PROMOTION AND ADOPTION OF CONSERVATION AGRICULTURE IN MOZAMBIQUE AND ZAMBIA By Philip Paul Grabowski A DISSERTATION Submitted to Michigan State University in partial fulfillment of the requirements for the degree of Community Sustainabil ity - Doctor of Philosophy 201 5 ABSTRACT PROMOTION AND ADOPTION OF CONSERVATION AGRICULTURE IN MOZAMBIQUE AND ZAMBIA By Philip Paul Grabowski The development of improved agricultural technologies has tremendous potential for improving the liveliho ods of smallholder farmers in sub - Saharan Africa. Conservation vulnerability to climate change. However, the benefits and challenges associated with reducing til lage vary by soil type and rainfall regime. Due to the complexity of both the livelihood strategies of resource - poor farmers and of their agro - ecological conditions, widespread adoption of any one form of CA is unlikely. Instead , technologies need to be a dapted to specific agro - ecological and socio - economic contexts. The first paper in this dissertation use s the case of conservation agriculture (CA) in Mozambique to obtain an in - depth perspective on the challenges researchers and development agencies face in using innovation networks that include farmers and input suppliers to improve the process of technology adaptation. The results show widespread agreement among researchers and program managers about the need to locally adapt CA due to the agro - ecologic al diversity of Mozambique . However, they also show farmers through Farmer Field Schools to adapt CA to the lo cal context. There is widespread agreement about the importance of establishing links across the value - chain, and lessons from nascent efforts to accomplish this are documents . The results also indicate that effective collaboration will require coming to terms with polarized disagreements on two key issues: the importance of emphasizing minimum tillage and the role of commercial inputs for CA. The second and third papers combine quantitative and qualitative analysis of farmers practices in Eastern Zamb ia . A survey was carried out with 245 farmers in 15 communities where CA adoption was expected to be relatively high. I n - depth interviews were carried out with 63 farmers and cotton company representatives . re mains low and disadoption is common. The main reasons farmers use minimum tillage are to improve their yields and to reduce their vulnerability to droughts. T here are also a number of challenges preventing more widespread use of the technology. T he increa sed effort needed for dry season land preparation is a key constraint . Households that have more available household labor were able to use MT on more of their land. D ry - season ripping is seen as too taxing for the oxen to make ripping services worthwhile . F armers who use ripping tend to be better - off, enabling them to invest in the new equipment and take the risk of a new technology. Lack of adequate information also limits the number of farmers using MT. Farmers who have never tried MT tend to be poorer and have more diverse livelihood strategies. The main conclusion is that farmers are not stuck in traditional practices but are carefully evaluating CA with the information they have available to them. Widespread adoption will require adapting existing te chnologies to overcome technical challenges and developing new ones to match a broader r ange of resource endowments. This process could be greatly improved by insights on the constraints and possible adaptations for the technologies . iv ACKNOWLEDGEMENTS First and foremost I want to acknowledge the tremendous support I have received from my advisor Dr. John Kerr. His dedication to supporting my studies and research has bee n exceptional. He has helped sharpen my thinking by listening patiently, asking the right questions at the right time , and humbly sharing his own experiences and insights. I would also like to thank the other members of my committee for their support thr ough the Ph.D. process. Dr. Kim berly Chung provided valuable guidance in my qualitative research and in considering how to make my research as useful as possible for th e respondents . I appreciate her practical advice on methods and her insightful comments on my writing. Dr. Laura Schmitt - Olabisi has helped me focus on the big picture and ask the hard questions about what is causing what. Dr. Steve n Haggblade oversaw my research assistantships and laid the foundation for all the research in Zambia. He als o provided me with a great example of humble leadership through many discussions over skype and bumping along Zambian roads. Man y other colleagues in Zambia and Mozambique made important contributions to this research. Heart - felt thanks to Stephen Kabwe at Indaba Agricultural Policy Research Institute (IAPRI) for his collaboration and support on the research in Zambia , and to Bordalo Mouzinho for all his hard work with Michigan State University in Mozambique . Special appreciation is also due to Dr. Cynth ia Donovan, Dr. Rafael Uaiene and Dr. Rui Benfica for their oversight of my work in Mozambique. This research would not have been possible without the funding and I would like to thank the many people who work behind the scenes to make that funding availa ble. T he C.S. Mott Group for Susta inable Agriculture provided me with funding that I was able to use for the field work in Zambia . Special thanks to Dr. Mike Hamm for his leadership there. T he United States v Agency for International Development missions in Lusaka, Zambia (USAID/Zambia) and in Maputo, Mozambique (USAID/Mozambique) provided funding for my assistantship and several preliminary trips under Michigan State University's Food Security Research Project (FSRP) and the Department Agricultural, Food, a nd Resource Economics at Michigan State University. Special thanks to Dr. Duncan Boughton and Dr. Eric Crawford for th eir leadership in bringing this funding to MSU . More people than can be named generously gave of their time by responding to surveys, par ticipating in interviews or helping me find the respondents I was looking for. In Mozambique, I would like to thank all of the CA researchers and program managers who participated, especially the CA working group. I n Zambia, I wish to thank the many farmer s and staff at NWK Agri - Services and Cargill Zambia for voluntarily participating in this research . In particular, I wish to recognize the valuable contributions made by Philip Ntitima and Kenan Bakasa from NWK, as well as Emmanuel Mbewe from Cargill. I wo uld also like to thank the enumerators Abraham Banda, Rachel Kabinda, Maurice Mitti, and Daniel Kashobondo for their hard work and dedication to quality research. Finally I want to express my deepest gratitude to my wife Christa for her love, patience and prayers throughout graduate school. She has always been an excellent research assistant and the best critic of my thinking. I am also grateful for he r loving care for our children , especially while I was busy or traveling . Words cannot express how much I appreciate her. vi TABLE OF CONTENTS LIST OF TABLES ................................ ................................ ................................ ........................ LIST OF FIGURES ................................ ................................ ................................ ...................... CHAPTER 1: INTRODUCT ION ................................ ................................ ................................ ... 1 REFERENCES ................................ ................................ ................................ ............................... 5 CHAPTER 2: USING FA RMER PARTICIPATION AND INNOVATION NETWORKS FOR CONSERV ATION AGRICULTURE ADAPTATION IN MOZAMB IQUE ................................ ................................ ........ 8 2.1. Introduction ................................ ................................ ................................ .............................. 8 2.2. Developing agricultural technologies in complex contexts ................................ ................... 10 2.2.1. Effective innovation for smallholders requires farmer participation .............................. 10 2.2.2. How participation is implemented affects the information flow ................................ ..... 12 2.2.3. Many agricultural problems cannot be solved at the farm level ................................ ..... 14 2.2.4. Researcher involvement in innovation networks requires institutional support ................................ ................................ ................................ ............................. 16 2.3. Background on Mozambican agriculture ................................ ................................ ............... 17 2.4. Methods ................................ ................................ ................................ ................................ . 19 2.4.1. Inventory of CA experiences in Mozambique ................................ ................................ 19 2.4.2. Literature review of the performance of CA across Mozambique ................................ .. 19 2.4.3. Survey of CA professionals ................................ ................................ ............................. 20 2.4.4. In - depth inte rviews ................................ ................................ ................................ .......... 21 2.5. Results ................................ ................................ ................................ ................................ .... 21 ........................... 22 2.5.1.1 Promotion and research of conservation agriculture ................................ ................. 22 2.5.1.2. The need to adapt CA to local conditions ................................ ................................ . 23 2.5.1.3. The context - specificity of CA performance ................................ ............................. 25 ................................ ............................ 27 2.5.3. Lessons from nascent efforts to link actors across the value chain ................................ . 30 2.5.4. Challenges anticipated for collaborative research to adapt CA ................................ ...... 32 2.5.4.1. Debating the emphasis on minimum tillage ................................ ............................. 32 2.5.4.2. The role of inputs in CA promotion ................................ ................................ ......... 34 2.5.4.3. Institutional support for collaborative CA adaptation ................................ .............. 37 2.6. Conclusion ................................ ................................ ................................ ............................. 38 REFERENCES ................................ ................................ ................................ ............................. 40 CHAPTER 3: DETERMINA NTS OF ADOPTION OF M INIMUM TILLAGE BY COTTON FARMERS IN EASTERN ZAMBIA ................................ .............. 46 3.1. Introduction ................................ ................................ ................................ ............................ 46 3.2. Conservation agriculture ................................ ................................ ................................ ........ 48 3.2.1. Minimum tillage ................................ ................................ ................................ .............. 48 i x xi vii 3.2.2. Conservation agriculture promotion among cotton farmers in Zambia .......................... 49 3.2.3. Agriculture in Eastern Zambia ................................ ................................ ........................ 50 3.2.4. Minimum tillage adoption in Zambia ................................ ................................ .............. 51 3.3. Data and methods ................................ ................................ ................................ ................... 52 3.3.1. Selection of respondents ................................ ................................ ................................ .. 52 3.3.1.1. Community selection ................................ ................................ ................................ 52 3.3.1.2. Respondent selection for in - depth interviews ................................ .......................... 53 3.3.1.3. Farmer selection for the survey ................................ ................................ ................ 54 3.3.2. Data collection ................................ ................................ ................................ ................. 54 3.3.2.1. Qualitative data collection ................................ ................................ ........................ 54 3.3.2.2. Survey data collection ................................ ................................ .............................. 55 3.3.2.3. Community - level data collection ................................ ................................ .............. 55 3.3.3. Data analysis ................................ ................................ ................................ ................... 56 3.3.3.1. Qualitative data analysis ................................ ................................ ........................... 56 3.3.3.2. Statistical data analysis ................................ ................................ ............................. 56 3.3.3.3. Sampling weights ................................ ................................ ................................ ..... 57 3.3.3.4. Cluster analysis to control for community fixed effects ................................ ........... 57 3.4. Results and discussion ................................ ................................ ................................ ........... 58 3.4.1. Adoption rates ................................ ................................ ................................ ................. 58 ................................ ................................ ........... 60 ................................ ......................... 62 3.4.3.1. Incentives ................................ ................................ ................................ .................. 62 3.4.3.2. Labor as the primary constr aint for basins ................................ ............................... 62 3.4.3.3. Equipment costs as a key constraint to ox - ripping ................................ ................... 63 3.4.3.4. Fear of over - exerting oxen as a barrie r to widespread use of ox - ripping ................................ ................................ ................................ ................. 64 3.4.3.5. Information needs as a barrier to adoption of ox - ripping ................................ ......... 64 3.4.3.6. Lack of motivation as a reason for non - adoption ................................ ..................... 65 3.4.3.7. Other reasons for disadopting ripping ................................ ................................ ...... 65 3.4.4. Transitions in land preparation methods ................................ ................................ ......... 66 3.4.5. Household - level regression results ................................ ................................ .................. 67 3.4.5.1. Labor availability as a determ inant of adoption ................................ ....................... 72 3.4.5.2. Wealth indicators as determinants of adoption ................................ ......................... 72 3.4.5.3. Livelihood strategy as a determinant of adoption ................................ .................... 73 3.4.5.4. Knowledge as a determinant of adoption ................................ ................................ . 73 3.4.5.5. Incentives as a determinant of basin use and disadoption ................................ ........ 73 3.4.6. Summary ................................ ................................ ................................ ......................... 74 3.5. Conclusion ................................ ................................ ................................ ............................. 75 APPENDICES ................................ ................................ ................................ .............................. 78 Appendix A: Survey instrument ................................ ................................ ................................ ... 79 Appendix B: Qualitative interview guide for distributors or buyers ................................ ............ 88 Appendix C: Qualitative interview guide for farmers ................................ ................................ .. 90 REFERENCES ................................ ................................ ................................ ............................. 97 viii CHAPTER 4: UNDERSTAN DING PART IAL ADOPTION OF MINI MUM TILLAGE BY COTTON FA RMERS IN EASTERN ZAM BIA .......................... 102 4.1. Introduction ................................ ................................ ................................ .......................... 102 4.2. Conservatio n agriculture in Zambia ................................ ................................ .................... 103 4.2.1. Conservation agriculture and minimum tillage ................................ ............................. 104 4.2.2. Conservation agriculture and cotton f armers ................................ ................................ 105 4.2.3. Cotton farmers in eastern province ................................ ................................ ............... 105 4.2.4. Adoption of MT in Eastern Province ................................ ................................ ............ 106 4.3. Methods and data ................................ ................................ ................................ ................. 107 4.3.1 Data collection ................................ ................................ ................................ ................ 108 4.3.1.1. Qualitative data ................................ ................................ ................................ ....... 108 4.3.1.2. Survey data ................................ ................................ ................................ ............. 109 4.3.2. Data analysis ................................ ................................ ................................ ................. 110 4.3.2.1. Qualitative data analysis ................................ ................................ ......................... 110 4.3.2.2. Statistical analysis ................................ ................................ ................................ ... 110 4.3.2.2.1. Modeling choice of land preparation method for each plot ............................. 111 4.3.2.2.2. Modeling intensity of adoption of minimum tillage ................................ ........ 112 4.4. Results ................................ ................................ ................................ ................................ .. 112 4.4.1. How farmers are using basins and ox - ripping ................................ ............................... 113 4.4.2. Conditions where farmers prefer using MT ................................ ................................ .. 115 4.4.2.1 . Crop - specific MT benefits ................................ ................................ ...................... 115 4.4.2.2. Plot - specific benefits and challenges of MT identified by farmers ........................ 116 4.4.2.2.1. Farmer ................................ .............. 116 4.4.2.2.2. Soil type affects the costs and benefits of MT ................................ ................. 117 4.4.2.2.3. Soil fert ility affects the magnitude of the benefits ................................ ........... 117 4.4.2.3. Plot characteristics associated with MT ................................ ................................ . 118 4.4.3. Constraints to incre ased MT use ................................ ................................ ................... 121 4.4.3.1. Information constraints: Incremental testing of MT as a reason for partial adoption ................................ ................................ ................................ ...... 122 4.4.3.2. Investment constraints - challenges obtaining rippers and oxen ............................. 123 4.4.3.3. Labor constraint the increased effort needed for MT as a reason for partial adoption ................................ ................................ ................................ ...... 123 4.4.3.3.1. Dry season labor issues ................................ ................................ .................... 123 4.4.3.3.2. Weeding labor issues ................................ ................................ ....................... 125 4.4.3.4. Quantitative analysis of household - level constraints ................................ ............. 127 4.5. Implications for CA promotion ................................ ................................ ............................ 131 4.6. Conc lusion ................................ ................................ ................................ ........................... 132 APPENDIX ................................ ................................ ................................ ............................... 134 REFERENCES ................................ ................................ ................................ ........................... 138 ix LIST OF TABLES - farm research ................................ .............. 13 Table 2: The broadening focus for agricultural innovation research ................................ ............ 15 Table 3: CA experiences by province ................................ ................................ ........................... 27 perspectives on the importance of commercial inputs ................................ ................................ ................................ ....... 35 Table 5: Community stratification by company and adoption level ................................ ............. 53 Table 6: Response rates of farmers by stratification category ................................ ...................... 54 Table 7: Mean values of the characteristics used to cluster communities ................................ .... 58 Table 8: Land preparation method use rates and ra tes of disadoption for cotton farmers in areas of medium and high adoption in Eastern Province ............................... 59 Table 9: Description of variables used in the household - level regression on MT use ................. 68 Table 10: Factors affecting household use of MT, multinomial logistic regression results with robust standard errors ................................ ................................ ................ 69 Table 11: Tabulation of land preparation methods for all plots in the dataset ........................... 109 Table 12 : Characteristics of clusters used in regression analyses ................................ ............... 111 Table 13: Percent of overall area under minimum tillage by crop for farmers who use some MT ................................ ................................ ................................ .............. 113 Table 14: Differences in plot size, preparation date and inputs by land preparation method for cotton and maize plots of farmers who use minimum til lage on some part of their land ................................ ................................ ........................... 115 Table 15 : Description of variables used in the plot - level regression of determinants of land preparation method for MT users ................................ ................................ ... 119 Table 16: Factors affecting plot - level use of MT, mu ltinomial logistic regression results with robust standard errors ................................ ................................ .............. 120 Table 17: Predicted probabilities of land preparation method from the results of t he multinomial logistic regression. ................................ ................................ ........... 121 Table 18: Description of variables used in the household - level regression ................................ 129 x Table 19: Average partial effects of household - and community - level factors influencing intensity of adoption of minimum tillage (MT). ................................ ..... 130 Table 20: Multinomial logisitic regression re sults of household characteristics and community characteristics associated with whether MT is used on cotton or maize or both. ................................ ................................ ......................... 137 xi LIST OF FIGURES Figure 1: Number of organizations promoting and researching CA in Mozambique by district ................................ ................................ ................................ ....................... 23 Figure 2: Researcher and program manager re opinion have CA researchers already succeeded in identifying a profitable form of CA that will lead to large - scale farmer adoption given proper extension efforts and minor adapt ................................ ..... 24 Figure 3: Percent of area under each land preparation method for the four largest crops ............ 60 Figu ................................ ................................ ...... 67 Figure 5: Odds ratio plot for economic indicator variables ................................ .......................... 71 Figure 6: Odds ratio plot for capacity indicators and other variables ................................ ........... 71 Figure 7: Decision tree summarizing the interactions between farmers' resources and priorities and land preparat ion decisions ................................ ................................ 76 Figure 8: Percent of area under minimum tillage (MT) by cotton farmers who use some MT ................................ ................................ ................................ ................ 114 Figure 9: Average area under minimum tillage ................................ ................................ .......... 122 Figure 10: Percent of area under minimum tillage (MT) for maize, cotton and other crops by cotton farmers who use some MT on any crop ................................ .... 135 1 CHAPTER 1: I NTRODUCTION Over the past 50 years agricultural yields have remained stagnant in Afri ca while the population has more than tripled (World Bank, 2007) . Agricultural area has expanded i n order to meet consumption needs at the national level (World Bank, 2007) but this has caused shortening of fallow periods and land degradation , which threatens the sustainability of production ( Morris et al., 2007; Todaro and Smith, 2009). Improved agric ultural technologies hold tremendous potential for national food security by increasing the productivity of smallholder farmers ( Pretty et al., 2011) . This would also reduce poverty by improving the livelihoods of some of the poorest people in the world (B arrett, 2010) . Though agricultural technologies led to dramatic yield increases in Asia during the green revolution , in Africa the diversity of agro - ecological contexts across the continent has inhibited the benefits from technological transfers because s pecific technologies have limited domains where they are applicable (World Bank, 2007) . Furthermore, the complex livelihood strategies of the rural poor make it difficult for agricultural researchers to develop appropriate technologies (Chambers, 1997) . D eveloping effective agricultural technologies for smallholder farmers in Africa requires including farmers in a process of localized adaptation in order to overcome context - specifi c challenges (Pretty et al., 2011) . This dissertation focuses on the need f or adapting conservation agriculture technologies to overcome adoption constraints in southern Africa. Conservation agriculture (CA) has been widely promoted to sustainably increase yields (Kassam et al., 2009) . The combination of three principl es (minimal soil disturbance, permanent soil cover, and incorporation of legumes through intercropping or rotations) has the potential to increase soil fertility (FAO, 2001) , increase water infiltration (Thierfelder and Wall, 2009) , and overcome labor bott lenecks through dry season land preparation (Haggblade et al., 2011) . Minimal soil disturbance in Zambia is accomplished 2 through three major forms of minimum tillage: digging basins with a hand hoe, using an ox - drawn Magoye ripper to open a rip line, or us ing a tractor drawn chisel plow to open a rip line. Despite promotion since 1996 , adoption levels remain low and promotional efforts have been criticized for narrowly pushing technological packages that are not well aligned with owments (Giller et al., 2009; Andersson and Giller, 2012) . Nevertheless, some farmers continue to use CA on small portions of their land (Arslan et al., , which suggests the possibility of adapting the technologies to over come the constraints. This dissertation aims to contribute practical guidance for this adaptation process by linking empirical observations of CA research, promotion and use with lessons from the literature on smallholder agricultural development. The se cond chapter , Using Farmer Participation and Innovation Networks for Conservation Agriculture Adaptation in Mozambique , analyze s the challenges researchers and development agencies face in using innovation networks that include the participation of farme rs, input suppliers and other stakeholders to improve the process of CA technology adaptation. It is based on an inventory of the experiences with CA in Mozambique, an on - line survey of CA researchers and projects managers, and phone interviews with key in formants who were implementing CA projects using Innovation Platforms (IPs) and Farmer Field Schools (FFS). With the aim of learning from history, t he results are framed in the context of the broader literature of participatory technology development. T he results indicate widespread agreement about the need for adapting CA tec hnologies to the local context with in Mozambique because of t he diversity of farmers and their conditions. S et ting up client - oriented participatory research efforts requires collab oration at two levels: between r esearchers, extension and the private sector across the value chain and between researchers and resource - poor farmers. The results indicate that collaboration in Mozambique will require researchers to be aware of and either resolve, or 3 learn to live with, polarized disagreements on two key issues: dedication to the CA components and the importance of commercial inputs. Furthermore, institutional arrangements for research management will have to shift so that researchers hav e the support and incentives to manage the complexity of an evolving research process that is tightly linked to non - these collaborative efforts are decentralized, such as through regional CA working groups. The third chapter Determinants of Adoption of Minimum Tillage by Cotton Farmers analyzes the factors that c ause some households to use , and others to disadopt or never have tried minimum tillage (MT ), the principle of CA that is typically emphasized first . The stud y focuses on areas where promotion has been adequate , and where complementary inputs and equipment are available on credit , in order to identify constraints that may require adaptive rese a rch . It is based on a survey of 245 farmers in 15 communities and in - depth interviews with 63 farmers and cotton company buyers. The results indicate that there are four main reasons that keep farmers from using MT: the cost of equipment for ox - ripping, the increased effort for dry season labor, the high levels of uncertainty associated with a dramatic shift in farming and a lack of motivation by those who are satisfied with their current yields. The main reasons farmers use MT are to improve their yield s and to reduce their vulnerability to droughts. The fourth chapter , focuses on identify ing constraints to MT use by analyzing T his study combines quantitative analysis of the 445 plots farmed by 81 MT farmers and qualitative analysis of in - depth interviews with 43 MT farmers who explained why they ch ose to use MT on some plots but not others . The results show that the increased ef fort needed for dry season land preparation (for both hoe farmers and animal traction farmers) is a key constraint preventing many households from using MT 4 on as much land as they would li ke. Households that have more available household labor and larger farm sizes were able to use MT on more of their land, showing an ability to overcome labor and investment constraints. Adaptive research is needed to overcome the challenges farmers face in using ox - ripping on large areas with an average team of oxen. The results also show that f armers are rationally analyzing where they used MT based on the benefits, costs and risks they anticipate. Perceived benefits are greater with maize or cotton, and the perceived risks are less on flat plots (due to concerns about erosion with MT) . MT can have large immediate benefits by making a degraded field productive (such as by breaking through hard pan or through the application of manure) so farmers selectively prefer those plots for MT . CA promotional efforts can be more e ffective by targeting areas where constraints are lowest and benefits are highest . The papers from Zambia provide many examples of how farmers are not stuck in traditional land preparation methods but are carefully evaluating the benefits and costs of MT given the information they have available to them. Similarly, the Mozambique paper shows that development agencies are actively adapting CA technologies constraints and priorities. There is potential for enhancing these adaptive e fforts if agricultural researchers engage with farmers and development agencies to solve practical problems related to adapting existing technologies . Widespread adoption of CA in southern Africa requires making the principles beneficial under a broader r ange of conditions and resource endowments. The key message from all three studies is that the process of developing agricultural technologies suitable for African smallholders could be greatly improved by valuable insights on the const raints and possible adaptations . The on - going challenge is finding creative ways to organize research and - making. 5 REFERENCES 6 REFERENCES Andersson, J .A. and S laims to Understanding Farmers and Contexts: A Literature Review of Conservation Agriculture (CA) Adoption among Smallholder Farmers in Southern Africa. Agriculture, Ecosystems and Environment 187: 116 - 32. doi:10.10 16/j.agee.2013.08.008 Andersson, J. and Giller, K. 2012. or Conservation Agriculture and the politics of its promotion in African smallholder farming. In: Sumberg, J. and Thompson, J. (eds.) Conteste d Agronomy: Agricultural Research in a Changing World . London: Earthscan. Arslan, A., N. McCarthy, L. Lipper, S. Asfaw, A. Cattaneo. 2014. Adoption and Intensity of Adoption of Conservation Farming Practices in Zambia. Agriculture, Ecosystems and Environme nt 187: 72 - 86. doi:10.1016/j.agee.2013.08.017 Barrett, C. 2010. Food Systems and the Escape from Poverty and Ill - health Traps in Sub - Saharan Africa. In The African Food System and its Interactions with Human Health and Nutrition , ed. P. Pinstrup - Anderson. Ithaca, NY: Cornell University Press. Chambers, R. 1997. Whose Reality Counts? Putting the First Last . London: Intermediate Technology Publications. Food and Agriculture Organization (FAO). 2001. The Economics of Conservation Agriculture. Rome: United Nati ons Food and Agriculture Organization. Giller, K., E. Witter, M. Corbeels, and P. Tittonell. 2009. Conservation Agriculture and Smallholder Farming in Africa: The Heretics' View. Field Crops Research 114 : 23 - 24. Haggblade, S., S. Kabwe, and C. Plerhoples. 2011. Productivity Impact of Conservation Farming on Smallholder Cotton Farmers in Zambia. Working Paper No. 47. Lusaka, Zambia: Food Security Research Project. Kassam, A., T. Friedrich, F. Shaxson, and J. Pretty. 2009. The Spread of Conservation Agricultu re: Justification, Sustainability and Uptake. International Journal of Agricultural Sustainability 7.4: 292 - 320. Morris, M., V. Kelly, R. Kopicki, and D. Byerlee. 2007. Fertilizer Use in African Agriculture: Lessons Learned and Good Practice Guidelines. Wa shington D.C.: The World Bank. Pretty, J., C. Toulmin, and S. Williams. 2011. Sustainable Intensification: Increasing Productivity in African Food and Agricultural Systems. International Journal of Agricultural Sustainability 9.1: 5 - 24. Thierfelder, C. and P.C. Wall. 2009. Effects of Conservation Agriculture Techniques on Infiltration and Soil Water Content in Zambia and Zimbabwe. Soil and Tillage Research 105: 217 - 27. 7 Todaro, M. and S. Smith. 2009. Economic Development (10th edition ). Boston: Pearson Addis on Wesley. World Bank. 2007. Agriculture for Development . World Development Report 2008. Washington, D.C.: The World Bank. 8 CHAPTER 2: USING FARMER PARTICI PATION AND INNOVATIO N NETWORKS FOR CONSERV ATION AGRICULTURE AD APTATION IN MOZAMBIQUE 2.1. Introduc tion The development of improved agricultural technologies has tremendous potential for improving the livelihoods of smallholder farmers in least developed countries (World Bank, 2007; Pretty et al., 2011). Sets of technologies such as conservation agricul ture (minimum tillage, mulching and crop rotations) have the potential to sustainably increase yields and However, the linear technology transfer approach (from resear ch stations to farmers via extension) has failed to produce technologies that most farmers can adopt where farming systems are integrated into complex livelihood systems and agro - ecological conditions are diverse (Bingen and Gibbon, 2012; Buhler et al., 20 02; Ekboir, 2002). In contrast, there are cases where high levels of adoption have been achieved in complex contexts when innovation networks are used to locally adapt agricultural technologies (Ekboir, 2003; Klerkx et al., 2012). An innovation network all ows interdependent actors (such as farmers, input suppliers, buyers, extensionists and agricultural researchers) who cannot meet their innovation objectives independently to come together to collectively facilitate the innovation process (Klerkx et al. 201 0). The need to adapt conservation agriculture (CA) to the local context of smallholder farmers has been well established in the literature (Erenstein et al., 2012; Giller et al., 2009; Wall, 2007). This is primarily because the benefits and challenges as sociated with reducing tillage are variable across soil types and rainfall regimes (Giller et al., 2009; Baudron et al., 2012, Ekboir, 2002). There are also a wide range of minimum tillage technologies (basins, jab - planters, ox - drawn rippers, tractor rippe rs), each with different labor, knowledge and financial requirements for effective use (Grabowski et al., 2014). Furthermore, CA adaptation 9 requires integrated efforts across the value - chain to ensure the availability of commercial inputs, such as equipmen t for minimum tillage or herbicides, and secure markets for surplus production (Giller et al., 2011; Wall, 2007). A common feature of successful CA programs for smallholders is the use of participatory research approaches with linkages across the value - cha in to adapt CA technologies to specific agro - ecological and socio - economic contexts (Ekboir, 2002). In contrast, CA promotion in southern Africa has been criticized as being overly prescriptive (Andersson and Giller, 2012) and lacking in the critical refle ction of evidence that is needed for effective adaptation (Whitfield et al., 2015). In this paper I use the case of CA in Mozambique to obtain an in - depth perspective on the challenges researchers and development agencies face in using innovation networks for improving local adaptation of CA. The experience from Mozambique is especially instructive because it has nearly 20 years of experience with CA but without the coordinated promotion of specific CA technologies seen in much of the rest of southern Afri ca. The study specifically aims to answer the following research questions: 1) To what extent do CA researchers and program managers perceive the need for adapting CA to the local context? 2) How are farmers currently involved in the process of adapting C A technologies to local conditions in Mozambique? 3) What obstacles hinder the use of innovation networks (including farmers, input suppliers and extensionists) in the technology adaptation process? I used the following data collection activities to ans wer these questions: 1) an inventory of CA research and promotional projects, 2) a review of the literature on the performance of CA across the country, 3) a survey of CA researchers and program managers, and 4) in - depth interviews with key informants impl ementing CA projects that involved farmers and actors across the value - chain in local adaptation. We take the need for participation as the starting point for the effective development of crop management technologies in contexts with high levels of comple xity. I outline the 10 evidence supporting the importance of participation and innovation networks by reviewing the literature in the next section. Next I present background information about agriculture in Mozambique followed by details about the research methods used. The results are organized in parallel to the four sections of the literature review. 2.2. Developing agricultural technologies in complex contexts In this section I review the literature about how to effectively develop agricultural technolo gies in developing countries where smallholder farmers have complex livelihood strategies and typically live in diverse agro - ecological environments. First, I summarize literature on why the participation of farmers is needed in these contexts. Next I summ arize how different levels of participation are appropriate for different types of technologies. In the third sub - section I focus on the innovation systems approach to agricultural change and the need for a holistic perspective that includes actors across the value - chain. Finally, I summarize what is known about the institutional challenges of implementing participatory agricultural research. Together this information from the literature provides a conceptual framework for the analysis of how CA is being ad apted to local contexts in Mozambique and the challenges faced in using innovation networks for that purpose. 2. 2.1 . Effective innovation for smallholders requires farmer participation The context - specific information about the agricultural problems of res ource - poor farmer to the researcher (von Hippel, 1994). Agricultural problem solving with resource - poor iverse agro - ecological - physical context is typically implicit knowledge gained by observation and not easily communicated; 2) the livelihood strategies of resource - poor farmers tend to be diverse and complex (Chambers, 1997), which 11 increases the amount of information that needs to be transferred; and 3) there tends to be a wide social and cultural gap between formal researchers and resource - poor farmers. The top - down way in which CA has been researc hed and the inflexible way in which CA has been promoted in southern Africa (Andersson and Giller, 2012; Grabowski and Kerr, 2014) provides a perfect example of how researchers and development agencies who are actively engaged with farmers dominate the flo w of information and fail to address the context - specific constraints to adoption. Management research suggests that problem solving should be carried out where the von Hippel, 1994). Since the mid - 1970s a variety of agricultural research methods have aimed to (Merrill - Sands et al., 1991). The Farming Systems Research (FSR) mo vement was characterized by on - farm participatory trials and emphasized interdisciplinary research to address the complex interactions of the farming system (Bingen and Gibbon, 2012). While FSR led to many important insights into the production constrain ts of smallholder farmers in Southern Africa, it did not lead to widespread adoption of the promoted technologies (Waddington, 1993). The disappointing performance of FSR is the direct result of three problems. First, most FSR projects were operating with low levels of farmer participation (Merrill - Sands et al., 1991). Second, FSR focused too narrowly on farm - level issues with little attention to the broader systems in which they were embedded (Bingen and Gibbon, 2012). Third, FSR projects did not have the institutional support needed for participatory research that was so drastically different from conventional research station - based approaches (Biggs, 1995; Merrill - Sands et al., 1991). The next section discusses the need for higher levels of participatio n, after which I turn to the need for a systems perspective on innovation followed by institutional support issues. 12 2. 2.2 . How participation is implemented affects the information flow Participatory agricultural research has been implemented in very differ ent ways. It is useful to characterize these approaches as a continuum (Table 1) based on the level of et al. , 1989). At the low end, researchers carry out on - farm trials simply by contracting farmers to run their experiments. At the next level, researchers consult with farmers about their needs, run experiments on their land and then consult with them about their observations. At the collaborative level, farmers are involved with researchers through all phases of the research. Finally, at the collegial level the complementarities in knowledge and skills (Biggs et al. , 1989). 13 Table 1 : - farm research Contractual Consultative Collaborative Collegial Description of roles Researchers to run experiments on their land Researchers consult farmers about problems and develop solutio ns Researchers and farmers collaborate as partners to design, implement and analyze research Researchers strengthen research/problem solving systems Level of interaction Minimal At beginning and end High and continuous Long term and spor adic Conditions where approach is most appropriate Technically complex technologies that are context sensitive but broadly used. Minor adaptations to technologies with complex technical consequences. Where both realities and the technical requ irements are complex. Where technical information needs are relatively low realities are very diverse. Examples Plant breeding for high yields Minimum tillage equipment Crop management technologies Trying out varieties, species or practice s Importance of information Low Low High High Importance of technical skills High High High Low Source : Adapted from Buhler et al., 2002 Table 5.1 and Biggs et al. , 1989 The level of participation that is most appropriate depends on the information needs of the problem addressed by the research. The contractual and consultative levels of information is relatively low. Researchers can impro ve the effectiveness of their problem solving with minimal interactions with farmers. On the other end, the collegial level is best 14 - depth knowledge of the context is es sential. The collaborative level is likely to be best when agricultural innovations require both researchers. Examples of technologies requiring such collaboration includ e crop management technologies, natural resource management issues and improved germplasm for non - yield traits (Fujisaka, 1994). In contrast, plant breeding for high yields can be effective for addressing the needs of resource - poor farmers with low levels of participation, as long as the research is client - (Witcombe, 2006). Conservation agriculture is a set of crop management technologies with complex interactions among the component s of minimum tillage, rotation with legumes and covering the soil with residues or mulch. Due to these interactions CA has been described as knowledge - intensive (Kassam et al., 2009) and its effective implementation requires high levels of farmer particip ation and on - going adaptation through collaborations between researchers and farmers (Ekboir, 2002). 2. 2.3 . Many agricultural problems cannot be solved at the farm level The scope of agricultural innovation has broadened over time as awareness of the impor tance of the wider system has increased (Table 2). This broader focus has been associated with a shift in boundaries starting with research being confined to single disciplines expanding to trans - disciplinary efforts that value non - academics as key contrib utors. With the focus on value - chains, partners such as input suppliers, output buyers and policy makers become part of the collaborative team for fostering CA innovation (Ekboir, 2003). 15 Table 2 : The broadening focus for agricu ltural innovation research Transfer of Technology Early Farming Systems Research Farmer - first and AKIS a Agricultural Innovation Systems Time period 1960s on 1970s and 80s 1990s on 2000s on Activities Supply technologies constraints Colla borate in research Partner to foster innovation Disciplines b Single - discipline Multi - disciplinary Inter - disciplinary Trans - disciplinary Scope Productivity increase Efficiency gains Livelihood system Value chains, policies and organizations Role of scie ntists Innovator Expert Collaborator One of many partners Goals Behavior change and technology adoption Overcome constraints, better fit in farming system Empowerment and better fit to livelihood system Increased capacity to innovate and adapt Source : Ad apted from Klerkx et al., 2012 Table 20.1 a Agricultural Knowledge and Information Systems b Multi - disciplinary research has separate disciplines working on the same issue relatively independently while inter - disciplinary research has several disciplines a ctively collaborating together and trans - disciplinary research includes non - professional researchers as part of the research team. Innovation networks use this Agricultural Innovation Systems approach and have become operationalized in international agricu ltural research centers through what are called Innovation Platforms (IPs). The platform aims to provide space for collaboration by creating a new forum where all stakeholders can interact to collectively focus on solving a common problem (Klerkx et al., 2012). A researcher or extensionist typically takes the role of for innovation, strengthening and broadening the composition of the network, and managing conflicts (Klerkx et al., 2012). However, unless farmers can effectively share information and truly set priorities, the innovation process will miss the potential benefits of their participation. Farmers that are 16 better organized will be better able to articulate their needs to researchers and other partners (Rajalahti et al., 2008), which suggests community mobilization may be a necessary first efforts. For example, a case study of an IP focusing on CA in Zambia documented benefits from coordination at the district level and harmonization of CA messages to farmers but no et al., 2011) . 2. 2.4 . Researcher involvement in innovation networks requires institutional support Effectively fostering organizational change to meet the needs of a more client - oriented participatory approach to agricultural research is a key struggle in implementin g the Agricultural Innovation Systems approach (Klerkx et al., 2012; Rajalahti et al., 2008) and was one of the major implementation challenges of the FSR approach (Merrill - Sand et al., 1991; Biggs 1995). For researchers to effectively implement the partic ipatory and collaborative processes outlined above, research systems must support them to: 1) cross boundaries (interact with farmers, NGOs, the private sector, other scientists), 2) focus on practical problem solving, not just publications, and 3) impleme nt an evolutionary research process. First, agricultural researchers will need to spend significant time with farmers and actors across the value - chain as well as with scientists in other disciplines or working on different crops. Case studies from FSR p rojects show that these linkages are easier where prioritized where research managers had firsthand experiences benefiting from crossing boundaries (Merrill - S ands et al., 1991). The commodity orientation of most agricultural research organization goes against the system perspective needed (Buhler et al., 2002), though interdisciplinary research planning can be used to ameliorate this challenge (Merrill - 17 Sands e t al., 1991). Becoming boundary crossers will require scientists to have skills in 2009). Next, researchers need incentives to engage in practical problem - solving, n ot just the production of peer - reviewed publications (Biggs, 1995; Klerkx et al., 2012). However, experience shows that employers and research funders are likely to continue using publications as their preferred performance indicator for scientists (Buhler et al., 2002). Nevertheless, this could be broadened to include indicators such as technical recommendations or new techniques. A study of agricultural research productivity in Mexico showed that increasing the number and intensity of interactions with fa rmers resulted in increases in both practical technical recommendations and publications (Rivera - Huerta et al., 2011). Finally, researchers also need the flexibility to carry out an evolutionary research process to be able to respond to the needs of farme rs and other partners. Facilitating innovation in complex systems requires skills in adaptive management where decisions are based on information from regularly scanning the environment (Klerkx et al., 2012). This flexibility can be facilitated through dec entralized planning of research (Biggs, 1995). 2. 3. Background on Mozambican agriculture At the end of the 16 - year civil war in 1992 Mozambique was considered one of the poorest countries in the world. Food aid was astronomical, national infrastructure wa s largely destroyed, the economy was at a standstill and it was nearly a failed state (Newitt, 2002). The UN and the World Bank supported the Mozambican government in pursuing free market economic policies with strong emphasis on international investment in large projects (Hanlon and Smart, 2008). These policies led to dramatic increases in Gross Domestic Product (GDP) and with a more stable economy and increased tax revenues, the Mozambican 18 government was able to invest in highly needed rural development projects. The combination of a stable currency, improved roads and communication systems and foreign investment in cotton and tobacco production led to increases in the welfare of the rural poor through greater market inclusion (Hanlon and Smart, 2008). Po verty levels decreased from 69% in 1997 to 54% in 2008 (Feed the Future, 2009). world (von Grebmer et al., 2013) with 8.1 million people undernourished, which is 38% of the popu lation (Bread for the World, 2011). It is estimated that 44% of children under 5 years old are stunted and 18% are moderately or severely underweight (UNICEF, 2011). The rise in food commodity prices since 2007 and 2008 has caused increased concern for na tional Mozambique (AGRA, 2009). There is tremendous agricultural potential in Mozambique but very poor performance in terms of both yield and total production. Only 12% o cultivated and only 4% of irrigable land is actually irrigated (Feed the Future, 2009). Maize is the largest staple food crop and it is largely grown for household consumption, with only 15% of production being marketed in 20 11 (Benfica et al., 2014). Smallholder maize yields have stagnated since the 1960s at only 1.4 tons/ha on average, though yields as high as 5 or 6 tons/ha are possible (Zavale et al., 2006). In 1996 Sasakawa Global 2000 introduced CA to Mozambique, in c ollaboration with the National Directorate of Agricultural Extension (DNEA), the Agricultural Research Institute of Mozambique (IIAM) and Monsanto (Nhancale, 2000). Early CA promotion was championed by the Projecto de Promoção Económica de Camponeses (PROM EC) in Sofala (Zandamela et al., 2006), as well as the FAO and DNEA who formed an extension - focused working group. 19 Since 2007, funding for CA has increased including substantial research projects for international agricultural research centers and promoti onal efforts of development agencies from by both government and non - governmental organizations (Nkala et al., 2011). In 2012 a CA working group was established, consisting of research, extension and NGO staff, with the mandate to develop a national progra m for increasing the impact of CA for smallholder promoting conservation agriculture (CA) to improve smallholder productivity based on its potential to sustainably ma nage soil fertility and decrease vulnerability to climatic events and overall climate change (Mozambique Ministry of Agriculture, 2010). 2. 4. Methods 2. 4.1 . Inventory of CA experiences in Mozambique By contacting a large majority of CA researchers and man agers of CA projects in the country, the combinations of CA principles and technologies that were being researched and promoted in each region were able to be identified . The inventory documented the efforts of 29 development organizations, 10 research org anizations and 5 private sector organizations actively promoting CA. Respondents were identified using a combination of key informants, asking respondents to identify potential contacts and internet searches. Researchers and project managers filled out a f orm to provide details of what CA principles and technologies they were using in their projects and who they partner with. The results of the inventory were tabulated by region and research and development efforts were mapped by district. 2. 4.2 . Literatur e review of the performance of CA across Mozambique A comprehensive review of CA literature in Mozambique was used to analyze the performance of CA technologies in each agro - ecological zone of the country. There is a limited amount of research published i n scientific journals on conservation agriculture in Mozambique. To obtain a more complete understanding of the experiences with CA, the 20 literature review also included gray literature including student theses, project reports and research presentations. T he literature was analyzed by selecting the benefits and challenges of CA in each document and then summarizing that information by agro - ecological zone. The For further information see Grabo wski and Mouzinho (2013a). 2. 4.3 . Survey of CA professionals A two round on - line survey of researchers and project managers experienced with CA in Mozambique was used to obtain their perspectives on the importance of specific technologies for achieving ea ch of the three CA principles with smallholder farmers. They were also asked about what was necessary to promote CA in a way that would result in wide - scale adoption by prioritizing lists of potential research, development and policy activities developed a t a previous workshop. A list of 43 individuals with a diversity of backgrounds was developed based on their experience with CA in Mozambique. Most of these individuals were researchers or development agency project managers, though a few were also from th e private sector and educational organizations. Thirty - five of the 43 CA professionals responded to at least one round of the survey (30 in round 1 and 25 in round 2 with 20 responding to both rounds). The survey was developed based on the Delphi methodol ogy (Turoff, 2002) where respondents express their opinions about a topic and explain their reasons for that opinion in the first round. These results are then summarized so that respondents can see the opinions and arguments of others. In the second round questionnaire respondents can adjust their opinions or clarify their arguments. In theory the rounds can continue until the results have stabilized either in consensus or entrenched disagreement. In this case only two rounds were possible in the given tim eframe. The questionnaires are available as supplemental files with this dissertation. 21 The closed ended questions were analyzed by tabulating responses and comparing differences between the first and second survey. Responses to open ended questions were co mpiled and summarized to represent the diversity of responses. Some quotes from these information see Grabowski and Mouzinho (2013b). 2. 4.4 . In - depth interviews I pr esented the results of the expanded inventory of CA projects, literature review of CA evidence and the survey results to the CA working group in Maputo with participation from a variety of other CA stakeholders. At this meeting the group identified some of the themes emerging from the reports, including the need for local adaptation of CA technologies and better coordination between researchers and development practitioners. To pursue these ideas further, I carried out in - depth interviews with CA researcher s and project managers involved in Farmer Field Schools or establishing linkages across the value chain. The interviews were recorded and transcribed. Summary statements were then developed on the theme of challenges faced in implementing a participatory a nd/or collaborative approach to CA adaptation. 2. 5. Results The results of this study are organ ized into four sections. First, I present evidence of the context - participation i n the CA adaptation process. Next, I participation in current research and promotion. In the third section I present lessons from nascent efforts to link actors across the value - chain for CA adaptation. Finally, I present the anticipated challenges of using innovation networks for CA adaptation. 22 2. 2. 5.1.1 Promotion and research of conservation agriculture CA promotional and research efforts are widespread across Mozambique, with active programs in at least 84 of the 128 districts of Mozambique (81 districts with promotion and 33 districts with research, Figure 1). The largest concentration of organizations for both research and extension is in Manica provi nce in the districts surrounding Sussendenga research station. Most CA promotion in Mozambique emphasizes minimizing soil disturbance as the first and most essential component. Manual CA systems of reduced tillage predominate, including basins (a grid of holes dug in an otherwise undisturbed field) and direct seeding. Animal - based CA systems are only promoted in areas where cattle populations are large such as parts of Manica and Gaza provinces. Of the 29 development agencies promoting CA with farmers, 16 of them promote the use of herbicides and inorganic fertilizers while the other 13 promote CA without commercial inputs. All of the research organizations use commercial inputs for their experimental trials. 23 Figure 1 : Number of or ganizations promoting and researching CA in Mozambique by district Source: Inventory of CA organization in Mozambique , 2011 Note: Does not include FAO and DNEA who report working in all provinces 2. 5.1.2. The need to adapt CA to local conditions Responde nts generally agreed that the existing CA technologies are not ready for widespread dissemination but need significant local adaptation. In the first round I asked respondents their opinion about if a profitable form of CA had been identified that would le ad to wide - scale farmer adoption given proper extension and minor local adaptation. Most sure. One respondent commented that it is not possible to develop a single for m of CA for the diverse agro - ecological zones, and that significant local adaptations were necessary. In the second round, I question - of respondents (27%) chose this new response category with another 55% saying that more research was needed (Figure 2). 24 - planters with herbicides were ready for wide - scale promotion. Figure 2 : CA researchers already succeeded in identifying a profitable form of CA that will lead to large - scale farmer adoption given proper extension efforts and minor adaptations to the local Source: Second round survey of CA researchers and professionals, 2012 The importance of adapting CA technologies to the local context is also highlighted in the responses to the importance of various minimum tillage technologies for CA. In both rounds, manual forms of minimum tillage were ranked as the most important, with respondents explaining that manual agriculture is predominant in Mozambique, so these are likely to be the ones that can lead to wi despread adoption over the short term. There were mixed opinions about the importance of basins, as some saw basins as too labor intensive, and inappropriate for sandy soils. Respondents emphasized that context - specificity is important, and in certain area s animal traction, and even tractor power, can be useful for smallholder farmers in Mozambique. 0 2 4 6 8 10 12 14 No. More research is required Yes Not sure Bad question - CA needs to be locally adapted Number of responses 25 - scale use of CA highlight the need for nologies to meet the needs of resource - poor farmers. Priority research activities included adoption/disadoption studies in different agro - ecological zones and socio - economic studies. This suggests the importance of understanding context - specificity, as wel perspectives and motivations. Respondents also prioritized more farmer - led development initiatives and long - term projects (greater than 5 years), as are often necessary for participatory projects. Dissemination activities that were prioritize d included the establishment of demonstration plots and training for extension workers (both public and private). Presumably these must come after appropriate CA technologies have been developed. In addition, respondents also emphasized the need for long - term agronomic and soil science research to better understand the subtle and hard to measure effects of implementing CA principles. This combination of prioritizing both localized adaptation and long - term scientific research highlights the importance of a participatory approach that draws on both the existing CA technologies. 2. 5.1.3. The context - specificity of CA performance A wide range of CA technologies have - ecological zones. The performance of CA, and thus its relative utility for smallholder farmers, depends on the agro - ecological context and how it fits with the dominant cropping system (Table 3). Across northern Mozambique many development agencies are promoting CA to increase yields of a diversity of crops. CARE is promoting CA with cassava in Nampula with an emphasis on mulching and intercropping with legumes as well as minimum tillage land 26 preparation. The Aga notable for its effective weed control by using grass cut from fallow lands for mulch (Dambiro et al., 2011), however, this is not possible in areas where livestock and fires dramatically reduce t he availability of dry season biomass. CA work in central Mozambique has focused on how to increase maize yields with researchers focusing on combining CA with fertilizer and herbicides, and NGOs promoting CA without these commercial inputs. One notable c hallenge is high termite activity on CA plots at Sussendenga research station (Famba, 2011). Despite high variability in rainfall between years, CIMMYT trials show long - term yield benefits from CA, except during poor rainfall years (Thierfelder and Nyagumb o, 2011). Research on maize - pigeon pea intercropping with CA found increased land productivity and reduced risk of crop failure (Rusinamhodzi et al., 2011). Nkala et al. (2011) emphasize how farmers in this region are actively redesigning CA packages to fi t their needs and assert that a participatory approach to adapting the technologies is needed. Southern Mozambique has lower rainfall than the rest of the country and CA has been promoted primarily as a strategy for water conservation. One project document ed low adoption of basins due to the challenge of them collapsing too easily in the sandy soils near the coast (Sampath, 2011). Research on CA in the especially arid interior of the south has focused on mulch and basins to improve rainfed crop yields by in creasing water availability, but the results are not conclusive. In this region, manual CA was resisted, in part because of rigid promotion of specific technologies fails to result in adoption. 27 Table 3 : CA experiences by province Province Region Primary agriculture system targeted Unique opportunities for CA Unique challenges for CA Cabo Delgado North Maize low input Mulching with grass Nampu la North Cassava Lack of research on CA with cassava Sofala Central Horticulture Manica Central Maize high input Adequate rainfall, Some animal traction Termites, input pries Tete Central Maize high input Input prices Gaza South Maize and cowpe a Animal traction Very arid Inhambane South Maize low input Sandy soils, arid Maputo South Horticulture Source: Inventory of CA projects in Mozambique, 2011 Note: There is n o data for the northern provinces Niassa and Zambezia 2. of participation in CA adaptation The on - farm research that was reviewed in the literature on CA in Mozambique is typically carried out at the consultative level. In most cases farmers are contracted to manager researcher - fields with feedback from farmers on the results. There were no cases of CA research implemented at the collaborative level of participation where farmers were involved with researchers in designing and implementing the research as well as interpreting the focus on producing peer - reviewed articles and from their lack of training in facilitating Nevertheless, collaborative levels of participation in the adaptation process were obse rved in some exceptional CA promotional efforts. These efforts used the Farmer Field School (FFS) approach to evaluate and improve the CA technologies promoted by their 28 development projects. In the FFS methodology a group of about 20 to 30 farmers particip ate in regular meetings in a field with an outside facilitator to compare and adapt promising agricultural practices (Waddington et al. 2014). While the institutions using this approach (NGOs and extension) have expertise in working with farmers to adapt t he technologies, their lack of formal agricultural research skills means they are generally lacking the ability to determine how such adaptations will affect long - term soil fertility. Interviews with program managers from four institutions (the National Union (UNAC) the National Agricultural Extension Directorate (DNEA), the Aga Khan Development Network (AKDN), and CARE) in Mozambique were used to document the challenges they face in implementing programs with higher levels of participation for CA adaptation. UNAC uses FFS and farmer - to - farmer visits to provide training in a broad range of sustainable agriculture practices. Through the interview it was clear that UNAC shows strong commitment to working with farmers to find immediate solutions to agricultural problems but it is the least committed of the four institutions to CA technologies. The program manager who was interviewed explained that the focus instead is on farmer empowerment, combined with environmental sustainability. FFS provides th e forum for fine - tuning the technologies to UNAC simply does not promote it. After promoting CA with a technology transfer approach since the mid - 1990s DNEA has shifted to making FFS the primary extension methodology for the nation. In the interview, it was explained that the goal is to have farmers decide on the curriculum. ma king comes largely at the planning and evaluation stages. The emphasis on a set amount of 29 is run once the curriculum is chosen. AKDN has been using FFS as its o nly methodology for agricultural training for over four years, with 7000 farmers in 248 groups learning about CA as well as basic agricultural concepts. The AKDN program manager explained that the farmers decide how to set up the experiments according to t heir own priorities and ownership of the process is emphasized. One of the key implementation challenges identified is the long timeframe required for developing the ownership of the group by its members and for training farmers in basic science. CARE is actively using FFS to carefully evaluate and adapt specific CA packages that ownership of how the trials are designed. The experimental comparisons are the same for all commu nities and chosen by the project managers to be able to compare the performance of specific technologies - such as different types of cover crop or different varieties of cassava. The main challenge identified by the program managers in the interview was t he increased amount of staff time required to implement FFS compared to their other agricultural training programs. These four experiences using FFS with CA provide a number of lessons about the potential and the challenges of using participatory research at the collaborative level to develop CA technologies. First, it is clear that in order to adapt CA technologies effectively, there needs to be a balance between commitment to specific technologies and commitment to of commitment to CA allows them to be highly many specific CA practices can provide useful information for adapting CA but requires 30 and their motivations to do so. Another lesson is that when FFS is established rapidly as a means of training it is less likely to provide the type of collaborative engagement with farmers needed in the early stages of technology development. Researchers would be better off working with a few groups, over a long timeframe, early on and then use larger numbers of grou ps for the fine - tuning of high potential technologies. Though it may seem that such an approach is not defensible when compared to the numbers that could be reached at the same cost using less intensive communication strategies, it is important to remember that once specific technologies are developed they can be spread with less focused effort to farmers in the same recommendation domain and these technologies would have a greater chance of being adopted. 2. 5.3. Lessons from nascent efforts to link actors across the value chain importance of an innovation systems perspective that links actors across the value - chain. One of the prioritized policy actions was ensuring that bo th input and output markets work better for smallholder farmers. In Mozambique value - chains are relatively weak and undeveloped. Many forms of CA require commercial inputs such as equipment, herbicides, improved seeds and chemical fertilizer. Even with low input forms of CA there is increasing evidence that systems so that farmers can respond to market demand (Benfica et al., 2014). Nevertheless, most CA promotional efforts in Mozambique either focus only on the farmer or try to improve one link in the value - chain, such as helping farmers market their actors across the val ue - chain in association with CA. SIMLESA (Sustainable Intensification of Maize and Legumes in Southern Africa) is a research project that uses innovation 31 platforms, and ECA (Empresa de Comercialização Agrícola) is a contract farming operation that links fa rmers with inputs, credit and markets. participation, the interviewee emphasized that the innovation platform aspect of the project is highly collaborative. Four active innova tion platforms link farmers with agro - dealers, NGOs and grain buyers to reduce bottlenecks in production across the value - chain. Farmers have been enthusiastic about this, and in three of the IPs they have identified the high costs of inputs and the challe nge of selling outputs as their main constraints. Information about prices was the key production constraint identified by farmers in the fourth IP. As a result of the IPs, agro - dealers have been learning how to provide products demanded by farmers and ov er time they have become motivated to participate in the forum for their own benefit. The output buyers however have been less enthusiastic because the organized farmers have been trained to negotiate for higher prices. This highlights the key challenge of motivating participation across the value - chain. Traders of agricultural outputs may be more motivated where they benefit from farmers being organized, such as where monitoring quality is important or where the timing of bulk sales requires coordination. The real potential for inducing innovation through improved coordination along the value - chain can be seen in the success of the contract farming firm ECA, which provides farmers with input loans at cost and coordinates linkages between input suppliers an d groups of farmers. The interviewees explained that the input suppliers are becoming more sensitive affordable. Because ECA has set up contracts with large - scale buyers of grain, it can provide farmers with a guaranteed price from the start of the season. This allows farmers to reduce the risk of investing in inputs. Variation in climate is a major production risk, and for this reason ECA has chosen to train all of its farm ers in CA through demonstration plots. One of 32 the technical challenges identified by the ECA interviewees has been the inability to retain mulch on fields through the dry season due to uncontrolled brush fires. 2. 5.4. Challenges anticipated for collaborat ive research to adapt CA Using an innovation systems perspective for developing CA technologies using participatory research will require collaboration at two levels: between researchers, extension and private sector actors across the value - chain and betwe en researchers and resource - poor farmers. Any collaborative effort on CA in Mozambique will have to be aware of and either resolve, or learn to live with, polarized disagreements on two key issues: the importance of emphasizing minimum tillage and the role of commercial inputs for CA. R esearchers will also need institutional support to carry out this collaborative effort. 2. 5.4.1. Debating the emphasis on minimum tillage During the inventory, it became clear that several NGOs promoting CA in Mozambique wer e not emphasizing minimum tillage, though it is clearly a defining feature of CA. To explore the range of opinions on this issues, in the first round survey I included the possibility of incorporating residues through tillage with the questions about the i mportance of various crop residue management practices. All respondents stated that maintaining incorporating residues (thus requiring full tillage) were polarized. Thirty fo ur percent of respondents said incorporating residues was very important and explained that it is much better for the soil than burning the residues. Another 30% said incorporating residues was not important and argued that tillage is incompatible with CA and should not be practiced. Because of this divergence, in the second round I asked respondents for their opinions regarding the benefits and challenges of promoting CA without emphasizing minimum tillage. In terms of benefits, respondents stated that t hese practices are more easily adopted because farmers can continue doing their familiar land preparation with tillage, but with the 33 added benefit of mulch. Mulching helps control weeds, reduce erosion and retain moisture, though it may also require additi onal labor. disturbance is not emphasized. One explained it this way: the land at the same time. It is about renewing and maintaining the soil structure. Respondents pointed out that promoting CA without emphasizing minimum tillage means ignoring the problems of erosion, loss of soil organic matter and the loss of soil structure associated with tillage. Others argued that the benefits from the other two principles (mulching and rotation with legumes) would be less than if minimum tillage were achieved as well. Furthermore, farmers would still have the wo rk of tilling and have an additional task of adding mulch. While minimum tillage need not be emphasized as the first CA principle (though it re is a risk in being overly prescriptive if CA is defined too narrowly, there is also a risk of the term becoming meaningless if every improved agricultural practice can be labeled as CA (Andersson et al., 2014). Instead, where minimum tillage is not poss ible for farmers, technologies other than CA can be promoted, even if the theoretical benefits are less. The disagreement on the importance of minimum tillage shows a divide between those researchers who focus on the hard system (sustainability of the soil ) and NGOs, and policies). The heart of the matter is the tension between what agronomic research 34 suggests as the best way to manage the soil, and what farmers are w illing and able to actually do, given their priorities and constraints. Researchers that study the hard system tend to use positivist reductionist paradigms of science (where it is assumed there is one universal pool of knowledge and that reality is best understood one element at a time). Their research focuses on how to overcome specific technical challenges. Researchers that focus on the soft system tend to have a constructivist holistic paradigm of science (where problems are ill - defined and multiple ty pes of knowing are valued). Simply recognizing these differences in scientific paradigms may help agronomists, social scientists, NGOs and extension collaborate more effectively (Eigenbrode et al., 2007). For example, promotion of reduced tillage with larg e - scale commercial farmers in Queensland, Australia failed to result in adoption even when the ideal agronomic practice had been developed using positivist science. Adoption followed only when scientists worked with extension to develop adult learning tool s to help farmers understand why the technology was necessary for improving their production (Hamilton, 1998). 2. 5.4.2. The role of inputs in CA promotion Divergent opinions about the importance of commercial input use with CA is another area of tension t hat can constrain collaboration. Most scientists and some development practitioners see fertilizer, herbicides and improved seeds (such as hybrid maize) as key tools for modernizing the smallholder sector. But other development practitioners see them as pr environmental sustainability and social equity. One third of the respondents (eight out of 24 in the second round) stated that CA without these inputs was not even fe asible. A few explained that the high C:N ratio of cereal mulch requires increased nitrogen fertilizer. Others pointed out that herbicides were needed at 35 the beginning to effectively control weeds without tillage. One stated that without chemical inputs: back to conventional tillage, which controls weeds and improves decomposition of crop residues and release of nutrients leading to higher Another five of the 24 respondents stated that low - input CA was feasible, pointing out that farmers do not have access to inputs so this is the only option available for most in Mozambique. Other respondents emphasized that they have observed CA benefits even without purchased inputs. Fer half the respondents, with a quarter saying they were very important, and another quarter saying they were not important at all (Table 4). While the average rating is neutra l, the wide spread of opinions is the primary concern. Table 4 : perspectives on the importance of commercial inputs Fertilizer Herbicide Seed 1. Not Important 5 4 6 2. 1 2 1 3. Somewhat Im portant 8 9 9 4. 5 3 0 5. Very Important 4 4 5 Rating Average 3.09 3.05 2.86 Response Count 23 22 21 Source: First round survey of CA researchers and professionals, 2012 Arguments against chemical fertilizers were that they are expensive or unavailabl e and there is some risk in not seeing the benefit on a bad rainfall year. Arguments for chemical 36 fertilizers emphasized how they work together with CA to show greater benefits (yields) for all the effort the farmer has put in to improving soil quality. A rguments against herbicides included the need for training and the fear of health and not available and better left out of the equation. It can only harm the environment they are highly beneficial for increasing labor productivity. Hybrid seeds were seen by some as irrelevant because of the good quality of open pollinated varieties (OPVs), though access to these seeds is not necessarily reliable. Others were more emphatic about their disapproval: Under no circumstances will this benefit anyone except the seed companies. Seed supply is probably one of the least developed links in Mozambican agriculture. It is CERTAINLY NOT TO THE POINT that farmers should be encouraged to Some who rated hybrid seed as unimportant clarified that it is not relevant to crops like cassava, though they did point out that improved varieties are needed there too. Those be for food - insecure smallholder farmers. This divergence of opinions is especially noticed in maize - based systems where fertilizers can dramatically boost yields, though inputs such as herbicides are relevant to cotton and cassava systems as well. There are many biological and economic arguments that can be made on both sides of this debate. Those who favor commercial inputs tend to focus on yield potential and the subsequent profits from marketing that production. In contrast, those who favor low - input CA emphasize self - sufficiency and environmental integrity. While there is growing recognition of the importance of agro - ecological approaches (IAASTD, 2008), biotechnology and 37 commercial interests ha ve dominated agricultural research in developed countries (Vanloqueren and Baret, 2009). It is important to recognize that these divergent opinions do not necessarily stem from scientific uncertainty, but reflect differences in values, priorities and worl dviews. A good first step for collaboration is helping all sides to listen and understand each other, realizing that effective collaboration does not require consensus on these issues, but rather respecting each From a pragmatic pers pective low - input CA technologies have the short - term advantage as long as weeds can be effectively controlled. In Mozambique, where commercial input use is low because it is largely unavailable and unaffordable, it is logical to start with technologies th at only require inputs that can reasonably be made available at an affordable price. Nevertheless, the value - chain perspective emphasizes that input availability and prices are not fixed, and collaboration can reduce the barriers to their use. 2. 5.4.3. In stitutional support for collaborative CA adaptation For the Mozambican Ministry of Agriculture to support a process of participatory collaborative agricultural innovation, it will have to face the institutional challenges of managing evolving research pro cesses that are tightly linked to non - research stakeholders. Organizational change from a hierarchical bureaucracy to an egalitarian learning - focused institution is essential but will require courageous leadership (Matta et al., 2005). The national CA work ing group is a good start at effective collaboration linking researchers, NGOs and extension together. This group has made the first steps in developing regional es. Collaborative efforts to locally adapt agricultural innovations would also be assisted by bringing the research and extension branches (IIAM and DNEA) into closer coordination. The challenges of achieving such coordination in other countries suggest t hat patience and 38 perseverance will be needed (Merrill - Sands et al., 1991; Biggs, 1995). Establishing effective two - way communication between research and extension requires creativity to join these efforts in their common goal of rural poverty alleviation (Biggs 1995). One of the barriers experienced in other countries is that of valuing extension less than research rather than recognizing the complementarity and interdependence of the two institutions (Buhler et al., 2002). One promising development is tha t the Platform for Agricultural Research and Technological Innovation in Mozambique is considering how to be jointly managed by both the extension and research branches of the Ministry of Agriculture. 2. 6. Conclusion There is widespread agreement that agr o - ecological diversity of Mozambique, and the context specificity of CA technologies, make it especially important to develop suitable CA technologies through local adaptation. However, farmer involvement in CA research in Mozambique is minimal as most on - farm research simply contracts smallholders to manage realized if their involvement in the research process utilizes their implicit knowledge to develop technologies r participation come through closer information sharing as when professional researchers collaboratively engage with the clients who will use the technologies. The Farmer Field School m ethodology appears to be an appropriate forum for this type of intensive collaborative engagement between farmers and researchers. The FFS methodology has been used in several CA promotional efforts in Mozambique, which provides an opportunity to comb ine t he technical skills of researchers with the skills of extension and NGOs to faci It is timely that the national extension directorate (DNEA) is up - scaling FFS as its primary extension methodology. However, for Farmer Field Sch ools to effectively function for participatory research on CA 39 technologies they must be implemented in a bottom - meaningful contribution to decision - making. Including innovation networks to integrate efforts across the v alue - chain is especially important for adapting CA technologies in Mozambique where commercial inputs are largely unavailable. Locally adapting CA technologies will be more likely to result in widespread adoption if linka ges across the value - chain can incr have skills in facilitating group processes, understanding multiple perspectives and resolving conflicts. Innovation net works offer the potential to accelerate the innovation process by integrating the efforts of researchers, development agencies, farmers and actors acr oss the value - chain. However, e ffectively using innovation networks for CA adaptation in Mozambique will r equire coming to terms with polarized disagreements on two key issues: the importance of emphasizing minimum tillage and the role of commercial inputs for CA. It will also require overcoming the institutional challenges of managing evolving research proces ses that are tightly linked to non - research stakeholders. Collaboration takes effort and the returns to this investment in developing specific technologies can be maximized if areas with relatively large recommendation domains are targeted. Researchers mus t develop strong links with advisory support organizations (NGOs and extension) from the beginning so that dissemination strategies become part and parcel of the technology development process. History shows that farmers around the world continuously inno vate and adapt technologies as well as they are able. Research and development agencies have the opportunity to support this process of adaptation with CA in Mozambique and improve the livelihoods of some of the poorest people in the world. 40 RE FERENCES 41 REFERENCES AGRA. 2009. Strategy for an African green revolution: Alliance for a Green Revolution in Africa. Retrieved from: http://www.agra - alliance.org/section/about/agrast rategy or Conservation Agriculture and the politics of its promotion in African smallholder farming. In: Sumberg, J. and Thompson, J. (eds.) Contested Agronomy: A gricultural Research in a Changing World. London: Earthscan. agriculture for small - scale farmers in Sub - Ecosyst. Environ. 187 (201 4) 1 10 ]. Agriculture Ecosystems and Environment 196, 21 23. performance of conservation agriculture and current smallholder farming practices in semi - F ield Crops Research 132, 117 - 128. Benfica, R., Boughton, D., Mouzinho, B. and Uaiene, R. 2014. Food Crop Marketing and Agricultural Productivity in a High Price Environment: Evidence and Implications for Mozambique. Research Paper # 76E. Maputo, Mozambique : Directorate of Economics, Ministry of Agriculture. Biggs , S., Correa, C., Rello, F., Bautista, R. , Sala verry, J., Mundlak, Y., ... Errázuriz, M. 1989. Resource - Poor Farmer Participation in Research: A synthesis of experiences from nine National Agricu ltural Research Systems . (No. A50 B593). CEE, México, DF (México) . Biggs, S. , 1995. Farming systems research and rural poverty: Relationships bet ween context and content. Agricultural Systems 47, 161 174. Bingen, J. and Gibbon, D. 2012 Early Farming System s Research and Extension experience in Africa and possible relevance for FSR in Europe. In: Darnhofer, I., Gibbon, D. and Dedieu, B. (Eds) Farming Systems Research in the 21st Century . Springer, New York. Bread for the World Institute. 2011. Our common int erest: Ending hunger and malnutrition, 2011 hunger report, 21st annual report on the state of world hunger.Washington D.C.: Bread for the World Institute. Retrieved from: http://file s.hungerreport.org/reports/2011/hunger - 2011.pdf Buhler, W., Morse, S., Arthur, E., Bolton, S., Mann, J., 2002. Science Agriculture and Research: A Compromised Participation, 1st edition . Routle VA. Chambers, R., Pacey, A., Thrupp, L.A., 1989. Farmer First: Farmer Innovation and Agricultural Research. Practical Action, London. 42 Chambers, R. (1997). Whose reality counts? Putting the last first , Intermediate Technology Publica tions, London. Conroy, C., Sutherland, A., 2004. Participatory technology development with resource - poor farmers: maximising impact through the use of recommendation domains. Overseas development institute (ODI). Agricultural research & extension network ( AgREN). D ambiro J., Xavier F., Vasco B. Azito M. 2011. Introducing Conservation Agriculture in the Quirimbas National Park of Cabo Delgado, northem Mozambique . Paper presented at the 5th World Congress of Conservation Agriculture, September 2011 Brisbane, Australia . Morse, W., Nielsen - Pincus, M., Stephens, J., Winowiecki, L., Bosque - Pérez, N.A., 2007. Employing Philosophical Dialogue in Collaborative Science. BioSc ience 57 (10), 55 64. doi:10.1641/B570109 Ekboir, J. 2002. Developing No - Till Packages for Small - Scale Farmers. In: Ekboir, J. (ed.) CIMMYT 2000 - 2001 World Wheat Overview and Outlook: Developing No - Till Packages for Small - Scale Farmers. Mexico, DF: CIMMYT. Ekboir, J. 2003. Research and technology policies in innovation systems: zero tillage in Brazil. Research Policy 32 pp 573 - 586. Erenstein, O., Sayre, K., Wall, P., Hellin, J., Dixon, J., 2012. Conservation agriculture in maize - and wheat - based systems in t he (sub) tropics: lessons from adaptation initiatives in South Asia, Mexico, and Southern Africa. Journal of S ustainable A griculture 36, 180 206. Famba, S. 2011. The Challenges of Conservation Agriculture to Increase Maize Yield in Vulnerable Production Sy stems in Central Mozambique, Dissertation from University of Natural Resources and Life Sciences, Viena,125pp; Feed the Future. 2009. Mozambique: FY 2010 implementation plan. Retrieved from: http://www.feedthefuture.gov/documents/FTF_2010_Implementation_Plan_Mozambi que.pdf Fujisaka, S., 1994. Will farmer participatory research survive in the international agricultural research centres? Gatekeeper Series No.SA 44 . (Lond on: IIED), pp. 3 - 14. Giller, K.E., Witter, E., Corbeels, M., Tittonell P. 2009. Conservation Agriculture and Field Crops Research 114: 23 34. Grabowski, P. and Kerr, J. 2014. Resource constraints and partial adoption of conservation agriculture by hand - hoe farmers in Mozambique. International Journal of Agricultural Sustainability 12 (1) 37 - 53. Grabowski, P., Haggblade, S. Kabwe, S. and Tembo, G. 2014. Minimum tillage adoption among commercial sma llholder cotton farmers in Zambia 2002 to 2011. Agricultural Systems 131, 34 - 44. 43 Grabowski, P., and Mouzinho, B. 2013a. Mozambique conservation agriculture inventory report. IIAM working paper 6E. Grabowski, P., and Mouzinho, B. 2013b. Prioritizing actio ns for conservation agriculture in Mozambique. IIAM working paper 5E. Hamilton, G. 1998. Co - learning tools: powerful instruments of change in Southern Queensland, Australia. In: Röling, N.G. and Wagemakers, M.A.E. (Eds .) Facilitating Sustainable Agricult ure - Participatory Learning and Adaptive Management in Times of Environmental Uncertainty. Cambirdge, UK: Cambridge University Press. Hanlon, J., and Smart, T. 2008. Do bicycles equal development in Mozambique? Woodbridge, UK: James Currey. Hobbs, P. R. 2007. Conservation agriculture: what is it and why is it important for future sustainable food production. Journal of Agricultural Science , 145 (2), 127 - 137. IAASTD, 2008. Executive Summary of the Synthesis Report of the International Assessment of Agricul tural Knowledge, Science and Technology for Development (IAASTD). Available on http://www.agassessment.org . agriculture: Justi International Journal of Agricultural Sustainability 7 (4), 292 - 320. Klerkx, L., Aarts, N., Leeuwis, C., 2010. Adaptive management in agricultural innovation systems: The interactions between innovation networks and th eir environment. Agricultural Systems 103, 390 400. Klerkx, L., van Mierlo, B. Leeuwis, C. 2012. Evolution of systems approaches to agricultural innovation: concepts, analysis and interventions. In: Darnhofer, I., Gibbon, D. and Dedieu, B. (Eds) Farming Sy stems Research in the 21st Century . Springer, New York. Matta, J.R., Kerr, J., Chung, K., 2005. From forest regulation to participatory facilitation: Forest employee perspectives on organizational change and transformation in India. Journal of Environ ment al Plan ning and Management 48, 475 490. Merrill - Sands, D.M., Biggs, S.D., Bingen, R.J., Ewell, P.T., McAllister, J.L., Poats, S.V., 1991. Institutional considerations in strengthening on - farm client - oriented research in National Agricultural Research Syste ms: Lessons from a nine - country study. Experimental Agric ulture 27, 343 373. Moore, Keith M., ed. 2009. The Sciences and Art of Adaptive Management: Innovating for Sustainable Agriculture and Natural Resource Management. Ankeny, IA: Soil and Water Conserva tion Society. Mozambique Ministry of Agriculture. 2010. Strategic Plan for Agricultural Development: PEDSA 2010 - 2019. Republic of Mozambique, Maputo. Retrieved from: http://fsg.afre.msu.edu/ mozambique/caadp/index.htm 44 Newitt, M. 2002. Mozambique. In: P. Chabal (Ed.), A History of Postcolonial Lusophone Africa . Bloomington, IN: Indiana University Press. Nhancale, I. 2000. Enhancing the maize productivity of small scale farmers through zero t illage practices in Manica province, Mozambique, Thesis from University of Cape Coast, Ghana; Nkala, P., Mango, N. and Zikhali,P. 2011. Conservation Agriculture and livehoods of smallholder farmers in central Mozambique, Journal of Sustainable Agriculture 35:757 - 779 International Journal of Agricultural Sustainability 9(1), 5 - 24. Rajalahti, R., Janssen, W., Pehu, E., 2008. Agricultural innovation systems: From diagnostics toward operational pract ices. Agriculture and Rural Development Department, World Bank. Rivera - Huerta, R., Dutrénit, G., Ekboir, J.M., Sampedro, J.L., Vera - Cruz, A.O., 2011. Do linkages between fa rmers and academic researchers influence researcher productivity? The Mexican case. Res earch Policy 40, 932 942. Rockström, J., Kaumbutho, P., Mwalley, J., Nzabi, A.W.M., Temesgen, M.L., Mawenya, L. Barron, J., Mutua, J., Damgaard - Larsen, S. 2009. Conserva tion Farming Strategies in East and Southern Africa: Yields and Rainwater Productivity from On - farm Action Research. Soil and Tillage Research 103: 23 - 32. Rusinamhodzi, L., Corbeels, M., van Wijk, M., Nyamangara, J., Rufino, M. and Giller, K. 2011. Produc tivity of maize - legume intercropping under no - till in central Mozambique: Challenges and Opportunities. Regional Conservation Agriculture Symposium. 8 - 10th February 2011. Johannesburg, South Africa. pp 21; Sampath, P. and Celaya, R. 2011. CARE Mozambique C onservation Agriculture Project Thierfelder, C., Rusinamhodzi, L., Ngwira, A.R., Mupangwa, W., Nyagumbo, I., Kassie, G.T., Cairns, J.E., 2014. Conservation agriculture in Southern Africa: Adva nces in knowledge. Renewable Agriculture and Food Systems FirstView, 1 21. doi:10.1017/S1742170513000550 Turoff, M. (2002) The Policy Delphi. In: Linstone, H and Turoff, M. (Eds.) The Delphi Method: Techniques and Applications. Retrieved from: http://is.njit.edu/pubs/delphibook/ch3b1.html UNICEF. (2011). Mozambique statistics. Retrieved from http://www.unicef.org/infob ycountry/mozambique_statistics.html van der Lee, F., Kayula, F., Makasa, V., Heemskerk, W. 2011. Conservation agriculture in Zambia. In: Nederlof, S. Wongtschowski, M., van der Lee, F. (Eds. ) Putting heads together: Agricultural innovation platforms in practice. Bulletin 396, KIT Publishers 45 Vanloqueren, G., Baret, P.V., 2009. How agricultural research systems shape a technological regime that develops genetic engineering but locks out agroecological innovations. Research Policy 38, 971 983. von Grebmer, K., D. Headey, C. Béné, L. Haddad, T. Olofi nbiyi, D. Wiesmann, H. Fritschel, S. Yin, Y. Yohannes, C. Foley, C. von Oppeln, and B. Iseli. 2013. 2013 Global Hunger Index: The Challenge of Hunger: Building Resilience to Achieve Food and Nutrition Security. B onn, Washington, DC, and Dublin: Welthungerhilfe, International Food Policy Research Institute, and Concern Worldwide. innovation. Management Science 40, 4 pp 429 - 439. Waddington, S.R., 1993. Overview of the current situation and previous impact of adaptive agricultu ral research in southern Africa. I n: Craswell, E., Simpson, J. (Eds.) Soil Fertility and Climatic Constraints in Dryland Agriculture: Proceedings of ACI AR/ACCAR Workshop held at Harare, Zimbabwe. Australian Centre for Internationa l Agricultural Research, pp. 7 14 . Waddington, H., Snilstveit, B., Hombrados, J.G., Vojtkova, M ., Anderson, J., White, H., 2014 . Farmer field schools for improving farming practi ces and farmer outcomes in low - and middle - income countries: a systematic review. Campbell Systematic Reviews 2014:6. Retrieved from: http://www.campbellcollaboration.org/lib/project/203/ Wall, P., 2007, Tailoring conservation agriculture to the needs of smallholder farmers in developing countries: An analysis of issues, In: M. Kang (Ed.), Agricultural and Environmental Sustainability: Considerations for the Future . Haworth Food and Agric ultural Products Press, New York pp. 137 - 155. Whitfield, S., Dougill, A.J., Dyer, J.C., Kalaba, F.K., Leventon, J., Stringer, L.C., 2015. Critical reflection on knowledge and narratives of conservation agriculture. Geoforum 60, 133 142. Witcombe, J.R., Gy awali, S., Sunwar, S., Sthapit, B.R., Joshi, K.D., 2006. Participatory plant breeding is better described as highly client - oriented plant breeding. II. Optional farmer collaboration in the segregating generations. Exp erimental Agriculture 42, 79 90. World Bank. 2007, Agriculture for D evelopment : World Development Report 2008, The World Bank, Washington D.C. Implications for improved maize seed adoption. Prepared for Pr esentation at the International Association of Agricultural Economists Conference, Gold Coast, Australia, August 12 - 18, 2006. Retrieved from: http://ageconsearch.umn.edu/bitstream/25 648/1/cp061100.pdf Zandamela, C.B., Wall, P.C., Nhancale, I.T. (2006) Mapeando o caminho para o futuro de Agricultura de Conservação em Moçambique, Agrodec, 58pp. Available at: http://www.iiam.gov.mz/index.php?option=com_content&view=article&id=176:sinte ses&catid=59&Itemid=111 46 CHAPTER 3: DETERMINA NTS OF ADOPTION OF M INIMUM TILLAGE BY COTTON FARMERS IN EA STERN ZAMBIA 3. 1. Introductio n There is increasing concern about the food security of smallholder farmers as southern Africa experiences unpredictable rainfall patterns and is expected to experience declining rainfall due to global climate change (Lobell et al. , 2008; Boko et al. , 200 7). Growing demographic pressure on farmland and the resulting reduced fallow periods are also causing land degradation, soil erosion and nutrient mining (Todaro and Smith , 2009; Morris et al. , 2007; World Bank , 2007). As efforts proceed to develop and in troduce new agricultural technologies to help mitigate the effects of climate change and land degradation in sub - Saharan Africa, much of the discourse focuses on conservation agriculture (CA) a set of management practices including minimal soil disturban ce, permanent soil cover and rotation with l egumes (FAO , 2001). CA has demonstrated the potential to increase agricultural productivity and food security while preventing erosion and maximizing ecological functions of the soil (Kassam et al. , 2009). Propon ents argue that it is the best means of sustainably managing the soil and reversing land degradation (Rockström et al. , 2009; Kassam et al. , 2009 ). They point to widespread adoption throughout the world 106 million ha worldwide according to Kassam et al. (2009) as evidence of its promise. However, adoption in southern Africa is still at less than 1% of arable land after 20 years of promotion (Hove et al. , 2011), causing many to question its suitability to smallholder farmers (Giller et al. , 2009). This has led to a polarized debate with CA advocates assert ing that low adoption levels are to be expected during the initial phase of an S - shaped diffusio n process (FAO , 2001), and coordinated an d policies are supportive (Frie drich and Kassam , 2011). Critics point out the 47 challenges farmers face in using CA, such as increased weed pressure, competing demands for crop residues and a lack of markets for legumes (Giller et al. , 2009). CA adoption st udies in southern Africa tend to focus on minimum tillage (MT), which is the component of CA that is typically emphasized during promotion. Some adoption studies evaluate program interventions (for example in Zambia Nyanga , 2012; Kasanga and Daka , 2013) but these types of studies are typically biased by focusing on adoption and ignoring the possibility of subsequent dis , 2014). Other studies analyze the determinants of adoption using econometric analysis of multi - purpose h ousehold surveys (Arslan et al. , 2014; Ngoma, Mulenga and Jayne , 2014), but they tend to , 2014). Giller et al. (2011) highlight the need f or mixed In t his study , this gap in the literature is addressed by combining econometric analysis of the determinants of adoption with qualitative data on the motivations b ehind decisions to use or not use MT Eastern Province, the country with the highest number of CA farmers in southe rn Africa and is often seen as the exemplar for CA adop tion (Hove et al. , 2011; Haggblade et al. , 2010). Cotton farmers in Eastern Province provide a sub - population where MT has been successfully promoted and adopted without heavy use of incentives. This allows for analy sis of e constraints to its use where it is know n that promotion has been adequate, the environment is reasonably suitable and adoption is more than a temporary response to material incentives. The study does not aim to estimate national or even provincial adopt ion levels. Assuming farmers make rational decisions about MT given their individual objectives and constraints, I hypothesize that labor, wealth, experience, and technical challenges with 48 The goal is to determine the perceived benefits and limitations of the MT technologies in order to focus innovation and adaptation. The results provide guidance o n how to overcome challenges related to CA adoption and have implications for sustainable intensification efforts more generally. 3. 2. Conservation a griculture 3. 2.1. Minimum t illage The three principles of conservation agriculture (minimal soil disturbance, permanent soil cover and rotating or intercropping with legumes) are complementary in that the overa ll benefits are greatest when they are all used together (Thierfelder et al. , 2013b). In practice, however, farmers prefer to adopt technological packages in a step - wise fashion starting with those that are most beneficial for their own specific situations (Byerlee and de Polanco , 1986; Kasanga and Daka , 2013). In southern Africa the principle of minimizing tillage has received the most attention, often with some neglect of the other two principles (Andersson and , 2014; Baudron et al. , 2007). In Za mbia three specific MT technologies have been promoted - hand hoe basins, ox - drawn ripping and tractor ripping (Grabowski et al. , 2014 ; Nyanga , 2012). Basins are dug in a precise grid and each hole is 20 centimeters (cm) deep, 30 cm long, and the width of a hoe blade. Farmers with animal traction can use a the locally engineered Magoye ripper to open up a trench 5 cm wide and 15 cm deep where the seeds can be sown (Kabwe, Donovan, and Samazaka , 2007). Where tractors are available, tractor - drawn rippers can be used. Minimizing soil disturbance offers two types of important benefits to farmers: improved soi l fertility and increased water - use efficiency. Improvements to the soil tend to be long - term and are the result of reducing the decomposition of soil orga nic m atter and preventing erosion (Rockström et al. , 2009; Verhulst et al. , 2010; Thierfelder, Cheesman, and Rusinamhodzi , 2013a). MT can increase water - use efficiency by improving water infiltration, 49 especially with mulch (Thierfelder and Wall , 2009) or b y breaking through a compacted layer of soil (Haggblade and Tembo , 2003). In addition, MT takes place during the dry season, enabling earlier planting than with plowing which only begins once the rains start. C otton and maize yields tend to increase with e arlier planting on the order of 100 to 200 kg per week (Haggblade and Tembo , 2003; Haggblade, Kabwe, and Plerhoples , 2011). Some of the most commonly identified constraints for minimum tillage include increased labor requirements for basins ( Grabowski and Kerr , 2014 ) and the challenge of weed control without soi l inversion (Giller et al. , 2009 ; Wall 2007 ; Gatere et al. , 2013 ). The retention of dry season biomass is problematic due to uncontrolled fires and grazing by free range livestock (Giller et al. , 20 09; Baudron et al. , 2007). Rotation or intercropping with legumes tend to be low at least in part due to low prices, high seed costs, high labor requirements and poor access to improved varieties (Snapp et al. , 2002). 3. 2.2. Conservation agriculture p romo tion among cotton f armers in Zambia Cotton farmers are the largest group of spontaneous CA adopters among smallholder farmers in Zambia (Haggblade and Tembo , 2003) . Private sector cotton companies , which provide inputs on contract to smallholder growers , h ave actively promoted CA since the late 1990s to increase yields and reduce losses from dry spells (Kabwe, Donovan, and Samazaka , 2007). T he initial emphasis was on basins, in part because cattle corridor disease had reduced the availability of animal trac tion at that time (Haggblade and Tembo , 2003). Starting i n the early 2000s the Magoye ripper was increasingly promoted with cotton farmers (Kabwe, Donovan, and Samazaka , 2007). NWK Agri - services (previously known as Dunavant) and Cargill are the largest c otton companies and strongest private sector promoters of MT ( Haggblade et al. , 2010). NWK uses a system of distributors who are lead farmers that provide training, distribute seed and chemicals, monitor fields and buy the harvest for 50 to 100 cotton farm e rs . They earn a 50 commission from their farmers with bonuses for high volume, yields, and loan repayment rates. Cargill on the other hand employs buyers who may or may not be farmers to carry out similar functions but oversee 200 to 500 farmers. NWK encoura ges each of its distributors to have a CA demonstration plot to use for training in the communities. Cargill buyers hold what are called cotton schools to train farmers on CA and cotton production practices. Both companie s have increased their efforts over the last few seasons as herbicides and equipment have become more available to farmers o n credit (Grabowski et al. , 2014 ). Approximately 64% of small - and medium - scale farmers in Eastern Province grow cotton (Tembo and Sitko , 2013). Cotton is a demanding crop in terms of labor and management with regular pest monitoring and pesticide sprays. On average cotton farming households have larger cultivated areas, own less cattle and earn a larger portion of their income through agriculture than households that do not grow cotton (Haggblade, Kabwe, and Plerhoples , 2011). In addition to the cotton companies, n umerous development agencies have promoted CA in Eastern Province since the mid - 1990s (Baudron et al. , 2007; Arslan et al. , 2014). 3. 2.3. Agriculture in East ern Zambia Eastern Province is a high agricultural potential region w here 24% of all households are smallholders (the highest rate in the country), many of whom a re foo d insecure (Siegel, 2008). The p rovince has a unimodal rainfall pattern with annual prec ipitation varying between 600 and 1200 millimeters between November and May. Though overall population density is relatively low, localized land scarcity exists, especially around large villages. Eastern Province can be divided into two major agro - ecologic al zones. The lower elevation valley zone has lower rainfall, higher temperatures and lower cattle populations because of 51 tsetse fly infestation. The upland plateau regions have greater population density and higher rainfall. This study focuses only on the plateau portion of the province . 3. 2.4. Minimum t illage a doption in Zambia - rainfall zones, national adoption rates remain low (Arslan et al. , 2014; Grabowski et al. , 2014 ). Two separa te surveys suggest that MT use in Eastern Zambia has been expanding gradually , though remaining below 15% of households (Arslan et al. , 2014 ; Ngoma, Mulenga, and Jayne , 2014). Data from cotton farmers in Eastern Province also showed an overall increase in MT use, most of which came from the uptake of ox - ripping while basin use rates we re relatively unchanged (Grabowski et al., submitted). Both promotion and adoption of CA are clustered geogra phically (Grabowski et al. , 2014 ; Kasanga and Daka , 2013) , which m ake it more difficult to precisely estimate province - level MT use (Grabowski et al., submitted) . MT adoption correlates with promotion and higher rainfall variability, suggesting that farmers use MT to reduce their vulnerability to an unpredictable climat e (Arslan et al. , 2014; Ngoma, Mulenga, and Jayne , 2014). For communities where cotton is grown, adoption correlates with greater herbicide availability, longer promotion and better demonstrations by lead farmers (Grabowski et al. , 2014 ). Adoption is ofte n temporary, particularly when development agencies provide material incentives to adopters. Arslan et al. (2014) report that in Eastern Province 88% of the 78 MT users in the sample in 2004 disadopted by 2008. The authors attribute disadoption to the expe ctation of free inputs to use CA and the discontinuation of those incentives, as has been documented elsewhere in Zambia (Ngoma, Mulenga, and Jayne , 2014; Baudron et al. , 2007; Haggblade and Tembo , 2003). 52 3. 3. Data a nd m ethods I used a mixed methods appro ach to gain an in - depth under standing of the factors affecting I used qualitative interviews to both document analyze the determinant s of adoption. I carried out the survey with 245 farmers in 15 communities where CA adoption was expected to be relatively high. I carried out in - depth interviews in the local language, Chinyanja, with 63 farmers and cotton buyers in 10 communities with va rying levels of adoption. I used thematic analysis t o analyze the qualitative data and econometric analysis to analyze the survey data. 3. 3.1. Selection of r espondents 3. 3.1.1. Community s election C ommunities were specifically chosen with relatively high levels of adoption by randomly selecting locations where lead cotton farmers reported MT use rates greater than the overall average where MT was practiced (census of lead cotton farmers described in Grabowski et al. (2014)). This stratification was done to enable analysis of the determinants of adoption where it was known that non - adoption was not simply the result of lack of promotion or the unsuitabi lity of the environment (Table 5) . All communities in this study are in plateau areas of Eastern Province. 53 Table 5 : Community s tratification by c ompany and a doption l evel NWK Cargill Strata Total Surveyed Total Surveyed High tractor ripping 5 4 0 0 High ox ripping 61 2 22 5 High basins 38 1 16 1 Medium MT 116 2 35 0 Low/Zero MT 40 8 0 102 0 Total number of groups 628 9 175 6 Source: Author s calculations using Survey data from NWK distributors and Cargill buyers, 2011 a re defined as twice the mean use rate (from the mean of communities where MT was used by at least one as less than half the mean use rate. 3. 3.1.2. Respondent s election for in - depth i nterviews I carried out in - depth interviews with 34 farmers, 18 distributors from NWK and 11 buyers and chairpersons from Cargill. In addition, I conducted three group interviews with a total of 122 farmers (69 ma les and 53 females) and ad hoc interviews with seven survey respondents. For all randomly selected communities I interviewed the distributor, buyer or chairperson who linked me to the farmers. I conducted most in - depth interviews with farmers in five com munities from the full range of adoption levels (high tractor ripping, high ox - ripping, high basins, and low adoption). In each community I selected farmers for in - depth interviews from each category of land preparation (plow, ox - ripper, hoe, basin user, tractor ripper or disadopter). I us ed group interviews during impromptu data collection opportunities where farmers had assembled to observe the survey process. The participants in the ad hoc interviews were identified as disadopters during the survey and I asked them to explain their experience and motivation for disadopting. 54 3. 3.1.3. Farmer s election for the s urvey D istributors and buyers at the community level were used as key informants to categorize farmers by their most distinctive (unique) land pr eparation method. Most farmers hoe some of their land so only those who exclusively hoe were categorized as hoe farmers. If farmers used any type of MT they were categorized by the MT technology used. If they used ox - ripping and basins they were categorize d as ox - rippers and if they used tractor ripping with any other type of MT they were categorized as tractor rippers. I then randomly selected up to eight farmers in each category from each community and invited them to participate in the survey. Table 6 : Response rates of farmers by stratification c ategory Stratum Selected Surveyed Response Rate Analyzed Actual Practice a Chairman 4 4 100% 4 - Hoe Farmer 52 35 67% 34 34 Plow farmer 62 40 65% 39 118 Basin farmer 22 14 64% 15 33 Ox - ripper 73 62 85% 56 45 Tractor plower 9 6 67% 6 2 Tractor ripper 31 17 53% 17 4 Disadopter 5 4 80% 4 - Unknown 83 63 76% 61 - Total 342 245 72% 236 236 Source: Survey of NWK and Cargill farmers, 2013 a This is how the farmers in the sampl e are categorized based on their actual responses 3. 3.2. Data c ollection 3. 3.2.1. Qualitative d ata c ollection Interviews with farmers focus ed on their farming practices for the previous season before asking about their experiences and perspectives on MT. I asked farmers who used MT about how they learned about it, what they had done previously, their motivation for using MT and their assessment of how MT performs compared to conventional tillage. Farmers 55 who had disadopted MT were asked to explain their re asoning for disadoption. I asked d istributors, buyers and community l eaders about community - level issues related to inputs, production and CA promotion in the area . These key informants also provided their ot using and disadopting MT in the area. I recorded and transcribed all interviews . The in - depth interview guides can be found in Appendices B and C . 3. 3.2.2. Survey data c ollection Four enumerators were hired to assist in conducting the survey using the local language, Chinyanja. Distributors a nd buyers invited the selected farmers t o meet in a central location and those who did not attend were visited at their home as time allowed. The survey response rate was 72%. The questionnaire included questions about farming practices for every plot cultivated during the 2012/13 rainy season. Household - level questions were used to understa nd the household composition, education level of the adults, total landholding, years of cotton experience, crops sold, sourc es of non - agricultural income, livestock ownership, and ownership of agricultural and household items. Farmers also were asked their opinions comparing hoeing to basins and ripping to plowing. The survey instrument can be found in Appendix A . T he total va lue of equipment at the time of deciding whether or not to use MT was calculated by including only equipment that had been owned since before the decision to adopt or disadopt was made. For those who never had used MT only equipment owned over three years was included , which is the average amount of time adopters have used MT. 3. 3.2.3. Community - level d ata c ollection Community - level data were collected from several sources. Distributors and buyers provided information about how long CA had been promoted an d about their own farming 56 I multiplied the percent of area they farmed with MT by the years they have used MT. The percent of farmers using animal traction was obtained from the ce nsus of lead cotton farmers (Grabowski et al. 2014). I obtained population density from the 2010 census information at the ward level and measured elevation using GPS. Some variables were generated by aggregating household - level responses in the communit y, such as average fertilizer application rate for maize and the percent of plots whose residues had been heavily grazed the previous dry season. 3. 3.3. Data a nalysis 3. 3.3.1. Qualitative data a nalysis The qualitative data includes transcripts, researcher notes and comments written on surveys . I coded this data using thematic analysis to facilitate retrieval of similar information across the data. In thematic analysis the researcher systemat ically examines textual data for each code and develops summary s tatements to concisely represent the diversity of responses (Miles and Huberman, 1994; Rubin and Rubin, 2005 ). I use quotes from respondents where possible to succinctly represent the perspectives presented in the data. In addition, numerical tabulations are used to clarify the level of agreement or disagreement about an issue. 3. 3.3.2. Statistical d ata a nalysis While other adoption studies typically explore adoption as a binomial variable (Arslan et al. 2014; Ngoma, Mulenga, and Jayne 2014; Nyanga 2012) this analysis allows for greater insight by differentiating adopters into those who use basins and ox - ripping and differentiating non - adopters into those who disadopted and those who never used MT. I used a multinomi al logistic regression to estimate how marginal changes in household characterist ics affect the probability of being in one of f our categories: 1) o x - ripper farmers (who may also use basins), 2) b asin farmers ( who do not use ox - ripping), 3) d isadopters (anyone who previously used MT bef ore the 2012/13 season) and 4) f armers who 57 have never used any type of MT. Tractor farmers were excluded from the analysis because tractor use among sampled farmers was too low (see Grabowski et al. (submitted) for more details) . A multinomial logistic regressio n estimates how a marginal change in the independent variables will affect the probabilities of fitting into any one category relative to another. The multinomial logistic model can be presented formally as: (equation 1) for land preparation categories m = 1 to J where b is the base category (Long and Freese 2001). 3. 3.3.3. Sampling w eights T o estimate the parameters of the population (cotton farmers in communities with relatively high adoption) , I weighted the observations b y the inverse sampling probability . Using weights in the analysis reduces bias when generalizing to the broader population and cannot be ignored when observations are stratified by the dependent variable (Elliott 2008) as was done here. Outlier weights in this study were trimmed to five times the median weight (following Pedlow et al. 2003). This ad - hoc way of trimming has been shown to be just as effective as more advanced trimming methods that use simulation and modeling (Chowdhury, Khare, and Wolter 2007 ). 3. 3.3.4. Cluster analysis to control for community fixed e ffects To control for unobserved heterogeneity across communities, I used cluster analysis drawing on key indicators of economic and environmental variation. I employed the k - means method of cl ustering to group communities into four groups according to five variables: population density, elevation, the percent of farmers using animal traction, the average fertilizer application rate to maize and the percent of plots where residues were heavily g razed (Table 7 ) . K - means is the most commonly used method for data clustering due to its 58 easy implementation and empirical evidence of its effectiveness (Jain , 2010). Dummy variables for the clusters were then used in the regression to control for fixed ef fects at the community level. Table 7 : Mean values of the characteristics used to cluster communities Population Density (people/square kilometer) Elevation (meters) Percent of farmers using animal traction Average fertilizer appli cation rate to maize (kg/ha) Percent of plots where residues were heavily grazed C l u ster 1 48.96 1006 76.6% 99.6 34.9% Cluster 2 140.53 1106 67.2% 308.0 29.6% Cluster 3 83.3 833 41.8% 273.6 32.0% Cluster 4 50.1 1012 65.6% 187.4 31.8% Source: Survey of NWK and Cargill farmers, 2013 3. 4. Results a nd d iscussion The results are organized to specifically draw on the strengths of both the qualitative and quantitative research methods. First I present quantitative data on the level of adoption to provide a clear idea of how farmers are using MT. Next, I summarize the qualitative analysis I present econometric analysis of the characteristics of those who use MT, disadopt MT and have never used MT, which provides statistical information about the relative importance of the constraints and motivations identified through the qualitative analysis. 3. 4.1. Adoption r ates Even in the surveyed areas where relatively high adoption wa s expected , 52 % of the farmers have never tried any form of MT and another 24% tried it previously but did not use it during the 2012/1 3 season (Table 8 ). Of the 24% of farmers who are using MT, about half are using ripping and half are using basins with a few using both. 1 Only an estimated 12% of cotton area and 20% of hybrid maize area were prepared using MT (Figure 3 ). Groundnuts 1 For details on how farmers in this dataset combine MT with the other principles of CA as well as other agron omic pract ices see Chapter 4 . 59 and other crops are only rarely planted on MT plots, which contrasts with another study that found over 20% of MT users with o ne fifth of their MT plots being planted to legumes (Kasanga and Daka , 2013). Table 8 : Land preparation method use rates and rates of disadoption for cotton farmers in areas of medium and high a doption in Eastern Province Proporti on of households 95% confidence interval Basins as only form of MT 12.8% (4.3% - 21.3%) Ox - ripping with or without basins 11.8% (6.1% - 17.5%) Both basins and ox - ripping 2.0% (0.1% - 4.9%) Disadopted all MT 23.7% (15.3% - 32.0%) Now only h oes 5.3% (0.6% - 10.0%) Now plows 18.4% (8.5% - 28.3%) Never used MT 51.7% (27.3% - 76.2%) Only hoes; never used MT 17.1% (10.5% - 23.7%) Plows; never used MT 34.6% (13.4% - 55.9% ) Total 100% Source: Survey of NWK and Cargill farmers, 2013. Note: The rows in italics are additional ways of grouping the data. 60 Figure 3 : Percent of area under each land preparation method for the four largest c rops Sourc e: Survey of NWK and Cargill farmers, 2013 3. dopting MT to improve their yields and reduce their vulnerability to drought. Fifteen out of 20 key informants (distributors or buyers) speci fically mentioned drought tolerance as a motivating factor. Now these days the rains are less and with MT 2 , even if the rains are less they rains were still falling. (Key informant 3) 2 Gamphani is the Chinyanja word commonly used for minimum and it is translated as M T for all quotes in this study. G amphani ya maenje (M T with hol es) and (MT with cattle) are translated as basins and ripping respectively . Tractor ripping 0.19% Ox - plowing 59.96% Ox - ripping 10.34% Hoeing 27.22% Basins 1.33% Direct seeding 0.96% Cotton Tractor ripping 5.61% Ox - plowing 63.83% Ox - ripping 12.37% Hoeing 15.22% Basins 2.42% Hybrid Maize Ox - plowing 65.1% Ox - ripping 1.5% Hoeing 33.3% Basins 0.1% Groundnuts Ox - plowing 52.07% Ox - ripping 8.03% Hoeing 30.25% Basins 5.56% Direct seeding 4.09% Local Maize 61 One NWK farmer explained his recent adoption saying , My friends were harvesting very well despite drought, it was resistant and it still grew very well and thrived, good maize, healthy maize. So I thought, let me take it as well. (Key informant 14) Early planting is a key aspect of MT that helps farmers achieve the goals of higher yields and drought tolerance (Haggblade and Tembo , 2003). Seven out of 20 key informants specifically mentio ned early planting as a motivation for using CA. A s one Cargill buyer put it: Those who have rippers and use this method, they recommend it because they plant early. Because for them by the time the rains come they have already done the ground work. So they can plant early, the weeding is done early. Th e production is higher moisture. So their crops, despite the dry spell, they still look very good. (Key informant 26) Several farmers explained that when they saw their yie lds declining with conventional agriculture they switched to MT to redress the situation. For example one farmer said, So what made me start using MT, for many years I had been making ridges [by hoe], but I was not finding food well enough. Harvests were down. So I tried MT and I found it. I harvested two ox - carts. (Farmer 33) From the survey data it is clear that cotton farmers in general believe that basins and ox - ripping result in higher yields, better soil fertility, better crop performance during dr ought year s and reduced erosion (Grabowski et al., submitted ). 62 3. sing MT 3. 4.3.1. Incentives It is important to note that among respondents current MT use is not closely associated with receiving material incen tives (such as cash, fertilizer, or food) , though some farmers did receive these for using MT in the past. O nly two farmers who were interviewed were receiving incentives from an NGO and in both cases this was for teaching about MT, not simply for using it . Both of them identified jealousy of the incentives as a reason for others to decide not to adopt MT. One farmer explained his reasoning not as jealousy but as a perceived injustice : What has made me not try MT, is that those leaders who are in front, t hey can write your name but then later they take the fertilizer. That makes it so that I would be basins, there is nothing for me to put in them. (Farmer 1) In three com munities distributors explained how disadoption was widespread once purpose of getting a bike. Then after that thing, that fu infor mant 29). 3. 4.3.2. Labor as the primary constraint for b asins Farmers identified labor a s the key constraint for more widespread use of basins. Of the 20 farmers who stated their reasons for disadopting basins, 12 of them said it was the hard work of digg ing the basins that made them stop. They started using animal traction. Another farmer stopped when his wife passed away. A fourth f armer stopped when he started having regular employment. A fifth stopped when she moved er 63 to do basins on a large area (Farmer 2). All of these show how changes in circumstances that alter the value of household labor directly impact the relative utility of using that labor to dig basins. Furthermore, the perceived increased effort needed for manual MT has kept many peopl e from trying basins. 3. 4.3.3. Equipment c osts as a key constraint to o x - ripping Many farmers expressed a desire to use ox - ripping but they said they were unable to afford the equipment and/or were unable to obtain oxen. M ost current basin users previously hoed and most current ripper farmers previous ly used oxen for plo wing and making ridges . This suggests that oxen ownership and animal traction experience are likely to determine if a household uses ripping or not. Altogether the total cash outlay required for the equipment is either $150 for the full set or $55 for the ripper attachment and chain if the farmer has an extra plow frame. With a mean per capita gross household income of only $390 for smallholders in Eastern Province (Tembo and Sitko , 2013) such an investment would require significant trad eoffs in other expenditures. Of the 11 farmers interviewed who disadopted ripping, six of them had borrowed the ripper. They explained that the lack of availability of the ripper led them to not rip in the 2012/13 season. As one farmer put it is busy using it. The time may go by when you are supposed to u se it. That is the main problem (Farmer 24). Another explained that the loss of one of his oxen prevented him from ripping in 2012/13. Of the 50 ripping farmers surveyed, 30 of them own a rip per. M ost of them bought their ripper on cr edit from the cotton companies but with low cotton prices it can be difficult for f armers to repay. L ow cotton prices resulted in loan defaults on 20% of the 40 ripper loans provided in three Cargill depots in 201 1, causing the company to reclaim the equipment. For this reason many farmers were cautious about taking on too large of a loan during the 2012/13 season. 64 While many farmers are interested in getting rippers on credit, they must be willing to take the ris k of a relatively large loan and their distributor or buyer must deem them creditworthy. The other ways farmers obtained a ripper was to pay cash or receive a ripper from an NGO . M any farmers without oxen made requests for oxen loans so that they too could start ripping. 3. 4.3.4. Fear of over - exerting oxen as a barrier to widespread use of o x - ripping Because ripping can be done throughout the dry season one would expect a well - developed rental market by those who own the equipment. However, ripping rental service provision is not common. One of the main reasons why those who own the equipment do not extensively rip for others with their oxen (as is common for plowing at the start of the rains) is that dry - season ripping is seen as too taxing for the oxen. A ripping, you know, it ruins the oxen... It is very dry below so they need to be strong (Key informant 21). Feed for oxen also tends to be running out in the dry season as the grasses and crop residues largely disappear du e to burning, tilling and grazing, so the oxen are at their weakest (Wall , 2007). The concern for oxen health has even kept some people from using ripping at all. A farmer who uses ox - ripping explained They think it is c ausing problems to their oxen; to make them dig deep like that is to (Farmer 19) 3. 4.3.5. Information needs as a barrier to adoption of o x - ripping While most farmers were aware of basins, ripping is a newer and less familiar technology to farmer s. T raining on r ipping was commonly requested and t he need for training , over and above equipment , is shown by the sur prisin g fact that 13% of those who invested in a ripper had never used it, despite owning it for over a year. One of these was a 65 distributor who explained that he needed more support in learning how to use the ripper effectively. Five of the fifteen farme rs (33%) who explained their motivation for adoption said that what persuaded them to use ripping or basins was the combination of receiving training and observing the benefits of MT in the fields of earlier adopters. These comments suggest that the uncert ainty of a new technology and the perception that MT is too challenging can be reduced through training that is accompanied by real life observation. 3. 4.3.6. Lack of motivation as a reason for n on - adoption Because farmers are primarily motivated to use M T because of concerns with drought or soil fertility it is logical then that those who are satisfied with their harvests are less likely to adopt MT. A n NWK distributor identified this as one of the key differences between MT users and non - MT users: Wha (Key informant 21) When asked why he had never tri ed MT, a hoe farmer responded, t really believe what we have heard. I am still interested in hoeing. When I plant on the ridge I see that the maize grows well and if the fertilizer D [compound] and Urea are there it will do well. (Farmer 16) Similar ideas were communicated by the non - adopting farmers who participated in the three group interviews. 3. 4.3.7. Other reasons for disadopting r ipping The five farmers interviewed who disadopted ox - ripping though they were ripper owners had various reasons for stopping . One explained that he c ould not use ripping without herbicides so he stopped when he was unable to obtain herbicides on credit 66 ripper , the weeds are many. But if we make ridges the weeds stay small. So that is why I used the ripper one year and then s topped and kept making ridges (Farmer 34). Another said that he only used the ripper with fertilizer, which he could not afford last year. He asserted that unless fertilizer was added he saw no difference between plowing and ripping. Two other farmers said that the y failed to rip in 2012/13 because the trained household member was busy during land preparation months. These reasons for discontinuing use of ripping show that disadoption can be an active rejection of the technology due to challenges such as la bor, weed s and erosion or a more passive and temporar y change brought on by the unavailability of equipment or trained household members. This range of reasons behind disadoption needs to be kept in mind in interpreting the characteristics of disadopters in the sta tistical analysis below. 3. 4.4. Transitions in l and p reparation m ethods Analyzing the frequency of transitions in land preparation methods (including disadoption) allows for a greater understanding of who is using each type of MT and why, what they were d oing previously, and what future adoption trajectories may be. In this study all MT users and disadopters were asked if they used oxen or hoed before they started using MT. The results show that most current basin users previously hoed and most current rip per farmers previously used oxen for plo wing and making ridges (Figure 4 ). This suggests that oxen ownership and animal traction experience are likely to determine if a household uses ripping or not. It is also striking that a larger proportion of disadopt ers previously plowed with their own oxen before they tried using MT. Many of these disadopters received incentives to dig basins, which suggests that they used MT primarily for the incentives. 67 Figure 4 : Illustration of f t ransitions to MT Source: S urvey of NWK and Cargill farmers, 2013, unweighted data including observations from low adoption communities. Note s : Pipe diameters are proportional to the number in each category with that number provided in parentheses. For simplicity, disadoption is not disaggregated into basins and ripping in the figure. Of the 46 disadopters who previously plowed, 38 used basins and 15 ripped (seven did both). Of the 28 disadopters who previously hoed, 24 used basins and seven ripped ( three did both). 3.4.5 . Household - level regression r esults I used multinomial logistic model to determine how explanatory variables affected the probability of households being categorized as follows: those who use ripping, use basins, disadopt all MT or n ever use any MT. In the final regression 215 observations had no missing values. The means, standard deviation and range of the explanatory variables are presented in Table 9 . The results are presented in Table 10 . Transition from Plowing to Minimum Tillage Transition from Hoeing to Minimum Tillage Unknown transitions between hoeing and plowing Plowi ng (251) Disadopters (46) Basins (12) Ox - ripping (37) Any Plowing without any MT (156) Hoeing (132) Hoeing only (49) Disadopters (28) Basins (42) Ox - ripping (13) 68 Table 9 : Descr iption of variables used in the household - level regression on MT use Mean Standard Deviation Minimum Maximum Household Level Variables Household Composition Female Headed HH (Y/N) 0.083 0.276 0.00 1.00 Age of head 42.44 14.03 17 81 Age squared 1997.4 1314.4 289 6561 Economic qualities of household Adults in hh / Operated Area 1.11 0.80 0.178 6.294 Total Operated Area 3.91 2.43 .98 16 Number of non - ag. income sources 1.00 1.00 0 5 Total fertilizer used (1000 kg) 0.26 0.35 0 2.6 Equipment value ($) at time of adoption/disadoption Oxen 1.50 1.77 0 8 Total Livestock (TLU) a 3.70 4.06 0 23.27 Capacity qualities of household Trained in CA (Y/N) 0.73 0.44 0 1 Ever received incentives for CA (Y /N) 0.10 0.31 0 1 Years of schooling 5.18 3.08 0 12 Cotton experience (years) 10.54 6.95 1 40 Community Level Variables Years CA has been promoted 5.23 2.81 2 13 Buyer CA practice 1.33 1.70 0 6.48 Cluster 1 0.22 0.42 0 1 Cluster 2 0. 23 0.42 0 1 Cluster 4 0.25 0.44 0 1 Source: Survey of NWK and Cargill farmers, 2013 a Tropical Livestock Units - cattle have a value of 0.7, goats and sheep 0.1, pigs 0.2, poultry 0.01 and donkeys 0.5 (Jahnke, 1982 ). 69 Table 10 : Factors affecting h ousehold use of MT, multinomial logistic regression results with robust standard e rrors Explanatory Variables Rip vs. Never Rip vs. Disadopt Basins vs. Never Basins vs. Disadopt Household Level Female Head ed HH (Y/N) - 2.29 ** (1.04) - 1.30 (1.12) 0.17 (1.09) 1.17 (1.17) Age of head 0.35 ** (0.15) 0.26 * (0.15) 0.13 (0.14) 0.04 (0.14) Age squared - 0.004 *** (0.00) - 0.004 ** (0.00) - 0.002 (0.00) - 0.002 (0.00) Economic variables Adults in hh / Operated Area 1.21 ** (0.51) 0.73 (0.52) 1.46 *** (0.45) 0.98 ** (0.44) Total Operated Area 0.46 * (0.27) 0.08 (0.23) 0.64 ** (0.27) 0.26 (0.23) Sources of non - ag. income - 0.81 * (0.48) - 0.92 * (0.50) - 0.68 ** (0.33) - 0.79 ** (0.35) Total fertilizer (1000 kg) 3.41 *** (1.17) 0.99 (0.68) 1.96 (1.35) - 0.46 (0.92) Equipment value ($) 8.99 *** (2.83) 5.58 *** (1.82) 5.62 ** (2.58) 2.21 (1.74) Oxen - 0.29 (0.30) - 0.48 (0.30) 0.01 (0.26) - 0.18 (0.25) Total Liv estock (TLU) a - 0.12 (0.12) 0.06 (0.13) - 0.24 ** (0.12) - 0.06 (0.11) Capacity variables Trained in CA (Y/N) 4.39 *** (0.98) 3.50 *** (1.10) 1.14 (1.15) 0.26 (1.12) Years of schooling 0.07 (0.12) - 0.06 (0.12) 0.38 *** (0.1 3) 0.26 * (0.14) Cotton experience (years) 0.05 (0.08) - 0.04 (0.07) 0.19 *** (0.07) 0.10 (0.07) R eceived incentives (Y/N) 0.79 (1.76) - 2.31 (1.74) 5.36 *** (1.05) 2.26 ** (0.90) Community Level Years CA promoted 0.20 (0.17) 0.06 (0.16) 0.06 (0.22) - 0.08 (0.24) Buyer CA practice 0.23 (0.43) 0.37 (0.44) - 0.57 (0.39) - 0.43 (0.38) Cluster 1 - 0.39 (1.12) - 0.05 (1.07) - 0.50 (1.36) - 0.16 (1.30) Cluster 2 0.74 (0.91) 1.60 * (0.85) - 0.29 (1.06) 0.57 (1.00 ) Cluster 4 - 1.78 (1.38) - 0.71 (1.24) - 0.83 (1.16) 0.24 (1.17) Constant - 17.07 *** (4.64) - 9.38 ** (4.41) - 10.82 *** (3.41) - 3.13 (3.48) Observations 215 Wald Chi 2 (57) = 177.00 Pseudo R 2 = 0.3903 Prob > chi 2 = 0.0000 Log pseudolikelihood = - 12,661.594 Source: S urvey of NWK and Cargill farmers , 2013 a Tropical Livestock Units - cattle have a value of 0.7, goats and sheep 0.1, pigs 0.2, poultry 0.01 and donkeys 0.5 (Jahnke , 1982). 70 One of the challenges wit h interpreting the results of this multinomial regression on four categories is that marginal effects for each explanatory variable must be considered for all pairs of comparisons. In Table 9 I present the coefficients for only the four most re levant compa risons . An odds ratio plot (also known as a factor change plot) makes it easy to visualize how a change in each variable (holding the others constant) affects the probability that a household will fall into any category relative to the others (Figures 5 an d 6 ) . For dummy variables the odds ratio between a pair of choices m and n can be X=1 the odds of a household being in category m versus n are expected to change by a factor of Z, holding all other variables constant. For continuous v ariables the odds ratio plots below show standard deviation changes. The o The interpretation for continuous variables is that when there is a standard deviation change in X the o dds of a household being in category m versus n are expected to change by a factor of Z, holding all other variables constant. In the following figures the household categories are represented by their first letter (B = Basins, R = Ripping, D = Disadopter, N = Never used MT). When there is a line between two letters there is no statistically significant difference in the probabilities of being in either category (p=0.1). The category of Never used MT is the base category so all the value of 1 on the top axis (which shows the odds ratio) and 0 on the of the base category it means that an increase in the explanatory variable leads to reduced probability of a household being in that category, thus requiring the odds ratio to be inverted. So when a category is at 0.1 on the top axis it means it is 10 times less likely than the base category. 71 Figure 5 : Odds ratio plot fo r economic indicator v ariables Source: Survey of NWK and Cargill farmers, 2013 Note: Household categories are represented by their first letter (B = Basins, R = Ripping, D = Disadopter, N = Never used MT). Figure 6 : Odds ratio pl ot for capacity indicators and other variables Source: Survey of NWK and Cargill farmers, 2013 Notes: Household categories are represented by their first letter (B = Basins, R = Ripping, D = Disadopter, N = Never used MT). 72 3.4.5 .1. Labor availability as a determinant of adoption The statistical analysis supports the qualitative finding that l abor is a key constraint to MT use. In addition to the challenge for land preparation with basins, labor is also higher for weeding in the absence of herbicides for any MT. A standard deviation increase in the number of adult workers per hectare of la nd farmed makes a household 4.3 times more likely to use basins and 3.3 times more likely to use ripping than to have never tried MT (Figure 5 ). Surprisingly, a standard deviation increase in oxen has no effect on the probability of a household falling into a ny of these categories (Figure 5 ). The fact that half the basin users also plow with oxen indicates that basin use is not mutually exclusive with oxen ownership. While hand - hoe farmers may be the best candidates to adopt basins (since plowing and ripping are not easy options for them) , they also have limited labor. 3. 4. 5 .2. Wealth indicators as determinants of adoption A variety of wealth indicators were used in this an alysis and several were significantly associated with MT adoption. Wealth allows farmers to invest in ox - ripping equipment, a major constraint identified in the qualitative analysis. Wealth also makes the household better able to risk trying out a new te chnology . Households with more farmed area are more likely to use basins and rippi ng than to have never tried MT. Similarly, households with higher value s of equipment at the time of adoption are more likely to use ripping or become a disadopter than to n ever try MT. Households that apply more fertilizer are also more likely to use ripping and, to a lesser degree, basins (Figure 5 ) . Total fertilizer use challenge f , 2014). The characteristics of the head of the household play an important role in the MT adoption decision and are directly related to the ability to invest and take risks . Female - headed households are 10 times less likely to use ripping than to have never tried using MT 73 and 12 times more likely to u se basins than ripping (Figure 6 ). Age of household head has a positive non - linear effect on the probability of a household using ripping. The combined marginal effect of age and age - squared is positive over the entire range ( being older makes one more likely to use ripping ) and peaks at age 41 . The interpretation is that older household heads have greater ability to invest in equipment, which is supported by t he insignifi cance of age for explaining adoption of basins. 3.4.5 .3. Livelihood strategy as a determinant of adoption H ouseholds without much land may not be able to focus on making it more productive if their small farmed area forces them to have diverse livelihood s trategies. Households with more non - agricultural income sources are less likely to use ripping or basins. Households with more diverse livelihood strategies may have too many non - farm responsibilities to focus attention on learning and implementing complex management practices such as MT. 3.4.5 .4. Knowledge as a determinant of adoption Conservation agriculture has been described as a knowledge - intensive technology (Kassam et al. , 2009) and as such training is expected to be important for adoption . These results show this to be especially true for ox - ripping. Mo st households in the sampled communities (64%) have received some training , though not necessarily in ripping , which is more technical and has been promoted less than basins. Nearly all households using ripping (97%) have received training, suggesting that without it adoption is highly unlikely. Farming experience and formal schooling are more important for explaining adoption of basins than training. 3.4.5 .5. Incentives as a determinant of basin use and disadoption While most farmers in this study did not receive incentives, those who did were typically basin farmers. If a household has ever received incentives for doing MT it is more 74 likely to use basins or disadopted than to have never tried MT. The effect on ripping is smaller, which fits with the histo ry of MT promotion - basins have been promoted longer and by more NGOs than ripping. The effect of incentives on disadoption matches well with the qualitative data, suggesting that once incentives are removed disadoption is common. Only 10% of disadopters said they received incentives, so the other challenges merit attention as well. 3.4.6 . Summary Cotton farmers in Eastern Zambia are generally aware of minimum tillage and think highly of its potential for improved yields, especially due to earlier plantin g, increased drought tolerance and efficient nutrient application. economic problems significantly limit the use of both basins and ripping. Significant labor limitations suggest that the use of basins by hand - hoe farmers is unlikely to take off exponentially . However, this does not mean that basins are not worth promoting as many farmers persist in using basins on small plots. F urthermore, a s Zambian smallholders face increasing land scarcity in pockets of higher population density there will be less opportunity to fallow and greater need to make land more productive, such as through the labor - intensive method of basins. For rip ping the primary constraints relate to the cost of investment in the equipment and the associated risks of investing in a new (to the farmer) technology. Ox - ripping rental services are not widely available, primarily because of concerns for the health of o xen while they are laboring to do dry season minimum tillage when feed is scarce. Improving the health of oxen may help reduce this challenge and the promotion of forage crops may aid in this while simultaneously reducing the competition between using resi dues for mulch or for feed (Giller et al. , 2011). The lack of effective demonstration and training on how to rip limits its 75 use among those who can afford the equipment. Training and personal observation of ripping appear to help farmers overcome the chall enges of learning a new land preparation technique. Farmers who have never tried MT have less land, use less fertilizer and have more non - agricultural income sources, which suggest they may be poorer and have more diverse livelihood strategies. Disadopter s tend to be better off than those who never tried MT. They are also less educated and have less labor than basin users and use less fertilizer than ripper users. These relationships and constraints are summarized in Figure 7 . 3 . 5. Conclusion The results of this study highlight three key lessons that can improve efforts to help smallhol der farmers mitigate the problems of climate change and land degradation through improved agricultural technologies. First, the diversity of smallholder farmers observed in this study should give caution to those seeking simple technological solutions to l ow productivity. Any particular technology is likely to be a good match for some farmers and bad match for many others because of the variation in their skills, assets, livelihoods strategies and agro - ecological conditions. Providing a basket of choices (C hambers , 1997) to farmers is much more likely to yield long - term change than promoting one or two narrow solutions. Second, adaptation of technologies should draw to expand the range of conditions where the technology can be used . The clear reasoning that farmers provided in this study about their decisions sho w that they are not stuck in tradition but wrestling with a complex set of potential benefits and challenges associated with MT. Including farmers in the technology developmen t process can make the most of what they have learned through their experiences and makes it more likely that technologies can be adapted to match with their actual needs and constraints. 76 Figure 7 : Decision tree summarizing the interactions between farmers' resources and priorities and land preparation decisions Household has reliable access to animal traction Has sufficient resources to buy ox - ripping equipment Yields Declining Ox - ripping and/or basins Ox - plowing Yields Declining Basins or ox - ripper rental Ox - plowing Has sufficient labor for digging basins Yields Declining Basins Maybe use basins Yields Declining Tiny area of basins Hand - hoe Wealthiest Middle wealth group Poor Poorest No Yes Yes N o Yes No Yes No Yes No Yes No No Yes 77 Finally, promotion of new technologies should not be based on top - do wn behavior chan ge messages but rather should follow adult education principles of facilitating Farmers using MT highlighted the importance of combining training with real life observations where their peers benefited from using the technologies. Also, p olicy makers considering subsidizing the use of technologies should be aware of the perverse effects this may have on adoption. Those providing incentives may be hoping to g et the ball rolling but may actually be distracting farmers from evaluating the technology. For sustainable intensificati on efforts to find effective solutions for smallholder farmers they will have to consider their diversity, include their participation in the technology . 78 APPENDICES 79 Appendix A : Survey instrument Company___________ Distributor/Buyer: _______________________ Survey Number:________ Survey of cotton farmer practices Date:_____________ Enumerator: _____________ A. Introduction Strata: ____________ Checked By: _____________ Introduce yourself by saying your name and that you are doing research for Michigan State University. Next read t he consent script. B . Getting to know the respondent Name:_________________________ _______ Village/Section: _______________ NRC#: _________________________________ Cell phone: _____________________________ 1. How old are you? _____________ 2. How many people live in your household? ____ 3. How many of them work in the field?____ 4. How many of them work like adults?_________ 5. Did you go to school? Yes No 6. What is the highest grade level that you complete d in school? ______ 7. Did your wife/husband go to school? Yes No 8. What is the highest grade level that your spouse completed in school? ______ 80 C . Understanding their farming practices 9. Do you live in a village or on a farm? Village Farm Other: Please describe ____________ 10. What crops do you grow? Prompt with: Any others? Any others? ( Check boxes column1) 11. On how much area did you plant _____ this year? (E ach crop, column 2 , if small: write < ¼ lima ) 12. This year how did you pre pare your fields? Prompt with: Do you plow or hoe? Do you rip or use basins? ( Check all that are used) Ox - plow Ox - ripping Hoeing Basins Tractor Plowing Tractor Ripping 13. How much of your cotton area did you use each land preparat ion method on this year? ( Continue to for all crops , columns 3 - 6 ) 10. Crop 11. Total Area 12/13 13.Ox Plowed Area 13.OX Ripped Area 13.Hoed Area 13.Basin Area 13.Tractor Plowed Area 13.Tractor Ripped Area cotton hybrid maize local maize groundnuts sunflower soya tobacco sweet potato cassava beans other ____ TOTALS: 81 14. On how many plots did you plant _________ (crop) using _________ (land prep method)? For e xample: On how many plots did you plant cotton and plow? C ontinue for each cell in Table 1 writing the crop and land prep type into the first two rows of the plot table - a column for each plot. Plot 1 Plot 2 Plot 3 Plot 4 P lot 5 Plot 6 Plot 7 Crop (2012/2013) Land Preparation Method 15. What size is that plot ? (hectares, acres or lima) 16. How much did you harvest from that plot? 17. Are there other crops that you planted with __________? ( Indicate which crop, If NO skip to 19) 18. How much did you harvest from those other crops? 19 . If the plot was ripped or had basins: How many consecutive years have you done ripping (or basins) on that plot ? 20 . What month was th e last land preparation before planting that field? Or: What month did you plow/rip/dig basins/ hoe ridges? 21. How much seed did you use to plant that field? (kg) 22. How was the weed pressure on that plot ? 23. How many times d id you weed that plot? (including banking) 24. Did you bank (kupantira/kufochera) that plot? 25. What did you plant on that plot the previous season? (2011/2012) 26. What happened to the crop residues on that plot before planting th is crop? 27 . If it was grazed: How much of the residues were consumed ? 28. What is the soil type on that plot? Sand, stones, loam, 8 2 Plot 1 Plot 2 Plot 3 Plot 4 P lot 5 Plot 6 Plot 7 Crop (2012/2013) Land Preparation Method clay (soil type) 29. What color is the soil on that plot? Red, Black, Brown, White (soil color) 30. What is the slope on that plot? OR: How does the water run off of it? 31. Do you do anything to reduce soil erosion on that plot? 32. What do you do (if anything) to reduce erosion? 33. Did you use fertilizer or manure on that plot? 34 . If Yes: How much fertilizer did you apply to that plot? 35. Did you use herbicide (weedkiller) on that plot? 36a. If Yes: How much herbicide? (liters or kg) 36b. What type of herbicide did you apply on that p lot? Faidherbia albida) on that plot? (fertilizer trees) Yes or No 38. Do you own, rent or borrow that plot? 39 . Do you think you will continue farming that plot 5 years from now? 40. If No: W hat makes you think that you may not be farming that plot in the future? 41. Which plot is your best plot? Which one is the next best? (Rank all plots) 96. Have you ever used CA on that plot? Yes or No 97. If Yes: How ma ny total years have you ever used CA on that plot? 83 42. Do you have a wetland garden? Yes No 43. If Yes: What is the total area of your dimba? ________ 44. How much area of land did you leave fallow during the previous season? ____ 45. How m uch land do you have with trees on it? ______ 46. Do you have any fields that you rented out last year? Yes No 47. How much area (if any) did you rent out to others during the previous season? ________ Animal traction If they plow or rip with oxen. O therwise skip to 52 48. How many oxen do you own?______ If zero skip to 50 49. If they own any oxen: For how long have you owned oxen?________ Go to 52 50 . How do you obtain oxen for plowing/ripping? pay/barter to rent them relative does it for me borrow from friend or relative 51 . If they rent oxen: Do you pay the full price up front? Or can you pay some after harvest? pay all up front pay all at harvest pay part up front, part at harvest other__ __ Cotton production 52. How many times did you spray pesticides on cotton during this season? _______________ 53. How do you decide when you need to spray? Scouting Every Two Weeks Not sure 54. Do you own a sprayer? Yes No (If no g o to 57 ) 55 . What type of sprayer? Ovaplus (battery) Jacto (knapsack) Other ________ 56 . For how long have you owned it? _____________ (Go to 58) 57 How do you get your fields sprayed? Pay/barter to rent Relati ve does it for me Borrow without paying 58. How long have you been growing cotton? __________ years 59. How many 15kg packets of seed did you plant last season? ________ 60. How much seed was left over last season? more than half about half ab out a quarter less than quarter none 61. How much cotton did you produce last season? ___________ (woolpacks or kg - circle ) Crop marketing 62. What crops do you sell? cotton maize groundnuts sunflower soya tobacco sweet potat o cassava beans other __________________ 63. Did you buy maize this year for home consumption? Yes No 84 64. If yes AND they sold maize: Which was larger, the total maize you sold or the total maize you bought? total maize sold was la rger total maize bought was larger equal Agricultural wage Labor 65. Did you hire any laborers for your fields last year? Yes No 66. Did you work as an agricultural laborer for others last year? Yes No 67. If No: Did your spouse work as an agricultural laborer for others last year? Yes No Non - agricultural economic activities 68. In addition to crop and animal production how else do you earn money for school supplies, soap, salt, etc. Processing food (cooking scones, fritters, etc.) Making or selling beverages (including alcohol) Small business trading (groceries) Wage labor not in agriculture Building/masonry Carpentry Basket weaving / mat making Sewing/Tailoring Brick molding Charcoal making Bicy cle repair/other repairs NGO work other (please specify ________________) 6 9. Were you or your spouse ever employed previously? Yes No 70. If Yes: What occupation did you have? _________________ 71. For how many years did you do that work? __ ____________________ Sellable assets 72 - 77. How many __________(livestock) do you own? 72.Cattle 73. Goats 74. Pigs 75. Poultry 76. Donkeys 77. Sheep 78. Do you own any of the following agricultural equipment? If yes, for how many years have you owned them? Ox - cart (Years owned ________) Cultivator (Years owned ________) Plow (Years owned ________) Ripper (Years owned ________) Ridger (Years owned ________) Treadle Pump (Years owned ________) Engine to pump water (Years owned ________) 79. How many working bicycles do you own? _______ 80. How many working radios do you own? _______ 81. How many working cell phones do you own?__________ 82. Do you own any of the following items? If yes, for how many years hav e you owned them? Motorbike (Years owned ________) Vehicle (Years owned ________) Solar panel (Years owned ________) Generator (Years owned ________) Maize mill (Years owned ________) Television (Years owned ________) 85 83. What type of house do you live in? Metal roof OR grass roof Fired bricks/cement block OR unfired bricks OR Other _________________ D. Conservation Agriculture ( If they already stated that they are using CA skip to 85) 84. Have you ever heard of farming without tilling the soil such as basins or ripping? Yes No If NO then SKIP to 109 - but really make sure they have never heard of it. 85. Who have you heard about basins or ripping from? ( Check all that are mentioned) Dunavant Cargill CF U (Conservation Farming Unit) Agricultural extension officer Some other NGO (please specify __________________________________________) Radio Written material Other _____________________________________ 86. How many of your friends and relat ives are using basins? ___________ 87. How many of your friends and relatives are using ripping (both ox and tractor)? _____________ 88. Have you ever been trained to do basins or ripping? Yes No 89. Have you ever received incentives for digging basins or using ripping? Yes No ( If they already stated that they are using Gamphani skip to 92) 90. Have you ever tried using basins or ripping on your own land? Yes No (I ) 91 . If Yes: How many years has it been since you last u sed basins or ripping? ___________ 92. How many agricultural seasons have you used basins? _______ 93. How many agricultural seasons have you used ox ripping? ___________ 94. How many agricultural seasons have you used tractor ripping? _________ 95. On how much total area did you use basins/ripping each year since you started? Year 1 ___________ Year 2 __________ Year 3 __________ ( Indicate units) 96. On which plots have you ever used CA? ( Record on PAGE 5!) 97. For how many years did you use it on eac h plot? ( Record on PAGE 5!) 98. What land preparation practices did you use before using basins/ripping? (Check all that were used) Hoeing Plowing with own oxen Plowing with hired oxen Other _______ 99. Which person was most influential in helping you learn about basins/ripping? Name: _____________________ Relation: Friend Relative Distributor/Buyer/Chairman Extension officer NGO field worker Other: __________________ 86 100. If they have ever used basins: Do you have a Cha ka hoe? Yes No If they have never used ripping SKIP to 106 101. If you want to rip a field that has ridges, do you rip on the ridge or in the furrow? Ridge Furrow I have never done that 102. Where you rip, do you weed with oxen or tractor, such as by using a cultivator or plow? Yes No 103. Where you rip, do you bank the crop mid - season? Yes No ( If No skip to 106) 104. If yes: Do you always bank the crop when ripping? Yes No ( If Yes skip to 106) 105. I f no: Why do you sometimes bank the crop when ripping? Only when there is excessive rainfall I only bank for particular crops ( List: __________________) Other: ____________________________________________ 106. How would you compare a field with one half under ox ripping and one half ox plowed? Ox Ripping Ox Plowing No difference know a. Which one would be more work? b. Which one would give you more trouble with weeds? c. Which one would produce more harvest? d. Which one would have better soil? e. Which one would do better on a dry year? f. Which one would do better on a wet year? g. Which one would have more erosion? 107. How would you compare a field with one half with basins and one half hoed? Basins Hoeing No difference know a. Which one would be more work? b. Which one would give you more trouble with weeds? c. Which one would produce more harvest? d. Which one would have better soil? e. Which one would do better on a dry year? f. Which one would do better on a wet year? g. Which one would have more erosion? 87 For tractor ripping communities only - otherwise skip to 109 108. How would you compa re a field with one half under tractor ripping and one half tractor plowed? Tractor Ripping Tractor Plowing No difference know a. Which one would be more work? b. Which one would give you more trouble with weeds? c. Which one w ould produce more harvest? d. Which one would have better soil? e. Which one would do better on a dry year? f. Which one would do better on a wet year? g. Which one would have more erosion? 109. Thank you so much for your time. You have answered many questions. Do you have any questions for me or have anything else you would like to say? (Take notes on questions and comments in the box below) 88 A ppendix B: Qualitative interview g uide for distri butors or buyers 1. How many years have you worked as a buyer/distributor?________ If they work for Dunavant: 2. Are you a Yield Coordinator Distributor (YCD)? Y/N a. If No, are you a Yield Group Leader (YGL)? b. If Yes, How many YGLs do you have among your farmers? ______ 3. Are you a buyer for Dunavant? Y/N If they work for Cargill: 4. Do you over see cotton schools? Y/N How many are active? ________ 5. For all respondents: 6. How many farmers do you oversee? ___________ a. How many packets of seed did they receive last season?____________ 7. What is the largest number of farmers you have ever overseen? ___________ a. What is the largest number of seed packets you provided in a single season? ________ Spray service provision and herbicides 8. Have your farmers been trained as spray service providers? Y/N a. How many of your farmers are providing the service of spraying? _________ 9. Were herbicides available to your farmers on credit last season? Y/N a. If yes, was there enough b. How many farmers received herbicides on credit last season? ____________ Tillage service provision and ox - ripper availability 10. Have your farmers been trained as tillage service providers? Y/N a. How many of your farmers are providing tillage services? ________ (total) i. Plowing __________ Ripping _____________ 11. Were ox - rippers available on credit to your farmers last season? Y/N a. How many farmers have ever received ox - rippers on credit? _________ b. How many farmers have bought their own ox - rippers? __________ c. Cost to rent a plow and oxen _______________ d. Cost to rent a ripper and oxen______________ Farmer density 12. How often do you visit each farmer in a season? __________ 13. How far do you have to travel to find the furth est farmer? _____________ 14. How many villages or clusters are the farmers you oversee grouped into? _________ Community level context 15. What are the economic opportunities for people in these areas other than agriculture? 16. What is the travel cost a farmer woul d ha ve to pay to obtain fertilizer? _____________ 17. How much does an average male earn working 8 hours of ganyu in this community?_________ 18. Do most farmers that you oversee live on farms or are they primarily in villages? 19. What are the main challenges your f armers face in getting higher cotton yields? 89 Conservation agriculture 20. How many years has conservation agriculture been promoted in these areas? __________ 21. What other organizations are promoting CA in these areas? a. For how long has each been operating? 22. W hat do you see as the primary motivation for your farmers to use CA? 23. What has made it possible for those farmers to use CA? 24. What are the main challenges your farmers face in using CA? 25. What do you think can be done to overcome those challenges? Work as a Buyer/Distributor 26. Can you describe for me your work responsibilities? 27. What are the main challenges you face in completing your work as a buyer/distributor? 90 Appendix C: Qualitative interview guide for f armer s Name:_________________________ Village: Date:_____________ A. Introduction Consent process - not a test, what they are actually doing B. Getting to know the respondent 1. How many people live in your household? ____ 2. How many of them are 15 years or older? ______ 3. Did you go to school (Y/N)? What is the highest grade level that you completed in school? __ 4. Did your wife/husband go to school (Y/N)? What is the highest grade level that your spouse completed in school? ______ 5. In what village and distr ict were you born? ______ (here, close by, far away) 6. How many years have you lived in this village? _______ C. Understanding their farming practices 7 : Can you describe for me the different agricultural activities you carry out on your farm? 8. What i s the total area that you farmed last season/this year? _____ 9. How much of that area was planted to maize? ___________ 10. How much of that area was planted to cotton? __________ 11. How much of that area was planted to groundnuts? ________ 12. How m uch land was planted to sunflowers? ________ 13. How much land was planted to beans or other legumes (cowpea, pigeon pea, soya)? _ 14. How much of that area was planted to other crops (cassava, tobacco, paprika, etc.)? _ 15a. How many wetland gardens do y ou have? _________ 15b.What is the total area of your dimba? ________ 15c. What crops do you grow there? _______ 16a. Did you intercrop your maize with legumes or other crops like sorghum or pumpkin? (Y/N) 16b. How much area was intercropped? _______ Other land use questions 17. How much area of land did you leave fallow during the previous season? ____ 17b. For how many years has it been fallow? ________ 18. How much of that fallow land was an improved fallow with leguminous shrubs? ____ 19. How much area (if any) did you rent in during the previous season? _____ 20. How much area (if any) did you rent out to others duri ng the previous season? ____ Inputs 21. How much fertilizer (if any) did you use on your maize durin g the previous season? ______ 22 . What types of fertilizer did you apply? ____D - compound ____Urea ____ Other ________________ ____ Other ________________ 23. What price did you pay for that fertilizer? _____ 24. What distance did you travel to obtain that fertilizer? _____ ___ 91 24b. What was the travel cost?_______ 25. Did you use any herbicide on your crops last year? Y/N 26. What type of herbicides did you use? __ Round up (Glyphosate, systemic) __ Pantera (Quizalofop - selective for grasses) __ Paraquat (cont act burning) __Metalachlor (pre - emergent - seed dormancy) __ Other (please specify _______________) 27. How much did you apply on each crop?___________ ___________________ Plots 28. How many plots of land did you farm? If it is all together, how is it divided? (See Table for recording) 29. On each plot/division - What size is it? 30. What crops did you plant this year? The previous year? 31. What inputs did you use? (kg and liters) 32. What land prep method? 33. Distance to home? 34. Ranking compared to other plots? Crop Rotation and legumes 35. How do you decide which field you will plant each crop in? 36. Do you practice crop rotation? Y/N 37. What is your experience growing maize in a field that previously had legumes? 37b. How does that compare with maize afte r cotton? _______________________ 38. What would make it easier to grow more legumes? __________________________ 39. Have you ever tried growing an improved fallow (tephrosia, gliricidia or sesbania)? Y/N 40. What results? 41. Have you ever planted Msangu trees (Faidherbia albida)? 42. What results? Erosion 43. Do you have any concerns about erosion in any of your fields? Y/N 44. What do you do if you are concerned about erosion? Mulching 45. What do you do with the maize stalks and the cotton stems in the field after you harvest? Cotton production 46. How long have you been growing cotton? __________ 47. With what company? ______________ Have you ever changed companies? Y/N 48. How are your yields this season? ___________ Last Season? _______ ________ How many packets of seed did you plant this season? __________ Last season? ____ How much seed was leftover this season? ________ Last season? ___________ 49. How many times did you spray pesticides during this season? _______________ 50. Do yo u spray pesticides on cotton yourself? Y/N Or do you hire someone to do it? Y/N 51. Do you own a sprayer? Y/N 52. If Yes: Do you provide spraying services to others? Y/N 92 53. What fee do you charge? _____________ 54. If No: Do you pay others to spray your fields? Y/N 54b. Or do you pay them to use their sprayer? Y/N 55. If Yes: How much do you pay? _____________ 56. If No: How do you get your fields sprayed? ____ Agricultural innovation 57. How have your farming practices changed over the past few years? 58. What new varieties or crops have you tried? 59. What new soil conservation measure have you tried? 60. What new pest management strategies have you tried? 61. What new weed control methods have you tried? Future Plans 62. How will you p repare your fields for the coming year? When will you start? Land Tenure Security: 63. Do you think you will be farming the same plots next year? Y/N 64. What about two years from now? Y/N 65. Five years from now? Y/N 66. (If No: What makes you thin k that you may not be farming these plots in the future?) Crop marketing 67. What crops do you sell? 68. Where do you sell them? 69. What price did you receive last year? Animal production 70 - 73. How many cattle/goats/pigs/poultry do you own? Cattle Goa ts Pigs Poultry 74. Do you use your cattle forPlowing? Y/N 75. Ripping? Y/N 76. Pulling a cart? Y/N 77. How many oxen are currently able to work in these tasks? _______________ 78. Do you ever rent out your cattle so that others can use them to plow? Y/N 79. How m uch do you charge? ___________________ 80. Do you ever rent out your cattle so that others can use them for ripping? Y/N 81. How much do you charge? ____________________ 82. How long have you been using them for land preparation? ______________ Agricultural wage Labor 83. Did you hire any laborers for your fields last year? Y/N 84. How did you pay them? 85. Did you work as an agricultural laborer for others last year? Y/N 86. How much does an average male earn working 8 hours of ganyu in this community?__________ N on - agricultural ec onomic activities 93 87. In addition to crop and animal production how else do you earn money for school supplies, soap, salt, etc. __ processing food and beverages (including alcohol) __ petty trading __ wage labor not in agriculture __ building/masonry __ carpentry __ basket weaving / mat making __ bicycle repair/other repairs __ NGO work __ other (please specify ________________) 88. What activities do you do between the start of the harvest and when the rains start again? 89. How much time do you spend on th e different activities? 90. Which ones are most important? D . Knowledge about CA (for non - CA users) 1. Have you ever heard of farming without tilling the soil? Y/N 2. If yes: What have you heard about it? a. Who did you hear about it from? b. What is your opinion abou t these ideas? i. What do you think the benefits would be? ii. What would be the challenges? iii. How do you think it would affect your yields? iv. How do you think it would affect the work you can carry out? c. Do you know anyone who uses these land preparation method s? i. What have you observed from these people? ii. How are they similar or different from you? iii. How is their land similar or different from yours? iv. Who approves of their use of basins/ripping? Who disapproves? v. Would anyone disapprove if you started using basin s/ripping? d. Do you think you know enough about it that you could try it if you wanted to? i. If not, who would you talk to in order to learn more? e. What would need to change for you to use basins or ripping? i. What would motivate you to try it out? 3. What do you t hink the effects of farming without tilling the soil would be? (check all that are mentioned) Less erosion More erosion Wetter soil Drier soil Softer soil Harder soil Higher yields Lower yields More weeds Less weeds More work (digging basins/ripping) Less work (not having to make ridges or plow) E . Minimum tillage (for CA users) 1. How many agricultural seasons (if any) have you used basins/ripping? ____ a. On how much area did you use basins/ripping each year since you started? 94 2. What land p reparation practices did you use before using basins/ripping? ( skip if not using CA) a. How much area of each? Hoeing _____ Plowing ________ 3. Can you explain to me how you learned about planting in basins/ripping? 4. Who is the first person you knew who used CA? a. Name: _____________________ Relation: ______________________ 5. Which person was most influential in helping you learn about basins/ripping? a. Name: _____________________ Relation: _______________________ 6. Where did you first learn about basins/rippin g? __Cotton company (specify which one ___________) __CFU (Conservation Farming Unit) __Some other NGO (please specify ______) __ Agricultural extension officer __ Radio __Written material 7. What have you heard about basins/ripping from outside organiz ations supporting your 8. What motivated you to try this method of land preparation? 9. How did you decide on which plot to try using basins/ripping on? 10. What was your attitude about basins/ripping when you first tried it? Did you thi nk it would work? 11. What made it difficult? What parts were easy? 12. Who approved of you trying basins/ripping? Who disapproved? 13. What prevents you from using basins/ripping on more area? 14. What is your general opinion about the benefits of the basins/ripping? 15. P lease describe to me how to use basins/ripping for agriculture? a. How do you go from a ridged field to a field with basins/ripping? b. Do you need any special tools? c. How do you decide where to dig the basins (or place the rip lines)? Do you measure? d. Do you have to dig the basins every year? Do you measure every year? e. What do you put in the basins/rip lines? f. Do you do anything differently depending on the type of soil in the field? Please explain. 16. How do you weed without turning over the soil? 17. How does weedi ng compare with traditional ridge tilling? 18. Who in your household has been most affected by any changes in weeding? 19. How do you bank your maize? 20. Do the basins/rip lines affect the water in your field? How? 21. When it rains hard how is a plot with basins/rippin g different from a plot with ridges? 22. When there is a dry spell how is a plot with basins/ripping different from a plot with ridges? 23. Can you share examples from the last few seasons about any noted differences between basins/ripping and other land preparati on methods? If they use ox - ripping (All the questions above and the following): 24. What maintenance do you need to carry out on your ripper? 95 25. How frequently do you sharpen it? a. Where do you get it sharpened? How much does that cost? 26. Where do you buy spare nut s and bolts? How much do they cost in a season? 27. Have you had to replace the beam or the tine since you started? 28. Do you rent your ripper to others? How much do they pay? a. Do you carry out the ripping or do they do it themselves? b. Do they rent your animals a s well? 29. Have you ever rented a ripper from someone else? a. Did you rent their animals as well? b. How much did you have to pay? If they use tractor ripping 30. Do you own or rent the tractor? 31. What is the cost of renting a tractor for ripping? If they own the t ractor: 32. How did you obtain the tractor? How long have you had it? 33. What other uses do you have for the tractor other than ripping? 34. How do you pay for the maintenance and cost of the tractor? 35. What do you do when it breaks down? If they use multiple CA land preparation methods: 36. How do these reduced tillage methods compare? a. What are the benefits of each? b. What are the challenges of each? c. Are there fields, crops or seasons where one was notably better or worse than the others? d. Which one would you like to use more of? What would you need to do that? F . Minimum tillage (disadopters) I hear that you have experience using basins or ripping and I would like to learn from you about that. 1. Please describe to me how you used basins/ripping for agriculture? 2. How many se asons did you use basins/ripping on part of your farm? 3. What land preparation method were you using before you started using basins/ripping? 4. Can you explain to me how you learned about planting in basins/ripping? 5. What technical support have you had in lear ning how to use basins/ripping? a. Who did that information come from? 6. Who is the first person you knew who used CA? 7. Which person was most influential in helping you learn about basins/ripping? 8. What have you heard about basins/ripping from outside organizati ons supporting your 9. What motivated you to try this method of land preparation? 10. How did you decide on which plot to try using basins/ripping on? 11. What was your attitude about basins/ripping when you first tried it? Did you think it would work? 12. What made it difficult? What parts were easy? 13. Who approved of you trying basins/ripping? Who disapproved? 14. What made you decide to stop using this method of land preparation? 96 15. What has changed since you stopped? a. Are yields different? b. Is y our labor in agriculture different? c. The total cultivated area? 16. What, if anything, would be needed for you to try it again? 17. What advice would you give to someone who is thinking about starting to use basins/ripping? 97 REFERENCES 98 REFERENCES An C laims to Understanding Farmers and Contexts: A Literature Review of Conservation Agriculture (CA) Adoption among Smallholder Farmers in Southern Africa. Agriculture, Ecosystems and Environment 187: 116 - 32. doi:10.1016/j.agee.2013.08.008 Arslan, A., N. McCarthy, L. Lipper, S. Asfaw, A. Cattaneo. 2014. Adoption and Intensity of Adoption of Conservation Farming Practices in Zambia. Agriculture, Ecosystems and Environment 187: 72 - 86. doi:10.1016/j.agee.2013.08. 017 Baudron, F., H.M. Mwanza, B. Triomphe, and M. Bwalya. 2007. Conservation Agriculture in Zambia: A Case Study of Southern Province. Published jointly by the African Conservation Tillage Network, Nairobi; the Centre de Coopération Internationale de Reche rche Agronomique pour le Développement, Paris; and the Food and Agriculture Organization of the United Nations, Rome. Byerlee, D. and E.H. de Polanco. 1986. Evidence from the Mexican Altiplano. American Journal of Agricultural Economics 68: 519. doi:10.2307/1241537 Chambers, R. 1997. Whose Reality Counts? Putting the First Last . London: Intermediate Technology Publications. Chowdhury, S., M. Khare, and K. Wolter. 2007. Weight Trimming in the National Imm unization Survey. In Proceedings of the Joint Statistical Meetings (July 29 - August 2, 2007, Salt Lake City, Utah), Section on Survey Research Methods, ed. Valerie Nirala. Alexandria, VA: American Statistical Association. Retrieved from: www.amstat.org/sections/SRMS/Proceedings/y2007/Files/JSM2007 - 000077.pdf Elliott, M.R. 2008. Model Averaging Methods for Weight Trimming. Journal of Official Statistics 24: 517. Food and Agriculture Organization (FAO). 2001. The Economics of Conservation Agriculture. Rome: United Nations Food and Agriculture Organization. Friedrich, T. and A. Kassam. 2011. Conservation Agriculture: Global Perspectives and Development. Presented a t: Conservation Agriculture Regional Symposium for Southern Africa, 8 - 10 February. Johannesburg, South Africa. Gatere, L., J. Lehmann, S. DeGloria, P. Hobbs, R. Delve and A. Travis. 2013. One Size Does not Fit All: Conservation Farming Success in Africa M ore Dependent on Management than on Location. Agriculture, Ecosystems and Environment 179: 200 - 207. Giller, K., E. Witter, M. Corbeels, and P. Tittonell. 2009. Conservation Agriculture and Smallholder Farming in Africa: The Heretics' View. Field Crops Res earch 114 : 23 - 24. 99 Giller, K.E., M. Corbeels, J. Nyamangara, B. Triomphe, F. Affholder, E. Scopel, and P. Tittonell. 2011. A Research Agenda to Explore the Role of Conservation Agriculture in African Smallholder Farming Systems. Field Crops Research 124: 46 8 - 72. Grabowski, P., S. Haggblad e, S. Kabwe, and G. Tembo. 2014 . Minimum tillage adoption among commercial smallholder cotton farmers in Zambia 2002 to 2011 . Agricultural Systems 131, 34 - 44. d oi: 10.1016/j.agsy.2014.08.001. Grabowski, P., Kerr, J., Haggbla de, S., Kabwe, S. , and G. Tembo. Su bmitted for initial review . Estimating adoption of minimum tillage: Evidence from cotton farmers in Eastern Province, Zambia. In: Kassam, A. and Mkomwa, S. (Eds) Conservation Agriculture for Africa: Building Resilient Far ming Systems in Changing Climate . CABI . Grabowski, P. and Kerr, J. 2014. Resource constraints and partial adoption of conservation agriculture by hand - hoe farmers in Mozambique. International Journal of Agricultural Sustainability. 12 (1) 37 - 53 . Haggblade, S., S. Kabwe, and C. Plerhoples. 2011. Productivity Impact of Conservation Farming on Smallholder Cotton Farmers in Zambia. Working Paper No. 47. Lusaka, Zambia: Food Security Research Project. Haggblade, S., G. Tembo, D. Kabore, C. Reij, O.C. Ajayi, S. F ranzel, P. Mafongoya, and F. Place. 2010. Sustainable Soil Fertility Management Systems. In Successes in African Agriculture , ed. S. Haggblade and P.B.R. Hazell. Baltimore and Washington D.C.: Johns Hopkins University Press and IFPRI. Haggblade, S. and G. Tembo. 2003. Development, Diffusion and Impact of Conservation Farming in Zambia. Working Paper No. 8. Lusaka, Zambia: Food Security Research Project. Hove, L., I. Kadzere, B. Sims, M. Ager, and J. Mulila - Miti. 2011. Conservation Agriculture Research and Development in Southern Africa: A Review of Achievements and Challenges in the Past 20 Years. Presented at: Conservation Agriculture Regional Symposium for Southern Africa, 8 - 10 February 2011 Johannesburg, South Africa. Jahnke, H. 1982. Livestock Productio n Systems and Livestock Development in Tropical Africa. Kiel, Germany: Wissenschaftsverlag Vauk. Retrieved From: Http://Pdf.Usaid.Gov/Pdf_Docs/PNAAN484.Pdf Jain, A. 2010. Data Clustering 50 Yea rs Beyond K - means. Pattern Recognition Letters 31. 8: 651 - 666. Kabwe, S., C. Donovan, and D. Samazaka. 2007. Assessment of the Farm Level Financial Profitability of the Magoye Ripper in Maize and Cotton Production in Southern and Eastern Provinces . Workin g Paper No. 23. Lusaka, Zambia: Food Security Research Project. 100 Kassam, A., T. Friedrich, F. Shaxson, and J. Pretty. 2009. The Spread of Conservation Agriculture: Justification, Sustainability and Uptake. International Journal of Agricultural Sustainabilit y 7.4: 292 - 320. Kasanga, J. and O. Daka. 2013. Broad - Based Survey to Establish Baseline Conditions and Collection of Data for Monitoring the Impact of the Second Phase of the Conservation Agriculture Programme (CAP II) Survey Results 2013 Final Report. I MCS Limited for CFU. Retrieved from http://conservationagriculture.org/cfu - research . Long, J. and J. Freese. 2001. Regression Models for Categorical Dependent Variables Using Stata . College St ation, TX: Stata Press, Stata Corporation. Miles, M., Huberman, A., 1994 , Qualitative Data Analysis , second ed ition, Sage Publications, Thousand Oaks, CA. Morris, M., V. Kelly, R. Kopicki, and D. Byerlee. 2007 . Fertilizer Use in African Agriculture: Lesso ns Learned and Good Practice Guidelines. Washington D.C.: The World Bank. Ngoma, H., B.P. Mulenga, and T.S. Jayne. 2014. What Explains Minimal Usage of Minimum Tillage Practices in Zambia? Evidence from District - Representative Data. IAPRI Working Paper No. 82. Lusaka, Zambia: Indaba Agricultural Policy Research Institute (IAPRI). Nyanga, P. 2012. Factors Influencing Adoption and Area under Conservation Agriculture: A Mixed Methods Approach. Sustainable Agriculture Research 1: 27 - 40. Pedlow, S., J. Porras, C REACH 2010. In Proceedings of the Joint Statistical Meetings, Section on Survey Research Methods, American Statistical Associatio n, August 3 - 7, 2003 , San Francisco, CA , ed. Curtis Brooks. Alexandria, VA: American Statistical Association. Retrieved from: https://www.amstat.org/sections/srms/Proceedings/y2003/Files/JSM2003 - 000939.pdf Rockström, J. , P. Kaumbutho, J. Mwalley, A.W.M. Nzabi, M.L. Temesgen, L. Mawenya, J. Barron, J. Mutua, and S. Damgaard - Larsen. 2009. Conservation Farming Strategies in East and Southern Africa: Yields and Rainwater Productivity from On - farm Action Research. Soil and Ti llage Research 103: 23 - 32. Rubin, H., Rubin I., 2005. Qualitative Interviewing: The Art of Hearing Data . Sage Publications , Thousand Oaks, CA. Siegel, P. 2008. Beneficiaries of Agricultura l Commercialization? World Bank, Africa Region Working Paper Series No. 113. Washington, D.C.: The World Bank. Snapp, S., D. Rohrbach, F. Simtowe, and H. Freeman. 2002. Sustainable Soil Management Options for Malawi: Can Smallholder Farmers Grow More Legum es? Agriculture, Ecosystems and Environment 91: 159 - 74. 101 Tembo, S. and N. Sitko. 2013. Technical Compendium: Descriptive Agricultural Statistics and Analysis for Zambia. IAPRI Working Paper No. 76. Lusaka, Zambia: Indaba Agricultural Policy Research Institu te (IAPRI). Thierfelder, C., S. Cheesman, and L. Rusinamhodzi. 2013a. Benefits and Challenges of Crop Rotations in Maize - Based Conservation Agriculture Cropping Systems of Southern Africa. International Journal of Agricultural Sustainability 11: 108 - 24. do i:10.1080/14735903.2012.703894 Thierfelder, C., T. Mombeyarara, N. Mango, and L. Rusinamhodzi. 2013b. Integration of Conservation Agriculture in Smallholder Farming Systems of Southern Africa: Identification of Key Entry Points. International Journal of Ag ricultural Sustainability 11: 317 - 30. Thierfelder, C. and P.C. Wall. 2009. Effects of Conservation Agriculture Techniques on Infiltration and Soil Water Content in Zambia and Zimbabwe. Soil and Tillage Research 105: 217 - 27. Todaro, M. and S. Smith. 2009. E conomic Development (10th edition ). Boston: Pearson Addison Wesley. Verhulst, N., B. Govaerts, E. Verachtert, A. Castellanos - Navarrete, M. Mezzalama, P.C. Wall, A. Chocobarc, J. Deckersb, and K.D. Sayre. 2010. Conservation Agriculture, Improving Soil Quali ty in Sustainable Production Systems. In: Food Security and Soil Quality , ed. R. Lal and B. Stewart. Boca Raton, FL: CRC Press. Wall, P. 2007. Tailoring Conservation Agriculture to the Needs of Small Farmers in Developing Countries: An Analysis of Issues. Journal of Crop Improvement 19: 137 - 55. World Bank. 2007. Agriculture for Development . World Development Report No. 2008. Washington, D.C.: The World Bank. 102 CHAPTER 4: UNDERSTAN DING PARTIAL ADOPTIO N OF MINIMUM TILLAGE BY COTTON FARMERS IN EASTERN ZAMBIA 4 .1. Introduction Currently Sub - - poor (121 million people) who live on less than 50 cents a day (Barrett , 2010). Most of these people live in rural areas and agriculture is their primary livelihood strategy (Barrett , 2010). T he inadequacy of conventional farming practices for long term food security can be seen both by the alarming levels of land degradation (World Bank, 2007) as well as the frequency of crop failures in an inc reasingly erratic climate (Boko et al. , 2007; Lobell et al. , 2008) . As smallholder farmers in Africa face th e dual challenge of responding to climate change and mitigating land degradation there is increased attention to the potential for improved agricultural technologies to increase ( Pretty et al. , 2011; World Bank , 2007) . However, r esource - poor farmers typically have diverse livelihood strategies and live in complex agro - ecological environments, mak ing it especi ally challe nging to find technologies that fit the needs and priorities of most farmers (Conway and Barbier, 1990; Chambers, 1997). is essential for sustainable intensification efforts to effectively d eliver technologies that can be used by the diverse types of smallholders (Pretty et al., 2011). Conservation agriculture is a set of technologies that holds tremendous potential for reversing land degradation and making smallholder farmers less vulnerabl e to climatic variability (Rockström et al. , 2009; Kassam et al. , 2009; Hobbs , 2007). The three principles of conservation agriculture minimal soil disturbance, permanent soil cover , and either crop rotation or intercropping with nitrogen fixing legumes (FAO, 2001) have been widely promoted across 103 southern Africa for well over a decade . However, adoption levels remain low , with less than 6% of cultivated area under CA CA claim that adheren ce to the three principles is non - negotiable for sustainable land management (for example CFU, 2007) . That position has drawn criticism for being overly prescriptive and counter - productive to participatory problem - solving (Andersson and Giller, 2012). The risk of promoting a s trict set of principles embodied in specific technologies is that by ignor ing it may achieve only limited adoption while missing opportunities to adapt the available technologies to b resources and priorities. This paper uses empirical observation of the cond itions under which farmers used MT on at least some of their land to identify the cons traints farmers face in using it on more of their land . The goal is to a ssist CA promoters and agricultural researchers in identifying possible ways to overcome those constraints by providing detailed feedback from the farmers on the performance of the technologies . The research focuses on smallholder cotton farmers in Zambia s Eastern Province , a population and location with significant levels of promotion, training and adoption . It complements Chapter 3, which analyzes household level determinants of MT use and disadoption. This chapter combines quantitative analysis of 445 plots farmed by 81 MT farmers , and qualitative analysis o f in - depth interviews with 43 farmers who have used MT to understand why t hey chose to use MT on some plots but not others and to gain insight on what prevents them from using MT on more of their lan d . 4. 2. Conservation a griculture in Zambia 104 4. 2.1. Conservation a griculture and minimum tillage There is experimental agronomic evidence from on - station research that the CA principles are complementary and additive in that the benefits are greatest when t he three principles are implemented together (Thierfelder et al . , 2013a). However, in on - farm situations most smallholder farmers in southern Africa find it extremely difficult to retain crop residues due to free - range cattle and uncontrolled burning (Umar et al., 2011; Thierfelder et al., 2013b). In addition, the area under legumes tends to be much lower than the recommended 30% of cropland, and uptake of agroforestry (as an alternative means of integrating legumes into the farming system) remains a challe nge (Umar et al., 2011; Umar , 2012). For these reasons in Zambia minimal soil disturbance is the principle of CA that is often promoted first and emphasized most. In Zambia, three specific MT technologies are being promoted to allow farmers to prepare the land wh ile minimizing soil disturbance - hand hoe basins, ox - drawn ripping and tractor ripping (Grabowski et al. , 2014; Nyanga , 2012). Basins require only a hoe and promoters recommend digging a precise grid of hole s 20 centimeters (cm) deep, 30 cm long, and the width of a hoe blade. Ox - ripping requires specialized equipment and promoters recommend using the locally engineered Magoye ripper to open a trench 5 cm wide and 15 cm de ep where the seeds can be sown (Kabwe, Donovan, and Sam azaka , 2007). Tractor r ipping also requires specialized equipment, which in Zambia is usually a two - tined ripper . Due to the limited availability, tractor ripping is rarely used and so is not part of this analysis. A fourth technology for minimizing soil disturbance is that of s with a hoe and planting. This farmer - led innovation, which I call direct seeding , was observed but is not the focus of this analysis. Direct seeding with jab - planters, sticks or specialized hoes has been promote d for CA 105 in Southern Africa, especially for maize (Grabowski and Kerr, 2014), but it is not common among cotton farmers in Zambia and was not observed in this study. 4. 2.2. Conservation agriculture and cotton farmers CA promotion in Zambia began in the mi d - 1990s with the creation of the Conservation rivate sector cotton companies have collaborated with CFU to actively promote CA with smallholder s. NWK Agri - services (previo usly known as Dunavant) and Cargill are the largest cotton companies and strongest private sector promoters of MT ( Haggblade et al. , 2010). NWK has been promoting basins since the 1990s and b oth companies were involved in promoting Magoye rippers in Easter n Province as early as 2002 (Kabwe, Donovan, and Samazaka , 2007) . Efforts have increased since 2011 as herb icides and equipment have become more available to farmers on credit (Grabowski et al. , 2014). Smallholder farmers access the inputs to grow cotton o n credit through contracts , with the costs deducted from payment at harvest time (Haggblade, Kabwe, and Plerhoples , 2011). NWK uses a system of distributors, who are lead farmers that link 50 to 100 farmers with the company by providing training, distribut ing seed and chemicals, monitoring fields and buying the harvest. They earn a commission on the production from their farmers with bonuses for high volume, yields, and loan repayment rates. Cargill on the other hand employs buyers who may or may not be far mers to carry out similar functions but overseeing 200 to 500 farmers. 4. 2.3. Cotton farmers in eastern province Eastern Province has a unimodal rainfall pattern with annual precipitation varying between 600 and 1200 millimeters between November and May. T hough overall population density is relatively low, localized land scarcity exists, especially around some large villages. 106 Eastern Province can be divided into two major agro - ecological zones. The lower elevation valley zone has lower rainfall, higher temp eratures and lower cattle populations because of tsetse fly infestation. The upland plateau regions have greater population density and higher rainfall. This study focuses only on the plateau regions. Of the 257,000 small - and medium - scale farming househol ds in Eastern Province in 2012, 97% grow maize and 64% grow cotton (Tembo and Sitko , 2013). Cotton is a demanding crop in terms of labor and management with regular pest monitoring and pesticide sprays. On average cotton farming households have larger cult ivated areas, own less cattle, and earn a larger portion of their income through agriculture than households that do not grow cotton (Haggblade, Kabwe, and Plerhoples , 2011). 4. 2.4. Adoption of MT in Eastern Province Despite the general lack of residues a nd rotations, a doption of MT has been expanding gradually in Easter n Province over the last decade (Arslan et al., 2014; Grabowski et al., 2014; Ngoma et al., 2014). The persistent use of MT suggests that the benefits outweigh the costs for certain farmers under certain conditions. MT adoption correlates with promotion and higher rainfall variability, suggesting that farmers use MT to reduce their vulnerability to an unpredictable climate (Arslan et al. , 2014; Ngoma, Mulenga, and Jayne , 2014). For communiti es where farmers grow cotton, adoption correlates with greater herbicide availability, longer promotion and better demonstrations by lead farmers (Grabowski et al. , 2014). Smallholder farmers in southern Africa typically adopt MT on only a portion of their land and use conventional tillage on the rest of their plots. Data from Eastern Province shows that adopters used MT on 54% of their land in 2008 (Arslan et al., 2014). One of the key problems with MT is that w eed control is difficult without soi l inver sion (Giller et al. , 2009 ; Wall , 2007 ; 107 Gatere et al. , 2013 ) , especially in the absence of a thick layer of mulch. Another commonly identified constraint to using MT on larger areas is the increased labor requirements for basins ( Grabowski and Kerr , 2014 ). The analysis of multi - purpose household survey data provides some statistical evidence for the reasons for partial adoption in Zambia. Households tend to use MT on more of their land where soils are depleted of nutrients , (Arslan et al. , 2014) if they are male headed, and if they are in areas where CFU operated (Ngoma et al. , 2014). In one analysis, f armers with more land used MT on a smaller proportion of their land (Arslan et al., 2014) but an other study found that those who farm more land use MT on more area in absolute terms (Ngoma et al., 2014). Together these suggest that environmental and economic conditions combine to affect the relative benefit of using MT on more land, though training may serve to reduce some of the constraints. 4. 3. Methods and d ata This research combines qualitative and quantitative methods to understand partial adoption of MT. The statistical relationships among economic and agronomic variables are supplemented by qualitative analysis of explanations about why they ch os e to use MT on some plots but not others. I used results of a nationwide census of lead cotton farmers (as described in Grabowski et al. , 2014 ) to stratify communities in Eastern Province by MT adoption levels . I then randomly selected 15 c ommunities where CA adoption was expected to be relatively high . Next I stratified farmers within each community by their use of various tillage technologies and randomly selected some from each group . The complete data includes 1178 plots fr om 245 farmers, of which 81 f armers use some form of MT . The complete survey methodology is det ailed in Chapter 3 . 108 4. 3.1 Data collection 4.3.1.1. Qualitative data The first author conducted i n - depth interviews in the local language, Chinyanja, with 34 farmers, 18 distributors from NW K and 11 buyers from Cargill. Of the 69 individuals interviewed, 43 had experience using MT on their own fields for at least one season. For each of the communities that were randomly selected I interviewed the associated distributor, buyer or chairperson that links the farmers to the cotton company. I carried out most i n - depth interviews in five communities purposively chosen to represent the range of experiences with MT technologies (high tractor ripping, high ox - ripping (one each for Cargill and NWK), h igh basins, and low adoption). Similarly, in each community I selected farmers for in - depth interviews from each category of land preparation (plow, ox - ripper, hoe, basin user, tractor ripper or disadopter) to understand the perspectives of all groups. performance of MT. Because of the cultural tendency to please outsiders and the dominant narrative from the cotton companies that CA is the best way to farm, there was a risk that farmers would withhold criticism of CA, even when directly asked. To avoid this bias the lead researcher operated independently from cotton company staff, emphasized the confidentiality of responses and presented the research goal in general terms of learning from farmers about how to improve production. Interviews with farmers initially focus ed on their farming practices for the previous season before asking about their experiences and perspectives on MT with the goal of building rapport and starting with questions about facts before asking for opinions. In the course of the interview I asked farmers who used MT how they learned about it, what they had done previously, their motivation for using MT, their assessment of how MT 109 performs compare d to conventional tillage and the factors that prevented them from using MT on more land. I asked d istributors, buyers and community leaders about general community - level issues related to inputs and production as well as the constraints to more widesprea d MT use in the area . I recorded and transcribed all interviews . The in - depth interview guides can be found in Appendices B and C of Chapter 3 . 4.3.1.2. Survey data Surveys were carried out with 245 cotton farmers across Eastern Province and detailed inf ormation was provided for the 1178 plots of land they cultivate ( Table 11 ) . F or this study plots were defined as portions of fields with one land preparation method and planted to one primary crop. I excluded p lots with tubers as the primary crop (cassava and sweet potato) from the analysis because all of them were prepared with conventional hoeing. Table 11 : Tabulation of land preparation methods for all plots in the dataset Land preparation method Conventional plots MT fa plots All plots Ox - plowing 483 230 713 Ox - ripping 0 80 80 Hoeing 210 85 295 Basins 0 54 54 Direct seeding 16 1 17 Total plots 693 449 1142 Farmers 160 85 245 Source: Survey of NWK and Cargill Farmers, 2013 I asked each farmer to pro vide details about the characteristics of all plots cultivated and their management practices on each plot. I also asked about household characteristics such as educational attainment , total landholding size, types of non - agricultural inc ome sources , lives tock owned and ownership of a variety of agr icultural and household items. I also asked f armers opinions about how they compare hoeing to basins and ripping to plowing. Distributors 110 and buyers at each location provided quantitative information about how l ong CA has been promoted and about their own farming practices. F or each plot f armers provided details about soil type (sand, clay or other) and soil color (black, bro wn, red, or white) . In general, white sand and red clay soils are lower in fertility than brown or black soils. Sandy soils tend to be easier to work and can be planted quickly with moderate rainfall but do not hold nutrients or moisture as well . Clay soils hold nutrients and moisture but are harder to work and require more rainfall before the y are ready to be planted. A wide range of textures lay between sand y and clay soil (i.e. clay - loam, sandy - loam, etc.) but I categorize these intermediate textures together in this analysis. The complete survey instrument can be found in Appendix A of Chapter 3. 4.3.2. Data a nalysis 4.3.2 .1. Qualitative data a nalysis I systematically examined t he qualitative data (interview transcripts, researcher notes and comments written on surveys) using thematic analysis. In thematic analysis the researcher categorizes the textual data using codes to facilitate retrieval of similar information across the data and then develops summary statements to concisely represent the diversity of responses (Miles and Huberman, 1994; Rubin and Rubin, 2005 ). I use quotes from respondents where possible to succinctly represent the perspectives in the data. In addition, I use numeric tabulation of responses to clarify the level of agreement or disagreement in the data. 4.3. 2 .2. Statistical a nalysis To focus on understanding partial adoption, the statistical analysis includes only data from the 81 respondent s who used some MT in 2012/13. To be able to generalize to the wider population (cotton farmers using MT in medium to high adopting communities) s ampling 111 weights were used for both regressions as out lined in Chapter 3 . I included d ummy variables for community clusters to control for unobserved heterogeneity across communi ties (see Chapter 3 for details and Table 12 for characteristics of clusters ). Table 12 : Characteristics of clusters used in regression analyses Cluster 1 Cluster 2 Cluster 3 Cluster 4 Crop area with sandy soil 4.7% 16.9% 10.5% 18.7% Crop area with red loam soil 3.7% 6.3% 7.9% 5.9% Crop area with clay soil 45.2% 17.2% 57.0% 48.7% Crop area on flat plots 37.2% 66.0% 41.2% 63.8% Crop area with plan to rest soil 1.6% 9.4% 0.0% 2.4% Crop area under a cash crop 77.7% 81.0% 81.9% 81.8% Crop area with secure tenure 100.0% 95.7% 99.5% 98.7% Percent of crop area planted to hybrid maize 42.0% 19.7% 29.4% 28.9% Percent of crop area planted to local maize 11.5% 19.1% 15.2% 18.5% Percent of crop area planted to cotton 30.6% 30.5% 36.6% 26.6% Population Density (people/square kilometer) 48.96 140.53 83.3 50.1 Elevation (meters) 1006 1106 833 1012 Percent of farmers using animal traction 76.6% 67.2% 41.8% 65.6% Average fertilizer application rate to maize (kg/ha) 99.6 308 273.6 187.4 Percent of plots where residues were heavily grazed 34.9% 29.6% 32.0% 31.8% Source: Survey of NWK and C argill farmers, 2013 4. 3.2 .2.1. Modeling choice of land preparation method for each plot Most farmers use more than one land preparation method in any given season. The 81 farmers using MT had a total of 449 plots prepared with one of the four primary land preparation methods (plowing, ripping, hoeing or basins) . I used dummy variables for soil types with loam soils that are not red colored as the base case. A multinomial logi stic model estimate s how a 112 marginal change in plot characteristics affect s the pro bability of using any given land preparation method to the others. The multinomial logistic model can be presented formally as: (equation 2 ) for land preparation categories m = 1 to J where b is the base category (Long and Freese, 2001). 4. 3.2 .2.2. Modeling intensity of adoption of minimum tillage facilitated by defining intensity of adoption as the proportion of their land under MT. I used a q uasi - maximum likelihood fractional logit estimator (Papke and Wooldridge , 1996 ) to model the the best choice for proportional responses that include the extreme values of zero and on e. It is more appropriate than a Tobit model, which assumes that there is a continuous latent variable for which negative values are being censored (Baum, 2008). For simplicity this model co mbines basin and ripping area. The average partial effects are pre sented after multiplying by 100 so that they can be interpreted as the effect on the percent of land under MT. 4. 4. Results In the results I first provide a brief summary of how farmers are using MT. Next I present analysis of how crop and plot characteri stics determine the conditions where farmers prefer MT over conventional tillage. In the third section I analyze the household - level constraints to more widespread use of MT. For both the plot - level and household - level analyses, I first present the quali results. 113 4.4.1. How farmers are using basins and ox - ripping Among farmers who use some form of MT, 30% of their area was ox - ripped, 8% was in basins and the o ther 62% was prepared wi th conventional tillage (Table 1 3 ). Farmers use MT on all types of crop s except tubers, but the overwhelming majority (84%) of MT area is planted to maize and cotton . Partial adoption of MT is the norm (Figure 8 ). Only one farmer us ed basins on all plots, totaling just one hectare. Another farmer used ox - ripping on 97% of the farm (9.5 ha), hoeing only a 0.25 hectare plot of sweet potatoes. Table 13 : Percent of overall area under minimum tillage by crop for f armers who use some MT All crops Cotton Hybrid Maize Local Maize Ground - nuts Other crops Percent crop area a with ox - ripping 30.1% 35.3% 31.0% 31.2% 14.2% 24.0% Percent crop area with basins 7.8% 4.5% 6.1% 21.6% 0.01% 4.8% Percent crop area conventional 62.1% 60.2% 62.9% 47.2% 85.8% 71.2% Total 100% 100% 100% 100% 100% 100% Source: Survey of NWK and Cargill farmers, 2013 a Percent crop areas are calculated using data from all farmers using some MT 114 Figure 8 : Percent of area und er minimum tillage (MT) by cotton farmers who use some M T Source: Survey of NWK and Cargill farmers, 2013 ton and maize differs according to the land preparation method that they used. Basin plots are smaller than hoed plo t s on average (Table 14 ), which suggests a labor constraint (see 4. 4.3 below). Plots prepared with ox - ripping and ox - plowing tend to be larger since the human labor constraint is overcome with animal traction. Fertilizer use is lower on hoed plots than plo wed plots, which is likely indicative of cash constraints that limit both the hiring of oxen and purchasing of fertilizer. Basin plots tend to receive more manure and less herbicides than ox - ripped plots , which again is likely related to the limited cash r esources available to farmers using basins . Ox - ripped plots are twice as likely to have herbicides applied than ox - plowed plots , which is understandable given the lack of physical weed control from tillage . 0 2 4 6 8 10 12 14 16 18 Frequency Percent Area under Minimum Tillage 115 Table 14 : Differences in plot size, preparation date and inputs by land preparation metho d for cotton and maize plots of farmers who use minimum tillage on some part of their land Basin plots Hoed plots Ox - ripped plots Plowed plots Number of plots 47 37 63 121 Mean plot size ( ha) 0.46 A 0.66 B 1.43 C 1.03 C Percent of plots fertilized 42.3% AB 33.4% A 49.8% AB 49.3% B Percent of plots with manure added 30.6% A 16.7% AB 3.8% C 7.0% BC Percent of plots with herbicides 0.3% A 3.9% A 47.6% B 17.2% C Percent of plots prepared be fore December 99.9% A 59.3% BC 86.8% AB 40.9% C Source: S urvey of NWK and Cargill farmers, 2013 Notes: Values are calculated using sampling weights in order to estimate the true population values.Values in a row that do not share a capital letter superscr ipt are significantly different (p < 0.1) from each other using a Wald Chi - squared test and linearized standard errors that incorporate the survey design 4.4.2. Conditions where farmers prefer using MT s where MT is preferred over conventional tillage provides insight into the constraints that must be overcome. Whether will need to be addressed before more wide spread adoption is likely. I first present the results of the in section 4.4.2.3 I quantitatively analyze to what degree these stated concerns affect how farmers use MT on their fields. 4. 4.2.1. Crop - specific MT benefits Groundnuts are the most commonly grown legume in Eastern Province but MT farmers planted them primarily on conventionally tilled plots. Several farmers explained in the interviews that they believed groundnuts would not produce well without adequate tillage . This concern stems from how groundnuts need to penetrate the soil with the fertilized ovary when pegging. In 116 contrast, several farmers said they had good success with groundnuts on MT plots. One of the key challenges with using MT on groundnuts , and other less prioritized crops, is that they tend to be planted later, once the maize and cotton has bee n planted . Farmers explained that if they do not plant a MT plot soon after the first rains they w ill have to till the soil or the weeds will take over. Unless MT becomes more widely used with groundnuts and other legumes, farmers will not be able to practice both crop rotation and minimal soil disturbance on most of their land. O ne of the primary rea sons farmers provided for adopting MT was to increase their yields through earl y planting. This benefit is especially important for cotton and local maize varieties, which take the full length of the rainy season to mature . For this reason most farmers us e MT with cotton and maize and carry out the land preparation before December so that they can plant with the first rains and avoid yield losses associated with late planting. 4. 4.2. 2 . Plot - specific benefits and challenges of MT identified by farmers By a sking farmers how they decided on which plot to use MT , a number of location - specific issues came to light. F armers considered the slope of the field, the soil type and the fertility of the soil in their decision. 4. 4.2. 2 .1. leads to e rosion Several farmers who use ripping expressed concern about erosion on plots that are not flat, even th ough they rip across the slope. One farmer expressed this concern as a request for technical advice: hen the rains come I am getting gul lies so now I need knowledge for using m (Key informant 28). Another farmer who used tractor ripping even completely di sadopted MT because his seed eroded out of the rip lines. Similarly, several farmers found plowing prefera f I rip there [on the slope], the water just goes but if I make ridges the water starts to stop (Key informant 24). The 117 lack of residue retention on most MT plots increases the potential for erosion. Residues on the surface reduce the impact of rain on the soil and can increase infiltration of water by preventing crusting of the soil. However, this concern about erosion was not shared by all respondents . Most farmers surveyed think erosion is greater with conven tional tillage than with MT . Several farmers who use ox - ripping explained how they made contour ridges in their ripped fields to prevent erosion. some small ridg es with oxen, just on top, to break the water from running (Key informant 25). This adaptation obviously creates an additional task for the farmer and disturbs the soil, thus reducing some benefits from MT. 4. 4.2. 2 .2. Soil type affects the costs and benef its of MT Most farmers said that ripping and basins could be used on any soil. However, two ripper farmers with sandy soil said tha t they could not use ripping because the furrow made by the r ipper would not be stable . If the rip line collapses before plan ting , the work of ripping was in vain. One of these farmers , who also had a red loam and a black loam soil, explained that ripping was more difficult in the red soil because it became so hard in the dry season. 4.4.2.2 .3. Soil fertility affects the magnit ude of the benefits Of the eight farmers who specified why they chose t o use MT on a particular plot, seven of them mentioned declining soil fertility on the plot as a reason for using MT there. For example, one farmer said, [I decided to ripn that field ] t o protect the soil. I saw the soil was not great. I thought I should just do ripping . The cotton was not doing well (Farmer 27). Similarly another farmer reported successful remediation of a degraded plot as follows: 118 I chose that one [plot] because it was going down but I learned about ripping and that it could do well, so I thought let me do that plot. So, even though that plot was without fertility, the crop did well. (Key informant 4) One of the mechanisms by which MT can significantly improve crop performance over the short term is by breaking through a compacted layer of soil. One respondent described his If you dig it [basins] , the roots are happy, because it is porous, it is nice. So I th ought I should dig th 4.4.2.3. Plot characteristics associated with MT T he statistical relationships among plot - level characteristics and land preparation method largely support the reasons farmers provided (Tab les 15 and 16 ). The plot characterist ics affect the decision to use MT independent of what crop is planted on the plot, as these are controlled for. A simple way to visualize these results is a table of the predicted probabilities that each land preparation method will be used on a plot given the crop planted on it (Table 17 ). Flat plots are not statistically more likely to have MT than sloping plots, so the concern some farmers have about erosion does not seem to widely impact behavior. The strong contrast bet ween predicted rates of basin use and ox - ripping use by soil type (Table 17 ) shows how they are being used very differently. O x - ripping is less likely on sandy soils and red loam soils , though not at a statistically significant differe nt rate from ox - plowi ng (Table 16 ). Thus there is some evidence that ox - ripping is used preferentially on plots with soil types that are neither too soft nor too hard. Basins are used more on s andy soils , which are easier to dig in the dry season than clay soils , which is logi cal given the effort required for manual dry season land preparation . However, it is also likely that sandy soils are lower in fertility and that basins are being used with 119 the intention of improving their productivity through the precise addition of manur e. T his interpretation support the qualitative results and is confirmed by t he coefficient on the dummy variable for degraded plots , which is negative a nd significant for basins, showing that farmers do not consider a plot to be degraded once basins have b een used there . Table 15 : Description of variables used in the plot - level regression of determinants of land preparation method for M T users Mean Standard Deviation Minimum Maximum Soil Type Sand (Y/N) 0.176 0.381 0.00 1.00 Clay (Y/N) 0.374 0.484 0.00 1.00 Red Loam (Y/N) 0.065 0.246 0.00 1.00 Plot Characteristics Flat (Y/N) 0.608 0.489 0.00 1.00 Soil Degraded (Y/N) 0.031 0.174 0.00 1.00 Secure Tenure (Y/N) 0.980 0.140 0.00 1.00 Crop Characteristics Crop is sold (Y/N) 0.706 0.456 0.00 1.00 Cotton Plot (Y/N) 0.247 0.432 0.00 1.00 Hybrid Maize Plot (Y/N) 0.203 0.402 0.00 1.00 Local Maize Plot (Y/N) 0.151 0.359 0.00 1.00 Community Level Variables Cluster 1 0.174 0.379 0.00 1.00 C luster 2 0.241 0.428 0.00 1.00 Cluster 4 0.285 0.452 0.00 1.00 Source: S urvey of NWK and Cargill farmers, 2013 120 Table 16 : Factors affecting plot - level use of MT, multinomial logistic regression results with robust standard errors Explanatory Variables Plowing vs. Hoeing Basins vs. Hoeing Ripping vs. Plowing Soil Type Sand (Y/N) - 0.408 (0.789) 1.04 * (0.557) - 0.297 (0.492) Clay (Y/N) - 1.652 * (0.958) 1.204 (0.827) - 0.762 (0.749) Red Loam (Y /N) - 0.432 (0.678) - 0.428 (0.485) 0.392 (0.627) Plot Characteristics Flat (Y/N) 0.662 (0.440) 0.0816 (0.449) - 0.306 (0.427) Soil Degraded (Y/N) 1.328 (1.279) - 18.05 *** (1.035) - 0.579 (0.693) Secure Tenure (Y/N) - 2.412 ( 1.629) - 0.886 (1.563) - 2.356 * (1.334) Crop Characteristics Crop is sold (Y/N) 1.151 ** (0.448) 0.451 (0.726) 0.464 (0.574) Cotton Plot (Y/N) 0.0343 (0.452) 0.533 (0.745) 0.66 (0.723) Hybrid Maize Plot (Y/N) 1.177 ** (0.57 5) 2.007 *** (0.619) 0.425 (0.605) Local Maize Plot (Y/N) 0.888 (0.542) 3.111 *** (1.015) 1.247 ** (0.487) Community Level Cluster 1 - 1.327 (0.999) - 1.234 ** (0.495) 0.802 (0.547) Cluster 2 - 1.654 * (0.957) - 0.0733 (0.504) 0.9 14 (0.565) Cluster 4 - 0.169 (0.826) - 0.548 (0.585) 0.909 (0.893) Constant 2.394 (1.922) - 1.215 (1.575) - 0.105 (1.543) Observations 449 Wald Chi 2 (39) = 5,585.65 Pseudo R 2 = 0.1620 Prob > chi 2 = 0.0000 Log pseudolikeli hood = - 27,800.9 Source: Survey of NWK and Cargill farmers, 2013 121 Table 17 : Predicted probabilities of land preparation method from the results of the multinomial logistic regression. Source: Survey of NWK and Cargill farmers, 20 13 Notes: P robabilities calculated assuming secure tenure and plot not degraded and holding cash crop and community variables at the mean 4. 4.3 . C onstraints to increased MT use In addition to these crop - and plot - specific factors that affect the performanc e of MT , there are household - area. From the in - ir partial use of MT fall into three categories of constraints: informational constrai nts , investment constraints , and labor constraints . In this section I start by summarizing the qualitative results for each of these 122 constraints. Next I quantitatively analyze the degree to which these constraints explain partial adoption. 4.4.3.1. Informa tion constraints: Incremental testing of MT as a reason for partial adoption Farmers who are new to MT typically test the technology on a small portion of their cotton or maize. One farmer explained, The ripper training I received just last year, so I wa s just trying it out to really learn it. So I saw that ripping is good and this year I will rip on a large area, pretty much all crops I will rip. It does very well. (Key informant 4) E xperienced MT farmers described how they went through a similar proces s of testing before using MT on most of their area. One farmer and CFU trainer who was in his eighth year using ox - ripping exp lained that he incrementally increased his ripped area from 0.25 ha in the first year to 2.65 ha in 2012/13. Surveyed f armers ave rage area under MT doubles from the fir st to the second year and increases by another 36% in the third year, after which the mean area plateaus at just under 2 hectares (Figure 9 ). Figure 9 : Average area under minimum tillage So urce: Survey of NWK and Cargill farmers, 2013 Note: Mean of 71 farmers for years 1 to 3; Mean of 22 farmers for >3 years 0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2 Year 1 Year 2 Year 3 > 3 years Hectares 123 The stories shared by veteran MT users about their early challenges with MT show the wisdom of this incremental approach. Several farme rs mentioned their frustrations learning to effectively apply herbicides either killing some of their crop or not killing the weeds, resulting in heavy competition before they managed to weed manually. Others described how , when they first started using MT , they watched their waterlogged crop suffer before seeking advice. They then learned the importance of quickly adding soil to the base of the plant so that yields are not affected . S ome farmers decide d to use both MT and conventional tillage to ensure t hat at least one plot does well rega rdless of the rainfall pattern. 4.4.3.2. Investment constraints - c hallenges obtaining rippers and oxen For those farmers who rent ripping services, their MT area is limited by the time they can access the oxen and equi pment. Several farmers who borrowed rippers and rented oxen explained that the challenge of coordinating these services limited the areas they could manage. One older man and ripper owner explained that he and his sons were unable to rip all of their maiz e because they shared the ripper among 6 households. Though the dry season is a long period (May through October), farmers typically only rip part of the day to avoid tiring their oxen. This also contributes to a reluctance to lend or rent out the equipmen t. 4.4.3.3. L abor constraint the increased effort needed for MT as a reason for partial adoption 4. 4.3.3.1. Dry season labor issues Dry season land preparation with MT is labor intensive, causing many farmers to be unable to implement it on all of their area. Many farmers recognize that the effort needed for preparing the land with MT is lower soon after harvest, before the soil has had a chance to dry out. However, they explain that this is a busy time for marketing and processing their harvest. As o ne distributor said, 124 when I am going all around buying cotton. So now I just di d one acre of maize with basins (Key informant 16). Another basin farmer who also does wetland gardeni ng in the dry season said that his area was limited because of other activ ities early in the dry season. There are a lot of things we normally do. So the time we are busy preparing fields, we cotton, whereby, it takes more time do that. So we are busy with other things. But MT 3 is better. (Key informant 14) Due to these issues many can not focus on dry season land preparation until October when the soil is drier and harder. While the labor - int ensive nature of digging basins is well documented, the increased effort needed for oxen relative to plowing is less well known. Several farmers who rip explained that they manage their ripping in small portions each day so as not to wear out their oxen: do it in the hot times. We do it first in the morning and then we do 10 lines, tomorrow 15 lines; that is all. Then we give them [the oxen] some time [to rest]. (Farmer 19) One day we can do 1 lima, or a half acre [0.2 to 0.25 ha], so that the cattle have enough time to get food; forage is difficult at this time. It is also the time when we are harvesting when we rip, so we do it little by little. Harvesting some, ripp ing some. (Farmer 10) Dry preparation is very difficult, the soil is very hard, we do it bit by bit. It may take a month or so. (Key informant 20) 3 is translated here as minimum tillage (MT) as it can refer to both basins and ox - ripping. 125 This concern for oxen health limits the area a farmer may rip on his own land. A careful oxen owner will not risk hurting the oxen by ripping a large area. One farmer explains this challenge as follows: 5 acres the health of my oxen was low from all of that work. So I was forced to stop so that I store up enough power to be able to plow well the other 10 acres. So this is only half a bread better than none, so that is why I plowed. (Key informant 18) Th e concern for oxen health also limits the availability of ripping rental markets, which prevents those who cannot afford the equipment from even trying ox - rip ping ( Chapter 3 ). Once the rains come, ripping is no longer recommended and most farmers do not at tempt it. Likewise, d igging basins after the start of the rains is not recommended , though it is occasionally practiced . There is some risk that digging basins in wet soils with high clay may actually compact the soil, thereby increasing waterlogging and reducing root growth. 4.4.3.3.2 . Weeding labor issues One of the primary functions of tillage is weed control and weed pressure is greater on MT plots (Wall , 2007; Giller et al. , 2009). Farmers may therefore limit their areas under MT so as to make sure t hey will not have to abandon plots that become overrun with weeds. This quote from a ripper farmer summarizes a perspective shared by many MT farmers that herbicides are a requirement: The challenge for using a ripper is that weeds, weeds do attack very fast, since you have just ripped in that furrow you will find the weeds attack you very much there. Once you 126 where you are doing ripping you need to have herbicides, some where somehow. (Key informant 20) One farmer stopped using ripping when he could not get herbicides on loan from the cotton company. Another farmer said that if the cotton company rejected his requests for herbicides on credit he would not be able to do r ipping, saying, informant 28). One farmer explained that he plows the fields where he notices the weed pressure is greatest and only rips on the one hectare where weeds are a bit less. Others explained that they ar e managing to use MT without herbicides but they have to put extra effort into weeding . One of the complementarities between CA principles is that substantial crop residue retention can provide a layer of mulch that will suppress weed growth. Unfortunately , in Eastern Zambia the frequency of free range cattle and uncontrolled fires make it extremely difficult for farmers to retain their residues. Over 80% of MT plots recorded in the survey had lost more than half of the residues to graz ing and over 25% to b urning . One reason why some farmers did not find weeds a challenge with MT is that they used conventional mid - season tillage to control weeds. Many ripper farmers (26 out of 49 responses) use banking (moving soil from the inter - row with a hoe or plow) on MT plots to control weeds. Others use an ox - drawn cultivator or hand hoe to weed between the rows. This mid - season tillage reduces the long - term potential benefits of minimizing soi l disturbance . Furthermore, fields that are banked are difficult to rip th e next year. O ver half the farmers rip between ridges the year after banking. Farmers explained that ripping in the inter - row is easier than ripping on a ridge and is necessary to cut sufficiently deep for improved infiltration of water. Unfortunately ripp ing the inter - row makes the oxen have to exert more effort than if the 127 rip lines stayed in the same place year after year. This contributes to the dry seas on labor constraint . 4.4.3.4. Quantitative analysis of household - level constraints To examine the st atistical significance of these constraints I used a fractional response model to estimate the average partial effects of household - and community - level variables on the proportion of cotton and maize area a farmer has under . The means, standard deviations and ranges for the variables in the regression are in Table 18 . The results generally support the importance of the constraints identified through the qualitative analysis (Table 19 ). Farmers use MT on more of their cotton where the distributor or buyer also uses MT on more cotton area, indicating that their example reduces the informational constraints . The lack of significance for the coefficient on years of experience with MT could be due to this variable only being important for the first three years, after which farmers reach an upper limit on MT (Figure 2). F armers who have more adult household members per area farmed tend to use MT on more area , which shows the importance of labor as a constraint . Also , farmers who operate larger areas use MT on a larger proport ion of their cotton and maize. Total o perated area is a wealth indicator (due to higher income from producing on more area). W ealth is generally associated with higher risk tolerance and also enables farmers to purchase herbicides and equipme nt. In addition, the regression results highlight a number of interesting issues that did not arise from the qualitative analysis. Farmers who sell more crops use MT on less maize area. This suggests that farmers who are less commercialized (i.e. produc ing primarily for their own consumption) are using more MT on more of their maize. These farmers may have lower opportunity costs of labor and greater need to produce sufficient food on their own land, thus 128 making MT relatively more attractive. Farmers w ho own more livestock use MT on less of their cotton and those who own more equipment use less MT overall. These may be related in that farmers who have more oxen and more plowing equipment may face fewer challenges preparing all their land soon after the rains, thus decreasing the need to use MT to achieve dry season land preparation. Farmers who live in d ispersed settlement s (i.e. not in villages) use MT on 32% less of their cotton area. Living in a dispersed settlement could potentially increase all t hree of the constraints identified above: information constraints are greater due to less opportunities to equipment; and weeding labor costs tend to be higher in areas with lower population density. Further research is needed to understand the relative importance of these factors in dispersed settlements. Another reason for lower percent of area under MT in dispersed settlements is that those farmers tend not to use MT on cotton, which may simply be caused by lack of extension by cotton companies in these areas (see Appendix for details). The percent of cotton area under MT is lower in cluster 1 which has the least sand and red loam (where MT tends to be used) an d the most clay soils (where MT is used less). Cluster 2 and 3 have larger proportions of cotton area under MT, possibly because they have the most flat area (see Table 1 2 for more details on cluster characteristics). 129 Table 18 : De scription of variables used in the household - level regression Mean Standard deviation Minimum Maximum Dispersed settlement (Y/N) 0.58 0.50 0 1 Buyer's area under MT (% ) 33.4 24.6 0 81.5 Years using MT 3.20 3.07 1 25 Number of crops sold 3.15 1.69 1 9 Total Fertilizer (kg) 410 426 0 2600 Total Livestock (TLU) a 4.80 4.91 0 23.27 Total Value of Equipment ($) 1226 2419 0 15,945 Adult HH members per ha 1.09 0.94 0.29 6.29 Total Operated Area (ha) 4.83 3.18 1.05 18.4 Cluster 1 dummy 0.19 0.39 0 1 Clus ter 2 dummy 0.22 0.42 0 1 Cluster 4 dummy 0.26 0.44 0 1 Source: Survey of NWK and Cargill farmers, 2013 a Tropical Livestock Units - cattle have a value of 0.7, goats and sheep 0.1, pigs 0.2, poultry 0.01 and donkeys 0.5 (Jahnke, 1982) 130 Table 19 : Average partial effects of h ousehold - and community - level factors influencing intensity of adoption of minimum tillage (MT) . Percent Maize Area under MT Percent Cotton Area under MT Percent of All Area under MT Explanatory variable APE S td. Error APE Std. Error APE Std. Error Dispersed settlement (Y/N) - 1.38 (12.33) - 32.18 *** (6.15) - 13.61 * (7.88) Buyer's area under MT (%) - 13.99 (27.05) 58.86 *** (11.71) 11.09 (18.16) Years using MT - 2.09 (1.94) 0.36 (0.74) - 0.69 (0.95) Num ber of crops sold - 7.10 *** (2.60) 1.31 (1.29) - 1.74 (1.27) Total Fertilizer (kg) - 0.001 (0.02) 0.004 (0.013) - 0.01 (0.02) Total Livestock (TLU) a 0.36 (1.37) - 3.37 *** (0.94) - 0.50 (0.94) Total Value of Equipment ($) - 0.004 (0.004) - 0.004 (0.00 3) - 0.005 * (0.003) Adult HH members per ha 13.59 (9.55) 11.38 *** (3.27) 7.35 ** (3.47) Total Operated Area (ha) 4.78 * (2.87) 10.82 *** (3.21) 5.07 *** (1.85) Cluster 1 dummy - 20.68 (13.55) - 83.05 *** (26.90) - 13.56 (8.65) Cluster 2 dummy - 9.76 ( 13.78) 29.30 *** (10.59) - 0.18 (9.28) Cluster 4 dummy - 5.23 (13.47) 19.16 ** (8.45) 1.44 (7.91) Observations 79 77 79 Log pseudo - likelihood - 2,294.56 - 1148.05 - 1827.55 Wald Chi - squared Chi 2 (12) = 13.82 p=0.3122 Chi 2 (12) = 31.77 p= 0.0015 Chi 2 (12) = 30.83 p=0.0021 Robust delta - method standard errors in parentheses. *** p<0.01, ** p<0.05, * p<0.1 Source: Quasi - maximum likelihood fractional logit estimator (multiplied by 100 for easier interpretation) using data fro m survey of NWK a nd Cargill farmers, 2013 a Tropical Livestock Units - cattle have a value of 0.7, goats and sheep 0.1, pigs 0.2, poultry 0.01 and donkeys 0.5 (Jahnke , 1982) 131 4.5. Implications for CA promotion st that agricultural extension services (from any source) that provide training on how to reduce erosion may support more widespread use of MT. Instructing farmers to use contour bunds between rip lines to prevent erosion may require training as well as ov ercoming negative associations with being forced to carry out soil conservation during colonial times. Achieving residue retention will require controlling free - range cattle and wildfires. The identification of three primary constraints to farmers using M T on more of their land (informational, investment and labor constraints) helps guide how CA needs to be adapted and promoted so that farmers can experience the yield benefits on more of their land. First, increased adoption will require reducing farmers technologies. Awareness of CA is widespread in Zambia so mass campaigns are likely to be ineffective. Education about CA will be more likely to help farmers use MT on more area in a shorter time key concerns, such as how to effectively manage weeds, what to do on MT plots in wet years and how to use MT on sloping land. Personal experience using MT and the presence of local opinion leaders using MT are important determinants of farmers using MT on more land. This suggests that CA promotional efforts should combine situated experiential learning - by - doing with efforts to link participants with MT veterans who can help farmers overcome their practical concerns with how to benefit from using more MT. In addition, effective learning about MT must be accompanied by increased ability to access ox - Equipment subsidies or longer - term loans may help reduce the investment cons traint by making 132 the equipment more available. T he concerns of oxen owners for oxen health indicate that the dry - season labor issue must be addressed together with the equipment availability issue. Finally, current MT technologies need to be adapted to ov ercome labor constraints. F armers minimize dry season labor costs by us ing MT on plots where the dry - season labor requirement is less but where MT is still effective: basins tend to be dug on sandy soils and ripp ing tends to be done on soils that are not t oo hard for the oxen but also not sandy (so t he furrow does not collapse) . mulch) has the potential to not only reduce weeding labor but also to reduce the ox - labor constraint associated wi th dry season ripping by reducing the need for banking, and thus enabling ripping in the same line each year. Improvements to livestock health and dry season forage availability could help an average team of oxen be able to rip through the dry season with out over - exertion. Easing this labor constraint is also likely to increase the availability of ripping rental services , which would help more farmers have access to the rippers and avoid the investment constraint . While technologies are being adapted, tar geting areas with lower constraints would make sense. Due to the unavoidable labor constraint of digging basins with a hoe, promotion may be more effective if it is targeted at labor - abundant, land - constrained households. Also t he short - term benefits of b asins tend to outweigh the costs on plots that are degraded or where compacted soil has significantly affected yields. Targeting basin promotion at areas with land degradation may therefore be a nother logical approach. 4.6. Conclusion Conservation agricul ture holds great potential for helping farmers become less vulnerable to rainfall variability. Theoretically CA would be adopted as a package of complementary 133 y. Over time the labor costs would decrease while the benefits to crop production increase. However in Eastern Zambia, where CA has been promoted for over a decade, the farming system of free - range cattle and widespread burning make residue retention ext remely difficult, which increases the challenges of weeds and erosion on MT plots. Furthermore, even where MT land prepar ation methods are used, the benefits of minimizing soil disturbance are lost due to the widespread use of tillage for weed control. The fact that many farmers are using MT despite these challenges, and without any expectation of direct incentives, shows that the short - term benefits of MT outweigh the costs at least for certain farmers and under certain condit sons for partial adoption and empirical observation of the factors associated with increased use of MT reveal three challenges that need to be addressed for making the benefits of CA available to more farmers on more of their land: 1) reducing uncertainty about the technology through experiential learning, 2) reducing barriers to entry by making equipment more available and more effective, and 3) adapting technologies or identifying new technologies that can improve production despite difficult to change co nstraints. Similar challenges are likely to be applicable to a wide range of agricultural technologies across many contexts. Realizing improvements in smallholder production through agricultural technologies will require systematically identifying and a ddressing these challenges. The results for guiding the efforts of development agencies and researchers. 134 APPEN DIX 135 Appendix: Intensity o f MT use by crop This section provides details of the differences between farmers who use MT on cotton and those who do not use MT on cotton. When the area under MT for each farmer is separated by crop, a striking result is a bimodal distribution with hig h peaks at complete adoption and complete non - adoption for cotto n and maize (Figure 11 ). Twenty farmers used MT on all of their cotton, while 39 used it on none of their cotton . A similar distribution of many farmers using MT on 100% or 0% of their maize a rea was also observed. Eight farmers (10%) used MT on 100% of their cotton and maize area. Figure 10 : Percent of area under minimum tillage (MT) for maize, cotton and other crops by cotton farmers who use some MT on any crop Sour ce: Survey of NWK and Cargill farmers, 2013 A simple comparison of means with a t - test shows some important differences. There is 0 5 10 15 20 25 30 35 40 45 50 Frequency Percent Area under Minimum Tillage Maize Cotton Other 136 farmers who use MT on cotton an d those who do not (42% compared to 23%). There is also a statistically significant difference in the proportion of farmers living in dispersed settlements (as opposed to villages) between those who use MT on cotton and those who do not. 76% of those who d o not use MT on cotton live in dispersed settlements, compared to 42% of those who use MT on cotton (57.6% live on farms on average) . Also a larger proportion of farmers who do not use MT on cotton are basin users (66% over 44%), which is statistically sig nificant. There is not a statistical difference in the size or proportion of maize area under MT between those who use MT on cotton and those who do not. However, this comparison of means across those with and without MT on cotton and those with and witho ut MT on maize can be misleading. There are really three distinct groups: those using it on cotton but not maize (14 farmers), those using it on both cotton and maize (29 farmers), and those using it on maize but not cotton (36 farmers). When these variab les are combined in a multinomial logistic regression (for the three groups mentioned above) it becomes clear that source of information about MT is the primary determinant of whether farmers use MT on cotton only or maize only (Table 20 ). T he influence of cotto n. L iving in a dispersed settlement and owning more livestock also having significant negative relationships with MT use on cotton. 137 Table 20 : Multinomial logisitic regression results of household characteristics and community characteristics associated with whether MT is used on cotton or maize or both. MT on cotton only vs. MT on cotton and maize MT on maize only vs. MT on cotton and maize Dispersed Settlement (Y/N) 1.034 (1.03) 4.004 *** (1.14) Buyer's area under MT (%) - 0.367 (2.15) - 11.3 ** (4.39) Received Incentives (Y/N) - 4.439 *** (1.48) 2.65 ** (1.35) Total Operated Area (ha) 0.131 (0.32) - 0.675 ** (0.32) Total Fertiliz er (1000kg) - 0.473 (0.00) - 2.43 (0.00) Total Livestock (TLU) a - 0.0999 (0.17) 0.419 * (0.23) Cotton Experience (years) - 0.0205 (0.13) 0.0513 (0.08) Community Level Variables Years CA Promoted 0.41 (0.27) 0.625 * (0.37) Cluster1 - 5.29 (4.06 ) 1.587 (1.98) Cluster2 - 2.144 (1.76) - 6.769 ** (2.92) Cluster4 - 1.546 (1.60) - 3.157 (1.98) Constant - 1.399 (2.52) 1.943 (2.17) Observations 77 77 a Tropical Livestock Units - cattle have a value of 0.7, goats and sheep 0 .1, pigs 0.2, poultry 0.01 and donkeys 0.5 (Jahnke , 1982). Farmers using MT on maize but not cotton tend to live in communities where the buyer or distributor uses MT on a smaller proportion of land, there is a longer history of CA promotion, and where tha t promotion included the use of incentives. Incentives were previously used by NGOs and government CA promotional programs but not by the cotton companies. In areas where buyers or distributors are not providing strong examples of MT use, it is likely that the primary source of information is from extension services (whether government or NGOs), which tend to emphasize MT for maize production. Farmers in such communities may not have enough information or encouragement to try using MT with cotton. In contra st, farmers using MT on cotton but not maize were significantly less likely to have ever received incentives for doing CA, which suggests their primary information source is th e cotton company. 138 REFERENCES 139 REFERENCES Andersson, J.A. and S . laims to Understanding Farmers and Contexts: A Literature Review of Conservation Agriculture (CA) Adoption among Smallholder Farmers in Southern Africa. Agriculture, Ecosystems and Environment 187: 116 - 32. doi:10.1016/j.agee. 2013.08.008 Andersson, J. and Giller, K. 2012. Conservation Agriculture and the politics of its promotion in African smallholder farming. In: Sumberg, J. and Thompson, J. (eds.) Contested Agronomy : Agricultural Research in a Changing World . London: Earthscan. Arslan, A., N. McCarthy, L. Lipper, S. Asfaw, A. Cattaneo. 2014. Adoption and Intensity of Adoption of Conservation Farming Practices in Zambia. Agriculture, Ecosystems and Environment 187: 72 - 86. doi:10.1016/j.agee.2013.08.017 Baudron, F., H.M. Mwanza, B. Triomphe, and M. Bwalya. 2007. Conservation Agriculture in Zambia: A Case Study of Southern Province. Published jointly by the African Conservation Tillage Network, Nairobi; the Centre de Coo pération Internationale de Recherche Agronomique pour le Développement, Paris; and the Food and Agriculture Organization of the United Nations, Rome. Baum, C. 2008. Stata Tip 63: Modeling proportions. Stata Journal 8 (2), 299 - 303. Boko, M., I. Niang, A. Nyong, C. Vogel, A. Githeko, M. Medany, B. Osman - Elasha, R. Tabo and P. Yanda, 2007. Africa. Climate Change 2007: Impacts, Adaptation and Vulnerability. Contribution of Working Group II to the Fourth Assessment Report of the Intergovernmental Panel on Clim ate Change , M.L. Parry, O.F. Canziani, J.P. Palutikof, P.J. van der Linden and C.E. Hanson, Eds., Cambridge University Press, Cambridge UK, 433 - 467. Chambers, R. 1997. Whose Reality Counts? Putting the First Last . London: Intermediate Technology Publicatio ns. Conservation Farming Unit (CFU). 2007. Conservation Farming and Conservation Agriculture Handbook for Hoe Farmers in Agro - Ecological Regions I and IIa. Lusaka: ZNFU, CFU. Conway, G., and Barbier , E . , 1990, After the Green Revolutio n: Sustainable Agricu lture for D evelopment , Earthscan , London . Food and Agriculture Organization (FAO). 2001. The Economics of Conservation Agriculture. Rome: United Nations Food and Agriculture Organization. Giller, K., E. Witter, M. Corbeels, and P. Tittonell. 2009. Conserva tion Agriculture and Smallholder Farming in Africa: The Heretics' View. Field Crops Research 114 : 23 - 24. 140 Grabowski, P., S. Haggblad e, S. Kabwe, and G. Tembo. 2014 . Minimum tillage adoption among commercial smallholder cotton farmers in Zambia 2002 to 2011 . Agricultural Systems 131, 34 - 44. doi: 10.1016/j.agsy.2014.08.001. Grabowski, P. and Kerr, J. 2014. Resource constraints and partial adoption of conservation agriculture by hand - hoe farmers in Mozambique. International Journal of Agricultural Sustainabilit y. 12 (1) 37 - 53 . Haggblade, S., S. Kabwe, and C. Plerhoples. 2011. Productivity Impact of Conservation Farming on Smallholder Cotton Farmers in Zambia. Working Paper No. 47. Lusaka, Zambia: Food Security Research Project. Haggblade, S., G. Tembo, D. Kabore , C. Reij, O.C. Ajayi, S. Franzel, P. Mafongoya, and F. Place. 2010. Sustainable Soil Fertility Management Systems. In Successes in African Agriculture , ed. S. Haggblade and P.B.R. Hazell. Baltimore and Washington D.C.: Johns Hopkins University Press and I FPRI. Haggblade, S. and G. Tembo. 2003. Development, Diffusion and Impact of Conservation Farming in Zambia. Working Paper No. 8. Lusaka, Zambia: Food Security Research Project. Hobbs, P. 2007. Conservation Agriculture: What Is It and Why Is It Important for Future Sustainable Food Production? Journal Of Agricultural Science 145.2 : 127 - 37. Jahnke, H. 1982. Livestock Production Systems and Livestock Development in Tropical Africa. Kiel, Germany: Wissenschaftsverlag Vauk. Retrieved From: http://Pdf.Usaid.Gov/Pdf_Docs/PNAAN484.Pdf Kabwe, S., C. Donovan, and D. Samazaka. 2007. Assessment of the Farm Level Financial Profitability of the Magoye Ripper in Maize and Cotton Production in Southern and Easter n Provinces . Working Paper No. 23. Lusaka, Zambia: Food Security Research Project. Kassam, A., T. Friedrich, F. Shaxson, and J. Pretty. 2009. The Spread of Conservation Agriculture: Justification, Sustainability and Uptake. International Journal of Agricul tural Sustainability 7.4: 292 - 320. Long, J. , 1997. Regression Models for Categorical and Limited Dependent Variables, 1st edition. Thousand Oaks , CA SAGE Publications. Long, J. and J. Freese. 2001. Regression Models for Categorical Dependent Variables Usi ng Stata . College Station, TX: Stata Press, Stata Corporation. Morris, M., V. Kelly, R. Kopicki, and D. Byerlee. 2007. Fertilizer Use in African Agriculture: Lessons Learned and Good Practice Guidelines. Washington D.C.: The World Bank. Ngoma, H., B.P. Mul enga, and T.S. Jayne. 2014. What Explains Minimal Usage of Minimum Tillage Practices in Zambia? Evidence from District - Representative Data. IAPRI 141 Working Paper No. 82. Lusaka, Zambia: Indaba Agricultural Policy Research Institute (IAPRI). Nyanga, P. 2012. Factors Influencing Adoption and Area under Conservation Agriculture: A Mixed Methods Approach. Sustainable Agriculture Research 1: 27 - 40. Papke, L. and Wooldridge, J. 1996. Econometric Methods for Fractional Response Variables with an Application to 401(k ) Plan Participation Rates. Journal of Applied Econometrics 11: 619 - 632. Pretty, J., C. Toulmin, and S. Williams. 2011. Sustainable Intensification: Increasing Productivity in African Food and Agricultural Systems. International Journal of Agricultural S ustainability 9.1: 5 - 24. Rockström, J., P. Kaumbutho, J. Mwalley, A.W.M. Nzabi, M.L. Temesgen, L. Mawenya, J. Barron, J. Mutua, and S. Damgaard - Larsen. 2009. Conservation Farming Strategies in East and Southern Africa: Yields and Rainwater Productivity fro m On - farm Action Research. Soil and Tillage Research 103: 23 - 32. Siegel, P. 2008. Beneficiaries of Agricultural Commercialization? World Bank, Africa Region Working Paper Series No. 113. Wa shington, D.C.: The World Bank. Tembo, S. and N. Sitko. 2013. Technical Compendium: Descriptive Agricultural Statistics and Analysis for Zambia. IAPRI Working Paper No. 76. Lusaka, Zambia: Indaba Agricultural Policy Research Institute (IAPRI). Thierfelder, C., T. Mombeyarara, N. M ango, and L. Rusinamhodzi. 2013a . Integration of Conservation Agriculture in Smallholder Farming Systems of Southern Africa: Identification of Key Entry Points. International Journal of Agricultural Sustainability 11: 317 - 30. Thier felder, C., Mwila, M., Rusinamhodzi, L., 2013 b. Conservation agriculture in Eastern and S outhern provinces of Zambia: Long - term effects on soil quality an d maize productivity. Soil and Tillage Research 126, 246 258. Todaro, M. and S. Smith. 2009. Economic Development (10th edition ). Boston: Pearson Addison Wesley. Umar, B.B., 2012. Reversing agro - based land degradation through conservation agriculture: Sustainable Agriculture Research 1, p72. Um ar, B.B., Aune, J.B., Johnsen, F.H., Lungu, O.I., 2011. Options for improving smallholder conservation agriculture in Zambia. Journal of Agricultural Science 3, p50. Wall, P. 2007. Tailoring Conservation Agriculture to the Needs of Small Farmers in Develop ing Countries: An Analysis of Issues. Journal of Crop Improvement 19: 137 - 55. 142 World Bank. 2007. Agriculture for Development . World Development Report No. 2008. Washington, D.C.: The World Bank. Zeller, M., Diagne, A., Mataya, C., 1998. Market access by sma llholder farmers in Malawi: Implications for technology adoption, agricultural productivity and crop income. Agricultural Economics 19, 219 229.