whereby each enumerator completed 3-4 interviews over the course of each day. The team used two days per EPA to collect survey data for 162 intervention households. Within each intervention household, enumerators attempted to speak with the farmer who had officially participated in Africa RISING, as indicated by the aforementioned participant list. If the intervention farmer was unavailable, enumerators interviewed another adult household 25 member who had thorough knowledge of the household’s agricultural production and of the intervention farmer’s management of his/her Africa RISING baby plot(s). Local control households were also sampled using a stratified random method. Within each study site in each EPA, comprehensive lists of all households were obtained from the AEDC, AEDO, and/or the village head person. Before randomly sampling from this list, participant names were crosschecked with the Wageningen University sample and also with the Africa RISING participant list. Households who had been previously surveyed or who were Africa RISING participants were eliminated from the local control sampling pool. After these households were removed from the pool, names were randomly selected from within each EPA using a random number generator. Twenty households were randomly selected from within each EPA as a primary sample, with 20-25 additional households selected as alternates in the event that primary households were unavailable. In each EPA, primary and alternate households were selected with an approximately equal distribution between the two Africa RISING mother trial site village clusters. In the field, key informants from each village cluster helped our team to determine the approximate location of the target households, whereby each enumerator completed 3-4 interviews over the course of each day. The team used one day per EPA to collect survey data for 81 local control households. Within each local control household, enumerators attempted to speak with the head of household, as indicated by the household lists provided by the AEDCs, AEDOs, and traditional authorities. If the household head was unavailable, enumerators were instructed to interview another adult household member (preferably the household head’s spouse, if available) who had thorough knowledge of the household’s agricultural production. This sampling method yielded 26 survey interviews across genders and household types, including responses from male heads of household, female heads of household, and female spouses within male-headed households. Early in the fieldwork period, household lists for Golomoti EPA were unavailable, which resulted in the creation of a last-minute random sampling framework that was neither understood nor consistently followed by the enumerators. This miscommunication resulted in enumerators using arbitrary, convenience selection methods (rather than systematic random methods) to sample local control households in Golomoti EPA. In order to rectify these sampling errors, field work was extended and local control households in Golomoti were re-sampled using household lists and the stratified random sampling methods previously described. One household that had been previously sampled (using the convenience method) was randomly selected from the list to be sampled again. This household was not revisited, but the all of the survey information they had provided during our first visit was kept. All other questionnaires from the first sample of Golomoti local controls were discarded and the information therein was excluded from the data analysis process. Instead, information from the second (random) sample was used to represent local control households in Golomoti EPA. Twenty households were sampled as a part of this group. Distant control households were sampled using a cluster sampling method (Vaske, 2008). Two distant control villages per EPA were selected to be a part of the survey. Within the distant control villages, a key informant (e.g. lead farmer) led three enumerators to the village center. From there, enumerators “drew” a Y-shaped axis through the village, whereby one enumerator would walk along each axis and sample one household every 50-100 meters. Distances between households were pre-determined according to the village size and layout (less distance between households in smaller, more condensed villages and more distance between households in larger, 27 more dispersed villages) so that all enumerators in a given village measured the same distance between sampled households. Using the Y-axis method, each enumerator surveyed between 3-4 households per distant control village. The team used one day per village to collect survey data for 81 distant control households. Within each distant control household, enumerators spoke with an adult household member who was present at the time of the interview and who had thorough knowledge of the household’s agricultural production (usually the household head or their spouse). In the event that both the household head and their spouse were present at the time of the interview, enumerators spoke with whomever seemed the most open to being interviewed. This sampling method yielded survey interviews across genders and household types, including responses from male heads of household, female heads of household, and female spouses within male-headed households. 3.3.4 Survey Data: Inspection, Entry, and Analysis During the household survey data collection period, the team met at the end of each field day to assign household identification numbers to the completed questionnaire instruments, check the questionnaires for errors, and code any unusual responses. Following these meetings, the completed questionnaires were inspected for omissions, outliers, and mismatching codes. Errors were corrected by enumerators on the following day. In cases where an enumerator made an irreversible error (e.g. neglected to ask the respondent a question), the enumerator returned to the household and gathered/corrected the relevant information. These daily checks helped to minimize data collection errors and also helped to ensure that all enumerators were interpreting and coding questions consistently. Finally, at the completion of the data collection process, enumerator errors in the data were tested for using an ANOVA test in SPSS. In this test, 28 enumerator identification codes were used as the grouping variables, and the means of the regression analysis variables were tested across all six enumerators. No significant differences between the means of any variables were found across the enumerators, indicating a low likelihood that enumerator bias has corrupted the data. At several points throughout data collection, enumerators worked in pairs and entered survey data into a Microsoft Excel spreadsheet. Data entry was completed during the field work period, which helped to reduce data entry errors as the information was still fresh in the enumerators’ memories. Entered data were checked for entry errors. After the fieldwork periods, a comprehensive codebook was created and the data were cleaned. Unusual codes or errors were checked against the questionnaire hard copies and corrected accordingly. Initial data analysis was performed using Microsoft Excel, specifically regarding demographic characteristics of the sample population, types of cropping patterns, frequency of experimentation, socioeconomic characteristics of innovative farmers, and farmers’ experiences with the Africa RISING project. The preliminary conclusions that were drawn from this initial analysis were presented at the Regional Africa RISING conference in Salima, Malawi, on July 29, 2013. Additionally, initial data analysis provided the basis for the purposive sampling framework which was used in the second, qualitative phase of fieldwork. This framework will be discussed in greater detail in the Section 3.4. Comprehensive data analysis was performed using the statistical software packages R (agronomic information) and SPSS (demographic information, experimentation data). Agronomic analyses included land use, soil management practices, fertility measures, and frequency of crops grown. The statistical software package SPSS was used to calculate descriptive statistics for household size, dependency ratios, productive and reproductive asset 29 ownership, farm characteristics (e.g. landholding size, number of fields, etc.), education levels, food security indicators, experimentation types and frequencies, and an experimentation classification structure. The household survey instrument contained the following questions to gauge farmers’ experimentation (Appendix 1): E11. What new crops did you grow this season (2012-2013) for the first time ever? E16. What new crop varieties did you grow this season (2012-2013) for the first time ever? E21. What new techniques or technologies did you try this season (2012-2013) for the first time ever? Follow-up questions prompted respondents to give details about the information source for each new crop, variety, and technique they reported (e.g. How did you learn about this new crop/variety/technique/technology?). Response categories were derived from three overarching information sources (Sumberg and Okali, 1997): 1. Institutions that actively promoted new things (e.g. AEDOs, Africa RISING or other non-profit project); 2. Peers/others who suggested new things or where farmers observed new things (e.g. family member, lead or other farmer, private distributors, social groups, radio); and 3. Independent ideas from the farmers’ own imagination. Based on farmers’ responses to these experimentation survey questions, a classification structure was created as a basis for further statistical analyses. Note that this classification structure is based on experimentation examples and their information sources, alone, and does not reflect 30 farmers’ socioeconomic standing or farm-level characteristcs. The experimentaiton classifications are:  Non-experimenters (n = 96): Farmers who did not report trying anything new in the 2012-2013 season;  Project participants (n = 145): Farmers who only reported trying something that had been actively promoted to them (e.g. by extension agents, intervention projects, etc.);  Followers (n = 64): Farmers who reported trying something that they had observed or had heard mention of (e.g. from peers, radio, family members);  Independent experimenters (n = 19): Farmers who reported trying something that was their own idea. This classification structure is based on a hierarchy which represents the magnitude of experimentation, where independent ideas represent the highest form of experimentation. Thus, independent ideas trumped suggested/observed ideas, which trumped simply following project recommendations. For example, a farmer who tried even one experiment that came from their own imagination was categorized as an Independent experimenter, regardless of the other types of experiments they tried. Likewise, a farmer who tried something they had heard on the radio and something that was part of an intervention project (but did not try anything from their own imagination) was categorized as a Follower. Below, Table 3 illustrates the distribution of Nonexperimenters, Project participants, Followers, and Independent experimenters according to their associated sample groups (e.g. intervention, local control, or distant control). 31 Table 3. Experimentation Classifications by Sample Group Group* Experimentation Classification Non-experimenters Intervention Local Control (n = 162) (n = 81) Distant Control (n = 81) Total Sample (n = 324) 12 (7.4%) 39 (48.1%) 45 (55.6%) 96 (29.6%) Project participants 124 (76.5%) 16 (19.8%) 5 (6.2%) 145 (44.8%) Followers 19 (11.7%) 23 (28.4%) 22 (27.2%) 64 (19.8%) Independent experimenters 7 (4.3%) 3 (3.7%) 9 (11.1%) 19 (5.9%) Total 162 (99.9%) 81 (100%) 81 (100.1%) 324 (100.1%) *Percentages calculated within Sample Groups Due to rounding, column totals may not equal 100% This structure was used as a basis for ANOVA tests which compared socioeconomic and farm-level characteristics of innovative farmers at the household level (Section 4.3). Additionally, chi-square tests were used to analyze intra-household, gendered decision-making and labor issues related to experimentation (Section 4.4). Finally, a binary logistic regression was used to estimate the probability that a farmer with certain socioeconomic and farm-level characteristics would experiment independently with an unfamiliar crop or technique (Section 4.6). The results of these statisical analyses will be discussed in Section 4.0. 3.4 Phase Two: In-depth Interviews 3.4.1 Target Population and Sampling Framework The target sample size for the second phase of field work was 20 farmers. The intended sample included both male and female farmers from all three sample groups, across every EPA, and from within both male- and female-headed households. This qualitative sample was drawn from the pool of farmers who had been previously surveyed during the quantitative phase of fieldwork, during which time data were gathered on experimentation both through the survey 32 questionnaire and through detailed field notes. Farmers were sampled for in-depth interviews using a purposive framework (Vaske, 2008), and only those farmers who reported experimentation with new crops, varieties, or technologies, either during the 2012-2013 season or in previous seasons were included in the qualitative sample. Primacy was given to interview those respondents who met the following criteria:  Those who tried at least one experiment independently (without being prompted by an agricultural intervention project or extension officer);  RISING participants who had tried a “baby trial” treatment to which they had never before been exposed;  those who experimented with unfamiliar legume crops and/or technologies (although cash crop, dimba, and other rainfed crop experiments were not excluded);  those who were experimenting with new crops and new technologies simultaneously;  those who were growing multiple experimental crops/varieties simultaneously;  those from whom detailed field notes had already been taken (so as to build on previously established rapport with these respondents and also to build on what was already known about their experimentation based on field notes). In order to incorporate a range of gender perspectives, both men and women were interviewed. Due to women’s close association with legume crops, however, women were given primacy in the interview sample so as to shed more light on their decision-making processes as they related to legume experimentation. Women in the sample were either female-heads of household or spouses within male-headed households. Additionally, respondents were purposefully drawn from five Extension Planning Areas (including Mtakataka, the distant control 33 site corresponding to Golomoti), from all three quantitative sample groups (intervention, local, and distant control households), and across a range of farm sizes (from 0.2 ha to >1.82 ha). Finally, in order to better understand farmers’ personal ideas of the success or failure of an experiment, the qualitative sample included farmers who had previously reported that their experiments had been “successful”, along with those who had reported that their experiments had “failed”. In total, the sampling pool contained 28 farmers (5 primary and 2 alternate farmers from each EPA, with Golomoti and Mtakataka combined for sampling purposes). Within each household, either the head of household or their spouse was interviewed, depending on who held the most responsibility for planting unfamiliar crops and managing onfarm experiments. Distribution of labor on experimental plots was determined by several questions on the initial quantitative survey, where respondents were asked “Who planted the [unfamiliar] seed?” and “Who managed the experimental [seed or technology]?”. To these questions, respondents could answer: household head; household head spouse; both household head and spouse together; or both. Respondents were chosen to participate in the in-depth interviews based on the information they gave on the household survey. After constructing the sampling framework, the names of primary and alternate farmers were taken to the field where the interview team met with the extension officer for the EPA, who helped to locate target households and introduced the field team to the farmers. Farmers in the sample were alerted several hours in advance that they could be selected for an interview. If farmers in the primary sample were unavailable, names were drawn from the alternate list. In total, in-depth interviews were held with a sample of 18 farmers (15 females and three males), with 10 in Dedza District and eight in Ntcheu District, meaning that the actual sample was slightly smaller than the target sample. Of the sample, 14 farmers were Africa RISING 34 participants, two were local control farmers, and two were distant control farmers. It is important to note, however, that although 14 interviews were held with Africa RISING participants, the interview conversation covered experiments that were undertaken independently as well as those that were conducted with prompting from Africa RISING. All interview participants were over the age of 18 and verbally agreed to participate in the interview after listening to an MSU IRBapproved informed consent script. All interviews were conducted at the respondent’s home, and a bottle of Coca-Cola and package of biscuits was provided to each respondent as compensation for participation. Interview content and respondent characteristics will be further discussed in the following sections. In addition to the 18 farmer interviews, four in-depth interviews were conducted with an Agricultural Extension Development Officer (AEDO) from each the four EPAs with an Africa RISING presence. All interviews were conducted in English (although the interpreter was present, in case of any misunderstandings) and each lasted approximately 45 minutes. These interviews provided a clear understanding of how the Africa RISING mother and baby trial experiments were managed and monitored, what lessons the AEDOs attempted to impart to project participants, which crop combinations were planted as part of the mother trial (and thus which options the participating farmers could choose from for their own baby trials), and any problems that participating farmers may have encountered (from the perspective of the AEDO leading the project). This information informed the interviews with Africa RISING farmers, and it also helped to check the validity of the information that farmers provided during interview conversations. 35 3.4.2 Data Collection: Interview Structure and Content In-depth interviews were conducted over a two-week period, from July 1-July 12, 2013. In total, the interview prompt contained 38 questions, although these questions were neither asked in a linear fashion, nor was every question asked to every respondent. The length and structure of any given interview conversation depended primarily on the respondent’s comfort, openness, and time constraints. On average, interviews lasted 1 hour and 15 minutes, with the shortest interview at 45 minutes and the longest at 1 hour and 45 minutes. Interviews were conducted in Chichewa, with the assistance of an interpreter. The interpreter had also been an enumerator during the household survey fieldwork phase, and thus she was already familiar with the research questions, the goals of the study, and the respondents. Translation was done in situ, where each question was asked first in English and then translated into Chichewa. As the interpreter became familiar with the interview topics and research goals, she occasionally asked probes in Chichewa without waiting for the prompt in English. Respondent answers were also translated from Chichewa to English in situ. To minimize recall errors, all interviews were audio recorded with the respondents’ permission. Question topics were related to: experimentation with unfamiliar crops, varieties, and techniques/technologies; management of experiments; motivations for trying something new; sources of information; ideas of “success” and “failure”; levels of satisfaction with experiments; intentions for future experiments; experimentation through Africa RISING and/or other agricultural intervention programs; and general ideas about on-farm experiments (e.g. To you, what does it mean to “experiment” with new crops, varieties, or techniques?) (Appendix 2). Most questions focused on experiments that were carried out in the 2011-2012 or 2012-2013 agricultural seasons, although respondents occasionally shared details of experiments that they 36 had conducted prior to 2011. The interview questions were written with contributions from Michigan State University faculty and Malawian agricultural experts. During the interviews, the interpreter used the phrase “try a new thing” in lieu of the word “experiment”, as there was no exact translation of “experiment” in Chichewa. Despite this linguistic difference, the conversational setting provided clarity of the idea of on-farm experimentation to farmers. 3.4.3 Qualitative Data: Translation, Transcription, and Analysis Every evening during the interview period, interviews were translated and transcribed. On average, 1 hour of interview tape took 3 hours to translate and transcribe. Each question, probe, and answer was recorded once, as if the conversation did not go through an interpreter. The questions and probes which were asked by the interpreter (in Chichewa) were only recorded in the transcription if they differed significantly from the original English, or if the interpreter asked them without first being prompted in English. During the translation and transcription process, an active effort was made to find the most appropriate English words to best represent the respondents’ ideas as they related to the study. For example, when a respondent used the phrase “trying something new” in Chichewa, the phrase was recorded as “experimenting” in the English transcript. After all of the interviews were translated and transcribed, the scripts were thematically coded and analyzed using the QSR NVivo 10 qualitative software analysis package. Themes related to experimentation with legumes, maize, other crops, and techniques/technologies, and included: management of experiments; plans for future experiments; motivations that drove experimentation; ideas of success and failure; satisfaction or dissatisfaction with an experiment; 37 persistence with a failed experiment; self-identification as an “experimenter”; theories as to why a particular experiment succeeded or failed; and memorable quotes (Appendix 3). To test for validity in the coding structure of the interview content, a second coder (who was not previously associated with the project) was trained to use NVivo 10 software and familiarized with the pre-established coding structure. The second coder analyzed and coded a subset of the full sample of interviews, six of the 18 total scripts (33%). This subset of scripts was chosen for the reliability analysis based on their clarity and representativeness of the whole sample. Subsequently, a test for inter-coder reliability was conducted using NVivo 10. Percentage agreement and Cohen’s Kappa coefficient (which takes into account the amount of agreement that could be expected to occur through chance) were used as indices for reliability. Overall percentage agreement between the coders was 96.4% (3.6% disagreement), and the Kappa coefficient was 0.4849, which indicates fair-good agreement between coders (NVivo10 for Windows Help, 2014). The following sections will explore on-farm experimentation among Malawian smallholders by drawing from both the qualitative and quantitative data sets. Quotations are used to help interpret quantitative findings related to experimental crops and technologies and the characteristics of innovative farmers. Additionally, quotations are used to shed light on the methods used by farmers when they try something new, and the drivers behind on-farm experimentation. These quotes help give a voice to innovators whose agricultural accomplishments have thus far gone unrecognized. Quotations were chosen according to their clarity and representativeness. 38 3.5 Respondent Characteristics 3.5.1 Household Survey Respondents The total sample size for the household survey was 324 respondents, and all respondents who were approached chose to participate in the survey. Below, Tables 4-7 illustrate some socioeconomic characteristics of the sample, disaggregated by geographic region and by sample group. Note that the column for Golomoti EPA also includes respondents who reside in Mtakataka EPA, but are considered distant control households corresponding to Golomoti for purposes of this study, and are thus combined with the Golomoti sample. This will be the case for all tables and figures presented in this paper. For the purposes of this study, a household was defined as a group of people who live together and share a common kitchen. Regarding gender of the household head, for households where a male lived or worked elsewhere and a female made the household and agricultural decisions more than half of the year, the household was defined as female-headed. If a male was present during the growing season, however, and made most of the agricultural decisions for the household (even if he lived elsewhere before and after the growing season), the household was defined as male-headed. 39 Table 4. Characteristics of Household Survey Population (by EPA) EPA Demographic Linthipe (n = 83) Golomoti (n = 80) Kandeu (n = 81) Nsipe (n = 80) Total Sample Male (n): 57 (69%) 60 (75%) 56 (69%) 58 (73%) 231 (71%) Female (n): 26 (31%) 20 (25%) 25 (31%) 22 (27%) 93 (29%) Avg. HH Size 5.2 5.1 5.2 5.1 5.1 Dependency Ratio 112 108 104 108 108 Avg. Farm Size (n = 288) 0.71 ha 0.83 ha 0.89 ha 0.97 ha 0.85 ha Avg. # of Fields 2.24 1.89 2.38 2.40 2.23 Avg. # of Tropical Livestock Units 0.50 0.35 0.76 0.48 0.52 Wealth Index [Range = 2-101] 15.4 15.2 16.8 17.5 16.2 Avg. # Months Food Supply 8.24 7.16 7.83 9.65 8.22 HH Head Gender: Total Sample N = 324, except where noted Table 5. Education of Household Head (by EPA) EPA HH Head Education Linthipe (n = 50) Golomoti (n = 44) Kandeu (n = 47) Nsipe (n = 36) Total Sample (n = 177) No Schooling 18 (36%) 13 (30%) 11 (23%) 3 (8%) 45 (25%) Some Primary 23 (46%) 24 (55%) 26 (56%) 27 (75%) 100 (57%) Completed Primary 4 (8%) 1 (2%) 3 (6%) 2 (6%) 10 (6%) Some Secondary 4 (8%) 2 (5%) 5 (11%) 2 (6%) 13 (7%) Completed Secondary 1 (2%) 4 (9%) 2 (4%) 2 (6%) 9 (5%) Due to rounding, column totals may exceed 100% 40 Table 6. Characteristics of Household Survey Population (by Sample Group) Group Demographic HH Head Gender: Intervention Local Control Distant Control (n = 162) (n = 81) (n = 81) Total Sample Male (n): 121 (75%) 49 (60%) 61 (75%) 231 (71%) Female (n): 41 (25%) 32 (40%) 20 (25%) 93 (29%) Avg. HH Size 5.2 4.8 5.2 5.1 Dependency Ratio 110 110 100 108 Avg. Farm Size (n = 288) 0.94 ha 0.80 ha 0.73 ha 0.85 ha Avg. # of Fields 2.42 2.10 1.98 2.23 Avg. # of Tropical Livestock Units 0.49 0.63 0.49 0.52 Wealth Index [Range = 2-101] 16.8 16.7 14.6 16.2 Avg. # Months Food Supply 8.80 8.12 7.15 8.22 Total Sample N = 324, except where noted Table 7. Education of Household Head (by Sample Group) Group HH Head Education Intervention Local Control (n = 82) (n = 45) Distant Control (n = 50) Total Sample (n = 177) No Schooling 14 (17%) 20 (44%) 11 (22%) 45 (25%) Some Primary 50 (61%) 19 (42%) 31 (62%) 100 (57%) Completed Primary 5 (6%) 2 (4%) 3 (6%) 10 (6%) Some Secondary 7 (9%) 4 (9%) 2 (4%) 13 (7%) Completed Secondary 6 (7%) 0 3 (6%) 9 (5%) Due to rounding, column totals may not equal 100% The total survey sample included 71% male-headed households and 29% female-headed households, which is a typical distribution for central Malawi. The majority (56.5%) of respondents had attended some primary school, but had neither completed primary nor began secondary school. Note that the educational data in Tables 5 and 7 only pertains to those interview respondents who were also the head of household (n = 177). Average household size 41 was 5.1 persons, and the average dependency ratio (number of economically inactive persons divided by number of economically active persons; as shown by number of dependents per 100 persons in the working-age population) was 108. Although the standard age groups in dependency ratio statistics for economically inactive persons are 0-14 and 15-64 (Findley, 2014), this study used the age groups 0-14 and 15-69, as persons in rural Malawi are often engaged in agricultural labor until later in life (R. Chikowo, personal communication, March 15, 2013). According to the World Bank Age Dependency Ratio data set (2014), the dependency ratio for the total population of Malawi in 2013 was 95, indicating that the survey population (which consisted wholly of persons who lived in rural areas, most of them farming on less than two hectares of land) contained a larger proportion of dependent persons than did the total population of Malawi in 2013. Farm size data (total hectares) was recorded only for farmers who worked three or less fields (n = 288; 88.9% of total sample), and it was found that the majority of these farmers held less than one hectare of land (μ = 0.85 ha per household). Farmers in the sample population had slightly smaller landholdings than were reported in the Republic of Malawi and World Bank Malawi Poverty and Vulnerability Assessment (2006), where farmers held an average of 1.2 hectares per household. For this study, a smallholder was defined as a farmer who held less than two hectares of land, meaning that at least 90% of the farmers in the sample could be identified as smallholders. Regarding average number of fields, farmers did not share a consistent sizebased method for breaking sections of land into “fields” or for breaking fields into “plots”. In general, however, a piece of land was divided into fields and further subdivided into plots. Farmers demarcated the land by agroecological factors such as soil type, topography, water holding capacity, cropping systems, etc. Therefore, field and plot level data were gathered based 42 on respondents’ definitions of a field or plot on their own farm. No assumptions should be made about the uniformity of field and plot sizes across farms. By farmers’ own definitions, it was found that on average farmers held 2.23 fields. Tropical Livestock Units (TLUs) can be interpreted as indicators of farming system types and livestock animal (productive) assets. The unit is a type of exchange ratio between livestock animals and is calculated by converting adult body weight into metabolic weight (Livestock, Environment and Development Initiative, 2005, as cited by Chilonda and Otte, 2006). Table 8 provides a list of coefficients that were used to convert total number of livestock animals into Tropical Livestock Units (FAO, 2005). Average TLUs for the survey population was 0.52 per household, which is in line with estimates from the Republic of Malawi and World Bank Malawi Poverty and Vulnerability Assessment (2006), where average TLUs were 0.53 per household. Table 8. Coefficients for TLU Conversion Livestock Species Cattle Pigs Goats Poultry Coefficient 0.7 0.2 0.1 0.01 Rabbits Excluded due to infrequency of ownership (n = 2) The Wealth Index is an asset-based measure of wealth, where both productive and consumer assets, along with housing materials, are assigned a numerical value. The values of each asset are then added together and the sum represents the relative wealth of a household. 43 Table 9. Asset Values used for Wealth Index Asset Type Asset Value Housing Material Fired brick walls 2 Unfired brick/mud walls 1 Corrugated iron roof 2 Thatch roof 1 Cattle (per animal) 3 Goats, Pigs (per animal) 2 Poultry (per animal) 1 Bicycle, Television (per unit) 3 Cell phone (per unit) 2 Radio (per unit) 1 Productive Reproductive Table 9 depicts the ranking system used for the Wealth Index that was used for this study, which was adapted from the Wealth Index used in a study of agricultural production and nutrition among farmers in northern Malawi in 2011 (Snapp et al., 2014). Farmers in our sample had Wealth Index scores ranging from 2-101. The average number of months that a household’s food supply was a rough calculation that farmers provided, based on the amount of maize they had harvested in the weeks before the survey. On average, households thought their maize supplies would last slightly more than eight months, although several farmers claimed that their supplies would be gone within the first month after harvest. Food security status varied across farmers, and was largely dependent upon climate, available resources, and landholding size. This figure, therefore, should be interpreted with caution. 3.5.2 In-depth Interview Respondents The total sample size for the qualitative interviews was 18 respondents, and all respondents who were approached chose to participate in the interview. Below, Table 10 44 illustrates some socioeconomic characteristics of the sample. Note that the terms household and household head gender were defined in the same way as during the household survey (see Section 3.5.1). The interview population included five respondents each from Linthipe, Golomoti/Mtakataka, and Nsipe EPAs, and three respondents from Kandeu EPA. Fourteen respondents were active participants in the Africa RISING project (and four were nonparticipants), although many of these respondents conducted at least one experiment independently of those that were promoted by Africa RISING. Female respondents were specifically targeted to participate in the qualitative interviews, which resulted in 15 female respondents and three male respondents. Female respondents, however, came from different household types, where six were heads of their own households, and nine were spouses within male-headed households. We did not interview any males from within a female-headed household. Respondents ranged in age from 25-54 years of age. The household size for most respondents ranged from three-six persons, and most respondents farmed one hectare of land or less. Likewise, the majority of respondents had 1.0 Tropical Livestock Units or less and scored in the lowest quartile of the Wealth Index. Table 10. Socioeconomic Characteristics of In-Depth Interview Population Linthipe 5 Golomoti/Mtakataka 5 Kandeu 3 Nsipe 5 Yes 14 No 4 EPA RISING Participant 45 Table 10. (cont’d) Male (MHH) 3 Female (MHH) 9 Female (FHH) 6 25 – 34 7 35 – 44 7 45 – 54 4 3 – 4 persons 6 5 – 6 persons 10 7 persons 2 ≤ 0.5 ha 4 0.6 - 1 ha 11 > 1 ha 3 0 – 0.1 6 0.2 – 1.0 10 > 1.0 2 2 – 10 7 11 – 25 8 25 – 40 2 > 40 1 Gender (by Household Type) Age Group Household Size Farm Size TLUs Wealth Index [Range = 2-101] Despite these general trends, however, some interview respondents came from large households (seven or more members), cultivated relatively large farms (> 1 hectare), and owned numerous reproductive and productive assets (high TLU counts and/or Wealth Index scores). Thus, the interview sample contained representatives from across a wide range of socioeconomic standings (within the Malawian smallholder population), which gave insight into the drivers, management practices, and decision-making processes related to on-farm experimentation across many perspectives. 46 4.0 Results The following sections will detail findings from the quantitative and qualitative strands of our study, and will address the research questions posed at the beginning of this paper. Note that as the survey and the interviews yielded a wealth of information related to on-farm experimentation beyond legume production, alone, the discussion will include trends related to general experimentation as well as those trends related to legume production, specifically. 4.1 “Experimentation” versus “Everyday Practice” As discussed in Section 2.0, scientists in the formal sector and experimenting smallholders often come from very different epistemological backgrounds, and they use different vernacular to describe their “experimentation” processes. From early in the design stage of this project, we expected that these differences might influence our study. In an attempt to overcome this epistemological and linguistic difference, we used a definition of “experimentation” that could be translated into Chichewa and still retain its meaning, where we explained to farmers that we were interested in any “unfamiliar crops, varieties, or techniques/technologies that they had tried for the first time ever”. Using this definition, we gathered survey data on 572 examples of experiments that took place in 2012-2013. The interview data, however, reveals that despite the large frequency of experiments reported in our survey, some examples may still have gone unrecorded due to miscommunications and the ways in which farmers conceptualized their own actions. Although all of the in-depth interview respondents were purposefully selected according to their propensity to try new things (information we learned from the survey data), and all of them fit our definition of an “experimenter”, many of these innovative farmers interpreted the 47 questions about “trying new things” to mean “trying new things with an intervention project”. Some of these farmers claimed that they had never experimented with new crops, varieties or technologies prior to joining an intervention project, but by the conclusion of the interview we usually discovered that these farmers had tried many new things on their own, but they were doing so without thinking of their actions as “experimental” or “innovative”: INT: Previously on your own, you’ve never tried anything new? 1140: No, previously we were just growing local maize. Then when we stopped that, now is when we’re growing maize for sale [hybrid]. INT: What about last season, when you weren’t working with RISING and you tried a new hybrid maize variety, and a new groundnut variety, and a new bean variety? 1140: Oh! It was experimenting? I thought experimenting would mean only working with these projects. INT: So before you started participating in these projects, did you try new things on your farm just on your own? 4110: We were only planting local maize. INT: You didn’t try any new crops or any new spacings just as your own idea before these projects? 4110: No. ***Later in the interview*** INT: Where did you get the idea to plant the three plots like that to compare [different crops]? 4110: It was my own idea. 48 These quotes illustrate that despite our efforts to make the terms “experimentation” and “trying new things” mutually understandable, it often required an in-depth conversation to surmount the epistemological and linguistic differences between Malawian farmers and formal scientists. Therefore, although the basic experimentation questions we asked on the survey instrument revealed 572 examples of farmer experiments, it is likely that many more examples were inadvertently omitted. If the assertions are true that farmers are constant experimenters (Chambers et al., 1989; Scoones and Thompson, 1994b; Warren et al., 1995) and that on-farm experimentation is so common as to be called “ubiquitous” by some studies (Rhoades and Bebbington, 1995: 306), then it is almost certain that some of the Non-experimenters in our sample actually were trying new things, but we failed to capture that information in a fixedresponse questionnaire. While quantitative data yield important socioeconomic and farm-level information about experimenters, we need to exercise caution when making inferences about experimenters using these household survey data. Qualitative data, therefore, are vital to the understanding of on-farm experimentation processes, and provide invaluable insight into the more nuanced aspects of smallholder experimentation. Overall, the household survey data and the in-depth interview data are the most insightful when interpreted in tandem, and therefore many of the following sections draw from both data sets. We turn now to the quantitative and qualitative data to explore the types of experiments that were reported, the characteristics of experimenters, and farmers’ motivations and methodologies for conducting experiments. 4.2 Types of Experiments To begin our analyses of on-farm experimentation, we needed to first understand what kinds of experiments farmers were conducting. We used the survey instrument to ask farmers 49 what unfamiliar crops, varieties, and techniques/technologies they experimented with in the 2012-2013 season, and from those questions we elicited 572 examples of experiments that farmers had tried. These examples came from 228 farmers (70.1% of the total sample). The frequency distributions of these experiments can be seen in Figures 4-7. Similar experiments were grouped together under a common theme (e.g. “land preparation experiments” include shifting ridges, using box ridges, and measuring the precise distance between ridges). Figure 4. Total Frequency (%) Distribution of Experiments New Crops New Varieties (Maize) New Varieties (Legumes) Plant Spacing Land Preparation Inorgainc Fertilizer Application Residue Management Other Fertility Measures Manure/Compost Application Weed Control Irrigation Mixed Cropping Sole Cropping Pest Control Planting Time N = 572 Examples 0 5 10 15 20 25 30 35 40 Figure 4 illustrates that 89% of experiments reported on the survey fell into three categories: new crops 34%; new varieties (maize and legumes) 40%; and plant spacing experiments 15%. Likewise, when the experiment examples were disaggregated—both according to sample groups of Intervention, Local control, and Distant control, and according to the experimentation classification categories of Project participants, Followers, and Independents (for definitions of these groups see Section 3.3.4)—similar trends emerged in the frequency 50 distributions (Figures 5 and 6). For reference, Figures 7 and 8 depict all of the new crop experiments that farmers tried, disaggregated by sample groups and experimentation classifications, respectively. Figure 5. Frequency (%) Distrubtion of Experiments (by Sample Group) New Crops New Varieties (Maize) New Varieties (Legumes) Plant Spacing Land Preparation Inorgainc Fertilizer Application Residue Management Other Fertility Measures Manure/Compost Application Weed Control Irrigation Mixed Cropping Sole Cropping Pest Control Planting Time Distant control (n = 57 examples) Local control (n = 56 examples) Intervention (n = 459 examples) 0 5 10 51 15 20 25 30 35 40 Figure 6. Frequency (%) Distribution of Experiments (by Experimenter Classification) New Crops New Varieties (Maize) New Varieties (Legumes) Plant Spacing Land Preparation Inorgainc Fertilizer Application Residue Management Other Fertility Measures Manure/Compost Application Weed Control Irrigation Mixed Cropping Sole Cropping Pest Control Planting Time Independents (n = 44 examples) Followers (n = 129 examples) Project participants (n = 399 examples) 0 5 10 15 20 25 30 35 40 Figure 7. Frequency (%) Distribution of New Crops (by Sample Group) Pigeonpea Soy Bean Cowpea Groundnut Common Bean Millet Tobacco Sorghum Cotton Cassava Irish Potato Rice Cocoyam (Taro) Distant control (n = 17 examples) Local control (n = 20 examples) Intervention (n = 157 examples) 0 5 10 15 20 25 30 35 40 45 50 55 60 65 70 52 Figure 8. Frequency (%) Distribution of New Crops (by Experimenter Classification) Pigeonpea Soy Bean Cowpea Groundnut Common Bean Millet Tobacco Sorghum Cotton Cassava Irish Potato Rice Cocoyam (Taro) Independents (n = 17 examples) Followers (n = 51 examples) Project participants (n = 126 examples) 0 5 10 15 20 25 30 35 40 45 50 55 60 65 70 Across all three experimentation classifications, farmers had a propensity to try new crops, varieties, and plant spacing techniques more often than other types of experiments. Note, however, that a larger percentage of Independent experimenters (18%) and Followers (25%) tried new maize varieties than did Project participants (14%). Additionally, Project participants tried more new leguminous varieties than either Independent experimenters or Followers (26% compared to 14% and 16%, respectively) (Figure 6). Likewise, a much larger percentage of Project participants tried pigeonpea (60%) than either Independent experimenters (30%) or Followers (33%) (Figure 8). This trend is reflective of the recommendations made by Africa RISING in 2012-2013, where the project encouraged its participants to grow specific legume crops (e.g. pigeonpea), and 76.5% of Project participants were working with Africa RISING in 53 that season (Table 3). The apparent popularity of crop, variety, and plant spacing experiments across experimenter groups will be further discussed in Section 5.2. In addition to the survey questions, we also used the in-depth farmer interviews to gain insight into the types of experiments farmers were trying. The quotes in Table 11 illustrate some of the experimentation themes we created based on survey and interview responses. Note that some themes (e.g. crop rotation) are detailed in Table 11, but are not included in the previous frequency distributions (Figures 4-6). This discrepancy exists because in the interviews, we elicited information about farmers’ past experiments (sometimes as far back as 11 years), while the surveys only gathered information about the 2012-2013 season. Table 11. Examples of Farmers’ Experiments from Interviews Experiment Type Examples from Interviews New Variety 2134: For groundnuts, I tried a new variety…Because we were told that this new variety of groundnuts yields more than the local variety. New Crop 3205: The new crop I’ve tried is pigeonpea that we received. Plant Spacing 3126: We wanted to compare. For this new variety, it’s something new, so if we plant it in a local way, it might not do well. That’s why we changed the plant spacings. Plant Spacing + Land Preparation 1217: We take a ruler then we use it to make sure the distance between planting stations is the same, and also the distance between the ridges is the same. Then we use a small plot and apply manure on that plot. Then we plant the seed using 1:1 technique. Then we see the difference from where we were planting 3:3. We like to see the difference. Land Preparation + Reside Management 4219: With the fertility of our soils, now we’ve started incorporating residues and we’ve starting using box ridges so that when there is a lot of rain, we don’t lose the water. Rather the box ridges should hold some of the water so in case of drought, the crops will survive. And also, incorporating crop residues traps the moisture in the soil. So then, when there is a lot of sun the crops don’t die. Then, we plant the crops we want on the fields. 54 Table 11. (cont’d) Residue Management 3205: The green residues, we incorporate them in the soil. Because we start harvesting groundnuts before maize. Then it’s like we’ve buried them while they’re fresh, so they rot, then they act like manure. Manure/Compost 4104: Yes. When I made the ridges, I was applying manure on the planting station which would later be used for maize. Application 4219: On the field where we planted maize this season, we can’t plant Crop Rotation maize again next season. Instead, we plant groundnuts on that field. Where we have grown tobacco this season, next season we grow maize. We change the fields. Sole Cropping 3205: …because previously we were just intercropping, in the same field maize, in the same field soya, and in the same field groundnuts. So last season was the first time to divide the field into 3: on one plot soya, on another plot maize, on the last plot groundnuts. It is important to note that the examples of farmers’ experiments we report are similar to those detailed in Scoones and Thompson (1994a), and the frequency distributions of the experiments in Figure 4 are similar to those reported by Sumberg and Okali (1997) and Kristjanson et al. (2012). 4.3 Innovative Farmers at the Household Level After gaining an understanding of the types of experiments that farmers were conducting, we continued our statistical analyses with several tests that would help to answer the research question: What kinds of farmers are experimenting with unfamiliar legume crops, varieties, and farming techniques related to legume production? To explore this question, we used a one-way analysis of variance (ANOVA) test to compare the means of several socioeconomic and farmlevel factors across four experimentation groups: Non-experimenters, Project participants, Followers, and Independents (for definitions of these groups, see Section 3.3.4). For this test, the independent variable was the experimentation group, and the grouping (dependent) variables 55 were: household size; dependency ratio; wealth index score; tropical livestock units; farm size; and number of fields held. As gender of the household head was a binary variable, it was not included in this analysis, but for reference this data can be seen in Table 12. The means of the grouping variables used in the analysis can be found in Table 13. Table 12. Gender of Household Head by Experimentation Groups HH Head Gender Nonexperimenters (n = 96) Project participants (n = 145) 26.8% 46.3% Male (%) 40.9% Female (%) 36.6% Due to rounding, row totals may not equal 100% Followers (n = 64) Independents Total (n = 19) (N = 324) 21.6% 15.1% 5.2% 7.5% 99.9% 100.1% Table 13. Grouping (Dependent) Variable Means by Independent Variable Groups Mean (S.D.) by Independent Variable Group Grouping (Dependent) Variables Household Size Dependency Ratio Wealth Index [Range = 2-101] Project Non-experimenters participants (n = 96) (n = 145) 5.03 (2.11) 5.36 (1.81) Followers (n = 64) 4.91 (1.72) Independents (n = 19) 4.47 (1.95) 1.05 (0.80) 1.13 (0.86) 1.09 (0.75) 0.91 (0.89) 14.85 (14.07) 16.39 (13.11) 17.44 (17.04) 17.58 (21.41) 0.62 (1.40) 0.44 (0.76) 0.56 (1.39) 0.52 (1.13) TLUs Farm Size 0.77 (0.47) 0.91 (0.72) 0.85 (0.49) 0.81 (0.50) (total ha)* 2.31 (1.20) 2.19 (0.92) 2.63 (1.61) Number of Fields 2.05 (1.29) *Note smaller sample size for this variable: Non-experimenters (n = 87); Project participants (n = 126); Followers (n = 59); Independents (n = 16) The overall ANOVA test revealed that none of the six grouping variables differed significantly across experimentation groups, although the descriptive statistics indicate that nonexperimenters scored lower on the asset-based wealth index than did experimenting farmers. As explained in Section 4.1, it is difficult to make inferences about a nuanced topic, such as 56 smallholder experimentation, with survey data alone. In light of the inconclusive ANOVA test, we turn now to the qualitative data from our farmer interviews to gain an understanding of the differences between experimenter groups. During the interviews, farmers commonly spoke of the ways in which the size of their landholdings (or the number of fields they held) affected their experimentation processes. Many farmers shared the opinion that beginning a new experiment (or scaling out a successful experiment) required more land, which often meant renting in additional field(s): 1105:…if you stick to planting on the same plot each year, you will suffer from hunger. But you should rotate, rent in more land, to experiment on that soil you’ve rented in to see the yields. 2134: Next season, I’m planning to grow on ¾ acres, when I find some more land. 2263: For next season, I want to rent in some more land, so that when we plant maize on the other field, on a separate field we will plant cowpea. 2132:…I paid money to rent in a field, then when we went, [my husband] said we should be tilling, then we sow the seed, then after the rains fell in February we transplanted the rice [experimental crop]. Likewise, some farmers who wanted to experiment with specific crops needed access to fields with particular characteristics (e.g. soil type, proximity to water source, etc.) in order to do so, and if they did not have access to those fields, they could not conduct (or repeat) their experiment: 2312: For mustard, we used the rented in plot. Then the owner took it back. Then we didn’t grow mustard again. 57 In the same spirit, a few farmers described how their limited landholdings resulted in the forced abandonment of an experiment: INT: So why did you decide not to plant the 8073 or the 8033 this season? 4104: I didn’t have enough land because I also wanted to plant the variety I received this season [from RISING]… INT: So you gave that land to the Africa RISING hybrid instead, this year? 4104: Yes. Not all farmers, however, abandoned their experiments due to land shortages. For some farmers, land constraints caused the modification, rather than the abandonment, of a planned experiment: 1140:…sometimes the field is not enough…So to avoid leaving out some crops, the crops should be intercropped. 4129: We first intercrop [the experimental crops] because the area is not large enough to plant each crop separately. So we feel that if we divided the plots, we’d harvest little maize. So we just intercrop [the experimental crops] in the same field. These farmer insights help us to understand that when challenged by limited resources— such as fields held—innovative farmers felt they had two options: abandon their planned experiment or make an adjustment to the experiment’s design. Therefore, while farmers who hold fewer fields may face more complications in their experimentation processes than farmers with more fields, fewer fields does not necessarily result in less experimentation. These results will be revisited in Section 5.3. 58 4.4 Innovative Farmers Within the Household While household data provide a wealth of information about the socioeconomic and agronomic characteristics of our experimentation groups at the macro level, we cannot fully understand the smallholder experimentation process without looking at the effects of individuallevel characteristics on the labor and decision-making processes of on-farm experimentation. As we posited that female farmers would be more likely to experiment with unfamiliar legumes than would male farmers, given that legume crops are traditionally planted and managed by women, we used statistical analyses to investigate the effects of gender on experimentation. In order to better understand the relationship between gender and experimentation, we used the survey instrument to ask farmers several questions about the distribution of labor during the initial planting of new crops/varieties and the initial implementation of new techniques/technologies during experiments in the 2012-2013 season (Appendix 1). A Chisquare test was used to determine the strength of the relationship between the experimenter’s gender and the type of experimental crop or variety that was grown (Table 14). 59 Table 14. Gender and Experimental Crops, Varieties, and Techniques Crops (n = 194): Legumes Cash Tubers Grains Male 32 (17.6%) 1 (33.3%) 2 (66.7%) 1 (16.7%) Female 66 (36.3%) 0 0 3 (50%) Both 83 (45.6%) 2 (66.7%) 1 (33.3%) 2 (33.3%) Other* 1 (0.5%) 0 0 0 Total 182 (100%) 3 (100%) 3 (100%) 6 (100%) Total Experiments by Gender† 36 (18.6%) 69 (35.6%) 88 (45.4%) 1 (0.5%) 194 (100.1%) Varieties (n = 226): Maize Male 6 (6.2%) Female 39 (40.2%) Both 51 (52.6%) Other* 1 (1%) Total 97 (100%) Legumes 11 (8.5%) 46 (35.7%) 71 (55%) 1 (0.8%) 129 (100%) 17 (7.5%) 85 (37.6%) 122 (54%) 2 (0.9%) 226 (100%) Male Female Both Other* Total 22 (14.5%) 48 (31.6%) 82 (53.9%) 0 152 (100%) Total Experiments by Gender Techniques (n = 152): Total Experiments by Gender *Represents laborer for whom no gender information was known (e.g. child) †Due to rounding, row total exceeds 100% The Chi-square tests revealed that the experimenter’s gender did not have a statisitically significant effect on the types of crops/varieties that were planted experimentally, or on the type of experimental technique that was attempted: crops: χ2 (9, N = 194) = 7.37, p > .05; varieties: χ2 (3, N = 226) = 0.81, p > .05; tech: χ2 (22, N = 152) = 31.21, p > .05. Despite the non-significant p-value of the Chi-square test, there are several important trends that emerged in the analysis. Firstly, we found that legume crop experiments were by far the most commonly reported among all respondents, where farmers said that 94% of the experimental crops they had tried in 2012-2013 were leguminous (note that leguminous crops were some of the most common experimental crops even among those farmers who did not participate in Africa RISING; see Figure 7). Likewise, farmers reported that 57.1% of the experimental varieties they tried in 20122013 were leguminous. It is also important to note the gender distribution of experimenters who 60 grew unfamiliar legumes. The majority of experimental legume crops and varieties that were grown in 2012-2013 were planted by both spouses together (crops: 45.6% planted by both spouses; varieties: 55% planted by both spouses). In instances where only one person reported planting an experimental legume crop, however, women planted over twice as many unfamiliar legume crops as men (36.3% compared to 17.6% planted by men). Similarly, women planted over four times as many experimental legume varieties as men (35.7% by women compared to 8.5% by men). This trend extends beyond experimental legumes: respondents reported that the majority of experiments (with crops, varieties, and techniques/technologies) were conducted by both spouses together. Those experiments that were conducted by only one person, however, were more often attempted by solo females than by solo males (crop experiments: 35.6% by females, 18.6% by males; variety experiments: 37.6% by females, 7.5% by males; tech experiments: 31.6% by females, 14.5% by males). Finally, it should be noted that both male and female respondents reported that for crop, varietal, and technical experiments, the majority of labor was undertaken by both spouses together, closely followed by solo women. For reference, these data can be seen in Table 15. 61 Table 15. Respondent’s Gender by Gender of Experimenter New Crop Experiments Respondent Gender: Male Female Total† Male 14 (21.2%) 22 (17.2%) 36 (18.6%) Female 26 (39.4%) 43 (33.6%) 69 (35.6%) Both 26 (39.4%) 62 (48.4%) 88 (45.4%) Other* 0 1 (0.8%) 1 (0.5%) Total 66 (100%) 128 (100%) 194 (100.1%) New Varietal Experiments Respondent Gender: Male Female Male 5 (6.4%) 12 (8.1%) Female 28 (35.9%) 57 (38.5%) Both 45 (57.7%) 77 (52.0%) Other* 0 2 (1.4%) Total 78 (100%) 148 (100%) Total 17 (7.5%) 85 (37.6%) 122 (54.0%) 2 (0.9%) 226 (100%) New Technical Experiments Respondent Gender: Male Female Both Other* 6 (10.9%) 17 (30.9%) 32 (58.2%) 0 Male 16 (16.5%) 31 (32.0%) 50 (51.5%) 0 Female 22 (14.5%) 48 (31.6%) 82 (53.9%) 0 Total *Represents laborer for whom no gender information was known (e.g. child) †Due to rounding, row total exceeds 100% Total (100%) (100%) (100%) Although we did not find a statistically significant relationship to support the notion that women are more likely to experiment with legumes than men, the frequency distributions in Table 14 still provide us with valuable insight into the gendered division of labor associated with on-farm experimentation. 4.5 Farmer Motivations and Methodologies 4.5.1 Initiating an Experiment To address our second research question (What motivates farmers to experiment with unfamiliar legume crops, varieties, and farming techniques related to legume production?), we held in-depth interview sessions with 18 innovative farmers. During the interviews, we asked farmers about their motivations for trying new things, in general, and for trying specific crops, 62 varieties, and techniques/technologies (Appendix 2). Respondents discussed many different motivations behind each of their experiments, and often farmers tried something new for multiple reasons and with several goals in mind. The motivations described by farmers fell into three overarching categories, two of which were modified versions of experiment categories used by Sumberg and Okali (1997) in their study of on-farm experimentation: proactive and reactive motivations. Additionally, we created “external” motivations as a third category. The first half of this section addresses farmer motivations for all experiments, and in the second half we discuss differences in motivations between experiments, specifically between maize varietal experiments and other crop/varietal experiments (e.g. legumes). Proactive experiments included those that were driven by a farmer’s desire to create a positive change in her life circumstances or farm system, for example to increase food production, generate household income, improve soil fertility, or maximize land use. Reactive experiments were those that were prompted by a farmer’s response to unexpected circumstances, such as climate change, pest or disease problems, or access to resources. Lastly, external experiments were those where a farmer was invited to try something new by an influential source (e.g. intervention project, extension officer, etc.). Note that while reactive motivations sometimes drove new experiments, it was also common for reactive factors (such as limited resources) to cause farmers to adjust an experiment or abandon it altogether. Unlike proactive and external motivators that primarily drove new experiments, reactive factors could also act as barriers to experimentation. According to our qualitative coding structure, farmers spoke of proactive experimentation 162 times, reactive experimentation 65 times, and external experiments 77 times. As was previously stated, however, farmers in our sample were usually motivated by a combination of 63 factors that spanned one or more of the motivation categories. It was not uncommon, for example, for one experiment to be identified as both proactively and externally driven, where a farmer tried something because it would benefit her and because it was suggested by an extension officer. In Table 16, we use farmer quotes to provide examples of experiments in all three motivation categories, as well as experiments that were driven by multiple motivations. Table 16. Farmer Motivations for Experimentation Motivation Examples from Interviews Proactive 1105: I tried this because there wasn’t enough food for my household. I have small children who are orphans, so if I don’t work hard, I’ll have problems with raising the kids. 2134: We try new things because we’d like to compare the benefits of the crops we previously grew and the crops we’re currently growing. INT: Why do you want to try zero tillage? 2263: To reduce labor. Because land preparation [with hand hoes] is labor intensive. 2301: I like to be like these other people who do not lack things. We should not just rely on getting help from other people, but we should be self reliant. That’s why we experiment. 3126: [So that] The village should be developed. And also the households should be food secure. 4129: For soya, most people who grow it can sell it and get lots of money. When you harvest a lot, you can sell it and use the money for other household needs. Reactive 1140: We try new things because of the changes in the rain. That’s why we stopped planting those crops which are hard to grow when the rains aren’t enough, and go instead for those crops which still grow well with less rain. INT: So why did you decide to try a new variety of hybrid? 2312: Because DK 8033 [usual variety] was no longer available. 4134: I would try something [new] if I had enough resources. But the main problem here is fertilizer, because the prices of fertilizer have risen very high. But if we had enough fertilizer, we could experiment. 4219: We weren’t happy with the prices for which we sold the cotton. That’s why this year we only grew tobacco. 64 Table 16. (cont’d) 2263: Experimenting, sometimes the AEDO tells us to do it like this, like this, External like this. Then we go to our fields and practice what he advised us. INT: So do you think World Vision and your experience with them made you more brave to try new things, like new crops? 2301: Yes. They say wealth is in the soil. 3105: The AEDO told us that there’s also pigeonpea here, and you need to plant it. We didn’t refuse that. We received the seed just to try it. 4110: We received it from Africa RISING, so I wanted to see its yields. 2134: For us to grow soya, they told us that we may get two main benefits. One Multiple Motivations is making soya porridge from soya flour, and the other is getting money from the sales. So when we received the soya seed, we decided to grow it to make soya porridge for the children, and it’s nutritious. And the remaining produce, to sell. [External, proactive] 3126: Because of the problems with the local varieties. We wanted to compare the new varieties and the old varieties to see which one will yield more, and which one would benefit us the most. [Reactive, proactive] 4219: Because when we were conducting our tobacco meetings, we were told that a farmer shouldn’t just rely on one crop. For example, if you grow maize and you rely on it to eat and to sell, it won’t work well. You need to grow more than one crop, so that if one crop isn’t selling well, you can try the other crop. And also, if you rely on one crop, when the rains aren’t good you’ll suffer a lot because you won’t have food for your household. [External, proactive] Just as it was common for one experiment to be driven by multiple motivations, as illustrated in Table 16, it was also common for a farmer to conduct different experiments for different purposes. In other words, we did not encounter any farmers whose every experiment was only influenced by one type of motivator. Farmer motivations were thus impossible to categorize by socioeconomic status or farm-level characteristics because the drivers of experimentation are complex and vary from case to case and farmer to farmer. We did, however, identify a trend in experiment types, specifically in the motivations behind maize varietal experimentation. Among those farmers who tried hybrid maize for the first 65 time or who tried a new variety of hybrid maize, many of their experiments seemed to be motivated by farmers’ reactions to changes in rainfall patterns: 1109: The things that are changing are, for example, we used to grow local maize. But with the way the rains are coming, because it [local] matures late, it’s different from hybrid maize, which is still doing well with the way the rains are coming. 1217: Sometimes [the rain is] erratic. Sometimes it stops early. So when you plant hybrid, it still does well even if rain stops early. INT: So why did you decide to try and plant hybrid maize last season for the first time? 3205: Because I saw that this local variety wasn’t doing well. INT: How? 3205: It’s not doing well because of the changes in the weather patterns. So we decided that it would be better to grow hybrid maize. 4134: Sometimes [we experiment] because of the way the rains are coming. For instance, some years we plant local maize, and then we see that the maize doesn’t do well with those rains. That’s why we change the variety, to try one which would do better with those rains. And also because maybe when we change, we may get better yields and sell some to get money. In contrast, those farmers who planted new non-maize crops and/or varieties (most of these experiments involved legumes) often identified motivations that were not related to climate 66 change or rainfall patterns, such as income generation, food production, improving the farm system, etc.: 1140: This year, for us to grow those crops, we were told they would help to solve some household problems. For example, if soya yields a lot and you sell it, you may use that money to address some other problems. That’s why we decided to grow soya and cowpea. INT: So why did you decide to plant cowpea and pigeonpea? 1217: Crop diversity. When you diversify crops, you get money from all those crops when you harvest a lot. INT: And why did you want to plant cowpea? 2263: Money! 4134: I wanted to use it [pigeonpea] to help with the relish problem. And also because when the leaves fall on the soil, they increase the soil fertility. Regarding experimentation with new techniques/technologies (applied to both maize and non-maize crops), farmers reported a wide range of motivators, including climate change, improving the farm system, reducing labor, etc.: 2134: The difference is that for zero tillage, it reduces labor. We don’t spent so much time cultivating the field. Weeds don’t grow quickly because they’re hindered by the maize stalks. It’s only maize which grows, not the weeds. Then we just go and pull up the 67 weeds from the maize field. And also when there’s too much sun, it doesn’t penetrate easily, and the field remains moist for a longer period of time. 2263: We were told that with zero tillage, after you lay the maize stalks, you just plant. Even if the rain isn’t enough, the maize won’t wilt because the soil is still moist. 4110: Because for example, if the rain stops for 2 weeks, during that time, the field will still be moist and the crops will still grow well where you’ve made box ridges. The crops don’t wilt because the field stays moist. 4134: Incorporating crop residues is very beneficial because it increases soil fertility. The soil doesn’t lose its fertility when you incorporate those residues. 4219: Because a field where we grew tobacco, groundnuts, or soya, is fertile. It’s like we’ve increased the soil fertility [by growing those crops]. Then, if the next season, you grow maize on that field, the yield will be high. These examples indicate that farmers in our sample frequently experimented with hybrid maize varieties as a reaction to changing rainfall patterns, and with non-maize crops and varieties to meet various other goals (income, household nutrition, farm health, etc.). These farmers also indicated that their technical experiments were driven by a wide range of factors including climate change, improving farm health, and others. These qualitative findings are compounded when we draw from the survey data, where 86.1% of farmers said they had noticed changes in 68 the rains over the last 20 years, and out of those farmers, 67.2% reported that the changes they noticed included: less rain; erratic rainfall; and a tendency for the rains to stop before the crops had matured. When taken in combination, these results indicate that farmers are not only noticing climate change, but they are actively experimenting with maize varieties and with new techniques/technologies in an effort to mitigate undesirable changes. The implications of these results will be discussed further in Section 5.2. 4.5.2 Designing an Experiment After we gained an understanding of the motivators that drive farmers to initiate an experiment, we needed to further explore the methods used by farmers when conducting an experiment. This section relates to our final research question: How are farmers managing their experimental crops, varieties, and techniques? During the interviews, we asked farmers detailed questions about the ways they set up their experiments (Appendix 2). These questions related to the size of the experiment (e.g. plot size, amount of seed, etc.), the use of a comparison or “control”, and the separation of new crops/varieties from familiar ones. In general, our results surrounding an experiments’ size are aligned with those of similar studies (Rhoades, 1989; Sumberg and Okali, 1997), where most farmers started an experiment on a small scale, either planting a small amount of seed or using a small tract of land: 1105: We start with small quantities [of seed]. 1121: We just plant on a piece of land, like a bed, to try. INT: And how big was the area where you tried the zero tillage this year? 2301: It was one bed…It was just experimenting, so it was small. 69 4110: I tried the new variety on a small plot. INT: So the first time that you tried it, were you nervous that it wouldn’t work? 4219: Yes, we doubted it. And we only tried it on one field. Likewise, our findings related to farmers’ use of a comparison or “control” were similar to those found by Sumberg and Okali (1997), where some farmers consciously compared an experiment to a control plot during the same season, other farmers compared an experiment to what Sumberg and Okali called a “historical control” (farmers’ detailed knowledge of the past performance of a crop or technique after years of experience), and still other farmers did not use any obvious control—historical or otherwise. Illustrative quotes can be found in Table 17. Table 17. Farmers’ Use of a “Control” in Experiments Control INT: So besides the different variety, did you plant it just the same, with the 1:1 spacing and the fertilizer application? 1217: There were no differences. INT: Demeter. The way you grew it, was it the same way [as the old variety], or was there some differences? 2312: It was just the same. No control INT: Did you also plant the local variety [cowpea] as 1:1? 2134: No, for local I planted 3:3. INT: So the first time you planted it [hybrid], did you plant it the same as you planted your local maize? 3205: No. For this one [hybrid], we planted 1:1. INT: Did you plant the old varieties and the new varieties in the same way? Like both as sole crops, both the same spacing? 4219: They were different. To better understand the ways in which farmers implement an experiment, we also asked respondents about their planting methods—whether experimental crops/varieties were separated 70 from familiar crops, or whether they were planted together. Again, farmers elicited mixed responses, where sometimes new and old crops were separated (so farmers could see the new crops’ benefits) and sometimes new and old crops were planted together (so farmers could analyze the performance of an experimental crop as an intercrop). Related quotes can be found in Table 18. Table 18. Farmers’ Separation of Crops Separate experimental crops from familiar crops 1105: I start planting it as a sole crop, so that I should see it. 1109: When we plant them on the same plot, we wouldn’t notice the crop that we’re very interested in. 1217: I wanted to see the benefits of the new variety, and compare them to those of the previous variety. 3205: We divided the field. We planted hybrid on one side and local maize on the other side. Plant experimental crops together with familiar crops 2134: We wanted to compare the yields to see which would do best—as a sole or an intercrop. INT: And why did you decide to plant it in those 3 different ways: one as a sole crop, one intercropped with cowpea, and one intercropped with pigeonpea? 4134: I wanted to see the yields when we intercropped the different crops. 4219: In the past, we planted it [an experimental crop] as a sole crop. But now, with the way things are changing, we sometimes start with intercropping. INT: For the first time? 4219: Yes. So we can see if the intercrops do well or not. These differences in design across farmers (and even across experiments) indicate that although farmers use a variety of management techniques when trying an experimental crop, variety, or technique/technology, whichever method they use has a definite purpose and farmers have specific goals in mind when they implement an experiment using certain methods. We will propose a potential explanation for these methodological variations in the Section 5.3. 71 4.5.3 Repeating, Adjusting, or Abandoning an Experiment After we had learned more about why farmers initiated experiments and the methods they used to conduct experiments, we turned to the decision-making process of farmers after the conclusion of an experiment. To understand farmers’ attitudes and their decision-making processes, we asked questions during the interview about farmers’ assessment criteria (“success” or “failure”) and their attitudes of satisfaction (or dissatisfaction) with a completed experiment (Appendix 2). The following quotes demonstrate how those assessments and attitudes (together with a farmer’s resources and social situation) shaped farmers’ intentions for future experimentation and helped determine if they would repeat, adjust and repeat (i.e. scale out, scale back, or make a change), or abandon their original experiment. Through our conversations with interview respondents, we learned that the relationship between success (failure), satisfaction (dissatisfaction), and future intentions is not direct, but is mediated by factors such as resource availability. Thus, dissatisfaction with an experiment does not necessarily result in the abandonment of that experiment. Likewise, satisfaction with an experiment does not necessarily result in its repetition. Regarding farmers’ assessment criteria, interview respondents defined an experiment as a “success” if the experimenter gained something from it (e.g. food, income, knowledge, etc.), and a “failed” experiment as one that did not meet the farmer’s expectations or desired outcomes. These definitions are represented by the quotes in Table 19. 72 Table 19. Ideas of “Success” and “Failure” Success INT: And the whole thing, would you say that was a successful experiment? 2134: Yes. INT: Why would you say it was successful? 2134: Because we were able to compare the yields. Across the 4 plots, we were able to compare the yields to see which did better. And we found out that the 2:2 had a better yield than the 3:3. 4104: I saw that the maize yielded well, unlike just planting without using anything. Planting without anything, you get nothing. But planting with manure, at least you get something. INT: So do you think that this was a successful experiment that you tried? 4104: Yes, it’s a good technique because if you don’t have enough money to buy fertilizer, you can just use manure. INT: So do you think that that new spacing was a successful experiment…? 4110: Yes. INT: And why do you think so? 4110: Because previously, we were just planting plants one here, the other one over there [with large spaces between plants], without following any strategy. Instead of planting many seeds in a row, we were only planting a few seeds per row. Where before we were planting 3 seeds, now we’re planting 6, and now we’re harvesting more. Failure 1105: It has proved to be a failure because I’ve tried it twice [without harvesting anything]. 2301: …This season, I didn’t harvest anything. But last season, I got 3 bags. This season, ah! Nothing. Not even a bag. 3205: For the pigeonpea, we just planted it, and now the goats are eating it. So we haven’t seen any benefits from it. 4134: This season we didn’t harvest anything, and we won’t harvest anything, because the goats are eating the crop, as I already said [during the survey], because this variety is late maturing. So now goats and cattle have eaten up the pigeonpea. INT: So it’s all gone? 4134: Yes, we haven’t gotten anything. Farmers’ attitudes (satisfaction or dissatisfaction) did not predictably correspond with their experimental outcomes. Successful experiments were consistently associated with feelings of happiness or satisfaction, but farmers did not always equate failed experiments with dissatisfaction. On the contrary, it was common for an innovator to be satisfied with the 73 outcomes of a failed experiment if she felt that she had learned something in the process, or that the experiment’s failure could be attributed to another factor beyond the experimental crop or technique, itself (e.g. weather, personal health, etc.): INT: So were you happy that you planted pigeonpea? 1217: Yes. Although we didn’t eat anything. INT: Even though you’ve harvested nothing because of goats? 1217: Yes, the seed is appealing. That’s why we did not eat it [before planting it], but we were happy we grew it. 2301: I found it to be a good technique. Only the problem is that the rain stopped early. The stalks were healthy, indicating that we could have had large yield. But the rains stopped early, then the maize wilted. INT: So were you satisfied that you tried this this season? 2301: Yes, I’m very happy. 3126: No. Since the beginning we got nothing. INT: So are you happy that you tried that this year? 3126: Very much! INT: Why?! 3126: This just happened because of the rains. INT: Were you happy that you planted it? 4129: Yes. 74 INT: Why were you happy, even though you had no yield? 4129: Because I tried to grow it, just the way my friends did. It didn’t work well because I was in the hospital. At the conclusion of any experiment, an innovator had to determine if they would repeat, adjust and repeat (i.e. scale out, scale back, or make a change), or abandon their original experiment. A number of factors influenced this decision, including the interplay between the assessment (success or failure) of the experiment and the attitude (satisfaction or dissatisfaction) of the experimenter, as illustrated in the above quotes. Additionally, the future of an experiment was determined in part by a farmer’s access to resources. During the in-depth interviews, farmers identified certain resources as being influential during the experimentation process, including: landholdings (both total farm size and number of fields); input availability in local markets, from intervention projects, or through government subsidies (e.g. seed, fertilizer, pesticides, etc.); household income to purchase agricultural inputs; and available labor (which was closely related to health issues). Overall, farmers expressed that their choice to continue or abandon an experiment varies with every experiment, which is consistent with the theoretical frameworks supporting this study (Nitsch, 1990; Schön, 1983). There are several important trends, however, that emerged during the interviews. Firstly, although successful experiments always resulted in feelings of satisfaction, a failed experiment could result in either feelings of satisfaction or dissatisfaction. Secondly, when an experiment was deemed successful, a farmers’ access to resources had a great deal of influence over the farmer’s decision to continue or abandon an experiment. For example, one farmer had the desire to scale out (“adjust and repeat”) a successful experiment, but due to resource constraints (landholding size, in this case) she was only able to “repeat” the experiment: 75 INT: And will you plant more next year than you did this year? 1140: The field is just the same [size], every season we grow there so we won’t increase the area. The only difference is just the yield. Like for different seasons, we get different numbers of bags. But the area is just the same. INT: So if you had more land, would you want to plant more of this? 1140: Yes, I would increase the area. Likewise, another farmer learned through experimentation that she preferred one variety of groundnut over another, and because she had the resources (available seed) to plant her favorite variety the following season and she felt she had gained all she could from the comparison experiment, she only planted one variety the next season (and thus abandoned the comparison experiment): INT: And so that season did you plant the local and the hybrid? 4110: Yes. … INT: And you found that the yield from the hybrid was much better? 4110: Yes. INT: Any other differences that you noticed? 4110: The hybrid variety doesn’t spread across ridges. But for the local variety it spreads so much! So that it’s even difficult to harvest. INT: Anything else that was different? 4110: No, the yields only. It just spreads but it doesn’t yield that much. … INT: And so this season, did you decide to plant just the hybrid? 76 4110: Yes, hybrid only. INT: So are you happy that you tried the local that one season? 4110: Yes, because I’ve seen the bad and good sides of it. Resource availability was thus a crucial element of many farmers’ decision-making processes following successful experiments. In the event of a failed experiment, however, a farmer’s decision-making process was not dependent on her access to resources, alone, but on the interplay between her attitude (satisfaction or dissatisfaction) and her access to resources. A farmer who was dissatisfied with her failed experiment and had access to resources was not likely to “repeat” or “adjust and repeat” the experiment. It was common for dissatisfaction to lead to abandonment even if the innovator had access to the resources to retry an experiment, because farmers who were dissatisfied felt they had gained nothing (e.g. yield, knowledge, etc.) from an experiment and thus did not want to try again: INT: So did you decide to plant it again this year? 3126: No, we didn’t plant it. INT: Ah, why didn’t you grow it again? 3126: Because of what happened last year; we didn’t clearly see any benefits. And also because the soil type doesn’t suit well with cotton. A farmer with limited access to resources, however, sometimes had less freedom to abandon a failed experiment, despite their dissatisfaction, because repeating the experiment (in the hope that it would turn out better after a second attempt) was a safer option than abandoning it, as in the case of free or subsidized seed, for example: 77 INT: So do you expect to have any harvest from them? 1105: No, we won’t harvest anything. INT: So are you glad you planted pigeonpeas this year? 1105: No, I’m not happy. INT: So will you plant them again next year? 1105: If we receive it, we will plant it. INT: If you receive the seed next year, what will you do differently for the pigeonpeas? 1105: If the first rains will come early, then we will plant early. INT: So why did you decide to try a new variety of hybrid [last season]? 2312: Because DK 8033 [previous variety] was no longer available. But I liked it [DK8033] because it yielded a lot. INT: It was no longer available? 2312: Yes. And then because this [new Demeter variety] was for free, we then decided to just get it to cover the whole field. But it yields less than DK 8033 yielded. INT: So this season, did you go back to the first hybrid you were growing, or did you grow Demeter again? 2312: No, Demeter again. INT: Because the other seed still wasn’t available? 2312: Yes. Regardless of whether a farmer had the desire to repeat an experiment or abandon it, their decisions were largely influenced by their access to productive resources. Lack of access to resources is commonly seen as a barrier to experimentation, but in the case of farmers who are 78 forced to try unfamiliar seeds season after season—perhaps due to their dependence on volatile subsidies or their participation in revolving intervention projects—the lack of access to resources can actually be a driver of experimentation, although these experiments are not independently motivated. The influence of resource availability on experimentation will be further discussed in Section 5.3. 4.6 Drivers of On-Farm Experiments After considering the inconclusive results of the ANOVA test between experimenter groups (Section 4.3) and the insights gained from the qualitative data surrounding the influence of resources on experimentation (Sections 4.3; 4.5.3), we came back to the quantitative data and conducted a regression analysis to predict experimentation likelihood among our survey respondents. Binary logistic regression analyses are commonly used in studies that aim to understand the likelihood of technology adoption (e.g. Barungi et al., 2013), and as adoption is closely related to experimentation, we found a binary logistic analysis to be the most fitting regression for our data. Our dependent variable was binary, where Non-experimenters = 0 and Independent experimenters = 1, where experimentation was a function of: number of fields planted in 2012-2013 season; total field area; exposure to extension information; household head gender; Tropical Livestock Units; Wealth Index score; and the interactions of several of these variables. These predictor variables can be found in Table 20. 79 Table 20. Description of Predictor Variables Used in Logistic Regression Analysis Variable Name HH Head Gender Variable Description Gender of household head Measure 0 = Male 1 = Female Total Fields Farm Size Total fields held Total farm size (ha) – only for farmers holding 3 fields or less Count Hectares TLU Wealth Index Score Extension Tropical livestock units held Score on asset-based wealth index Extension advice received in 2012-2013 Weighted count Weighted count 0 = No 1 = Yes HH Head Gender_ Total Fields HH Head Gender_ Wealth Index Score Gender of household head by Total fields held Interaction Gender of household head by Score on assetbased wealth index Interaction Wealth Index Score_ Total Fields Score on asset-based wealth index by Total fields held Interaction Wealth Index Score2 Total Fields2 Score on asset-based wealth index (squared) Total fields held (squared) Interaction Interaction The binary logistic regression formula was expressed as: where logit(p) is a binary indicator variable that equals 1 if a farmer planted an unfamiliar crop or tried an unfamiliar technology independent of an agricultural intervention and zero otherwise; a is a constant of the equation; and b is the coefficient of the predictor variables. As the logistic regression was meant to analyze the drivers of independent experimentation, Followers and Independent experimenters (from the aforementioned classification structure) were coded as 1 (n = 241), and Non-experimenters and Project participants were coded as zero (n = 83). A test of the full model against a constant only model was statistically significant, indicating that the predictors as a set reliably distinguished between Independent experimenters 80 and Non-experimenters, where χ2 (11, N = 288) = 33.04, p = .001. Although Nagelkerke’s R2 indicated a relatively weak relationship between prediction and grouping (.159), prediction success overall was 74.3% (96.7% for non-experimenters and 10.7% for experimenters). The Wald criterion demonstrated that the following variables are significant determinants of experimentation: wealth index score (p < .05); extension advice (p < .001). Additionally, HH head gender displayed marginal significance (p = .056). These results can be found in Table 21. Table 21. Logistic Regression Analysis of Experimentation Independent Variable HH Head Gender Total Fields Farm Size TLU Wealth Index Score Extension HH Head Gender_ Total Fields HH Head Gender_ Wealth Index Score Wealth Index Score_ Total Fields Wealth Index Score2 Total Fields2 Constant B - 1.745 -.100 -.174 -.638 -.111 -1.160 S.E. .911 1.205 .322 .353 .048 .300 Wald 3.666 .007 .293 3.265 5.357 14.944 Sig. .056 .934 .588 .071 .021 .000 Exp(B) .175 .905 .840 .528 .895 .313 .472 .462 1.047 .306 1.603 .036 .030 1.391 .238 1.036 .034 .018 3.551 .060 1.035 .001 -.042 .446 .001 .303 1.174 3.362 .019 .144 .067 .890 .704 1.001 .959 Model χ2 (11) = 33.038, p = .001 Pseudo R2 = .159 N = 288 Note: The dependent variable in this analysis is coded so that 0 = non-experimenter and 1 = experimenter. For the significant predictor variables of wealth index score and extension, EXP(B) values were less than 1.0, indicating that as these predictor variables are raised by one unit, the 81 odds ratio becomes smaller, and therefore the likelihood of experimentation decreases. In other words, those households that had fewer assets or received no extension advice were more likely to experiment than households that had more assets or received some extension advice in the previous season. Regarding the marginal significance of household head gender, the EXP(B) value of this variable indicates that members of male-headed households were more likely to experiment independently than are members of female-headed households. This result was not altogether unexpected, given the nature of experimentation as—in part—an individual-level process, regardless of the gender of the household head (see Section 4.4).The implications of these results will be further discussed in Section 5.1. 4.7 Summary of Results Out of a total of 324 farmers surveyed, 228 (70.1% of total sample) reported conducting at least one experiment in the 2012-2013 season. Those 228 farmers elicited 572 examples of experimental crops, varieties, and/or techniques that they had tried both independently and through an intervention project. We learned in Section 4.2 that 89% of the reported experiments from 2012-2013 had involved new crops, varieties, or plant spacing techniques, and this trend held true for all three experimenter groups (Participants, Followers, and Independents). In Section 4.3, an ANOVA test revealed no significant differences in socioeconomic or farm-level means between experimenters and non-experimenters, or between farmers who conducted different types of experiments. These inconclusive findings were supplemented with qualitative data, where many farmers shared the opinion that starting a new experiment (or scaling out a successful experiment) required more land, which often meant renting in additional field(s). These insights helped us to understand that when challenged by limited resources—such 82 as fields held—innovative farmers felt they had two options: abandon their planned experiment or make an adjustment to the experiment’s design. Section 4.4 highlighted legume experiments, and we learned that 94% of experimental crops that farmers had tried in 2012-2013 were leguminous, and 57.1% of experimental varieties that farmers had tried in 2012-2013 were leguminous. In this section, we also examined the relationship between gender and legume experimentation, and found no statistically significant differences between men and women and their propensity to experiment with legumes. The frequency distributions from this test illustrated that the majority of experimental legume crops and varieties that were grown in 2012-2013 were planted by both spouses together (crops: 45.6% planted by both spouses; varieties: 55% planted by both spouses). When only one person reported planting an experimental legume, however, women planted over twice as many unfamiliar legume crops as men (36.3% compared to 17.6% planted by men) and over four times as many unfamiliar legume varieties as men (35.7% by women compared to 8.5% by men). Regarding all experiments (not just those involving legumes), similar trends emerged. The majority of experiments (with crops, varieties, and techniques/technologies) were conducted by both spouses together. Those experiments that were conducted by only one person, however, were more often attempted by solo females than by solo males: crop experiments: 35.6% by females, 18.6% by males; variety experiments: 37.6% by females, 7.5% by males; technical experiments: 31.6% by females, 14.5% by males. As we learned in Section 4.1, not all farmers thought of their actions as “trying new things” or “experimenting”, and the inconclusive ANOVA test in Section 4.3 also illustrated the difficulty of measuring experimentation quantitatively. The best way we found to overcome these challenges was to hold in-depth conversations with farmers. Therefore, Section 4.5 used 83 interview data to explore the motivations and methodologies behind smallholder experimentation. In Section 4.5.1, we learned that experimentation was driven by many motivators, including proactive, reactive, and external forces, or sometimes a combination of all three. Maize varietal experiments were more commonly driven by reactive forces (especially changes in rainfall patterns) than anything else, and other varietal/crop experiments were driven primarily by proactive and external forces (such as income generation, nutrition, participation in an intervention project, etc.). Technical experiments were driven by a range of motivators including, but not limited to, climate change and variability. These data suggested that farmers tried new maize varieties and some technical experiments because they were concerned about climate change. These findings were corroborated by the survey data, where 86.1% of farmers said they had noticed changes in the rains over the last 20 years, and out of those farmers, 67.2% reported that the changes they noticed included: less rain; erratic rainfall; and a tendency for the rains to stop before the crops had matured. When taken in combination, these results indicated that farmers were not only noticing climate change, but they were actively experimenting with maize varieties and new techniques in an effort to mitigate any undesirable changes. Whereas Section 4.5.1 dealt with the motivations behind experimentation, Section 4.5.2 focused on a farmer’s management strategy once they had decided to conduct an experiment. Through examination of the qualitative data, we learned that most farmers conducted their experiments differently, where the only common practice among respondents was the propensity to start an experiment on a small scale. Regarding the use of a “control” or comparison, we found that some farmers used a simultaneous control, some farmers used a historical control, and some farmers did not use any kind of comparison in their experiments. Likewise, we found 84 that where some farmers tried experimental crops/varieties on a separate plot from their traditional crops, some farmers preferred to intercrop unfamiliar crops/varieties with their traditional ones. Overall, we found that although farmers used a variety of management techniques when trying an experimental crop, variety, or technique/technology, whichever method a farmer used had a specific purpose and reason behind it. Section 4.5.3 looked at the end results of farmers’ experiments, specifically regarding the interaction between a farmer’s assessment of an experiment (success or failure), attitude about the experimental outcomes (satisfied or dissatisfied), access to resources, and future intentions for the experiment. In general, successful experiments were associated with feelings of happiness or satisfaction, and the future of a successful experiment was closely tied to a farmer’s access to resources. Failed experiments, contrastingly, could result in either satisfaction or dissatisfaction, and the future of a failed experiment relied upon both a farmer’s attitude and their access to resources. After considering the differences between experimenters, the motivations that drive a farmer to trying something new, and the emergent relationship between resources and experimentation, in Section 4.6 we built a model to predict the likelihood of experimentation and tested it using a binary logistic regression analysis. Prediction success overall was 74.3% (96.7% for non-experimenters and 10.7% for experimenters), and the Wald criterion suggested that those households that had fewer assets or received no extension advice in the previous season were more likely to experiment than households that had more assets or received some extension advice in the previous season. The logistic regression results run contrary to some of the findings reported in Sections 4.3 and 4.5. These apparently contradictory findings, as well as the previously outlined results and their implications, will be discussed in detail in Section 5.0. 85 5.0 Discussion and Recommendations for Further Research 5.1 Farmers Who Experiment In total, our study identified 228 experimenting farmers, or 70.1% of our total sample. While we cannot claim, therefore, that all farmers try new things as a matter of course, we can discern that smallholder experimentation is widespread, especially when we consider that many examples of experimentation may have been inadvertently omitted from our study due to the epistemological and linguistic differences that exist between smallholders and researchers in the formal sector. We wanted to understand the differences across experimenting farmers, specifically the gendered division of labor on experimental plots. Intra-household frequency distributions illustrated that the majority of experiments (with crops, varieties, and techniques) were conducted by both spouses together, but that those experiments that were undertaken by only one person were more often attempted by solo females than by solo males: crop experiments: 35.6% by females, 18.6% by males; variety experiments: 37.6% by females, 7.5% by males; technical experiments: 31.6% by females, 14.5% by males. Note that these trends hold true when even experiments are disaggregated by crop categories (e.g. legume experimentation), where the majority of experimental legume crops and varieties that were grown in 2012-2013 were planted by both spouses together (crops: 45.6% planted by both spouses; varieties: 55% planted by both spouses). In instances where only one person reported planting an experimental legume crop, however, women planted over twice as many unfamiliar legume crops as men (36.3% compared to 17.6% planted by men), and over four times as many experimental legume varieties as men (35.7% by women compared to 8.5% by men). 86 These data call into question the commonly-held notion that legumes are “women’s crops”, as legume experimentation was reportedly conducted by both spouses in the majority of cases. As a whole, however, when labor was divided between spouses, experimentation (with legumes, but also in general) was undertaken mainly by women. Even in instances where both spouses conducted experiments together, it is likely that women’s priorities were as influential as men’s in shaping the experiment. We will discuss some of the motivations that may have been driving these women to experiment in Section 5.2. Next, we differentiated between experimenters according to the source of their ideas, where those farmers who had only tried something new as part of a project or at the advice of an extension officer were called Project participants (n = 145), those farmers who tried at least one experiment that they had seed from a peer and replicated were called Followers (n = 64), and those farmers who had tried at least one experiment that had spawned from their own minds were called Independents (n = 19). Approximately one-third of our sample population reported participation in an agriculture or non-profit project in 2012-2013, and 55% of our sample population reported receiving extension advice in the same season. Due to this effectual saturation of new ideas into a very small and densely populated area, it was thus difficult to measure truly independent experimentation, which is why we used the aforementioned classification structure. We measured group differences across the four experimentation classifications using an ANOVA test. The results of this test demonstrated that experimenting farmers came from a wide range of socioeconomic backgrounds and farm types, and that there were no statistically significant differences across experimenters, or between farmers who tried something new in 2012-2013 and those who did not. This inconclusive statistical finding led us back to the in-depth 87 interview data, where we learned that many farmers felt their capacity to experiment hinged on their access to land. If their access to land was limited, many farmers felt that they either had to abandon their experimental plans or creatively adjust them if possible (e.g. intercrop an experimental crop with a traditional crop instead of planting each as sole crops). Finally, we wanted to gain a better understanding of a farmer’s likelihood to try something new without being guided by an intervention project or extension officer. To accomplish this goal, we used a binary logistic regression test where farmers who had tried at least one experiment on their own (either replicating a peer’s experiment or trying out their own idea) were coded as “1”, and farmers who did not experiment or only experimented with project/extension guidance were coded as “0”. Unlike the ANOVA, the regression did not measure all types of experimentation. Rather, the regression attempted to look more closely at those experiments that farmers conducted independent from intervention projects/extension advice. Due to the skewed values of the binary dependent variable in our logistic regression model (non-experimenters: n = 241; independent experimenters: n = 83), the results of this analysis should be interpreted with caution. Future analyses of this experimenter classification structure would benefit from using a multinomial logistic regression to estimate the determinants of group membership across all four experimenter categories, rather than only using a binary structure. That being said, the binary regression yielded some interesting results that warrant discussion. The logistic regression demonstrated that the likelihood of independent experimentation was greater for households that owned fewer assets or received no extension advice in the previous season. Those households that held more assets or received some extension advice in 88 the previous season were less likely to experiment, or they only tried new things as they had been advised by an intervention project or extension officer. The logistic regression results were surprising, in that they contradicted our theory (which was grounded in literature) that farmers who have fewer assets and/or less physical capital (e.g. landholdings, livestock, etc.)—and who would thus have less resilient farm systems and livelihoods—would be less likely to experiment with new crops, varieties, and technologies for want of resources (e.g. “experimental” plots of land) and/or for fear of the opportunity cost that might accompany a failed experiment. Why did the regression analysis conclude, then, that resource-poor farmers would be more likely to experiment independently, while farmers with greater access to resources would be less likely to experiment independently? Perhaps these results reflect the tenacity of resource-poor farmers: If they have the desire to create a change or they need to solve a problem in their farm system but they do not have the resources or they cannot access an extension agent, then they will help themselves by experimenting independently. We spoke with many farmers who had problems with poor soil fertility, for example, but did not have the capital to purchase large amounts of fertilizer so they actively experimented with more accessible alternatives to build up their soil fertility (e.g. manure, compost, crop rotations, etc.). Although experimentation is certainly influenced by resource availability (as we will discuss in Section 5.2), it is not dependent upon it. During the in-depth interviews, we heard from several resource-poor farmers who attempted to cope with the changing rainfall patterns not by purchasing hybrid seed (which they could neither find in local markets nor afford to buy), but by experimenting with planting times and/or seed spacings to maximize their crops’ water use efficiency. These farmers conducted independent, limited-resource experiments because they had 89 no other way to improve their farming systems. In light of these qualitative findings, it is not altogether surprising that the logistic regression told us that farmers with fewer assets and less extension advice were more likely to conduct independent experiments. Along the same lines, perhaps those farmers with more assets or greater extension access were less likely to experiment independently because they had access to “expert” advice and they preferred to follow those recommendations as opposed to experimenting on their own. Unfortunately, the logistic regression does not tell us whether access to extension advice encouraged expert-guided experimentation, or inhibited independent experimentation among smallholders. We also need to consider the coding of the dependent variable in the logistic regression. Remember that those farmers who were not experimenting or who were only experimenting as part of an intervention project/with extension advice were coded as “0”, and those farmers who were experimenting independently were coded as “1”. Perhaps the regression results thus reflected the unintended exclusion of farmers with fewer resources and less extension access from intervention projects such as Africa RISING. This interpretation is made more valid by the proximity of the Africa RISING mother trial plots to main roads—where farmers with more assets (e.g. bicycles, cell phones) would be better able to travel to the mother trial plots to participate in work days. Additionally, Africa RISING participants were recruited by extension agents, so those farmers who previously had regular contact with an extension agent would have been more likely to hear about the project than those farmers who did not have access to extension. If the regression results are indeed a reflection of the socioeconomic status or extension access of intervention project participants, then it would seem that these projects are missing their mark by inadvertantly excluding resource-poor and/or information-poor farmers. 90 One aspect of the logistic regression remains to be addressed: prediction success overall was 74.3%, but there was a stark difference between prediction success of non-experimenters (96.7%) and prediction success of experimenters (10.7%). Why does this disparity exist? Personality type and natural curiosity are two factors that undoubtedly influence a farmer’s propensity to experiment, but that would be difficult to measure and that we neglected to measure in this study. These personal characteristics might have confounded the predictive ability of our logistic regression, which could explain the disparity in the prediction success percentages. Additionally, as we will discuss in Section 5.2, the motivations that drove every experiment were different, and farmers often had multiple goals in mind for every experiment they conducted. Experimentation probabilities may vary not only by socioeconomic and farmlevel characteristics, but also by personality traits and individual motivations. These nuances in experimentation likelihood cannot easily be captured quantitatively, which lends even more value to the range of mixed methods (i.e. survey data, in-depth conversations, field observations) used in this study. Ultimately, we have learned throughout this study that household surveys are useful tools for gleaning demographic information, but they do not yield the most reliable data when the subject of study is nuanced, intricate, and highly individualized, such as smallholder experimentation. It is difficult, therefore, to extrapolate concrete conclusions based on statistical analyses using our household survey data, such as the binary logistic regression and ANOVA tests. The qualitative insights we gained from the in-depth interviews are thus of critical importance in helping us to understand the ways in which farmers try new things, their experimental priorities and preferences, and their motivations for experimenting. 91 5.2 Experimental Crops, Varieties, and Techniques Overall, we found that 89% of experiments reported in 2012-2013 fell into three categories: new crops, 34%; new varieties (maize and non-maize), 40%; and plant spacing experiments, 15%. When experiment examples were disaggregated according to the experimentation classification categories of Participants, Followers, and Independents, similar trends emerged in the frequency distributions. Across all three groups, experimenting farmers had a propensity to try new crops, varieties, and plant spacing techniques more often than other types of experiments. It is important to note that the examples and frequencies of farmers’ experiments we reported are in line with the findings of similar studies of farmer experimentation (Scoones and Thompson, 1994a; Sumberg and Okali, 1997). Additionally, we found that 94% of the experimental crops and 57.1% of the experimental varieties that farmers had tried in 2012-2013 were leguminous. These data illustrate that regardless of the source of an experimenter’s ideas, smallholders had specific interest in experimenting with leguminous crops and varieties primarily, and with maize varieties secondarily (42.9% of experimental varieties tried in 2012-2013 were maize). Also remember from Section 5.1 that the majority of on-farm experiments were being conducted either by both spouses together, or by women alone, meaning that women’s priorities likely had a large impact on what kinds of experiments were being tried. When taken in combination with our qualitative data, these similarities across experiments can give us insight into farmers’ interests, concerns, and priorities. From our qualitative data, we learned that many farmers were reactively motivated to experiment with new maize varieties, especially in relation to changes in rainfall patterns. When we also consider that 86.1% of farmers said they had noticed changes in the rains over the last 20 years, and out of those farmers, 67.2% reported that the changes they noticed included: less rain; 92 erratic rainfall; or a tendency for the rains to stop before the crops had matured, we can understand why almost half of the varietal experiments reported were maize-based. Most farmers in our study had noticed rainfall changes and were concerned about how their livelihoods would be impacted by those changes. In light of their observations, concerned farmers actively experimented with maize varieties (and new techniques, in some instances) in an effort to mitigate the undesirable effects of climate change on their staple food crop. Contrary to maize, other crop and varietal experiments (which primarily consisted of leguminous plants) were found to be driven primarily by proactive and external forces (such as income generation, nutrition, participation in an intervention project, etc.). Where maize is the staple crop in central Malawi and farmers were primarily concerned with maintaining maize growth in the face of climate change, many farmers reported experimentation with legume crops in order to meet other, more diverse goals. Some farmers tried cowpea because it seemed to be a viable cash crop alternative to cotton or tobacco. Other farmers reported that after trying soya for only one season, they had already begun to notice positive diet-related changes in their children. Still others grew pigeonpea to promote soil fertility in the face of increasing synthetic fertilizer costs. Farmers who experimented with legumes were motivated by a wide range of priorities and concerns. Additionally, experimenting farmers (especially women) seemed very interested in trying out new legume crops or varieties if they thought those new plants would meet multiple goals. 5.3 Managing an Experiment from Start to Finish Just as farmers expressed various motivations that drove their experimentation, most farmers used different methods to conduct their experiments, and each method was driven by a 93 specific goal. Some farmers who wanted to compare a traditional variety to an experimental one grew the two simultaneously, using the same plant spacings, fertility measures, and so on, so as to have a stable comparison or “control” by which to measure the growth of the experiment. Other farmers based their assessment of new techniques on their past experience with traditional techniques, a sort of “historical control”. Still other farmers did not use any obvious control, and inserted experimental crops directly into their existing farm system, for example intercropping a new crop with a traditional one, so they could see how unfamiliar plants would interact with the system as a whole. Overall, farmers used a variety of different methods when trying new things, and behind every method was an intentional decision and a desired outcome. One of the only similarities across experimenters was that many farmers preferred to start their experiments on a small scale. These findings indicate that farmer experimentation is not haphazard, but rather onfarm experiments were carefully planned and implemented so that the farmer could gain the most knowledge (along with other benefits) from only one trial with a new crop, variety, or technique. After the conclusion of an experiment, farmers go through a complex decision-making process to determine whether they will repeat an experiment, adjust and repeat it, or abandon it. Most farmers reported being satisfied with the outcomes of their experiments, regardless of whether the experiment succeeded or failed (according to farmers’ own definitions), and many farmers only expressed dissatisfaction with an experiment if they felt they had gained absolutely nothing from it (e.g. knowledge, good yield, income, etc.). Depending on the interplay between a farmers’ assessment of their experiment, their attitude about the experiment, and the resources available to them, they would decide whether to repeat, adjust and repeat, or abandon the experiment. 94 The interviews we had with farmers validated our logistic regression findings, where a farmer’s decision to experiment was largely related to their access to resources such as land, inputs (e.g. seed), and household income. Some farmers who were satisfied with their experiments did not have the resources to try them again, at least not without making some adjustments to the experimental design. Similarly, some farmers who were dissatisfied with their experiments were so dependent upon subsidies or intervention projects that they had to repeat the experiment (e.g. grow the undesirable crop again), because free seed was better than nothing. Farmers who had access to adequate resources, however, were the most at liberty to repeat, adjust and repeat, or abandon an experiment depending on their assessment of and attitude toward a new crop, variety, or technique. These results tell us that general experimentation (i.e. the decision “to try or not to try”) is not dependent upon resources, because even a farmer who has extremely limited resources can try a new crop through an intervention project, or can experiment with different plant spacings or planting times without overreaching their means. The decision-making process after the completion of an experiment, however, is strongly influenced by a farmer’s resources, as is a farmer’s ability to try a specific experiment (e.g. for a farmer who sees her neighbor trying a new crop, but cannot afford to purchase the seed). We therefore need to better understand how resource-poor farmers shape their experiments from season to season compared to farmers with available resources. 5.4 Smallholder Experiments and the Theory of Reflection-in-Action Schön’s theory of Reflection-in-Action (1983) asserts that experimentation and decisionmaking after an experiment’s conclusion are dependent upon the interaction of a complex situation, an innovator’s knowledge and experience, and the innovator’s perceptions of the 95 experimental outcomes. Based on the interactions of these factors, the theory posits that after dissatisfactory experiment, a farmer may critique her experimental design, make adjustments, and attempt the experiment again. Alternatively, she may abandon the experiment. In the case of a satisfactory experiment, a farmer may choose to scale-out the innovation and/or repeat it in subsequent seasons. While these theoretical conclusions are not incorrect, they are over-simplified. We have seen that there are several other important factors that shape a farmer’s decision-making processes prior to, during, and after an experiment, including: access to resources (physical capital such as seed and land; social capital such as involvement in an intervention project or access to extension advice) and priorities and concerns (which are reflective of gender, considering that women participated in over 80% of crop experiments, over 90% of varietal experiments, and over 85% of technical experiments either alone or with their spouses). Farmers’ physical capital, social capital, and gender wield significant influence on their knowledge base and past experiences, and in turn over their experimentation processes. Likewise, farmers’ goals and priorities will influence their perceptions of an experiment’s outcome. The experimentation process, therefore, cannot be generalized for all persons as it is in Schön’s (1983) theory of Reflection-in-Action. Rather, the decision-making process associated with every experiment is dependent upon the characteristics of the experimenters, themselves. 5.5 Implications for Development The primary aims of this study were to gain an understanding of the characteristics of experimenting smallholders, to learn why they try new things, and to discover how they prefer to conduct experiments. Throughout the course of the project’s design, fieldwork, analysis, and dissemination of results, we have attempted to recognize and congratulate smallholders for the 96 progress they have made through agricultural experimentation. By studying on-farm experimentation, we have also learned several important lessons that we hope will influence future research and development projects in Malawi and the surrounding region, so that development practitioners might be design more effective and sustainable interventions around current farmer practices and preferences. Firstly, we learned that the likelihood of experimentation that was independent from interventions and extension was higher for male-headed and resource-poor households. While these results suggest tenacity among independent experimenters, they also suggest that those resource-poor farmers were experimenting independently because they had been inadvertently overlooked by intervention projects and extension agents. If this is the case, the harbingers of “expert knowledge” could be leaving behind some of the most vulnerable and marginalized smallholders in Malawi. Future projects should actively and deliberately include a certain population of resource-poor farmers for the benefit of both parties—farmers who had less were more likely to try new things on their own, and if these farmers were to be involved in an intervention project they could bring fresh insight and experience to share with other participants. We also learned that women were actively involved in more than 80% of all reported experiments in 2012-2013, and therefore women’s concerns and priorities were likely crucial in shaping the types of experiments that were conducted on-farm. Many of those farmers who experimented with new maize varieties did so out of concern about climate change, and many of those farmers who experimented with legumes were hoping to meet multiple goals by growing leguminous crops. These findings indicate that before promoting a certain crop or variety in any given area, projects should firstly take inventory of the priorities and concerns of local farmers 97 (especially women), and only after these priorities are taken into consideration should intervention projects introduce new crops, varieties, and/or techniques. Finally, we learned that in their own experiments, farmers use an array of intentional but varied methods. Most farmers preferred to try new things on a small scale, so perhaps intervention projects should follow suit. The structure of Africa RISING’s mother-baby trials is a sound example of using small experimental plots that are appealing to many farmers. Otherwise, however, many farmers differed in their use of a “control” and their intercropping or solecropping preferences, depending on the goals they wanted to meet. What would happen if a project introduced a new crop or technique and then left farmers to experiment with it using their own methods? Perhaps an altogether new innovation would be born from the experience, or perhaps the exercise would facilitate farmer-to-farmer learning rather than encouraging dependence on outside knowledge. Designing a project that integrates formal science with local knowledge is neither quick nor easy. It is likely, however, that projects which incorporate the methodologies used by farmers in their own independent experiments will be better equipped to help farmers achieve their goals while simultaneously respecting their achievements. It is clear that smallholders have a great capacity for experimentation, and their knowledge, experience, preferences, and priorities—if properly understood and incorporated—could be benefit both future agricultural development projects and their participants. 5.6 Limitations and Suggestions for Future Research Although this study yielded an ample amount of quantitative and qualitative data and provided extensive insight into smallholder experimentation in Malawi, there were limitations to the study and there are questions about farmer experiments that have yet to be answered. 98 Through a combination of qualitative and quantitative data analysis, we learned that experimentation is influenced by a farmer’s productive resources such as landholdings, input availability and access, asset ownership, and access to information. While we have a basic understanding of these relationships, we need to learn more about the experiments of resourcepoor farmers compared to those farmers who have better access to resources. For example: Do resource poor farmers try more technical experiments than crop/varietal experiments because seed is expensive or difficult to access? Likewise, do farmers with available resources try more crop/varietal experiments because they have the means to access seed? Is there a level of wealth where experimentation shifts from dependent (i.e. conducted in partnership with an intervention project or extension officer) to independent? These economic questions are important to consider in future research of on-farm experimentation. Additionally, many respondents in our study were asked to detail experiments from several months or years prior to the interviews, and therefore recall error (along with the epistemological and linguistic differences addressed in Section 4.1) may have resulted in the omission of some cases of experimentation from our study. To circumvent similar issues in future experimentation studies, we first suggest conducting in-depth interviews with multiple self-identified “non-experimenters” to understand whether these farmers really were not trying new things, or if they were experimenting but failed to report their attempts due to a miscommunication. Unfortunately, we did not interview any non-experimenters in this study, and their perspectives may have provided some insight into the barriers or challenges of experimentation. We also suggest that future experimentation studies implement a mixedmethods longitudinal design that begins by recording farmers’ intended experiments (and their hopes and goals for those experiments) before planting, measures the progress of and 99 modifications to those experiments throughout the growing season, details the post-harvest outcomes (both agronomic and attitudinal), and finally records farmers’ modifications to their experimental designs in the following season. Such a study would provide a comprehensive picture of on-farm experimentation from start to finish, and would drastically reduce recall error and miscommunications between researchers and farmers. An in-depth, long-term study would also help us to better understand if the experiments that farmers are trying are true representations of their interests, concerns, and priorities, or if most of these common experiments are the result of convenience, more than anything. In other words, are farmers commonly experimenting with new legume varieties, for example, because they want to grow a new legume, or because a new legume was available to them and was the only thing they were able to try? While our study focused mainly on experiments that farmers have already conducted, there is much to be learned about farmer priorities by asking farmers about the experiments they wanted to try but could not. Finally, we learned during our fieldwork that farmers in central Malawi, are bombarded with new ideas from all directions—extension officers, radio advertisements, subsidy programs, seed distributors, intervention projects, and peers, to name only a few information sources. Due to this effectual saturation of new ideas into the region, it was difficult to measure truly independent experimentation. This concern leads us to wonder what this study would have found, had it been conducted in a less information-saturated place, or in a less densely populated place where ideas do not flow as rapidly between areas. Previous studies have found that farmers worldwide are actively experimenting, just like those farmers in central Malawi, but what kinds of crops, varieties, and techniques are farmers in less densely populated areas trying? By 100 conducting similar studies across the world, we can better understand and support on-farm innovations and farmer capacities on a global scale. 5.7 Conclusions This study has vividly illustrated that smallholder farmers in central Malawi are capable of agricultural experimentation, and are actively working to improve their livelihoods and farm systems by trying new crops, varieties, and/or techniques on their farms. These experiments are deliberately planned and executed by tenacious smallholders whose decisions are influenced by their gender, their access to information, and their available resources. Women and resourcepoor farmers are particularly rich repositories of local knowledge based on a multitude of onfarm experiments. The methods, motivations, and accomplishments of experimenting smallholders, however, are underrepresented in the existing body of agriculture and development literature. This research provides valuable insight on the socioeconomic characteristics of innovators and the drivers of smallholder experimentation, and these insights will bolster the relatively scant literature that currently surrounds agricultural experimentation. Additionally, the findings of this study attempt to validate the experimentation processes of innovative smallholders to the research, extension, and development communities. Malawian farmers are trying new things in conjunction with experts, but they are also bolstering their own expertise by experimenting independently. The details of what smallholders are trying, why they want to try new things, and how they prefer to conduct experiments provide valuable insight into the decision-making processes and priorities of Malawian farmers. Future intervention projects that are designed around these decision-making processes and build on farmers’ expertise and 101 priorities are likely to result in more relevant, readily adopted strategies for improving Malawian agriculture. Regarding extension opportunities, this study found that the most vulnerable smallholders (those who held fewer assets or received no extension advice in the previous year) were more likely to experiment than households that were less vulnerable (e.g. had more assets or received some extension advice in the previous season). The relevance of extension information could therefore be strengthened if extension agents worked to target the most vulnerable populations (including women within male-headed households), learn about their experimentation processes, and incorporate that local knowledge into future extension projects and share it among other farmers who are not experimenting. Most importantly, the findings of this study demonstrate the numerous ways in which Malawian smallholders are working to solve their own agricultural problems through the creative combination of local knowledge and new tools and information. By widely disseminating our findings, we will encourage innovative smallholders in their efforts and also make known the accomplishments of these innovators to non-experimenting farmers in the same areas. We hope that by encouraging experimentation among smallholders, those farmers who are hesitant to experiment independently (i.e. without the guidance of an extension agent, intervention project, etc.) will begin to see the innovators in their own communities as accessible resources for agricultural advice and collaboration. As agriculture is the lifeblood for the majority of Malawians, the country’s current agricultural situation—low yields, poor soil fertility, and overcrowded arable land—warrants immediate, creative solutions. Fortunately, agricultural experts are actively working to sustainably increase yields, boost soil fertility, and improve household nutrition. These experts 102 include extension agents, international research and development programs, and innovative smallholder farmers. The successful collaboration of these expert groups has great potential to yield solutions to the challenges facing Malawian farmers, but first the ideas and innovations of smallholders must be understood, validated, and integrated into the development paradigm. 103 APPENDICES 104 APPENDIX A Household Survey Instrument 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 APPENDIX B In-depth Interview Question Guide 124 Farmer Interview Questions HHID: Date of Interview: ________________ Oral Consent Received?_____________ District: ___________________ EPA: ___________________________ Sub- Village: ____________________ Head Village: _____________________ GPS Coordinates: (S) _______________________; (E) ________________________ Name of Respondent: _____________________________________________________ Gender of Respondent: ______________ Age of Respondent: __________________ Household head? _________________ Household type (MHH/FHH): __________ Did Respondent participate in Africa RISING this season? _____________ “Experimentation” Processes and Attitudes M: First, I want to talk with you about the ways in which you try out new things on your farm. 1. To you, what does it mean to “experiment” with new crops, techniques, or technologies? 2. Do you do “experiments” on your farm? 3. When you want to try out something new on your farm, how do you do it? a. Why do you do it this way? 4. Why do you try new things on your farm, and how often do you try something new? 5. (For RISING participants): Besides Africa RISING, have you ever participated in another agriculture project? a. Tell me about your participation with Africa RISING and with the other project: 6. (For non-RISING participants): Have you ever participated in an agriculture project? (If “No”, skip to next section) a. Tell me about your participation with this project: 125 7. Before you started participating in RISING/__________ agriculture project, did you try new things on your farm? a. If so, how did you go about trying new things? 8. Did working with RISING/__________ agriculture project motivate you to try new things on your farm more often? 9. Did working with RISING/___________ agriculture project give you any new ideas of how to go about trying new things on your farm? 10. Do you think you farm differently now than before you started participating in RISING/ ________agriculture project? Crops: Last Season (2011-2012) thru This Season (2012-2013) thru Next Season (2013-2014) M: These questions are about the crops/varieties you grew last season for the first time. 11. Did you plant any new crops/varieties last season for the first time ever? 12. Why did you decide to plant this crop/variety? Possible probes: Where did you hear about it? How much did you know about it prior to planting? How did you get seed for it, and how much did you plant? 13. Tell me about how you planted, managed, and used this crop/variety last season: Possible probes: Where, when, and how did you plant? Tell me about your management practices and any changes in those practices: How did you use it post-harvest? 14. Overall, were you pleased with the crop/variety? 15. Did you decide to grow it again this season? (If “No”, skip to next section after probe) a. Why or why not? 126 16. Tell me about how you planted, managed, and used this crop/variety this season: Possible probes: Did you do anything differently this season? How did those changes seem to affect the crop/variety? How did you get seed for it, and how much did you plant? Where, when, and how did you plant? Tell me about your management practices and any changes in those practices: How did you use it (or how do you intend to use it) post-harvest? 17. Overall, were you pleased that you grew the crop/variety again? 18. Are you planning to grow it again next season? (If “No”, skip to next section after probe) a. Why or why not? 19. Will you do anything differently next season? Crops: This Season (2012-2013) thru Next Season (2013-2014) M: These questions are about the new crops/varieties you grew this season for the first time. 20. Did you plant any new crops/varieties this season for the first time ever? 21. Why did you decide to plant this crop/variety? Possible probes: Where did you hear about it? How much did you know about it prior to planting? How did you get seed for it, and how much did you plant? 22. Tell me about how you planted, managed, and used this crop/variety this season: Possible probes: Where, when, and how did you plant? Tell me about your management practices and any changes in those practices: How did you use it (or how do you intend to use it) post-harvest? 23. Overall, were you pleased with the crop/variety? 24. Are you planning to grow it again next season? a. Why or why not? (If “No”, skip to next section after probe) 127 25. Will you do anything differently next season? Techniques and Technologies: This Season (2012-2013) thru Next Season (2013-2014) M: These next questions are about the new techniques/technologies you tried on your farm for the first time this season. 26. Tell me about the new techniques or technologies you tried this season for the first time: 27. Why did you decide to try this technique/technology? Possible probes: Where did you hear about it? How much did you know about it before you tried it? 28. What was the result of this technique/technology? Possible probes: Did it work as you expected or cause any unexpected problems? Would you say this technique/technology was “successful”? 29. Overall, were you pleased with the technique/technology? 30. Are you planning to try it again next season? (If “No”, skip to next section after probe) a. Why or why not? 31. How will you adjust it for next season? Techniques and Technologies: Not new to the Respondent, but novel or unorthodox in the area (The following questions only apply to certain Respondents) M: From the survey we did with you a few weeks ago, I learned that you are using a very unique technique/technology on your farm, and that you’ve been using it for several years. I’d like to ask you a few questions about that technique/technology. 32. Please describe the technique/technology in detail: 33. For how long have you been using this technique/technology? 34. Why did you first decide to try this technique/technology? 128 Possible probes: Where did you hear about it? How much did you know about it before you tried it? 35. Over the years, how have you adjusted it? 36. Why have you kept using this technique/technology? 37. Have you tried using this technique/technology on other parts of your farm/with other crops? 38. Have you taught this technique/technology to anyone else? M: This is the end of the interview. Thank you for taking the time to speak with me today. Do you have any questions? 129 APPENDIX C Qualitative Coding Audit Trail 130 Table 22. Qualitative Coding Audit Trail Node Name Description Themes Thematic coding done by hand Dissatisfaction R was NOT satisfied with the results of an experiment Experimentation Types Different examples of experimenting--this node is mainly a heading folder, as specific experiments will be detailed in subnodes Instances where R experimented with legume crop or variety (this node is where instances of legume experimentation are listed, but NOT methods, source of ideas, etc.) Results of legume experiments (how 2 compared to each other, the yield, etc.), also how the crop was used post-harvest Where R got idea for experiment (this node is mostly a place holder, as all sources will be coded in sub-nodes under this heading) R experimented with legume that they saw or that was suggested elsewhere and copied or adapted it (from peers, lead farmers, family, cheifs, etc.) Experiment was Rs own idea (including radio) Legumes Results- Legumes Source- Legumes Copy- Legumes Independent- Legumes Promoted- Legumes Maize Results- Maize Source- Maize Copy- Maize Independent- Maize Promoted- Maize Other crop or variety Results- Other Source- Other Copy- Other Independent- Other R experimented with legume that was actively promoted by an AEDO, non-profit (e.g. RISING), subsidy program, etc. The experiment was placed in their hands! Instances where R experimented with maize crop or variety (this node is where instances of maize experimentation are listed, but NOT methods, source of ideas, etc.) Results of maize experiments (how 2 compared to each other, the yield, etc.) Where R got idea for experiment (this node is mostly a place holder, as all sources will be coded in sub-nodes under this heading) R experimented with maize that they saw or that was suggested elsewhere and copied or adapted it (from peers, lead farmers, family, cheifs, etc.) Experiment was Rs own idea (including ideas they got from radio but never knew anyone personally who had tried it) R experimented with maize that was actively promoted by an AEDO, non-profit (e.g. RISING), subsidy program, etc. The experiment was placed in their hands! Instances where R experimented with other non-legume and nonmaize crop or variety (this node is where instances of other crop or variety experimentation are listed, but NOT methods, source of ideas, etc.) Results of other crop or variety experiments (how 2 compared to each other, the yield, etc.) Where R got idea for experiment (this node is mostly a place holder, as all sources will be coded in sub-nodes under this heading) R experimented with other crop or variety that they saw or that was suggested elsewhere and copied or adapted it (from peers, lead farmers, family, cheifs, etc.) Experiment was Rs own idea 131 Table 22. (cont’d) Promoted- Other Tech Results- Tech Source- Tech Copy- Tech Independent- Tech Promoted- Tech Failure Memorable Quotes Methods R experimented with other crop or variety that was actively promoted by an AEDO, non-profit (e.g. RISING), subsidy program, etc. The experiment was placed in their hands! Instances where R experimented with technique or technology (this node is where instances of tech experimentation are listed, but NOT methods, source of ideas, etc.)----- IF R experimented with a new crop or var AND a new tech, code to both places (2 experiments!) Results of tech experiments (how 2 compared to each other, influence on the yield, etc.) Where R got idea for experiment (this node is mostly a place holder, as all sources will be coded in sub-nodes under this heading) R experimented with tech that they saw or that was suggested elsewhere and copied or adapted it (from peers, lead farmers, family, cheifs, etc.) Experiment was Rs own idea (including radio) R experimented with tech that was actively promoted by an AEDO, non-profit (e.g. RISING), subsidy program, etc. The experiment was placed in their hands! R claims that an experiment failed (and reasons why they think so); R didn't plant a crop again or use a tech again because the result was poor To pull for thesis RISING Experimental methods used -- field size, amount planted, sole or intercrop, comparitive planting techniques between 2 plants Reasons Rs name for why they experiment (or barriers to experimentation) R discusses working with Africa RISING Satisfaction R was satisfied with results of experiment Self-identify WFC Fields Answers to the question: Do you do experiments on your farm? [yes and no answers should be coded here] and also Did you do experiments before working with an agriculture project (e.g. RISING)? [again, include both yes and no answers] R claims that an experiment was successful (and reasons why they think so) Rs theories about why things are the way they are, why an experiment resulted the way it did, etc. Word Freq Count of “field” or “field”-related words in all interviews WFC Land Word Freq Count of “land” or “land”-related words Motivation Success Theories 132 WORKS CITED 133 WORKS CITED Asfaw, S., Shiferaw, B., Simtowe, F., Lipper, L. (2012). Impact of modern agricultural technologies on smallholder welfare: Evidence from Tanzania and Ethiopia. Food Policy 37: 283–295. Barungi, M., Ng’ong’ola, D.H., Edriss, A., Mugisha, J., Waithaka, M., Tukahirwa, J. (2013). Factors Influencing the Adoption of Soil Erosion Control Technologies by Farmers along the Slopes of Mt. Elgon in Eastern Uganda. Journal of Sustainable Development 6(2): 925. Balée, W. (1994). Footprints of the Forest: Ka’apor Ethno-botany – The Historical Ecology of Plant Utilization by an Amazonian People. New York: Colombia University Press. Bentley, J.W. (2006). Folk experiments. Agriculture and Human Values 23: 451-462. Bezner Kerr, R. (2005). Food Security in Northern Malawi: Gender, Kinship Relations and Entitlements in Historical Context. Journal of Southern African Studies 31(1): 53-74. Bezner Kerr, R. (2008). Gender and Agrarian Inequality at the Local Scale. In S.S. Snapp & B. Pound (Eds.), Agricultural systems: Agroecology and rural innovation for development (pp. 279-306). Burlington, San Diego, London: Academic Press. Bezner Kerr, R. & Chirwa, M. (2004). Participatory Research Approaches and Social Dynamics that Influence Agricultural Practices to Improve Child Nutrition in Malawi. EcoHealth 1 (Suppl. 2): 109–119. Bezner Kerr, R., Snapp, S., Chirwa, M., Shumba, L., Msachi, R. (2007). Participatory Research on Legume Diversification with Malawian Smallholder Farmers for Improved Human Nutrition and Soil Fertility. Exploring Agriculture 43: 437-453. Box, Louk. (1989). Virgilio’s theorem: a method for adaptive agricultural research. In R. Chambers, A. Pacey and L. Thrupp (Eds.), Farmer First: farmer innovation and agricultural research (pp. 61-67). London: Intermediate Technology Publications. Chambers, R., Pacey, A., and Thrupp, L. (Eds.). (1989). Farmer First: Farmer innovation and agricultural research. London: Intermediate Technology Publications. 134 Chilonda, P. and Otte, J. (2006). Indicators to monitor trends in livestock production at national, regional and international levels. Livestock Research for Rural Development, 18(8). Retrieved January 2, 2014 from http://www.lrrd.org/lrrd18/8/chil18117.htm Creswell, J.W. and Plano Clark, V.L. (2011). Designing and Conducting Mixed Methods Research. Los Angeles: SAGE Publications, Inc. Feder, G., Just, R.E., & Zilberman, D. (1985). Adoption of agricultural innovations in developing countries: A survey. Economic Development and Cultural Change 33(2): 255-298. Ferguson, A.E. (1992). Differences among women farmers: Implications for African agricultural research programs. In Proceedings of a Workshop on Social Science Research and the CRSPs (pp. 47-62). Ferguson, A.E. (1994). Gendered Science: A Critique of Agricultural Development. American Anthropologist 96(3): 540-552. Ferguson, A.E., and Mkandawire, R.M. (1993). Common Beans and Farmer Managed Diversity: Regional Variations in Malawi. Culture & Agriculture 13: 14-17. Findley, S.E. (2014). Demography Learning Module: Measure of the Total Population Structure and Size. Retrieved from http://www.columbia.edu/itc/hs/pubhealth/modules /demography/populationRatio.html Food and Agriculture Organization. (2005) FAOSTAT data [Data set]. Retrieved from http://faostat.external.fao.org/default.jsp Gilbert, R.A. (2004). Best-Bet Legumes for Smallholder Maize-Based Cropping Systems of Malawi. In Eilittӓ, M., Mureithi, J., Derpsch, R. (Eds.), Green Manure/Cover Crop Systems of Smallholder Farmers: Experiences from Tropical and Subtropical Regions (pp. 153-174). Dordrecht: Kluwer Academic Publishers. Kristjanson, P., Neufeldt H., Gassner, A., Mango, J., Kyazze, F.B., Desta, S., Sayula, G., Thiede, B., Förch, W., Thornton, P.K., Coe, R. (2012). Are food insecure smallholder households making changes in their farming practices? Evidence from East Africa. Food Security, 4: 381-397. IFAD (International Fund for Agricultural Development). (2010). Rural Poverty in Malawi. Retrieved on April 17, 2013, from http://www.ruralpovertyportal.org/country/ home/tags/malawi 135 Livestock, Environment and Development Initiative. (2005). Tropical Livestock Units. Retrieved from http://www.virtualcentre.org/en/dec/toolbox/Mixed1/TLU.htm#What Millar, D. (1994) Experimenting farmers in northern Ghana. In I. Scoones and J. Thompson (Eds.), Beyond Farmer First: Rural people’s knowledge, agricultural research and extension practice (pp. 160-165). London: Intermediate Technology Publications. Misiko, M. and Tittonell, P. (2011). Counting eggs? Smallholder Experiments and Tryouts as Success Indicators of Adoption of Soil Fertility Technologies. Innovations as Key to the Green Revolution in Africa, 1137-1144. National Statistics Office of Malawi. (2008). Population and Housing Census: Spatial Distribution and Urbanization Report [data file]. Retrieved from http://www.nsomalawi .mw/images/stories/data_on_line/demography/census_2008/Main%20Report/ThematicR eports/Spatial%20Distribution%20and%20Urbanisation.pdf Nitsch, Ulrich. (1990). Computers and the Nature of Farm Management. Knowledge in Society: The International Journal of Knowledge Transfer 3(3): 67-75. NVivo10 for Windows Help. (2014). Run a Coding Comparison Query. Retrieved on June 12, 2014, from http://help-nv10.qsrinternational.com/desktop/procedures/run_a_coding_ comparison_query.htm Pannell, David J. (1999). Economics, extension and the adoption of land conservation innovations in agriculture. International Journal of Social Economics 26(7): 999-1014. Republic of Malawi and The World Bank. (2006). Malawi Poverty and Vulnerability Assessment: Investing in Our Future [Report]. Retrieved from http://fsg.afre.msu.edu/mgt /caadp/malawi_pva_draft_052606_final_draft.pdf Richards, P. (1989). Agriculture as a performance. In R. Chambers, A. Pacey and L. Thrupp (Eds.), Farmer First: farmer innovation and agricultural research (pp. 39-43). London: Intermediate Technology Publications. Rhoades, R. (1989). The role of farmers in the creation of agricultural technology. In R. Chambers, A. Pacey, L. Thrupp (Eds.), Farmer First: farmer innovation and agricultural research (pp. 3-9). London: Intermediate Technology Publications. Rhoades R. and Bebbington A. (1995). Farmers Who Experiment: An untapped resource for 136 agricultural research and development. In D.M. Warren, L.J. Slikkerveer and D. Brokensha (Eds.), The Cultural Dimension of Development: Indigenous knowledge systems (pp. 296-307). London: Intermediate Technology Publications. Schön, D. A. (1983). The reflective practitioner: How Professionals Think in Action. USA: Basic books. Schultz, T.W. (1975). The Value of the Ability to Deal with Disequilibria. Journal of Economic Literature 13(3): 827-846. Schulz, S., Honlonkou, A.N., Carsky, R.J., Manyong, V.M., Oyewole, B.D. (2003). Alternatives to Mucuna for Soil Fertility Management in Southern Benin: Farmer Perception and Use of Traditional and Exotic Grain Legumes. Exploring Agriculture 39: 267-278. Scoones, I. and Thompson, J. (Eds.). (1994a). Beyond Farmer First: Rural people’s knowledge, agricultural research and extension practice. London: Intermediate Technology Publications. Scoones, I. and Thompson, J. (1994b). Knowledge, power and agriculture—towards a theoretical understanding. In I. Scoones and J. Thompson (Eds.), Beyond Farmer First: Rural people’s knowledge, agricultural research and extension practice (pp. 16-32). London: Intermediate Technology Publications. Sharland, R.W. (1995). Using Indigenous Knowledge in a Subsistence Society of Sudan. In Warren, D.M., Slikkerveer, L.J., Brokensha, D. (Eds.), The Cultural Dimension of Development: Indigenous knowledge systems (pp. 385-395). London: Intermediate Technology Publications. Snapp, S., Kanyama-Phiri, G., Kamanga, B., Gilbert, R., Wellard, K. (2002a). Farmer and Researcher Partnerships in Malawi: Developing Soil Fertility Technologies for the Nearterm and Far-term. Exploring Agriculture 38: 411-431. Snapp, S.S., Rohrbach, D.D., Simtowe, F., Freeman, H.A. (2002b). Sustainable soil management options for Malawi: can smallholder farmers grow more legumes? Agriculture, Ecosystems and Environment 91: 159-174. Snapp, S.S. and Silim, S.N. (2002). Farmer preferences and legume intensification for low nutrient environments. Plant and Soil 245: 181-192. Snapp, S.S., Blackie, M.J., Donovan, C. (2003). Realigning research and extension to focus on farmers’ constraints and opportunities. Food Policy. 28: 349-363. 137 Snapp, S.S., Blackie, M.J., Gilbert, R.A., Bezner Kerr, R., Kanyama-Phiri, G.Y. (2010). Biodiversity can support a greener revolution in Africa. Proceedings of the National Academy of Sciences 107(48): 20840-20845. Snapp, S., Ota, V., Bezner Kerr, R., Shumba, L., Dakishoni L., Mhango, W. (2014). What is the role of crop diversity in ecosystem services? Manuscript in preparation. Stolzenbach, A. (1994). Learning by improvisation: farmer experimentation in Mali. In I. Scoones and J. Thompson (Eds.), Beyond Farmer First: Rural people’s knowledge, agricultural research and extension practice (pp. 155-159). London: Intermediate Technology Publications. Sumberg, J.E. and Okali, C. (1997). Farmer’s Experiments: Creating Local Knowledge. Boulder, Colorado: Lynne Rienner. Vaske, J. (2008). Survey Research and Analysis: Application in Parks, Recreation and Human Dimensions. State College, PA: Venture Publishing, Inc. Warren, D.M., Slikkerveer, L.J., Brokensha, D. (Eds.). (1995). The Cultural Dimension of Development: Indigenous knowledge systems. London: Intermediate Technology Publications. The World Bank. (2014). Age dependency ratio (% of working age population): Malawi [Data file]. Retrieved from http://data.worldbank.org/indicator/SP.POP.DPND UNICEF (United Nations Children’s Fund). (2000-2009; 2006-2010). Malawi [statistics /factsheet]. Retrieved on April 8, 2013, from http://www.unicef.org/infobycountry /malawi_statistics.html 138