DETERMINANTS OF SUSTAINABILITY OF COMMUN I TY SEED BANKS IN NICARAGUA: A DURATION ANALYSIS APPROACH By David John DeYoung A THESIS Submitted to Michigan State University in partial fulfillment of the requirements for the degree of Agricultural, Food, and Resource Economics Master of Science 2015 ABSTRACT DETERMINANTS OF SUSTAINABILITY OF COMMUN I TY SEED BANKS IN NICARAGUA: A DURATION ANALYSIS APPROACH By David John DeYou ng A ccess by small holder producers to seeds of improved bean varieties remains a constraint in many parts of the world . In respon se to this development challenge, t his study examines the salient features of the Community Seed Bank (CSB) models implemente d in Nicaragua from 2010 to 2014, through support from the Bean Technology Dissemination (BTD) project and analyzes the determinant factors contributing to the ir sustainability. CSB level data w ere collected from 154 CSBs through a survey and from project reports. Using the duration analysis technique, several determinants of sustainability mentioned in the available body of literature are confirmed. Namely, the CSBs that produce quality seed, recover seed production costs, have experienced leadership, o perate formally as a group by documenting decisions, and have access to productive assets operate longer than CSBs that lack these characteristics. This study also confirms the importance of not only building seed production capacities of the CSBs but als o seed marketing and administrative capacities. Seed marketing training was found to reduce the failure rate of CSBs that reported low seed yield in their first year of operation. Intensity of CSB operation was positively associated with i ncreased risk of failure . Of the t hree types of CSBs implemented , individual seed banks failed 30% faster than CSBs administered collectively by community members. While the size of CSB membership did not impact time to failure, the CSBs that were more representative of the community did have lower failure rates. The study identifies important characteristics that should be considered in future promotions of community based or decentralized models of seed production of staple food crops such as beans. iii Dedicated to Jackie , Lydia, David, Ron and Fran iv ACKNOWLEDGEMENTS To my advisor, Mywish Maredia, thank you for your support, suggestions and insight throughout my studies at Michigan State University. To my committee members, Eric Crawford, Richard Bernsten and Karen Cichy, thank you for your comments to improve this document. To Ahsanuzzaman, thank you for your willingness to review my work and give several explanations and suggestions regarding durati on analysis. To the USAID - funded Dry Grain Pulses CRSP, now the Legume Innovation Lab, that administered the BTD project; specifically Irvin Widders and Luis Flores for their support and facilitation of communication with staff in Nicaragua. This research was made possible through support provided by the USAID - funded Feed the Future Innovation Lab for Collaborative Research on Grain Legumes (previously referred as the Dry Grain Pulses Collaborative Research Support Program) under the terms of Cooperative A greement No. EDH - A - 00 - 07 - 00005 - 00 and the Associate Award Cooperative Agreement No. AID - OAA - LA - 10 - 00007. To the staff of INTA that coordinated the BTD project in Nicaragua, thank you for submitting the requested data and answer ing my questions about the Co mmunity Seed Banks. I would like to thank Francisco Pav ó n and Aurelio Llano for coordinating and accompanying me around the country on my trips to Nicaragua, Delia Cortez for data entry, and Danilo Montalvan for your coordination of project resou rces. Ad ditionally, thank you t o the regional program directors, extension workers, promoters and CSB members for completing and submitting the survey. v To my wife Alcira Jackeline , thank you for love and support. Thank you for encouraging words written on napkins in my lunchbox, your unconditional love, and sharing your life with me. You are my ayuda id ó nea . To my daughter Lydia , thank you for your joy in the mornings and excitem ent to see me when I return home at night. You are a blessing to my life and I cherish you greatly. To my son David ant icipated in January, I love you and anticipate your birth and watching you grow. To my parents, thank you for your encouragement and sup port (emotional, financial and spiritual) throughout my life. vi TABLE OF CONTENTS LIST OF TABLES ................................ ................................ ................................ ....................... viii LIST OF FIGURES ................................ ................................ ................................ ....................... ix KEY TO ABBREVIATIONS ................................ ................................ ................................ ......... x CHAPTER 1: INTRODUCTION ................................ ................................ ................................ ... 1 1.1 Bean Production in Nicaragua ................................ ................................ ...................... 1 1.2 Efforts to Increase Yield in Nicaragua ................................ ................................ .......... 3 1.3 Objectives of the study ................................ ................................ ................................ .. 5 1. 4 Research Questions ................................ ................................ ................................ ....... 5 1.5 Organization of the study ................................ ................................ .............................. 6 CHAPTER 2: CONCEPTS, CONSTRAINTS AND CHALLENGES OF A SUSTAINBLE SEED SYSTEM: A REVIEW OF THE LITERATURE ................................ ....................... 8 2.1 Sustainable Seed System ................................ ................................ ............................... 8 2.2 Formal and Informal Seed Systems ................................ ................................ ............ 10 2.3 Community Based Seed Production Schemes as a Type of Informal Seed System ... 12 2.4 Sustainability of Community Based Seed System: Contributing factors ................... 14 2.4.a. Profitability ................................ ................................ ................................ ...... 15 2.4.b. Market demand ................................ ................................ ................................ 16 2.4.c. Seed quality ................................ ................................ ................................ ...... 19 2.4.d. Links to formal system ................................ ................................ .................... 20 2.4.e. Training ................................ ................................ ................................ ............ 20 2.5 Scope of Sustainability ................................ ................................ ............................... 21 CHAPTER 3: STUDY SETTINGS: THE BEAN TECHNOLOGY DISSEMINATION PROJECT AND THE COMMUNITY SEED BANK MODEL IN NIC ARAGUA ........... 23 3.1 The BTD Project in Nicaragua ................................ ................................ ................... 23 3.2 Criteria for Establishing a CSB ................................ ................................ ................... 26 3.3 Variation of CSBs in BTD Project ................................ ................................ .............. 27 3.3.a. Regional Differences in CSBs ................................ ................................ ......... 28 3.3.b. Organizational Differences of CSBs ................................ .............................. 30 CHAPTER 4: METHODOLOGY ................................ ................................ ................................ 33 4.1 Theoretical Framework ................................ ................................ ............................... 33 4.2 Empirical Estimation Strategy ................................ ................................ ................... 38 CHAPTER 5: DATA AND DESCRIPTIVE STATISTICS ................................ ......................... 44 5.1 Data Sources ................................ ................................ ................................ .............. 44 5.2 Descriptive statistics of variables included in the model ................................ ........... 45 5.2.a Characteristics of Individual Banks ................................ ................................ .. 48 5.2.b Organizational Structure of CSBs ................................ ................................ .... 48 5.2.c Community Characteristics ................................ ................................ .............. 49 vii 5.2.d Leadership Characteristics ................................ ................................ ................ 49 5.2.e Land Use in CSB Seed Production ................................ ................................ ... 49 5.2.f Labor use for seed production ................................ ................................ ........... 50 5.2.g Assets and access to facilities used for seed production ................................ .. 51 5.2.h Human capital in seed production ................................ ................................ .... 52 5.2.i Output and efficiency indicators in Seed Product ion ................................ ........ 53 CHAPTER 6: DURATION ANALYSIS RESULTS ................................ ................................ ... 56 6.1 Non - parametric duration analysis ................................ ................................ .............. 56 6.2 Parametric and semi - parametric duration analysis ................................ .................... 59 6.2.a Full and preferred duration analysis models ................................ ..................... 60 6.2.b Interaction Terms ................................ ................................ .............................. 66 6.2.c Test of Proportionality Assumption ................................ ................................ .. 67 6.3 Duration analysis of a dataset without Individual CSBs ................................ ........... 67 6.4 Frailty models to remove heterogeneity ................................ ................................ ..... 68 6.4.a Individual Frailty ................................ ................................ ............................. 68 6.4.b Shared Frailty ................................ ................................ ................................ ... 69 6.4.c Additional Potential Source of Bias ................................ ................................ . 71 CHAPTER 7: CONCLUSIONS, POLICY IMPLICATIONS AND OPPORTUNITIES FOR FUTURE RESEARCH ................................ ................................ ................................ ........ 73 APPENDICES ................................ ................................ ................................ .............................. 84 APPENDIX A: English Version of the Survey ................................ ............................... 85 APPENDIX B: Descriptive Statistics by INTA Administrative Region ....................... 100 APPENDIX C: Interaction Ter ms in Three Distributional Forms and Interpretation ... 102 APPENDIX D: Differentiation of Types of CSBs: With and Without Individua l CSBs ................................ ................................ ................................ ................................ . 108 APPENDIX E: Three Options for Removing Heterogeneity Effects ............................ 112 APPENDIX F. Results of Duration Analysis Accounting for Fourth Year .................. 115 WORKS CITED ................................ ................................ ................................ ......................... 117 viii LIST OF TABLES Table 3.1: Comparison of Community Seed Bank Types implemented in the BTD Project ...... 3 2 Table 5.1: Total number of CSBs targeted for the survey versus those that completed the su rvey and included in this study ................................ ................................ ................................ .... 4 5 Table 5.2: Distribution of s urveyed CSBs by region and type ................................ .................... 4 5 Table 5.3: Summary Statistics of variables used in the duration analysis: Differences across types of CSBs in community level characteristics, membership and operating procedures 4 7 Table 5.4 Summary statistics of variables used in the duration analysis: Differences across types of CSBs in seed production inputs ................................ ................................ ........................ 5 2 Table 5.5 Summary statistics of variables used in the duration analysis: Differences across types of CSBs in seed production output indic ators ................................ ................................ ....... 5 4 Table 6.1: Summary Results of Duration Analysis of CSBs ................................ ........................ 6 3 Table 6.2: Summary of Decision Statistics to Determine Best Model ................................ ......... 6 5 Table 6.3: Summary Statistics for Testing Coefficients of Omitted Variable .............................. 6 6 Table 6.4: Log Normal (AFT) Duration Analysis results with Heterogeneity Removed ............. 70 Table B.1: Summary statistics of variables used in the duration analysis: Difference across INTA Administrative Regions ................................ ................................ ............................ 100 Table C.1 Duration Analysis with Interaction Variable ................................ .............................. 102 Table D.1 Weibull Model D uration Analysis without Individual Seed Banks ............................ 108 Table D.2 Weibull Model Duration Analysis With Quality and Network Variables .................. 110 Table E.1 Log Normal Duration Analysis with Heterogeneity Removed ................................ ... 113 Table F .1 Comparison of Results Without Fourth Year and Right Censored Data ..................... 115 ix LIST OF FIGURES Figure 3.1: Diagram of Seed Production and Distribution in BTD Project Nicaragua ................ 25 Figure 3.2: Map of INTA Administrative Regions ................................ ................................ ........ 28 Figure 6.1: Kaplan - Meier Survival Curve of CSB Failure ................................ ........................... 5 6 Figure 6.2: Kaplan - Meir Survival Estimates by CSB Type ................................ ......................... 5 7 Figure 6.3: Log Log Kaplan - Meier Curves by CSB Type ................................ ............................ 59 Figure 7.1: Hazard Functions of CSBs with Training at Four Seed Yield Levels ........................ 7 4 Figure 7.2: Hazard Functions of CSBs without Training at Four Seed Yield Levels .................. 7 5 Figure 7.3: Hazard Functions of CSBs with and without Training at Four Seed Yield Levels .... 7 6 Figure 7.4: Hazard Functions at two Repayment Rates by Type of CSB ................................ .... 7 8 Figure C .1: CSB Hazard Curves with Training ................................ ................................ .......... 104 Figure C .2: CSB Hazard Curves without Training ................................ ................................ ..... 10 5 Figure C .3: CSB Hazard Curves with and w ithout Training ................................ ...................... 10 6 x KEY TO ABBREVIATIONS AIC Akaike I nformation Criterion AFT A ccelerated F ailure T ime Model BIC Bayesian Information Criterion BTD Bean Technology Dissemination Project CIAT International Center for Tropical Agriculture CSB Community Seed Bank FAO Food and Agriculture Organization of the United Nations INIDE Instituto Nacional de Informaci ón de Desarrollo INTA Instituto Nicaragüense de Tecnología Agropecuaria KM S Kaplan - Meier Survival Curves KM Kilometers LB Pounds MAGFOR Ministerio Agropecuario y Forestal MCC Millennium Challenge Corporation MZ Manzanas NARS National Agricultural Research Systems NGO Non - Governmental Organization PCA Princip al Component Analysis PESA Program of the Framework for Food Security and Nutrition PH P roportional H azards Model QDS Quality Declared Seed xi QQ Quintales (equal to 100 pounds) RAAS South Atlantic Autonomous Region UNI S EM Unidad de Semilla USAID United States Agency for International Development 1 CHAPTER 1 : INTRODUCTION Common bean (Phaseolus vulgaris) is widely produced and consumed in Nicaragua and is strategically important for food and nutritional security of both the rural and urban poor. Long - term investment in research by the national program in collaboration with international researchers has re sulted in the development and release of many disease resistant bean varieties with a potential to increase bean grain yield in the country. However, access by small holder bean producers to these improved bean varieties developed through the research sys tem remains a major constraint due to the lack of a private sector led seed multiplication and dissemination system. Consequently, bean yields remain low, contributing to food insecurity and limiting the potential of beans to be a profitable cash crop. 1.1 Bean Production in Nicaragua Nicaragua is the largest bean producing countr y in Central America. The increase in bean production seen by this country over the past two decades is due to an increase in area cultivated rather than an increase in yield, w hich was estimated to be 643 kg per hectare for the country in 2011 (Quiroz Cortez el al. 2009, MAGFOR 2009, Schmidt et al. 2012). Large shares of bean producers in Nicaragua are smallholder farmers . 1 In 2011 , 64 % of producers cultivate d beans on 1 Compared to many developing countries in Asia and Africa, the definition of smallholder farmer in Nicaragua based on the size of land holding may seem out of range. However, there are limitations to focusing only on land size as a definition of smallholde r. Even a land size plus family labor index fails to capture investment capabilities, market integration and regional socioeconomic environmental factors (Berdegué et al. 2011 and Hallensleben 2012). Furthermore, the decision to exclude subsistence farme rs (focused on non - farm income) and commercial family farms (less than 3 permanent non - family workers) from the smallholder category will impact policy decisions (Berdegué et al. 2011). Definitions with an upper threshold of 2 hectares are common outside of Central America , but Berdegué and Fuentealba (2011) and Carmagnini (2008) point out that given the conditions of rural Nicaragua, a family cannot maintain its sustenance on less than 5.6 hectares (8 manzanas). MAGFOR considers farmers to be small scale if they cultivate less than 50 manzanas for all crops (Hallensleben 2012). In Guatemala, families with less than 10 manzanas (7 hectares) are considered subsistence ((Berdegué and Fuentealba 2011). In Honduras, the 2007 - t category was farmers were those with less than 5 hectares (7 manzanas) (INE 2008). 2 less t han 20 m an z anas ( equivalent to 34 acres or 14 hectares) 2 , 50 % on less than 10 manzanas and 34 % of farmers cultivate d beans on less than 5 m anzanas ( INIDE 2012 ). One way to increase bean yields by smallholder farmers is to u se certified seeds of improved v arieties. (Remington 2002). However, according to the Instituto Nicaragüense de Tecnología Agropecuaria (INTA), due to the lack of availability and access to certified seeds, the use of this type of quality seed has remained low even among farmers who rec eive training on implementing practices to increase yield ( MAGFOR 2009 , Sain 2011 , Carter el al . 2012). The Nicaraguan Ministry of Agriculture estimates that in 2008 - 09 agricultural season only 6.2% of bean production area was planted with certified seeds, and in the past seven years, even though the use of certified seeds has increased, it has never surpassed 15% of area planted to beans in Nicaragua ( MA GF OR 2009, UNI S EM personal communication). Due to low profitability, private seed companies have had lit tle interest in marketing certified bean seed directly to farmers. Instead, this private sector led formal system has focused on selling the certified seed to government agencies and NGOs to feed into their free or highly subsidized seed distribution progr ams. This model of seed production by for - profit private sector and its purchase and distribution to the farmers by government and NGOs at less than the economic price is not sustainable over the long run. Moreover, this approach only reaches a limited su bset of bean producers in the country ( MAGFOR 2009). In general, the lack of a sustainable seed multiplication and dissemination system has developing new va rieties (Tripp and Rohrback 2001). As a result of th e seed system constraint, most farmers end up using grain (either saved on - farm or purchased from the market) as planting 2 One manzana is equal to 0.7 hectares and equal to 1.7 acres 3 ized at the harvest (in combination with other constraining factors, including, lack of fertilizer and pesticides, poor growing conditions, etc.). 1.2 Efforts to Increase Yield in Nicaragua limitations in increasing bean productivity, the government of Nicaragua has started several initiatives to improve access to quality seed by farmers in rural areas. One of these initiatives is to promote the community seed apt a seeds , or Quality Declared Seeds (QDS ) of basic grains, including beans. The community seed bank s (called Banco s Comunitarios de Semilla, in Spanish) is a formalized, but not a legally registered, organization that operates on the principles of self - hel p, whereby community members come together to produce seeds to meet their own current needs, save seeds for future seed security, and sell excess seeds to generate revenues to cover production costs . The CSB oversees community - level production, marketing, distribution, and storage of quality seeds (i.e., apta seeds). These apta seeds (i.e., QDS) are produced from registered by th e farmers under the aegis of a community organization with technical guidance from INTA and distributed to other farmers within or outside the community. This model of Community Seed Banks provides an opportunity to reach large numbers of smallholder be an farmers with quality seed of improved varieties. 3 was initiated in 2010 by the Dry Grain Pulses CRSP 3 Strategic Investment in Rapid Technology Dissemination: Commerciali zation of Disease Resistant Bean Varieties in Guatemala, Nicaragua, Honduras and Haiti U.S. Agency for International Development under the Feed the Future initiative through an Associate Award grant 4 through funding from USAID with the aim of introducing technologies (i.e., improved variet ies) that increase bean productivity to a large number of rural households , the CSB model was the most logical choice of the BTD project for seed dissemination in Nicaragua. The national bean research program of Nicaragua (INTA) was the in - country partner of the BTD project, and through its network of regional offices, it played an important role in supplying the registered seed stocks of improved bean varieties to community seed banks and provided technical assistance to ensure that the seeds produced by t he seed bank meet some minimum quality standards as planting materials. Within the broad class of community seed banks, there were three types of CSB models implemented across the country two communal CSBs were managed by community members but seed product ion took place either on community managed land (the classic CSB model) or individually managed land (parceled CSB), and one completely managed by individuals (individual CSBs). From 2011 to 2013, more than 234 communities 4 received support to establish a CSB in Nicaragua from the BTD project , and the project was able to reach 16,065 beneficiaries ( 23 % of farmers cultivating bean on 10 MZ or less) through this community based approach of seed production and distribution. In 2011, an estimated 5,365 farmers rec eived seed produced by the CSBs representing 8 % of small holder bean farmers cultivating bean on 10 MZ or less in Nicaragua . In 2014 , when the BTD project ended, several of these CSBs that were established by INTA had ceased to exist (i.e., fa iled) and some had survived and continued to receive funding to Michigan Stat e University f rom October 2010 to March 2014. One of the objectives of this project was to implement sustainable bean seed systems with local farmer involvement/ownership so as to ensure long - term availability of quality seed of improved bean varieties to resource - poor farmers in the four project countries. 4 In these 234 communities 501 CSBs were established according to BTD project reports. To be counted as a different CSB a unique promotor had to appear for a CSB that had not previously been included i n previous years reports. Some communities had multiple CSBs because of the parceled CSB structure (in Pacifico Norte, one community had 16 CSBs) while other communities are listed as having multiple CSBs after year 1 because a second or third CSB was est ablished with a different promoter. 5 support from INTA beyond year 3 of the BTD project. This mixed result on the survival status of the CSBs at the end of the BTD project offers an interesting opportunity to study the factors influencing t he sustainability of the community based seed production model. Studies that have looked at similar experiences in Africa indicate that community based seed production is unsustainable due to low demand for the seed and an inability of the community base d organization to cover production costs (Tripp and Rohrbach 2001). Yet, there are few alternate models that address the constraint of lack of availability of seeds of improved varieties for staple food crops, such as legumes, to smallholder farmers in dev eloping countries. Understanding the factors that influence the success or failure of different models of seed multiplication and dissemination (including community based seed systems) that can meet the seed needs of a large number of smallholder farmers t hus remains an important research question. 1.3 Objectives of the study The objective of this study is to use duration analysis methodology to understand the factors that determine the sustainability (or failure) of the community based model of CSBs used i n Nicaragua under the BTD project. Specifically, the study: (1) Gives an overview of different CSB models used in the BTD project in Nicaragua; (2) Characterizes the differences and similarities between the CSB models; (3) Analyzes the determinant factors of sustaina bility (or failure) of the CSBs as a collective group and by CSB model; and (4) Provides recommendations for improving future implementation of CSBs in Nicaragua and elsewhere. 1.4 Research Questions This study achieves these objectives by addressing the follo wing research questions: 6 (1) What are the characteristics of the Nicaraguan CSBs that participated in the BTD project in terms of their: a. membership and leadership profile; b. community characteristics, asset ownership and prior experience in collective action; c. se ed production quality, quantity, distribution and repayment rates; and d. duration of participation in the BTD project? (2) What are the factors that determine the success or failure of the CSBs in the BTD project? a. Do the factors vary across different Duration A nalysis models? b. What variations in determinants of sustainability are attributed to the different CSB models observed in the BTD project? c. Which of the three CSB models in the BTD project is more sustainable, if they are different? (3) What improvements can be made for future implementation of CSBs to make them more effective and sustainable? 1.5 Organization of the study The study is divided into seven chapters. Chapter 2 describes the concepts and principles of a sustainable seed system and seed security. Pr evious examples of CSBs and lessons learned from these experiences are presented along with theoretical determinants of CSB sustainability. Chapter 3 describes the BTD project and the bean seed sector in Nicaragua. The three variations of CSB models imple mented under the BTD project are described and their 7 differences and similarities are explained. This chapter also brings out the regional differences in the history of seed production activities and their effect on the design and operationalization of th e CSB models across the country under the BTD project. Chapter 4 provides the analytical framework and methodological models for studying the determinants of sustainability. A model of adoption and withdrawal from participation in non - traditional markets is considered and its application for the current study is explained. Finally, the methodology of duration analysis is explored. Chapter 5 presents the descriptive statistics of the data after describing the data collection process. The variables of inte rest are described and differences in descriptive statistics between the three CSB models are explored. Chapter 6 presents the results. Several possible duration analysis models are considered initially. After finding the best model and exploring interac tions terms, a separate model is considered excluding individual CSBs due to their structural difference from the other two communal models. CSB and regional heterogeneity, possibly from unobserved variables, are tested and removed as needed from the fina l model. Chapter 7 concludes the thesis with the policy implications. Limitation of the study and recommendations for future research are also provided. 8 CHAPTER 2 : CONCEPTS, CONSTRAIN T S AND CHALLENGES OF A SUSTAINBLE SEED SYSTEM: A REVIEW OF THE LITERA TURE The model of CSBs, which is the focus of this study, assumes that their presence in the communities will meet needs of the farmers in terms of quantity, quality and diversity of seeds, while generating the resources to be economically sustainable and continue to meet those needs in the future. Before we analyze the factors contributing to the sustainability of this CSB model, it is important to clarify several concepts and terminologies associated with sustainability, seed system and community seed ba nks. 2.1 Sustainable Seed System S associated with the development, multiplication, processing, storage, distribution and marketing dia et al. 1999). It encompasses the entire spectrum of the seed value chain that can exist at the community (village), municipality (district), department (province), country or regional levels. A sustainable seed system is characterized by a set of players, infrastructure, policies and guiding principles that provide framers with quality seed in the right quantity at the right time, place, and price (van den Burg 2004). In the context of CSBs, it is the ability to meet a nd doing so while recovering all production costs. Additionally, sustainability implies that a seed production and distribution system is profitable beyond the end of a project (Sperling et al . 2013). Often, these principles of sustainability are reduced to the three components of seed security: availability, access and utilization (quality) (Sperling et al. 2011). As Remington et al . 9 (2002) indicate, these same parameters have been used to assess food security although recent publications have added stab ility as a fourth component. Availability of seed is determined by two conditions, sufficient quantity and service. Service is defined as both providing the seed prior to the planting period (timing) and within a realistic distance to the farmer s home (l ocation) (Longley et al . 2002). income and assets can be used to obtain seed through sale, barter, or loan, then the farmer can access available seed. A loan or lending mechanism that allows farmers to obtain the seed and repay from their harvest is one way to increase access , which is more amenable to a community - based seed production and distribution model than a private sector led seed system (Longley et al . 2002) . F inally, utilization is divided into two conditions, quality of seed and diversity of varieties. Quality can be measured in terms of genetic value, purity and seed viability (Se n timela et al. 2004). A germination test is a n indicator of seed quality. The diversity of varieties refers to providing the varieties appropriate for the region that the farmers prefer either for yield potential, resistance to stresses, grain characteristics (size, weight, taste, and cooking time) or a combination of these factors (Longley et al . 2002). In order to meet the needs of a community and continue in the following year, a CSB must be able to cover its expenses. Length of operation is a good measure of economic sustainability as it indicates that revenues generated are su fficient to cover the costs needed for its continued operation (Wiggins and Cromwell 1995) (Witcombe et al . 2010). According to Sperling et al. (2013) continued operation without any project or external support is the true measure of economic sustainabili ty. 10 2.2 Formal and Informal Seed Systems Within a seed system a distinction is made between formal and informal system. A Sperling et al. 2013). Variety bre e ding and seed development is most often coordinated by National Agricultural Research Systems (NARs) in collaboration with universities, international research centers and possibly NGOs. In the case of highly profitable crops or seed types (i.e., hybrid seeds), the private sector develops new varieties that are registered with government authorities prior to release for sale. For such crops, even if the varieties are developed by the public sector, often private enterpr ises purchase the registered seed (which are also called basic or foundation seed in some countries) to then be used to grow seed for commercial sale. Government programs with a goal of disseminating new varieties or t o provide seed aid after natural disa sters are also part of et al. 2013). An informal seed system, also called a traditional or farmer system, lacks gov ernment - (Almekinders 2000). Seed is obtained by farmers from his or her own harvest or exchanged through barter, sale or gift with other farmers (Bentley et al. 201 1). Although no certification process takes place, quality is ensured by trusting their production methods or those of known seed producers in their area. In the worst case scenario, farmers purchase grain for sale at a local market and must judge the ph ysical attributes to determine the quality of using the grain as seed (Maredia et al. 1999). 11 Informal seed system remains the main source of seeds for farmers around the world contributing between 90 - 100% of seed depending on the crop (Rubyogo 2007). Othe r studies have found similar evidence across countries and crops and place the percent of seed coming from informal sources at 60 to 85% for most staple crops and nearly 99% for neglected and underutilized crops (Shrestha et al. 2013). Under certain condi tions, seeds sourced from the informal system can effectively meet the seed needs of the farmers. There is no need to have a formal seed system to meet 100% of seed need for all the farmers in all the seasons. For example, for most self - pollinated crops th e use of saved seed from previous harvest is a common practice. This is because the genetic quality of the seed of self - pollinated crops such as beans, rice, and wheat does not deteriorate from one generation to the next. For such crops if there are no wid espread seed borne disease s , the informal seed system co - exists with the formal seed system even in more matured agricultural systems. For example, Almekinders (2000) reports that in Europe informal seed systems make up the majority of seed supplied for al l crops. Greece, Germany and the Netherlands had area sown with seeds supplied from the informal sector at 90, 50 and 25% respectively ( Almekinders 2000) . In the traditional maize and bean farming systems of Central America, use of seed from the formal sy stem is low. Wierema et al. (1993) found that only 6% of the interviewed farmers reported obtaining maize seed from the formal system in Nicaragua while in Costa Rica and Honduras the use of the formal system was at 2 and 13% respectfully. For beans, 13% of Nicaraguan farmers used the formal system compared to 21% in Costa Rica and 6% in Honduras. A representative survey of the seed recipients of the BTD project in Nicaragua in 2012 revealed that only 32% had easy access to certified seeds of bean and 54 % of farmers had never used them (Maredia et al. 2014). The same study interviewed seed recipients of the BTD 12 project in Honduras and Guatemala in 2013 where respectively, 44 and 19% of farmers reported easy access to certified bean seed while 21 and 57% respectively stated that they had never used certified bean seed (Maredia et al. 2014). The main source of seed within the informal system is saved seed from the previous sed their own seed. Similarly, 79% of farmers in Honduras and 72% of farmers in Nicaragua used their own seed. For maize, 79% of farmer s in Costa Rica, 75% of farmers in Honduras and 81% of farmers in Nicaragua reported using their own seed. When a farm er does not have enough saved seed the complementary portion is obtained through other informal system sources such as family, friends, the grain market or local seed producer (Tripp 1997) (Bentl e y et al. 2011). Data from Maredia et al. (2014), reveal tha t 54% of bean plots planted in May 2012, by the Nicaraguan farmers who received seed from the BTD project in 2011, used saved seed from the purchased at a local market ( A ). 2.3 Community Based Seed Production Schemes as a Type of Informal Seed System To address the constraint of lack of availability of quality seed, NGOs, community leaders and government agencies have looked for ways to ensure seed security through community based seed production schemes (Shrestha et al. 2004). These schemes are known by different names in different countries, such as Village Seed Bank s , Farmer Seed Enterprises, Seed Savers Networ k s, Smallholder Seed Enterprises and Community Seed Banks (the term used in the current study) , but the basic element common across these schemes is the organization of community members within a geographical boundary that are focused on producing seed of desired crops and specific varieties. The organization of seed production can vary from a 13 commu nal approach to having one or few community members specializing in seed production. Post - harvest activities of seed treatment and marketing, however, are community orientated in each model. In their typology of CSBs, Lewis and Mulvany (1997) noted five ty pes of CSBs. De facto seed banks occur regularly within communities to spread the risk of an individual farmer losing seed among all farmers in the community. The de facto bank also allows farmers to obtain additional quantities of seed if they wish to e xpand area planted to a certain crop. Community seed exchange is a second type of seed bank with a formalized sale, lending or trade system. Additionally, regional seed fairs, most notably those in the Andean region, allow exchange of seed and ideas amon g different regions. A third type of CSB is the organized CSB that multiplies both traditional and improved varieties. While the link to the formal seed system is stronger in this type of CSB , the dependence on outside funding and lower equity of access to the poor farmers networks focus on conservation of seed varieties existing in the community. Proponents of biodiversity fall under this category of CSB and may create an organize d CSB (the third type of CSB) in order to multiply seeds of local landraces. Finally, ceremonial seed banks are controlled by community leaders and focus on traditional varieties and institutions. There is little or no overlap between the informal and fo rmal seed systems in this final CSB type. Lewis and Mulvany (1997) divide the third CSB category (organized banks) into four subcategories distinguished by the type of seed multiplied and reason for multiplication. Improved varieties are multiplied either for seed relief purpose or with the goal of dissemination of new varieties. Traditional varieties are multiplied either for conservation and biodiversity motives or to increase access to these varieties. Within the context of the current study, the 14 CSBs are multiplying improved varieties as part of a dissemination effort. Lewis and Mulvany (199 7 ) conclude that such banks increase seed security but may not be equitable to all as the main benefits are realized by participating farmers. The authors did no t reach a conclusion about the economic sustainability of this form of CSB. Sentimela et al. (2004) do address the contribution to seed security and economic sustainability of organized CSBs multiplying three types of seed: certified seed, quality declared (or Apta ) seed, and farmer varieties. Also included in the comparison by Sentimela et al . (2004) are contracted community level certified seed producers that offer a mutually beneficial agreement for both the seed producer and organization purchasing the seed, most likely the government or NGO. Sentimela et al . (2004) doubt CSBs multiplying farmer varieties will be standards that incur additional costs an d the selling price of such seed is often the same as grain prices. Finally, Sentimela et al. (2004) consider the multiplication of certified and Apta seeds (i.e., QDS) to be project based and unsustainable in the absence of an actor assuming the financia l and technical support role of the organization that launched the CSB. Without the project, sourcing registered seed must be addressed and producing Apta seed (i.e., QDS) faces low selling prices compared to certified seed. 2.4 Sustainability of Communit y Based Seed System: Contributing factors Community based seed production occurred through the 80s and 90s, but began receiving criticism for being unsustainable at the turn of the century. While Tripp and Rohrbach (2001) made the early claim that there w ere no examples of a successful CSB and that it is untenable, David (2004), Se n timela et al . (2004), Van Mele et al . (2011), and Witcombe et al. (2010) continue to present examples of community based seed production that they consider sustainable. 15 Currentl y, the literature is focused on combining the strengths of the formal and informal seed systems to form an integrated seed system (Sperling et al . 2013 and Louwaars et al . 2012). The focus is on cost effective, decentralized seed production of quality see d by producers who are strategically partnering with actors from the formal seed systems while using innovative marketing strategies (Sperling et al . 2013). Although not explicitly stated, the scale of production of these decentralized seed producers is s ignificantly larger than most CSBs. Despite differing opinions o n CSBs and the new focus on integrated systems, the re is broad agreement on many important factors that need to be in place for a community based seed production system to be economically sust ainable. For simplicity these factors are grouped into the following categories -- profitability, marketing, quality production, links to formal sector, and training. 2.4.a. Profitability As stated earlier, for a CSB to be sustainable, it must charge a pric e for its seed that covers production costs. Van den Burg (2004) suggests a simple formula for setting the price of seed compared to the price of grain. When seed is grown in the same field as grain, a 5 to 10% increase is warranted but if different inputs are used in a different plot, at least a 20% markup is required. Seed treatment and packaging should add 50 - 100% to t he price of grain. Charging a price that is more than double the price of grain is complicated when farmers do not distinguish between grain and seed (Rubyogo 2007). From the CSB oppon view point , the price obtained by most CSBs often do not cover th e full cost of production, seed inspection and technical support , and thus require to sustain their existence (Tripp and Rohrbach 2001). 16 H owever, as shown in Kenya by Katungi et al. (2011) , profitability is achievable in a com munity based model . A study of 30 bean seed producers across three districts found net revenue of 36% of total income. At the same time, a comparison seed company that incurred costs from irrigation, processing, storage and field inspection to obtain cer tification had net revenue of 61% due to the higher price charged for certified seed. For CSBs , price alone does not ensure profitability as repayment rates determine total revenue . Lemessa (1994) found repayment rates to be around 50% in Ethiopia, and Wi ggins and Cromwell (1995) reported similar rates among 84 CSBs in Mali but much lower ( between 20 and 40% ) among 16 CSBs in Sudan. Additional costs of enforcement to ensure repayment must be considered if the CSB is lending seed. CSBs may experience profi tability similar to early adopters in the technology adoption lifecycle model (Carletto et al. 1999). For example, CSBs formed in the early stages of its introduction in an area are more likely to experience favorable conditions such as external support, limited market competition and demand for new varieties that may not exist for CSBs that start at later stages or for the same CSB in subsequent years. Integration with formal seed system actors can be one mechanism to continue to innovate and identify gr owth opportunities and ways to remain profitable (Sperling et al. 2013). 2.4.b. Market demand Demand for seed, and the reasons for seed demand, must be identified prior to the start of seed production , preferably using a market potential survey (van den Bu rg 2004). Demand is often limited within a small community ( d ue to limited number of farmers and low frequency of seed replacement, which both translates into low quantity of seed demand), so CSBs must extend beyond their local clientele if they are to be successful (Tripp and Rohrbach 2001). As David 17 (2004) found in Uganda, many community members were one time buyers of seed from the community seed producers. By extending beyond the community, a seed producer will have a larger population of potential c Marketing and market development are important elements of CSB sustainability as it seed organization beyond its com munity. Tripp and Rohrbach (2001) attribute community based seed production failures to Documented experiences highlight several creative market development strategies used by organized seed production schemes. Participation in fairs, schools, door to door advertising, farmers meetings, and agricultural shows are some obvious examples but one group of seed producers in Uganda even authored a song to promote a new variety (David 2004). Partnerships are also formed within established organizations with broad local, district or even national coverage such as churches, NGOs and other social groups (Monyo et al . 2004). Van den Burg (2 004) suggests negotiating with owners of fields located in high traffic areas, such as bus stops, to use for seed production . The CSB can then use the fields as demonstration sites to show the benefits of quality seed of improved varieties in the form of increased resistance to disease and higher yield. Community based seed producers should not, however, be responsible for all aspects of marketing new varieties. Strong relationships with research stations, described below, should provide the communication and marketing opportunities for new varieties developed by the NAR S and approved by the seed authorities (Rubyogo 2004 , Sperling 2013). 18 Part of market research should identify the different seed package sizes and varieties demanded (David 2004 , David et a l . 1997). Providing seeds in varying package sizes to meet the varying needs of both small and large scale farmers as well as farmer groups making collective purchases, has been shown to be an effective tool for improved variety dissemination in Rwanda an d elsewhere (Rubyogo 2004 , Sperling et al . 2013). In general David (2004) found that when demand is high and regular , and the CSB has technical supervision, the producers can be successful. Also, where diseases are common, seed producers are able to show the benefits of clean seed and secure a market for their product. However, lack of access to resources or credit to cover the cost of inputs needed for seed production can be come a potential constraint in such an environment (David 2004 , Van den burg 2004). Several characteristics of beans affect the market demand for its seed and serve as a disadvantage for the success of a CSB. Ideally, the crop of choice for a CSB would have a high varietal deterioration rate (to maintain demand) , hig h multiplication rate (to produce large volumes of seed) and low carryover of seed based diseases (to minimize seed quality concerns of clients) (Rubyogo 2007 , David 2004). As proof that beans are not the first choice of the private sector (and thus also a concern for CSBs investigating market demand), c ase studies of seed enterprises across nine African nations noted that small seed enterprises focused on hybrid maize seed or vegetable seed because it could not make a profit competing with farmer - saved se ed of a crop such as beans . The only exceptions were enterprises with government support or linked to larger grain producers (Bentley et al . 2011). 19 2.4.c. Seed quality Seed q uality involves genetic value, purity and seed viability (Se n timela et al. 2004 ) . The assurance or guarantee of seed quality provided to farmer s varies by type of seed system. In the formal system, a government agency provides monitoring and a government issued label certifies quality , but in the informal system, the reputation of the seed producer guarantees quality (Sperling et al . 2013) . When CSBs are formed as part of a project, seed quality concerns can arise after the NGO or the technical experts end their monitoring services for the seed production activities (Se n timela et al . 2004). Q uality incentives must be established for the individuals responsible for seed production to maintain seed quality after the project support formally ends (Bänziger et al . 2004). Additionally, there are several issues that affect production, but not necessarily seed quality. Group production schemes run the risk of neglected care of the group production plot as individuals tend to their personal plots prior to working on the communal plot (David 2004). While the conflict of prioritizing the hou sehold plots over the CSB or business plots was true of all groups, David (2004) noted it was most present in female groups. Female participation in community organizations is also affected by traditional attitudes of gender roles held by 49% of men in Ni caragua (Ortega et al . 2005). Trust formed by previous work in groups should be considered prior to forming CSBs. In some cases, individual seed production may be favored to communal production but care must be taken in collective post - harvest decision ma king to ensure fair compensation among group members (David 2004). 20 Seed production is labor intensive and requires additional labor compared to bean grain production (Katungi et al. 2011). Farmers with resources should be targeted for new CSBs as they are more likely to avoid bottlenecks in seed production, practice crop rotation, use fertilizers and have land suitable for seed production (David 2004). A source of transportation (animal or truck), drying space, storage silos, packaging facilities and fina ncial management are additional assets needed to ensure seed quality (Van Mele et al . 2011). 2.4.d. Links to formal system Tripp and Rohrbach (2001) have criticized the CSB movement because it is often completely disconnected from the formal seed sector. Improvement has been made since the first NGO supported CSBs and the most successful seed producers in the informal sector have strong contact with the formal sector (Se n timela et al . 2004 , Van Mele et al 2011 , Witcombe et al . 2010 , Katungi et al. 2011). It is important for CSBs to identify where they will obtain registered ( or foundation ) seed of new improved varieties (Bänziger et al . 2004). For most developing countries and crops without a thriving private seed enterprise, the only option is the public seed authorities of the NARSs. For CSBs, this presents an opportunity to build ties with breeders and begin participatory plant breeding practices if they have not already begun in a given country (Rubyogo 2004 , Sperling et al . 2013). Thus the business model of a CSB should go beyond seed production and sale to include a strategic partnership with the closest research station ( Sentimela et al. 2004) 2.4.e. Training Training is needed for new seed producers to produce quality seeds and be able to sell tho se seeds . As pointed out by Tripp and Rohrbach (2001), CSBs need to develop markets, not 21 just multiply seed. Thus, CSBs need to be trained not only in agronomy and science of seed production, but also in business management, marketing, and accounting to g ain skills in managing the seed production as an enterprise (Bänziger et al . 2004). For farmers and farmer groups unaccustomed to such practices, these new requirements can be especially challenging. Having prior business experience can thus be an import ant asset for the success of a community based seed enterprise (David 2004). The success of three rice seed producer groups in Nepal was attributed to its focus on marketing and good management, and these groups were trained in this aspect by experts. Wit combe et al. (2010) found that nearly a decade after these groups were formed, two of the groups had expanded their market, and the third group had ended seed production activities but had continued their business operations by diversifying to specializati on in rice milling. 2.5 Scope of Sustainability In conclusion, there is an important issue related to sustainability of community based seed system discussed in the literature that needs to be highlighted. This is the issue of the scope of sustainability. Economic sustainability is a noble goal, but occasional investment s from donors and public sector support may be needed to ensure CSBs continuing to provide community services . Some definitions of economic sustainability, such as that used by Van Mele et al. (2011), allow for external support, while Sperling et al. (2013) require seed producers to be profitable without subsidies or external project funding. In cases of chronic, but not annual seed crisis, community based seed production has offered a n imp ortant contribution to a sustainable system than annual free seed distribution. Wiggins and Cromwell (1995) have documented NGO work with Village Seed Banks and point out that in the Sahel, the system worked for several years until rain failure depleted t he seeds in the seed banks. After ten years, 22 the banks again needed external support to replenish their seed stocks but they were not donor dependent every year. 23 CHAPTER 3 : STUDY SETTINGS: THE BEAN TECHNOLOGY DISSEMINATION PROJECT AND THE COMMU NI TY SEED BANK MODEL IN NICARAGUA The Bean Technology Dissemination project (BTD) , funded by USAID and implemented by the Feed the Future Innovation Lab for Collaborative Research on Grain Legumes (formerly called the Dry Grains Pulses Collaborative Research S upport Program) focused on improving bean productivity in four countries: Haiti, Guatemala, Honduras and Nicaragua . The BTD project promoted yield improving technologies such as improved bean varieties and the inoculant rhizobium through the National Agri cultural Research Systems in each country. Over the three years (2011 - 2013) of the project, 543 metric tons of improved varieties of seed were disseminate d to 102,047 farmers in the four countries (Maredia et al. 2014) . The partner organizations in each country chose different production and distribution strategies to disseminate seeds of improved bean varieties to the target numbers of farmer beneficiaries. 3.1 The BTD Project in Nicaragua In Nicaragua, the partner organization was the Instituto Nicaragüense de Tecnología Agropecuaria (INTA). INTA staff members chose to build on the Community Seed Bank (CSB) models previously implemented in Nicaragua for local seed production by the Food and Agriculture Organization (FAO) in the Pacific South reg ion and the Millennium Challenge Corporation (MCC) in the Pacific North region. Figure 3.1 provides a diagram of the seed production and distribution system used in the BTD project in Nicaragua. INTA, through its seed unit UNISEM, produced or contracted t he production of registered seed and provided it to the five INTA regional offices. Each regional office was responsible for establishing 40 CSBs using the criteria discussed below. INTA 24 technicians, or extension workers, supplied 80 pounds of registered seed, inputs and technical support to each CSB to establish a 1 manzana (0.7 hectares) seed production plot. After post - harvest seed treatment, each CSB was responsible for selecting 50 farmers to receive a seed loan. Farmers were to receive 20 pounds of seed and repaid the loan with 40 pounds of grain or its cash equivalent at harvest. As discussed below, the Centro Sur region decided to select individual seed producers instead of establishing communal CSBs in year 1. Under this scheme, the extension workers received delivery of the A pta seed (i.e., QDS) produced by the CSBs (individual seed producers), transported the seed to selected farmers in other communities and facilitated the repayment of the cash equivalent of 40 pounds of grain after harvest . 25 Figure 3.1: Diagram of Seed Production and Distribution in BTD Project Nicaragua Source: Updated from Reyes et al. (2014) 26 T he CSB model fit s production to and organiz ing bean farmers in bean producing communities. Extension workers are assigned to several communities to promote the field work of INTA. In each community, the extension worker identif ies a member of the community to serve as a promoter , who receives tra ining from INTA technical staff to promote improved technologies, including new variety seeds, through the establish ment of demonstrati on plot s . Under the BTD project, t he extension worker wa s responsible for providing the inputs, specific technical knowl edge on seed production and educational material s to the promoter and members of the community seed bank . Each CSB was given a target to reach a certain number of beneficiaries with its bean seed dissemination efforts. Thus, t he BTD project added a community organizational element in addition to the normal production focus of the INTA/ extensionist /promoter strategy. 3.2 Criteria for Establishing a CSB A manual describing the establishment of a CSB in Nicaragua has been developed through col laboration between the Nicaraguan Ministry of Forestry and Agriculture (MAGFOR) and the Program of the Framework for Food Security and Nutrition (PESA) of the FAO. The guide was published after the beginning of the BTD project but provides the structure a nticipated by INTA. It is part of a technical assistance series and intended for extension workers (FAO - PESA 2011). According to this guidelines document, the following conditions must be met for a community to be considered for INTA to establish a commun ity seed bank: a) favorable seed producing conditions, b) limited financial services including support from other organizations, c) limited access to quality seed, d) interest among farmers to organize a CSB, and e) farmers with limited resources. 27 To estab lish a CSB, INTA identifies community leaders as well as potential CSB members. Together with INTA staff, the economic conditions and seed supply needs are identified in the community. In the first meeting the community members are given an understanding of the advantages and disadvantages of working together or individually. According to the CSB guidelines/manual, the CSB establishment criteria are to have at least 12 members who meet the following conditions: a) reside in the community, b) have experien ce growing beans (or the crop for which the seed will be grown), c) have a reputation of being honest and responsible, d) willing to participate in group work, e) be receptive to the line of work, f) willing to take the initiative and g) willing to try new technologies. Each potential member should formally apply to join the CSB as a member at a meeting, and a forming charter should be written and signed by all the members. A board of directors is to be formed that includes a president (coordinator), a vi ce president, secretary, treasurer, leader of production, attorney ( fiscal ) and additional members ( vocales ). Although the CSB is not a legally registered organization, there are several policy and operational procedures recommended to be in place that giv es a CSB a formal organizational structure. For example, the CSB is to be operated under the established rules of the internal by - laws. A written loan policy needs to be developed after forming the charter. Production and training plans are also develope d to guide the actions of the CSB. Finally, an evaluation and follow up plan is needed to monitor progress and also to ensure repayment of seed. 3.3 Variation of CSBs in BTD Project Despite the well - defined CSB criteria given in the PESA document, there w ere two important differences among the CSBs established in the BTD project and included in this study. 28 Figure 3.2 provides a map of the five administrative regions of INTA comprised of two or more continuous departments and Table 3.1 compares the three d iffering type of CSBs. Figure 3.2: Map of INTA Administrative Regions Source: Aurelio Llano. Additional detailed information available at: http://www.inta.gob.ni/index.php/cobertura/mapa 3.3.a. Regional Differences in CSBs The mountainous regions of Centro Norte and Las Segovias are located near the border with Honduras. Thirty seven CSBs from the Centro Norte region are included in this study and all are classified as classic CSBs. The Las Segovias region is unique in that it contains CSBs of 29 all t hree types: fifteen classic, eleven parceled and three individual seed banks (definition of each type is provided below). The coverage of the Centro Sur region included parts of the South Atlantic Autonomous Region (RAAS) in the BTD project (see Figure 3. 2). Forty CSBs from the Centro Sur region are included in the study. Communities such as Santa Lucia in the department of Boaco have historically produced bean seed for sale to other parts of the region. Due to the seed production experience and conduci ve seed growing environment in the northwestern part of the region and given the adverse conditions for seed production due to humidity (at least during the primera agricultural season) in the southern and western part of the region, INTA staff in Centro S ur chose to work with individual seed producers in the BTD project. Instead of organizing CSBs according to the guidelines or criteria above, apta seed (i.e., QDS) was produced with individual seed growers and transported by INTA to farmers in other parts of the region. The decision met a need for quality improved varieties of seed in the southwestern part of the region. The Pacifico Norte region was the targeted area of the Millennium Challenge g incomes of rural farmers, bean farmers organized in cooperatives to collectively process and market the grain produced. The cooperatives participated in evaluation studies of different bean varieties to identify the best varieties for their region. Whe n the BTD project began, INTA regional staff formed the CSBs for seed production from among the members of the cooperatives. As a result, the CSBs choose to grow apta seed (i.e., QDS) in multiple plots instead of a larger communal plot. Although producti on was distributed among many individuals, the seed was collected, processed and distributed as an organized group. 30 Program of the Framework for Food Security and Nutrition (PESA) that included CSBs in its pilot project. The INTA staff in the region chose to only establish (or continue ongoing support of) 20 CSBs within the BTD project. The production plots were double the size (2 manzanas) compared to the majority of CSBs in other regions and INTA provide d double the inputs to each CSBs to ensure seed production and distribution to the regional goal of reaching 2,000 farmers each year. 3.3.b. Organizational Differences of CSBs The second important difference among CSBs is the type of CSB. For the current study, CSBs were self - identified as one of the three types defined below. Classic CSBs resemble closely to the structure described in the manual above and was defined for this study as follow s : The CSB is made up of several members of the community (part ners or members) and bean seed is grown in an area of approximately one manzana (possibly only one field) with one promoter. The members of the CSB make decisions about which seed variety to use (beginning in 2012), what input to use, who should receive s eed (loan), and how to secure repayment of grain (loan repayment). David (2004) pointed to trust and previous work experience in groups as reasons to use the classic model and in their absence, to choose an individual option. In this study, the individua l seed banks were defined as follows : The CSB is made up of one individual who grows the seed and makes decisions, sometimes with the help of INTA, as to who should receive the seed (loan). Additionally, parceled CSBs were formed and resemble the structure David (2004) mentioned allowing a farmer, or farmers here, to focus on production and incorporate a 31 collective process to post - harvest activities. The definition of parceled CSB used in this study is as follows: The CSB is made up of several members of t he community (partners or members) and bean seed is grown on several lots with several promoters. The members of the CSB make decisions about which seed variety to use (beginning in 2012), what input to use, who should receive seed (loan), and how to se cure repayment of grain (loan repayment). Table 3.1 provides a comparison of the salient features across the three types of CSB s . Although, there are differences in several aspects across the three types of CSB s , it is important to note the similarities th ey share (Table 3.1). All t he CSBs are producing apta seed (i.e., QDS) from fresh registered seed of improved varieties obtained from INTA 5 . Additionally, they are all operating within the BTD project management structure and received similar input package s and technical assistance from INTA technicians. Within these three broad categories of CSBs, there are differences in the characteristics of CSBs in terms of membership and leadership profiles, operating procedures, seed production and distribution outcomes, and number of years different CSBs survived in the BTD project. In other words, all the CSBs were not implemented as per the manual/guidelines established by the FAO and not all the CSBs received support from the BTD project (i.e., survived) for all three years. In this study we exploit these differences across and within the three types of CSB models to understand the association of heterogeneity in characteristics with the survival outcome. 5 growing it from registered seed. 32 Table 3.1: Comparison of Community Seed Bank Types implemented in the BTD Project Type of Community Seed Bank Classic Parceled Individual Organization Several members with one seed plot Several members with several seed plots One member with one seed plot Seed Production Seed used Registered Seed provided by INTA Registered Seed provided by INTA Registered Seed provided by INTA Size of Seed Plots 1 mz except for Pacifico Sur Region (2 mz) Size varies, but most common sizes are 0.5 mz and 1 mz Usually 1 mz Inputs (Fertilizer, pesticides, seed packing bags / supplies) Provided by INTA Provided by INTA Provided by INTA Seed Distribution Who makes the decision on how to use the seed produced Usually members of CSB Usually members of CSB (seed producers might have more voice) INTA Who receives the seed produced Community Members Community Members Unknown recipients outside the community Who is responsible to collect repayment Usually members of CSB Usually members of CSB (seed producers might need to take initiative) INTA Form of repayment Usually grain, but barter and cash equivalent accepted Usually grain, but barter and cash equivalent accepted Usually cash 33 CHAPTER 4: METHODOLOGY 4.1 Theoretical Framework The methodology of duration analysis, also known as survival analysis, aims at understanding the factors that explain the time that passes (i.e., duration) before a certain event occurs (G reene , 20 12 ). In the current study, the event of interest is the end of the CSBs participation in the BTD project. Thus, during the project phase (2011 - 2014), the CSB can be in one of two states, (1) participating in the BTD project or (2) having withdrawn from the BTD project. The only way to leave the initial participation state, often called a spell in duration analysis literature, is to leave the BTD project. The decision to withdraw from the BTD project can be modeled using a utility model. apprenticeship assignments in Australia, the members of each of the i th CSB evaluate their collective utility after each year t of the BTD project from two alternatives j =1, 2. They could continue in the BTD project, j= 1 or withdraw from the BTD project, j= 2. The CSB a lso makes this decision with the tecnicos or extension workers from INTA. If an alternative project, say soil conservation, better met the needs of the farmers or community, the extension worker could suggest changing the CSB to a group of farmers impleme nting soil conservation practices. The decision is made as a group, although an individual may choose to leave the CSB in well as human capital assets associate The joint utility function of the CSB decision to withdraw from the BTD project is composed of two sub utilities, the first from the expected utility of the decision made for that 34 time period and the second from the expected utility of future benefits , both economic and noneconomic , from participating in the BTD project conditional on recognition of the future benefits. This is expressed in equation 1, (1) Where is the expected economic and noneconomic benefits for CSB i from choosing option j at time period t. is the composition of CSB i from choosing option j at time period t. Although it is easy to see how the benefits would differ under when choosing between j= 1 and j= 2, note that a specific attribute, such as number of group members, might be different in option than in because a group member might co ntinue only if the group decides to take up a different activity, say for example, conservation of local or land race varieties instead of producing improved varieties. Since CSBs are expected to maximize their utility, the CSB chooses j= 2, to discontinue the BTD project, when the expected utility of the alternative option is higher than the net expected utility of continuing. The second sub utility u 2 captures an additional component of the CSB decision for awareness or recognition that low short - term ec onomic utility in the BTD project could be part of a learning process and that the knowledge gained from continued participation could lead to higher future economic utility. Likewise, social capital within the community could prove valuable in the future as well as the trust gained with government employees. This is seen in the case of Nicaragua, where a few selected CSBs continued with other sources of financial and technical support when the three year BTD project ended. While the above model is helpfu l conceptually, it does not lead directly to econometric specification, in part because enumerating future economic utility of is difficult in a pilot 35 project. Additionally, given the uncertainty of future profit, the model is difficult to operati onalize to explain the determinants of CSB sustainability. The model developed by Carletto et al. (1999) to explain the technology adoption decision and entrance into non - traditional markets is more useful in modeling the determinants of CSB sustainability . This model includes two analytical components -- the analysis of time to adopt a new technology or market, and the duration of participation in the nontraditional market duction of A pta (i.e., QDS) new technology as done by Carletto et al. (1999) is not applicable in the current study because the decision to adopt the technology (i.e., product ion of A pta (i.e., QDS) bean seed using the CSB model) and the new market (seed sales to neighbors/local community) had already been adoption analysis, i.e., the duration of participation in the nontraditional market before withdrawal, is however, applicable and used as a basis to model the determinants of CSB sustainability in this study. - househol d choice model, the risk - averse decision makers of a CSB, will maximize their utility and thus choose to withdraw from the BTD project if the change in utility, , from leaving the project is positive. This utility function is expressed as a function of several factors (explained below) that positively or negatively influence s the change in utility: (2) The factors included in equation 2 build on the model from Carletto et al. (1999) and the review of the literature discussed in Section 2. A represents land assets, which are associated 36 with access to credit, adoption of technology and slower withdrawal from adopted technology through the risk factor. CSBs that produce seed on land own ed by members are expected to be able to obtain financing should it be needed to cover emergency inputs to prevent crop loss. Bean seed harvest is particularly vulnerable to post harvest losses in the event of a rainy harvest season. Likewise, the more t han proportional decline in absolute risk aversion associated with increased land assets indicates that land assets are expected to be associated with longer participation in the BTD project. The size of the CSB is an indication of available labor L (per u nit of land asset) that its members can supply. Hired labor is assumed to be less efficient and thus increase production costs. The expected price p x of seed sales is negatively associated with withdrawal from the BTD project. The price comes from repay ment of seed loans to community members and thus is not dependent on grain prices. The expected price would have a one year lag in our model. Three sources of capital are negatively related to withdrawal from the CSB. Farm productive assets FK include so urces of transportation (pick - up trucks, mules, oxen, horses) and tools (backpack sprayers, grain/seed drying areas, silos for seed storage). Human capital assets HK include leadership age and experience as well as gender and education. Social capital as sets, SK are measured by the type of CSB (individual vs community based), share of CSB members that are related of other members and operational formality (i.e., number of meetings, written by - laws, recording minutes of the meeting, having a seed productio n and marketing plan). Training received by the CSB leadership and members could fall either under human capital (representing knowledge or education) or social capital (i.e., operational formality and connectivity to INTA). 37 Two time measurements are incl uded in the model. The years of previous collective organization of CSB members prior to CSB formation and the survival time between entering BTD project and withdrawing from the project. Finally, the village or community effects include measure s of remoteness such as distance to city (market), infrastructure, and public services. Regional effects are also included in the model to include regional level heterogeneity in the approaches used by the INTA regional offices to operatio nalize the CSB model. Since we are interested in the survival time in years, equation 2 is manipulated to express t s as a function of the independent variables. (3) While the body of litera ture on small - scale community - based seed production does n ot provide empirical research on the determinants of survival of the specific model of CSBs, evidence in the literature does highlight several factors that may constrain or encourage the sustainabil ity of such a community based seed production model. Based on the theory and information collected through interviews and literature review, the possible determinants of CSB survival, and thus sustainability, is expressed in the following e quation of t s : t s = ( years of operational experience prior to the BTD project , number of CSB members, community members (including CSB) at least partially decided use of seeds (dummy), number of monthly meetings, percent of member attendance, meetings minutes are recorded (dummy), written bylaws (dummy), percent of CSB member with immediate family member in CSB, percent of CSB member with extended family member in CSB, distance to paved road, travel time in private vehicle to municipal city, predicted Princip al Com ponent Analysis (PCA) index score, CSB president older than 30 (dummy) , and promoter are the same person (dummy), seed produced on CSB member land (dummy), seed produced on rented land (dummy), labor from CSB member (dummy), hired labor (dummy), number of silos, received silo from BTD project (dummy), access to: backpack sprayer (dummy), drying area (dummy), animal or vehicle for transportation (dummy), trai ned in CSB formation and organization (dummy), trained in seed marketing (dummy), yield, percent of seed 38 production distributed, , number of beneficiaries (per mz of seed production), percent of beneficiaries repaying, recovery rate, CSB supplied variety d emanded (dummy), region of INTA (dummy for each region), type of CSB ). The dependent variable t s is measured in years and calculated by the duration or survival of a CSB in the BTD project. 4.2 Empirical Estimatio n Strategy Duration analysis has differen t names in different disciplines including survival analysis, event history analysis, transition analysis, lifetime analysis and failure - time analysis (Guo 2010). Regardless of the name, duration analysis focuses on a specified distribution of the populat ion conditional on independent variables, called covariates, collected at the beginning of the study (Wooldridge 2011). In the current study, the duration of CSB operations in the BTD project is analyzed as a sequence of conditional probabilities that the CSB continues after time period t given that it has already survived until time period t. T is a non - negative random variable that denotes the time to failure of the CSB. Following Cleves et al. (2010), the cumulative distribution function is expressed a s, ( 4 ) By reversing the cumulative distribution function, the survival function S is: ( 5 ) At t =0, or prior to the end of the first year of the BTD project, all of the CSBs are operating and thus the function is equal to one. Therefore, the survivor function reports the probability that no CSB has failed prior to t. As t increases, and as CSBs fail , the function approaches zero. In this study, as in most studies using duration analysis, T is treated as a continuous random variable. As such, its density function, f(t) is obtained from S(t) or F(t) : ( 6 ) 39 This d ensity function forms the basis for estimating the hazard function, h(t), which has many names , but an appropriate name for this study is the age - specific failure rate. Also known as the instantaneous rate of failure, h(t) gives the probability of CSB fai lure during a given interval, here measured in years, conditional on the CSB having survived until the beginning of the interval, divided by the width of the interval: ( 7 ) By defining the accumulated risk up to time t as the cumulative hazard function H(t) , ( 8 ) it is possible to define any of the four functions (cumulative distribution function, survivo r function, hazard function and density function) of probability distribution of failure time if one is already given. (9) (10) (11) As mentioned earlier, duration analysis is interested in estimating hazard f unctions conditional on covariates. Two important classes of models are used to analyze the effects of covariates on survival time. The first is the proportional hazards (PH) model. The main assumption that cannot be violated in a PH model is that every hazard function is proportional to the baseline. The hazard function of PH models take on the form , (12) 40 with time invariant and positive function proportionally and multiplicatively shifting individual hazard functions away from the time dependent positive baseline hazard (Wooldridge 2011). Following the notation of Cleves et al. (2010), is parameterization as and the hazard function conditional on covariates for the PH models is (13) The semiparametric Cox Model and the parametric Weibull distribution and Gompertz models are PH models considered in this study. In each of these models, a well - defined baseline hazard can be added to equation (13) above to obtain the hazard function conditional on the covariates. The Cox model is a semiparametric model because it s baseline hazard does not assume a parametric specification or distributional assumption . That said, the effects of the covariates are parameterized to establish a baseline survivor function. Said differently, it is assumed that the covariates shift the baseline survivor function. The hazard rate for the jth CSB at time t conditi onal on covariates X j is given by (14) where j are the regression coefficients to be estimated. The equation is identical to equation (13) because the Cox model does not specify a baseline hazard function. Therefore, in estimations of Cox models no constant or intercept is computed. In statistical programs such as STATA, however, Cox - adjusted survival estimates can be computed. One of the benefits of the semiparametric model is apparent when the distribution function is depicted graph ically . While an assumed distribution will have a smoothed curve beginning at 1 when time is zero and approaching 0 as time increases, the Cox allows for a step 41 function (as it is not bound to a specific distribution) that resembles plots such as Figure 6.1 below . A second PH model is the Gompertz model. Like all of th e models considered below, the Gompertz model is a parametric model. Parametric models use maximum likelihood estimators as they allow time, the dependent variable, to assume non - normal parametric distributions (Guo 2010). The parametric distributions ar e defined by parameters that can be calculated and once specified, used to obtain the survival and hazard functions. The baseline hazard for the Gompertz model is (15) After combining the above with equation (13) the conditional hazard function is given by (16) and the conditional survivor function is (17) For the Weibull model, the baseline hazard function is (18) Thus using equation (13) above, the conditional hazard function is (19) and the conditional survivor function is (20) When the parameter p is greater than one, p >1, the hazard rate is increasing and when p <1, the hazard rate is decreasing as time passes. An interesting feature of the Weibull distribution is its relationship with the exponential distribution. When the p parameter is one, th e Weibull distribution reduces to the exponential distribution. Thus the exponential distribution 42 assumes constant hazard rate over time. The Weibull distribution not only can be a (PH) model, but it can also take the form of an accelerated failure time (AFT) model, the second class of models considered in this paper. While PH models provided results allowing for the comparison of hazards between CSBs of differing covariates, AFT models provide results allowing for a comparison of survival times. Specifi cally, how covariates accelerate the time that passes between the beginning of the study and the time of failure or CSB withdrawal from the BTD project. Following Cleves et al . (2010) AFT models or ln (time) models follow the parameterization: (21) but instead of assuming a distribution of , a distribution is assumed for (22) and since (23) When , then time passes faster so the event occurs sooner and thus time is accelerated. Likewise, when time passes slower so the event occurs later and time is decelerated. For the lognormal regression, (24) For the log logistic distribution model, and (25) 43 The parameters, , estimated in the AFT models give the proportional change in duration (survival) time given a one unit change in the explanatory variable, all else held equal. 44 CHAPTER 5 : DATA AND DESCR I PTIVE STATISTICS 5.1 Data Sources The data for this study comes from the survey of community seed banks conducted in 2012 in Nicaragua. The survey was conducted by the extension workers and promoters who were trained by the author in March 2012, during a round of regional training sessions for the BTD project. The extension workers and promoters were also trained in IRB requirements of protection of human subjects and how these requirements would impact the data collection process. The survey instrument used for this study captured the gen eral characteristics of the CSBs, their membership profiles, and seed production experience. An English version of the final instrument is included in Appendix A. 6 The regional coordinators of the BTD project collected the completed surveys from the exte nsion workers and submitted them to the national INTA office in Managua for data entry. All the 207 CSBs that had participated in the BTD project in 2011 - 2012 (i.e., the first year of the BTD project) were targeted for this survey. By June 2012, 154 CSBs had submitted the completed surveys. Table 5.1 presents the total number of CSBs targeted for the survey versus included in this study by region. Multiple attempts were made by the author through requests to national INTA employees to obtain the surveys from all 207 CSBs, however, there remained 53 non - respondent CSBs. Table 5.2 presents the distribution of the surveyed CSBs by region and type of CSB. 6 Unfortunately, the INTA leadership of the Centro Sur region circulated a draft version of the survey to the CSBs instead of a final version. As a result, certain variables regarding quality of the seed produced and feedback from beneficiaries (clients) were not obtained for the CSBs from this region. Since most of these CSBs were collected, packaged and distributed th e seed, responses to these questions would have been skipped in the survey for these types of seed banks that were common in the Centro Sur region. 4 5 Over the course of the BTD project, INTA staff submitted reports listing the CSBs that continued to ope rate in the final two years of the project as well as a selected group of CSBs that were supported in the 2014 agricultural season (after the official end of the BTD project). Information on which CSBs failed during the BTD project years 2 and 3, and whic h continued after the end of the BTD project is used in the duration analysis to determine which characteristics of the CSB are associated with variations in the CSBs failure rates and time of operation. Table 5.1: Total number of CSBs targeted for the su rvey versus those that completed the survey and included in this study Region Total number of CSBs targeted Number of CSBs that returned the completed survey (sample size for this study) Centro Norte 41 37 Centro Sur 40 40 Las Segovias 44 29 Pacifico Norte 62 28 Pacifico Sur 20 20 Total 207 154 Table 5.2 : Distribution of surveyed CSBs by region and type Region Type of Seed Bank All Classic Parceled Individual Centro Norte 37 0 0 37 Centro Sur 0 10 30 40 Las Segovias 15 11 3 29 Pacifico Norte 7 21 0 28 Pacifico Sur 13 7 0 20 Total 72 49 33 154 5.2 Descriptive statistics of variables included in the model Tables 5.3, 5.4 and 5.5 provide summary statistics of the variables included in the duration analysis. Since one of the objectives of this study is to characterize the differences and similarities between the CSB models, the statistics are presented by the three types of CSBs classic, p arceled and individual. The dependent variable in duration analysis is calculated as the 46 duration of time (i.e., number of years) from a starting point (in this case the start of the BTD project) to the occurrence of an event ( in this case the withdrawal of a CSB from the BTD the BTD project) is lowest for Individual CSBs at 1.6 years and highest for Parceled banks at 2.2 years ( Table 5.3). Individual banks as a group had a statistically different weighted average years of survival at a 10% level than the other two CSB types. 47 Table 5.3 Summary Statistics of variables used in the duration analysis: Differences across types of CSBs in c ommunity level charac teri stics, membership and operating procedures Type of Community Seed Bank Classic Parceled Individual TOTAL # of Observations 72 49 33 154 Mean Years participation in BTD 2.07 a 2.22 a 1.61 2.02 # of Years (% Yes) 1 Year 38.89 ~ 26.53 ~ 69.7 ~ 41.56 2 Years 29.17 ~ 30.61 ~ 12.12 ~ 25.97 3 Years 18.06 ~ 36.73 ~ 6.06 ~ 21.43 4 Years 13.89 ~ 6.12 ~ 12.12 ~ 11.04 CSB Organizational Structure # Years operation at beginning of BTD 0.24 a 0.31 a 0 0.21 # of CSB members 9.35 a 7.51 a 1 6.97 CSB or Community members had voice in use of seed produced ( % Yes) 81.94 a 79.59 a 45.45 73.38 Number of monthly meetings 1.41 a 1.37 a 0 1.09 % of CSB members attending meetings 82.38 a 89.38 a 0 66.96 Meeting Minutes Recorded ( % Yes) 54.17 a 61.22 a 0 44.81 CSB has written bylaws ( % Yes) 54.17 73.47 0 48.70 % of CSB members with Immediate family members in CSB 40.71 a 31.42 a 0 29.03 % of CSB members with Extended family members in CSB 22.86 11.03 a 0 a 14.20 Community Characteristics Distance to paved road (KM) 14.21 a 12.52 a 8.24 a 12.40 Travel time to Municipal Seat in private car (minutes) 25.86 a 25.57 a 28.64 a 26.36 PCA of Community Level Development 1 - 0.10 a 0.30 a - 0.23 a 0 Leadership Characteristics President older than 30 ( % Yes) 95.83 a 83.67 96.97 a 92.21 President's Gender ( % Male) 90.28 a 87.76 a 84.85 a 88.31 Promoter's Gender ( % Male) 87.50 a 85.71 a 84.85 a 86.36 President is Promoter ( % Yes) 65.28 a 48.98 a 100 67.53 5.58 a 7.29 b 5.73 ab 6.16 Notes: 1 PCA means Principal Components Analysis Types of CSBs that share a letter are not significantly different at the 10% level ~ indicates that a significance tests across groups has not be en performed for these variable 48 5.2.a Characteristics of Individual Banks There are several statistically significant differences in the descriptive statistics for individual banks compared to the other two types of CSBs (Table 5.3 and 5.4). As mentioned above, one such variable is the duration of BTD participation, the depende nt variable in this study. Years of operation prior to the beginning of the BTD project was zero for individual banks, not from lack of seed production experience, but from not previously using the CSB operational structure. 7 The percent of promoters that are also the presidents of their CSBs, percent of CSB members related to another CSB member, and the variables related to CSB meetings are not relevant to individual seed banks due to the nature of this type of seed bank ( Table 5.3). Training related to formation and organization of the CSB also did not occur for individual banks ( Table 5.4). Additionally, less than half of the individual seed banks reported having a voice in the decision of use of seed (Table 5.3) due on plan within the Centro Sur region described in Chapter 3 . 5. 2 .b Organizational Structure of CSBs Comparing classic and parceled CSBs in regards to indicators of organizational structure , there w ere only tw o variable s that w ere statistically different (T able 5.3) . A higher percentage of extended family members participated in c lassic CSBs than parceled CSBs and a higher share of parceled CSBs ha d writt en bylaws than the classic CSBs. Written bylaws are a good indicator of both following the CSB formation protocol given by INTA but also the organizational disciple of administering and planning required in a business orientated enterprise with several partners. No significant difference was detected in the number of meetings per month, average att endance per meeting and documenting minutes of the meetings between the two types of banks. 7 The survey instrument did not capture years of seed production experience, the relevant variable of inter est for individual seed producers. 49 5.2.c Community Characteristics No statistically significant differences were found between the three types of CSBs in the index of community development, 8 dis tance to paved road and travel time to town in a private vehicle. While these community level characteristics did not differ between types of CSBs, there were regional differences between the five INTA administrative regions presented in Table B.1 of Appe ndix B. 5.2.d Leadership Characteristics In rural Nicaragua, age is closely associated with farming experience as urban to rural migration is rare. Most rural Nicaraguans would state that they have been farmers since they were born. Likewise, with age co mes responsibility as community leadership roles are less likely to be held by youth. The presidents of over 95% of classic and individual banks were older than 30 years of age. Only 84% of presidents of parceled CSBs were older than 3 0 years representin g a statistically significant difference compared to the other two types of CSBs. While there was no statistical difference between the three types of CSBs in gender of the n than classic CSBs. On average presidents of parceled CSBs completed more than 7 years of formal education while classic CSB presidents completed less than 6 years. The years of education of presidents of individual banks were not statistically differen t than the other two types of CSBs. 5.2.e Land Use in CSB Seed Production Table 5.4 provides a summary of seed production inputs by types of CSBs . There was a large and statistically significant variation in ownership of land used for seed production betw een 8 A principal component analysis (PCA) of 14 community level characteristics (reflecting the infrastructure and amenities present/absent in the community) was used to generate this index of community development. Lower n umbers represented communities with less access to public services and amenities. See section B9 of the survey included in the Appendix for a complete list of community level characteristics comprising the PCA index. 50 the three types of CSBs. While 81% of classic CSBs used land owned by a CSB member, the share was much lower among parceled (61%) and individual (15%) CSBs. Around a quarter of the classic and parceled CSBs rented land and were not significantly diffe rent, while only 3% of individual CSBs rented land. Some CSBs with multiple seed plots reported using CSB member land and renting it from another farmer, while some individual CSBs did not report using either source of land. Other sources of land possibl y used by individual CSBs could be borrowed land (without paying rent) or a crop sharing agreement where the land owner provides the land and the seed bank (producer) provides the labor with the agreement of splitting the harvested production. Unfortunatel y, these other options were not included in the survey, and thus it is difficult to determine if and how many of the CSBs that did not report using their own land or rented land used these alternate sources of land or whether they simply did not respond to that question (and thus we potentially have the problem of missing data). 5. 2 . f Labor use for seed production In terms of labor input, as expected, the classic CSBs reported statistically higher use of CSB members for labor than the parceled and individual CSBs. While low use of hired labor by classic banks was expected because CSB members provided labor, it was surprising that individual seed banks did not hire labor and instead mostly relied on household members for labor input. The similariti es of seed production between the parceled and individual banks explain the lack of statistically significant difference in the use of CSB member labor. In the case of individual banks, the operator him/herself is in charge of production but for parceled banks they could have access to additional CSB members providing labor. 51 5. 2 . g Assets and access to facilities used for seed production The CSBs also vary in terms of assets and access to facilities (Table 5.4). Number of silos is an important variable f or CSBs because an additional silo provides the opportunity to store an additional variety of seed or grain while still maintaining the varietal distinction. Parceled CSBs had a significantly higher number of silos compared to classic and individual CSBs. Although the share of parceled banks that received silos from the BTD project was higher than the other two types, the difference is likely due to existing silo ownership prior to the BTD project. Some silos were inherited by CSBs from the MCC project i n the Pacifico Norte region and 43% of parceled CSBs are in the Pacifico Norte region. Backpack sprayers are an important tool used to apply inputs. Although there was no statistical difference between the three types of CSBs, o ver 76% of classic CSBs and 53% of parceled CSBs had this tool, while only 9% of the individual banks reported access to a backpack sprayer. If one CSB member had the tool, it is assumed that the CSB would have access to that tool. Therefore, the individu al banks might be at a disadvantage compared to the classic and parceled if they are unable to borrow the tool from neighbors. An area for drying seed is used during post - harvest treatment of seed. Twenty percent of parceled CSBs had access to a drying ar ea while only 3% of parceled CSBs and no individual banks indicated having access to a drying area ( Table 5.4). A source of transportation is important to increase efficiency of moving plants or grain from the field to the drying area. There was no statis tical difference between CSB types for the percent of CSBs with access to animals or a pickup for transportation. 52 Table 5.4 Summary statistics of variables used in the duration analysis: Differences across types of CSBs in seed production inputs Type of Community Seed Bank Classic Parceled Individual TOTAL # of Observations 72 49 33 154 Land Land used for seed production (MZ) 1.19 a 1.00 a 0.99 a 1.09 Seed produced on CSB member land ( % Yes) 80.56 61.22 15.15 60.39 Seed produced on rented land ( % Yes) 23.61 a 28.57 a 3.03 20.78 Labor CSB members provided labor ( % Yes) 90.28 69.39 a 69.70 a 79.22 Hired workers provided labor ( % Yes) 12.50 a 57.14 0 a 24.03 Assets/facilities # of Silos 1.83 a 3.39 1.09 a 2.17 CSB received Silo from BTD Project ( % Yes) 30.56 a 67.35 12.12 a 38.31 CSB has access to backpack sprayer ( % Yes) 76.39 53.06 9.09 54 .55 CSB has access to seed/grain drying area ( % Yes) 2.78 a 20.41 0 a 7.79 CSB has access to transportation (pickup/mule/ horse/ox) ( % Yes) 73.61 a 83.67 a 90.91 a 80.52 CSB has access to animal transportation (mule/ horse/ox) ( % Yes) 73.61 a 83.67 a 69.70 a 75.97 Human Capital CSB trained in Formation and Organization ( % Yes) 88.89 a 95.92 a 9.09 74.03 CSB trained in Seed Marketing ( % Yes) 25.00 ab 36.73 b 12.12 a 25.97 CSB trained in Seed Production ( % Yes) 90.28 a 95.92 a 90.91 a 92.21 Notes: Types of CSBs that share a letter are not significantly different at the 10% level 5. 2 . h Human capital in seed production As mentioned above, few individual seed banks received training on CSB formation and organization while the levels of this training were not significantly different for classic ( 89% ) and parceled (96%) CSBs . Training on seed marketing, a skill deemed very important for sustainability as per the literature, was received by 37% of parceled banks, 25% of classic banks and 12 % of individual banks. There was no statistically significant difference between the 53 c lassic CSBs and the other two types of banks, but the share of parceled banks receiving this training was significantly higher than individual banks (Table 5.4). Finally , seed production training was received by 92% of the CSBs and did not significantly d iffer among the types of CSBs. 5. 2 . i Output and efficiency indicators in Seed Production As reported in Table 5.5, there are also differences and similarities in the output and efficiency indicators across the three types of CSBs as measured by yield, number of beneficiaries, loan repayment and meeting the varietal diversity needs of the communities served by these banks in the first year of their operation under the BTD project . On average, CSBs of all three types used one manzana (0.7 hectares) to pr oduce a little over 1200 pounds of seed per manzana and marketed seed to just fewer than 29 (clients) beneficiaries per manzana. There was no statistical difference in potential yield 9 among the three types of CSB. The similarity in size, number of clien ts served, and yield can be attributed to recommendations by the BTD project and a similar input package distributed across types of CSBs. Despite the similarities, parceled CSBs disseminated a statistically larger percentage of their seed production ( 58%) to beneficiaries than classic CSBs (43%). The percent of production distributed to beneficiaries of individual CSBs did not differ than the other two types of CSBs. At the BTD project level, repayment terms for beneficiaries were established to be two po unds of grain for each pound of seed received from the CSB. The two - for - one scheme was to differentiate the value of seed compared to grain and recover production costs. The importance of ensuring repayment was not always understood by INTA staff and ext ension workers in all regions as noted in Table B.1 of Appendix B. The Centro Norte region, where 51% of the 9 Potential yield is the highest yield achieved by a CSB during the BTD project. For CSBs that withdrew from the BTD project after Year 1, yield and potential yield are the same. 54 Classic CSBs are located, had an extremely low compliance rate of only 1% due to a regional implementation of a one pound of grain repayment for e ach pound of seed received . The classic the parceled and individual banks (Table 5.5). Similarly, the recovery rate of seed was only 49% for classic CSBs and 7% for the Centro Norte region compared to 76% among all CSBs. Table 5.5 Summary statistics of variables used in the duration analysis: Differences across types of CSBs in seed production output indicators Type of Community Seed Bank Classic Parceled Individual TOTAL # of Observations 72 49 33 154 Output and Efficiency Indicators Year 1 Yield (QQ/MZ) 11.05 a 14.07 a 12.78 a 12.38 Yield potential (QQ/MZ) 15.75 a 17.67 a 14.00 a 15.98 % of production distributed to beneficiaries 43.37 a 57.89 b 50.18 ab 49.45 # Beneficiaries per MZ Seed Production 25.16 a 30.86 a 33.45 a 28.75 % of Beneficiaries Fully Repaying (2lb per 1lb) 23.76 55.74 a 54.04 a 40.43 Recovery rate (repaid/seed distributed) 0.49 0.88 a 1.18 a 0.76 CSB supplied variety demanded ( % Yes ) 29.17 63.27 a 57.58 a 46.10 Notes: Types of CSBs that share a letter are not significantly different at the 10% level In year 1 of the BTD project, the CSBs had few options of varieties to offer to their communities. Most CSBs received registered seed of only one variety, INTA Rojo 10 , because that was the main variety available from INTA and it was considered to be widely adapted to different conditions across the country, was resistant to diseases, and fetched a decent market 10 A little over 83% of CSBs received INTA Rojo. Other varieties produced by CSBs in year 1 of the BTD project were INTA Sequia (12% of CSBs) and INTA Matagalpa (5% of CSBs). 55 price for grain 11 . But the availability of one main variety for seed production throughout the country implied that the location specific demand for diverse bean varieties was not met by all CSBs. Less than a third of classic CSBs offered a variety that was among the highest demanded seed varieties in their community. Supplying a demanded variety was higher among parceled (63%) and individual (58%) CSBs (Table 5.5). 11 The price for grain of a category of land race or Criollos varieties call Rojo Seda is largely assumed to be higher than INTA Rojo, however, no market prices were collected for this study. 56 CHAPTER 6 : D URATION ANALYSIS RESULTS 6.1 Non - parametric duration analysis A non - parametric technique of duration analysis is helpful to visually observe the data and different subgroups of the data (Kleinbaum and Klein 2012). The no n - parametric technique used in this study is the Kaplan - Meier S urvival (KMS) curves. Additionally, the log - rank test is used to test if multiple KM S curves are different. Figure 6.1 gives the KM S survival probabilities for each year of the BTD program. Figure 6.1: Kaplan - Meier Survival Curve of CSB Failure The horizontal axis represents the years of the BTD project. For example, t= 0 is at the beginning of the BTD project and t= 1 is the end of year 1, t= 2 is the end of year 2, and so on. The agricultural year in Nicaragua begins with the rain in May, thus t= 0 is May 2011, and t= 1 is 57 May 2012 when the first year of the BTD project ended and the second year began. The vertical axis represents the percent of CSBs still operating, thus at t= y - value is 1 because none of the CSBs had failed. After year 1 of the BTD project or between t= 1 and t= 2, the function has the value of 0.58 because 58% of the CSBs remained in the BTD project. The speed of leaving the BTD project (or ending the spell) is highest at the beginning and slows near the end. Of interest in this study are the differences between the three types of CSBs. KM S curves for each of the three types of CSB are given in Figure 6.2. Figu re 6.2: Kaplan - Meir Survival Estimates by CSB Type After year one, the curve for parceled CSBs is higher than the curves for the classic and individual CSBs. These curves indicate that initially, parceled CSBs have the best probability of sustainabili ty. However, after year 3, the curves for classic and individual CSBs are higher than parceled CSBs. These results indicate that the speed of failure is high initially for individual 58 seed banks, but slows down considerably in subsequent years. There app ears to be little change in the speed of failure of the parceled CSBs over the analysis time. The KM S curves in Figure 6.2 indicate that the survival functions are not equal for all three CSB types. The log - rank test confirms with Chi squared(2)=4.96, p - v alue 0.0835, so at a 10% level we reject the null hypothesis that the surviv al functions of the three types of CSBs are equal. A second common test with the same null hypothesis is the Wilcoxon test and it confirms the results of the log - rank test at a hi gher level Chi squared(2)=10.84, p - value 0.0044. Since regional heterogeneity is anticipated, we can test for difference in survival curves while controlling for regional differences using a Wilcoxon test stratified by the five INTA regions. The results reveal the same conclusion, Chi squared(2)=6.11, p - value 0.0471, the null hypothesis, that the three survival functions are equal, is rejected at a 5% level. As described in Chapter 4, three proportional hazard (PH) models are considered, the semi - parametric Cox model and two parametric models, the Weibull and G omper tz. The PH model assumption is that th e hazard rates are constant and proportional to each other between all CSBs for the entire analysis time. A visual test by graphing the log - l og KM S survival estimates of each CSB type plotted against the log of time, given in Figure 6.3, provides evidence of violations of the PH assumption. 59 Figure 6.3: Log Log Kaplan - Meier Curves by CSB Type Figure 6.3 reveals that the three survival curv es are not parallel and thus the PH assumption is violated. To confirm or reject this result, a post estimation test using Schoenfeld residuals will be used after fitting a Cox PH model. Figure 5.3 also indicates that the exponential distribution is not appropriate because the slope of the curves appear to differ from - 1. The p - parameter of the estimated Weibull model will be able to confirm this result. 6.2 Parametric and semi - parametric duration analysis To fit the model, first all of the variables considered important from the literature review are included. Three PH models, the Cox, Weibull and Gompertz, and two AFT models, the log - normal and log - logistic are estimated. Comparison of the models to dete rmine the best model is performed by comparing the estimated log - likelihood (higher is better) and two post - estimation information criterion (lower is better) 60 Bayesian information criterion (BIC). Addit ionally, the significance levels of the coefficients are considered when comparing different models. 6.2.a Full and preferred duration analysis models Table 6.1 provides summary information of the conditional probability of CSB failure for the five full an d preferred models of duration analysis. The sign of the relationship between the independent variables and survival is given. A positive sign for the PH models indicate that the hazard ratio is less than one (one unit increase in the explanatory variabl e decreases the hazard of failure) while a negative sign indicates a hazard ratio greater than one (one unit increase in the explanatory variable increases hazard of failure). In AFT models, a positive sign indicates a positive coefficient and thus a dela y in failure (one unit increase in the explanatory variable accelerates or stretches out time) while a negative sign indicates a negative coefficient and thus faster CSB failure (one unit increase in explanatory variable decelerates or condenses time). The models give consistent results of the relationship between the dependent and independent variables when significant. There were few examples of variables found to be significant in the PH models but not in the AFT models. Four variables that are statist ically significant in the PH models but not in the AFT models are: the dummy variables for Centro Sur community voice in the seed use decision. Likewise, the dumm y variable for Pacifico Norte was significant in two AFT models but not in the HP models. The relationship between different characteristics of the CSB and its survival predicted by these models is generally aligned with expectations. Yield, repayment rat e, number of silos, all had a positive relationship with the survival variable. The number of beneficiaries per unit of 61 land used for production had a negative relationship with survival, possibly due to limitation in farmers, they were forced to lend to unknown farmers who may not repay at the same rate due to lower leve ls of trust. The relationship of survival and voice of community member in the seed use decision was also found to be negative. The relationship of survival and share of family members in the CSB membership was negative possibly due to community percepti ons that the CSB is exclusive to one clan or that poor management occurred because leadership roles were assigned for rea sons other than qualification . Travel time to the municipal seat (major city of local government with commerce and market) had a positi ve relationship with survival indicating that CSBs further from cities had higher survival rates. The opposite result was expected as an assumed decline in frequency and quality of supervision from INTA was predicted due to the increase in travel time. H owever, the results may be attributed to the fact that communities furthest from a commercial town may have low access to certified seeds and thus an increased interest in the CSB and greater support for having a local source for quality seed. The relation ship between the negative. While this was unexpected, it could be attributed to seed producers who do not use their own land know the soil quality limitations and seek soil of better quality than their ow n for seed production. Also, CSBs that used member land might have had fewer resources to gain access to quality land for seed production. From Table 6.1, the p - parameter on the Weibull models is not equal to 1 at a 5% level and thus we can officially r ule out the exponential model as the hazard rates are not constant. Likewise, the - parameter on the Gompertz model is statistically different than 0 at a 5% level, 62 confirming that the hazard rates are not constant. Since the shape parameters on the Weib ull models are greater than 1 and those of the Gompertz model are positive, the probability of CSB failure increases with time. Table 6.2 give s the summary statistics to determine the best model. Among the preferred models, the Weibull and Log Normal mode ls had the highest log likelihood. The Log Normal also had the lowest AIC and BIC among AFT models. The Weibull model had a lower AIC and BIC results than the Gompertz model among parametric HP models. While the Weibull model included the most variables that were significant at a 10% level, it was the Log Logistic model that included the most significant variables among AFT models. The Weibull for parametric PH and Log Normal for AFT will be the choice models to use for the final results. 63 Table 6.1: Summary Results of Duration Analysis of CSBs Cox Weibull Gompertz Log Normal Log Logistic Variables Full P referred Full P referred Full P referred Full P referred Full P referred CSB Organizational Structure # Years operation at beginning of BTD NS NS NS NS NS # of CSB member NS NS NS +* NS +* NS +** NS NS 1=Community had a voice NS - ** NS NS NS NS NS NS # of monthly meetings NS - * - * - ** NS NS - * NS % of member attending meetings NS NS NS NS NS 1=Meeting minutes recorded +*** +*** +*** +*** +*** +*** +*** +*** +*** +*** 1=written bylaws NS NS NS NS NS NS NS % CSB members with immediate family member in CSB - ** - *** - ** * - * ** - *** - *** - ** * - * ** - * ** - * * % CSB members with extended family member in CSB NS NS NS NS NS Community Characteristics Distance to paved road NS NS NS NS NS NS Travel time to city in private car +** * +** +*** +*** +** * +** * +** * +* * * +** * +** * PCA NS NS NS NS NS Leadership Characteristics 1=President's age> 30 +** * +** * +*** +*** +*** +*** +** * +* * * +*** +** * President male +*** +*** +* ** +* ** +* ** +* ** +* +** NS NS Promoter Male NS NS - * - * - * NS NS - * NS President is Promoter NS NS +* NS NS NS NS NS # Years Education of President +*** +*** +*** +*** +*** +*** +*** +*** +*** +** Land 1=seed produced on CSB member land NS - * - ** - ** * NS - ** NS - ** NS - * 1=land rented NS NS - * - * NS NS NS NS Labor 1=labor from CSB members NS NS - * NS - * NS NS NS NS NS 1=hired labor NS NS NS NS NS NS NS 64 Cox Weibull Gompertz Log Normal Log Logistic Variables Full P referred Full P referred Full P referred Full P referred Full P referred Assets/facilities # of Silos +* * * +*** +*** +*** +*** +*** +*** +*** +** * +*** Did receive Silo from BTD NS NS NS NS NS NS NS NS 1=backpack sprayer NS NS NS NS NS NS NS 1=drying area NS NS NS NS NS 1=mule/horse/ox access NS NS NS NS NS NS NS NS Human Capital 1=trained in CSB formation NS NS NS NS NS NS 1=trained in seed marketing NS - * NS NS - ** - ** NS NS NS Year 1 Yield (qq/mz) +*** +*** +*** +*** +*** +*** +*** +*** +** * +*** % of production distributed to beneficiaries NS NS NS NS NS beneficiaries/mz - ** - ** - * * - ** - * * - * * * - * - * * NS - *** % of beneficiaries repaying 2x1 lbs +** +*** +** +*** +*** +*** + * +** +* * +** * 1=Supplied variety demanded in community NS NS NS NS NS NS Region and Type of CSB Centro Sur dummy NS - * - * * - ** - * - ** NS NS NS NS Las Segovias dummy - ** - *** - *** - *** - ** - ** NS NS NS NS Pacifico Norte dummy NS NS NS NS NS NS NS +* NS NS Pacifico Sur dummy +** +** * +** +*** +** * +*** +*** +*** +*** +*** Parceled Bank dummy NS NS NS NS NS - ** NS NS NS NS Individual Bank dummy NS NS NS NS NS NS NS NS NS NS Constant NS NS NS NS - ** - *** - ** - ** Parameter p= 3.102 p= 3.022 9 Note: ***, **, and * indicate significance levels of 1%, 5%, and 10% respectively + and indicate whether the significant relationship is positive or negative NS indicates that the variable was included in the model but it is not significant 65 Table 6.2: Summary of Decision Statistics to Determine Best Model Cox Weibull Gompertz Log Normal Log Logistic Statistic Full P referred Full P referred Full P referred Full P referred Full P referred Log likelihood - 584.045 - 5 85 . 957 - 69.836 - 72.727 - 81.584 - 86.888 - 69.861 - 73.648 - 72.268 - 80.045 Restricted LL - 625.129 - 625.129 - 124.207 - 124.207 - 133.747 - 133.747 - 118.378 - 118.378 - 127.337 - 127.337 AIC 1246.090 1227.914 221.671 205.454 245.167 229.777 221.722 203.295 226.536 208.090 BIC 1364.53 1 1312 . 948 346.186 296.563 369.683 314.812 346.237 288.330 351.051 280.977 # of Significant Variables (at least 10%) 12 17 19 18 16 17 12 16 11 12 66 Table 6.3 provides the results of a likelihood ratio test that justifies the omission of non - significant variables with less than unity t - ratios in the preferred models. The null hypothesis is that the coefficients of the omitted variable are jointly not equal to zero. By failing to reject the null hypothesis, the preferred models have only removed a set of explanatory variables that jointly equal zero from the full mode l. Table 6.3: Summary Statistics for Testing Coefficients of Omitted Variable Statistic Cox (PH) Weibull (PH) Gompertz (PH) Log Normal (AFT) Log Logistic (AFT) Log likelihood a - 584.045 ( - 585.957 ) - 69.836 ( - 72.727 ) - 81.584 ( - 86.888 ) - 69.861 ( - 73.648 ) - 72.268 ( - 80.045 ) Number of restrictions b 1 1 11 13 13 1 7 Calculated chi2 3 .8 2 5 . 78 10 . 61 7. 57 15 . 55 P - value 0.9 748 0. 8875 0. 6435 0. 8709 0. 5556 Decision not rejected Not rejected Not rejected Not Rejected Not Rejected Notes a: The log - likelihood for the full model is presented first and the log - likelihood for the full model is included in parenthesis. b: The number of omitted variables are the same as the number of restrictions 6.2.b Interaction Terms One of the important findings in the literature t hat is not confirmed in our initial duration analysis is the importance of training in marketing. It is also anticipated that interacting the types of CSBs with variables of interest will indicate differences in sustainability between the types of CSBs. Since the literature has not indicated specific interactions to be included in the model, a search for all possible interaction was conducted. No interaction terms could be found to distinguish factors of sustainability between the types of CSBs. However, two interaction terms were discovered that are related to discussions in animal transportation were significant. The animal transportation variable is a more re stricted 67 version of the transportation variable used in previous models. Unfortunately, the interaction term is not significant in the lognormal model. A second interaction term was found for the dummy variable of seed marketing training and the continuou s variable of yield in year 1. The results of the three models w ith these two added interaction terms and a discussion of the results are given in Table C.1 of Appendix C. 6.2.c Test of Proportionality Assumption A test of the proportionality assumption of the Cox model using Schoenfeld residuals reveals that the model does not violate the assumption. The test results give a Chi squared ( 24) value of 28.47, and a p - value of 0.2409. We fail to reject t he null hypothesis of no proportionality and it is assumed that the model does not violate the proportional ity assumption. The test does provide results for each independent variable and the dummy variable for the Pacifico Sur region violates the proportionality assumption with a p - value of 0.0357. The hazard ratio describing the effect of a CSB located in this region is inappropriate if observed without the hazard ratios of other independent variables. One way to correct for this violation of the PH assumption is to allow the baseline hazard function to vary with the region. We accomplish this in the frailty section below. 6.3 Duration analysis of a dataset without Individual CSBs By the structure of individual banks, many potential determinants of sustainabil ity related to group organization are not relevant. For example, individual banks do not hold meetings or record meeting minutes and therefore, the presence of individual banks in the data may distort the influence of these variables. To ensure the result s are not distorted, survival regressions are obtained for a subset of the data that excludes individual seed banks. The sign on the hazard ratios does not change and thus there are no distortions to the variables of interest mentioned 68 above. Table D.1 i n Appendix D provides a comparison of the Weibull model with and without Individual CSBs. Since the parceled and classic CSBs were directly involved in the seed distribution process (unlike the Individual banks in the Centro Sur region where INTA distribu ted the seed) the number of communities reached (market size) and the feedback received from clients or seed recipients (quality) are available. Table D.2 in Appendix D provides the results that indicate quality, as measured by positive client feedback, h as a positive relationship with sustainability. The impact of market size on sustainability was found to be positive as CSBs that distributed seed to exactly two communities had a lower hazard ratio than CSBs that only distributed seed to one community. No significant difference was found, however, between CSBs that reached more than 2 communities with seed compared to CSBs that focused only on one community. 6.4 Frailty models to remove heterogeneity Duration analysi s comes from medical survival studies where frailty is a random (2012) example, two male smokers of the same age may have different failure times and different survival functions because one might be mor additional factors not capture d in the data that explain why they have different survival functions than CSBs with identical observed characteristics. 6.4.a Individual Frailty It has been assumed that each CSB , after controlling for all observable differences, are homogeneous. If they are not homogeneous, the results of the determinants of sustainability, or variables, in the duration analysis will be affected. Also, the share parameters could be wrong (we ha d concluded that risk of failure increases with time). Although heterogeneity can come 69 from misspecification of functional forms, it can also occur due to unobserved variation between CSBs. A frailty model includes an additional multiplicative term, with assumed mean of 1 and constant variance estimated as theta from the data. When the estimated frailty is greater than 1, than the CSB (or group in Shared Frailty below) has an increased hazard and decreased probability of survival compared to the other CS Bs (or groups in Shared Frailty below). A likelihood - ratio test that theta is equal to zero, indicating no heterogeneity, is preformed using the preferred log - normal model from above with a frailty distribution of gamma. No individual heterogeneity was d etected ; indicating that after controll ing for observable differences in CSBs , no individual CSB (or group in Shared Frailty below) experience any increased or decreased hazard of failure . 6.4.b Shared Frailty Although heterogeneity has been controlled for between CSBs in our preferred model, there still may be unexplained differences between the INTA administrative regions. Just as frailty models accounted for individual differences, shared frailty models account for differences in survival functions from unobserved factors. If we have not controlled for all of the differences between regions, once again our shape parameters and duration analysis results will be affected. We anticipate regional heterogeneity because of the unique seed production history of each region detailed in Chapter 3. While the previous models controlled for these differences through the fixed effects method, it is now necessary to test and adjust the preferred models for the presence of heterogeneity. 12 All three preferred model s i ndicate the presence of shared frailty. The log - normal model has the highest p - value, 0.082, of the three models indicating that at a 5% level shared 12 App endix E presents two alternative methods to shared frailty to remove the regional heterogeneity effects. 70 heterogeneity is not detected. However, due to the low p - values in the Cox and Weibull (frailty distribu tion of gamma) models (0.011 and 0.001 respectively), and given that at a 10% level the null hypothesis of no heterogeneity is rejected, it is assumed that shared heterogeneity is present. The results with regional heterogeneity effects removed are present ed in Table 6.4. Only the AFT model is displayed because the hazard ratios in PH models are affected by frailty. In PH models, the hazard ratios obtained are only the hazard ratios for t= 0 (Cleves et al 2010, Gutierrez 2002). Table 6.4 Log Normal (AFT) Duration Analysis results with Heterogeneity Removed Variables Coefficient Std. Err. 1=Meeting minutes recorded 0.251 *** 0.095 % CSB members with immediate family member in CSB - 0.231 0.143 Travel time to city in private car (minutes) 0.001 0.002 1=President's Age>30 0.420 *** 0.160 1=President is male 0.248 0.216 1=CSB has Horse, Mule or Ox 0.186 0.227 Interaction President male and CSB has animal - 0.148 0.238 0.020 * 0.012 1=seed produced on CSB member land - 0.088 0.087 # of Silos 0.059 ** 0.021 1=access to backpack sprayer 0.158 * 0.093 1=trained in CSB formation - 0.158 0.125 1=trained in seed marketing 0.258 0.191 Max Yield (yield potential) (qq/mz) 0.028 ** 0.005 1=trained in seed marketing*yield - 0.022 ** 0.010 # of beneficiaries/mz - 0.005 *** 0.002 % of beneficiaries repaying 2x1 lbs 0.251 ** 0.114 1=Parceled Bank - 0.109 0.107 1=Individual Bank - 0.290 * 0.164 Constant - 0.626 * 0.329 Shape Parameter 0.029 Shared Frailty Parameter 0.092 p - val=0.077 Log likelihood - 78.230 Restricted LL - 113.214 Frailty Distribution Gamma AIC 200.461 BIC 267.274 Note: ***, **, and * indicate significance levels of 1%, 5%, and 10% respectively 71 The results indicate that at an 8% confidence level we reject the null hypothesis that the shared heterogeneity parameter theta is equal to 0. Therefore, the shared frail ty model is the final and preferred model for our analysis of CSBs. The results from Table 6.4 reveal that an additional beneficiary for a CSB producing seed on a plot of one manzana, and all else held equal, decrease the time to failure by 0.5%. As repay ment compliance rates increase by 10%, time to failure is 2. 5 % longer. An additional silo delays time to failure by 5 . 9 % while CSBs with presidents aged 31 and older survive 42 % longer than CSBs with presidents aged 30 or younger. An additional year of C decreases failure time by 2%. Individual banks fail 29 % faster than classic CSBs. CSBs that meet, and record meeting minutes survive 2 5 % longer than CSBs that do not record meeting minutes. CSBs with access to a backpack spraye r survive 16 % longer than those without the device. 6.4.c Additional Potential Source of Bias Because continuation in the BTD program in each year is not only a CSB level decision but also decided by INTA regional or national staff or the extension worker (see Chapter 4 for additional details) one could argue that survival in Year 4 was an exogenous decision of the CSBs based on funding at best but also potentially political connections. To ensure that inclusion of year 4 is actual ly modeling BTD participation or survival and not a decision of INTA staff, 17 CSBs that had previously survived until Year 4 occurring in Year 3 reveal smal l changes in the coefficient values but minimal changes in significance level. Additionally, when the 4 th year was included by right - censoring the 17 CSBs in year 3, similar results were obtained. Table F.1 in 72 Appendix F provides the results of duration analysis of these two truncated data sets with heterogeneity effects removed and a comparison of the final model. 73 CHAPTER 7 : CONCLUSION S , POLICY IMPLICATIONS AND OPPORTUNITIES FOR FUTURE RES E ARCH Eight main results are found from the duration analysis of t he CSBs in Nicar agua, which were supported as part of the BTD project. Each has implications for future iterations of CSB projects but also larger scale seed enterprises focusing on the production and distribution of improved variety of bean seed. First of all, the analyses show that type of CSB does matter. Individual seed banks may provide a good contract farm ing option to NAR S and extension programs for meeting project driven seed requirements, but based on the evidence from this study , they do not provide a sustainable model for a community based seed production system . I ndividual CSBs , as implemented in the BTD project, failed 29 % faster than the classic CSBs . As against this , there was no statistically significant difference in the survival rate between the parceled and the classic CSB models. Furthermore, the number of membership had no effect on the sustainability of the CSBs. Recording the minutes of the meeting is found to be one of the determinants of sustainability and indicates the i mportance of formality of operations and documenting decisions within community groups in the longevity of community based seed organizations. Evaluating the hazard functions of parceled and classic CSBs ( similar to Figure 7.4 below ) indicate that the imp ortance of recording meeting minutes was the same for both types of CSBs. A second finding is that training on seed marketing is a determinant of CSB sustainability . As Witcombe et al. (2010) found in Nepal, marketing training was necessary to build deman d, establish partnerships and ultimately self - finance improved variety seed production. The results from the CSBs in Nicaragua suggest that in the first years of seed production, the impact of training is noted through an interaction term with yield. 74 The results indicate that a one unit (qq/mz) increase in yield for a CSB without marketing training and holding all else equal, will decrease failure time (and thus increase survival time) by 2.8%. The same one unit increase in yield for a CSB with marketing training and holding all else equal, will decrease failure time by 0.6%. For further discussion from earlier models of the implications of training, please see Appendix E. The implication is best seen graphically in the figures below. Figure 7.1: Hazard Functions of CSBs with Training at Four Seed Yield Levels Figures 7.1, 7.2 and 7.3 show the hazard rates at four levels of seed yield (0 qq/mz, 4 qq/mz, 8 qq/mz and 16qq/mz) for CSBs trained (Figure 7.1) in seed marketing and CSB s without seed marketi ng training (Figure 7.2). When the eight hazard curves are plotted together in Figure 7.3, it becomes clear that CSBs with training have lower hazard curves than CSBs with the same 75 level of yield but without training at levels of yield up to 15 qq/mz. Th e data from yield in year one of the BTD project reveals that 37% of the CSBs produced 8 qq/mz or less , 3 3% produced between 8 and 1 5 qq/mz, and 3 0% produced more than 1 5 qq/mz . Training reduces the variation in hazard functions and is clearly beneficial for CSBs with less than 15 qq/mz yield . Given the variation in seed production possible even with a package of technical inputs, seed marketing training is an important determ inant of success of CSBs. Figure 7.2: Hazard Functions of CSBs without Training at Four Seed Yield Levels 76 Figure 7.3: Hazard Functions of CSBs with and without Training at Four Seed Yield Levels Providing the variety demanded by the community had no effect on sustainability. These results are expected to be different in the long run as CSBs that do not offer varieties that are demanded will not be able to generate revenue due to lack of sales. In the context of the BTD project, however, the CSBs changed the seed varieties available to farmers and a subsequent change in demand is expected. Calculations by the author from the survey data of beneficiary farmers of the BTD project (Maredia et al. 2014) found that 85% of the CSB seed recipients had no t previously used the variety they received 13 . It is not surprising, therefore, that failure rates were the same for CSBs that did and did not supply a variety demanded in their communities. 13 Just less than 85% of farmers receiving INTA Rojo, 72% receiving INTA Matagalpa and 96% of farmers receiving INTA Sequia reported planting the variet 77 Third, cost recover y is often mentioned in the literature as nece ssary for sustainability. Two necessary business skills of CSBs offering seed purchase through loans to farmers is to judge the probability of repayment of each farmer at the moment of seed lending and successful reduction of seed loan delinquency through loan collection. Plotted for each type of CSB in Figure 7.4 are the hazard functions at 50% repayment and 95% repayment. The hazard function curves of all three CSB models shifted down proportionally with the increase in repayment rate from 50% to 95%. While the time to failure of parceled and classic CSB s where not statistically different as indicated in the results in Table 6.4, the difference is noted by comparing the two types of seed bank at the same repayment level. The classic CSBs have lower ha zard function curves than parceled CSBs. In fact, the parceled CSBs need ed a repayment rate of 95% to have a lower hazard function curve tha n the classic CSBs with a repayment rate of 50% holding all else equal . 78 Figure 7.4: Hazard Functions at two Repayment Rates by Type of CSB While increases in loan repayment rate have a positive relationship with sustainability, the number of clients (per unit area of seed production) had a negative effect on sustainability. The results indicate that CSBs fac e increased risk of failure when they lend seed to more farmers per unit of operation (i.e., when the intensity of seed operation is very high). The faster time - to - failure of CSBs with more clients per unit of land used for seed production indicates oper ational deficiencies and diminishing capacity of CSBs to manage a large number of clients. This result again emphasizes the importance of seed marketing training as well as training in business operations to increase the operational efficiencies of CSBs a nd their survival rates. Fourth, q uality of seed produced is important. The results of duration analysis of classic and parceled CSBs presented in Appendix D reveal that CSBs with positive feedback on the 79 quality of seed have a 43% lower hazard ratio than CSBs that received mixed or negative feedback. The results also indicate that land tenure has no effect on sustainability. One possible interpretation of this result is that the quality and suitability for seed production of the land are more important criteria for choosing land for seed production than choosing based on land owned by a CSB member or renting land. Unfortunately, CSBs could not provide reliable data on important seed quality indicators such as germination test results, humidity at time of storage and seed purity after post - harvest treatment (the percent of seeds free of lumps or divots, fungus, germinated, contrasting or seed of other varieties) 14 . Thus, we are not able to evaluate the importance of seed quality as measured by these indica tors. But as a good practice guideline, it is important that technicians and seed producers are trained on the importance of producing seed that meet these quality criteria in future promotion of decentralized seed production. Fifth, number of silos is an important determinant of CSB sustainability. On average CSBs had just over 2 silos but only 38% of CSBs received silos in the first year of the BTD project. Although transportation of silos to remote communities might present considerable challenges when roads are in poor condition or do not exist in mountain communities, silos are a necessary asset for seed producing organizations and efforts to coordinate their delivery are rewarded in the form of longer time to failure in CSBs. Sixth, CSBs with high co ncentrations of immediate family members (defined here as parents, children or siblings) had higher hazard ratios than CSBs with few or no immediate relatives in the Weibull proportional hazards models. Although no effect was found in the final heterogene ity removed model, e fforts should be made to form CSBs that are more representative 14 See Arraya and Fonseca (2007) for details of seed purity criteria in Central America. 80 of the community by including members from different families rather than more members from a few families. This will increase the stakeholder base within a CSB, which can increase the community support as the bank will be viewed as an equitable source of seed for the entire community . The impact of extended family members (cousins, uncles and aunts, nieces and nephews) on sustainability was non - significant in all models in dicating that in a village community where such relationships are likely to exist and perhaps unavoidable, this should not be a source of concern when forming a CSB. Seventh, experienced leaders are an important determinant of survival. As stated earlier, age is often associated with experience and for the CSBs in the BTD project with presidents older than 30 year of age, failure occurred 33% later in time than CSBs lead by younger presidents. When plotted to compare differences in types of CSBs ( similar to Figure 7.4), the effect of the promoters, indicating that youth leadership in seed production and implementation of new technologies can be as effective as their older peers and should not be discouraged. Female headed CSBs face additional challenges. Although no difference was found in the final model , there was evidence in earlier ( albeit less robust models ) th at female led CSBs have higher failure rate s and can benefit more from transportation assets. While this finding should not discourage policymakers implementing a project like BTD from including female leadership, it should be considered in the planning process, and efforts should be made to facilitate access to readily available and appropriate transportation and other assets. The identification of needed assets and their acquisition methods ( i.e., renting, cash purchase and financing the purchase) should form part of the initial training. 81 Finally, a comprehensive n eeds assessment by extensi on worker or supervision staff should precede the implementation of project supported CSBs. Evidence of the importance of liquidity and access to (or ownership of) asset s su ch as backpack sprayers and silos was found to be significant in the final model . Additionally, yield was a determinant of sustainability and thus liquidity to purchase inputs at the onset of disease or presence of pests to prevent crop failure should be a consideration to increase the viability and sustainability of a CSB. The PESA guide to C S Bs did not consider access to productive assets and financial services as important for choosing a community suitable for a CSB. The PESA guide lists the opposite a s a community level condition for implementing a CSB in a given community as it should have little or no technical and financial assistance from other organizations. While the goal of reaching the most needed communities is noble (and demand for quality s eed may be highest in such communities), project budgets will need to include purchase of such productive assets when they are not accessible in the community. In conclusion, the results confirm much of the literature regarding factors contributing to the sustainability of community based seed production including the importance of training (seed marketing and business skills), ownership of productive asset (especially silos), experience of leadership, cost recovery, quality and quantity of seed produced, a nd operational formality in the form of conducting meetings and documenting decisions made at meetings with minutes. The two communal CSBs, supported during the BTD project, provided a production and delivery model that lasted longer than individual banks . The policy implication of these results is that CSBs present a more sustainable dissemination channel of improved variety seed to farmers than small scale contract - based seed production by individual farmers. 82 There remain several opportunities for futur e research on CSBs. First of all, a follow up survey and interviews with the leaders of the 154 CSBs included in this study and the INTA technicians and regional/national staff involved in the promotion of CSBs in Nicaragua can provide many missing pieces of information to explain the factors that went into their decision to continue and/or discontinue a CSB beyond project support. Secondly, reliable production cost data has been difficult to obtain despite efforts by the author and others . A study of co mmunity based seed production costs similar to Katungi et al. (2011) that value all aspects of seed production of Apta seed (i.e., QDS) in Nicaragua is needed to obtain a clear picture of the benefits to the CSB member s and community as a whole. Finally, better knowledge of the determinants of purchase of replacement seed is needed to understand demand. While lack of access and affordability are often cited as the reason for low use of improved varieties, farmers that have technical training in bean grain production and increased resources from grain sales still have low rates of improved variety use in Nicaragua ( Sain 2011, Carter e t al. 2012). The literature is replete with studies that look at determinants of adoption of improved varieties (or decision to replace traditional/local varieties with new/modern varieties) (Feder et al. 1985, Mwangi et al. 2015). However , similar studies are needed to understand determinants of farmer behavior regarding replacement seeds post - adoption. Such information can help guide researchers, extension agents, policy makers and NGOs to better design sustainable seed production and distribution models in a developing country context such as Nicaragua. Finally, it is important to identify the li mitation to this study. Only the CSBs established in the first year of the BTD project that responded to the survey were considered for this study due to the data collection process. A more robust analysis would include the CS Bs that began in the second and third year of the BTD project as well . Additionally, this study ended tracking the 83 CSBs at the end of the BTD project in 2014. INTA employees insist that some CSBs continued to operate with or without external support afte r 2014. Information about all of the CSBs survival or status in years following the BTD project was not available to include in this study. Had this information be en readily available, it would provide valuable additional information regarding CSBs sust ainability beyond the years of the BTD project. Given these limitations, the results and conclusions of this study should be used in the context of survival of a CSB with in the BTD project timeframe only. 84 A PPENDI CES 85 APPENDIX A: English Version of the Survey INSTRUCTIONS: Meet with the people primarily in charge of the Community Seed Bank (CSB) leadership such as the president, vice president, secretary, treasurer and/or promoter. If at least four members are not available, ask to schedule a meeting when they will be available. Read the Consent Statement (page 2) and answer any questions group member have before continuing with the survey. Most questions will give number codes for the possible answers. Please clearly record the code on the line provided. Some questions require specific prices or quantities. Please refer to CSB records whenever possible to confirm the amounts that CSB members are reporting. Section G requests a list of seed bank members. If the secretary has a list of CSB members, it answering the information about each member. [If the bank only consists of a producer (usually cases like this only occur in the South Central region ) then the producer is the only person to meet with and answer questions personally. The CSB is the person, so if the question relates to the support received by the CSB, it is actually asking for the support received by the individual.] Thank you for particip ating in this survey. 86 INSTRUCTIONS : The consent statement should be read by an INTA staff member who will be mainly responsible to collect the information. Read the following Consent Statement to the CSB members that will be providing the information solicited through an interview and a record keeping booklet. Since the information will be collected over time through multiple group or one - on - one meetings, the consent should be obtained only at the first meeting when all of the key CSB members are prese nt. The USAID - funded Dry Grain Pulses CRSP at Michigan State University, in collaboration with INTA is conducting a study on documenting the experience gained in implementing the Community Seed Bank model of bean seed production in Nicaragua. As part of t his study, we are interviewing a few key members of all the CSBs in Nicaragua and systematically collecting information using a structured questionnaire and a record keeping booklet. The type of information we are collecting includes community setting, the membership of CSBs, its operation, and bean seed production practices, costs and outcomes. Some of the information will be collected by asking you some questions either as a group or individually and some will be collected by requesting you to keep a reco rd in the booklet provided. The completion of this study will require several short meetings or visits by me throughout the Primera 2012 season to collect and document data on seed production activities. Your participation in this study is voluntary . Your refusal to participate or to withdraw from the study carries no penalty or loss of any benefits . You are free to not answer any of the questions I will ask. However, I hope that you will answer all my questions and provide the information requested, as yo ur responses will help us better understand the CSB model and to derive lessons for improving it here in Nicaragua and elsewhere. Any information about individuals collected from you will be kept private and included in reports only in aggregate analysis without the Do you have any questions about this study? < Pause and respond to any questions raised, then continue with the following statement > uld like to start the interview. By answering my questions, you indicate your willingness to voluntarily participate in the study. During the course of this study, if you have any questions/concerns, you may contact Dr. Mywish Maredia, at Michigan State Un iversity by phone: 001 517 353 6602 or Aurelio Llano the local contact for information. Consent statement was read and discussed, and the CSB members agreed to be interviewed (check one) 1=Yes . 2=No 87 A. INTA Information A.1. INTA Region (write one)______ [1] Centro Norte [2] Centro Sur [3] Las Segovias [4] Pacifico Norte [5] Pacifico Sur A.2.a. INTA Technician Name in charge of this seed bank: _______________________ A.2.d. Was this INTA Technician the interviewer? ( circle one ) [1] Yes [2] No If answer to A.2.d. was YES, skip to question A.4. ____________________ A.3. A.3.d. Interviewer is (write one)_______ [1] INTA Staff [2] Seed Bank Promoter [99] Other (specify)__________________________ A.4. Date of the interview: Day_____ Month_______ ___ Year 2012 . B. General Information of Seed Bank B.1 Name of the Community Seed Bank (CSB) _______________________________ B.2. Name of Village/town of CSB operation ______________________ B.3. Name of Villages/towns where 2011 seed (loan) recipients live and approximate travel time: operates (If you do not know the name of the community or it is not possible to travel to the community walking, please include a note to explain the situation) a. Community Name _______________________ 1. Travel time: __________minutes walking b. Community Name _______________________ 2. Travel time: __________minutes walking c. Community Name _______________________ 3. Travel time: __________minutes walking d. Community Name _______________________ 4. Travel time: __________minutes walking e. Community Name _______________________ 5. Travel time: __________minutes walking B.4.a. Name of the Municipality where CSB is operating ___________________________ B.4.b. Name of Municipal Capital City _________________________ B.4.c. How many minuets must you walk from the community of CSB operation (B2) to the closest bus stop to board a bus traveling to the Municipal Capital City _______________minutes walking. ( If the bus stop is in the community, answer will be 0 minutes ) 88 B.4.d. Once you have boarded the bus at the closest bus stop, what is the travel time from closest bus stop to the co mmunity of CSB operation (B2) to the Municipal Capital City in bus in minutes : _______________minutes B.4.e. Frequency of bus route from (closest bus stop to) the community (B2) to Municipal Capital City B.4.e.1 Trips per day __________ B.4.e.2 Days p er week ____________ ( If the buses run Monday to Saturday, the answer will be 6. If the buses run Monday through Sunday, the answer will be 7 ) B.4.f. How many minuets must you walk from the community of CSB operation (B2) to the closest point that a car can get to your community? _________minutes walking. ( If a car can enter your community, the answer will be 0 ) B.4.g. Travel time from (the closest point that a car can get to) your community of CSB operation (B2) to Municipal Capital City in car in m inutes _______________minutes B.5.a. Name of the Department ___________________________ B.5.b. Name of Department Capital City _________________________ (If Department Capital City is the same as Municipal Capital City, go to B.6.) B.5.c. How many minu ets must you walk from the community of CSB operation (B2) to the closest bus stop to board a bus traveling to the Department Capital City _______________minutes walking. ( If the bus stop is in the community, answer will be 0 minutes ) B.5.d. Once you ha ve boarded the bus at the closest bus stop, what is the travel time from closest bus stop to the community of CSB operation (B2) to the Department Capital City in bus in minutes : _______________minutes B.5.e. Frequency of bus route from (closest bus stop to) the community (B2) to Department Capital City B.5.e.1 Trips per day __________ B.5.e.2 Days per week ____________( If the buses run Monday to Saturday, the answer will be 6. If the buses run Monday through Sunday , the answer will be 7 ) B.5.f. How many minuets must you walk from the community of CSB operation (B2) to the closest point that a car can get to your community? _________minutes walking. ( If a car can enter your community, the answer will be 0 ) B.5.g . Travel time from (the closest point that a car can get to) your community of CSB operation (B2) to Department Capital City in car in minutes _______________minutes B.6. Distance from community of CSB operation (B2) to closest paved or brick road _____ KM ( answer will be 0 if community boarders or is crossed by paved road) B.7.a. Name of city or community where CSB would sell harvested grain___________________ 89 B.7.b. What is the condition of the road from the community (B2) to the city mentioned above? ________ [1] Paved in good condition [2] Paved in poor condition [3] Some paved some unpaved [4] Unpaved in good condition [5] Unpaved in poor condition [6] Impa ssable at times during the rainy season (river crossings) B.8 In this village, what is the main bean producing season (write one)? ____ [1] Primera [2] Postrera [3] Apante 90 B.9. Which of the following services are available in the village (B2)? B9a B9b B9c Service ID Is this service currently available in your community/village? [1] YES [2] NO If NO, what is the distance to the closest available service center? KM Access to electricity 11 Access to water service (network) 12 Access to wells 13 Access to radio 14 Access to television network 15 Access to cell phone network 16 Access to telephones (landlines) 17 Access to health centers 18 Access to private bank services 19 Access to community/rural banks 20 Is there a primary school in this village? 21 Is there a secondary school in this village? 22 service office in this village? 23 Are there any NGOs providing agricultural - related services in this village? 24 B.10. Which of the following options best describes your CSB? _________ [1] The CSB is made up of several members of the community (partners or members) and bean seed is grown in an area of approximately one manzana (possibly only one field) with one promoter. The members of the CSB make decisions about which seed variety to use (beginning in 2012), what input to use, who should receive seed (loan), and how to secure repayment of grain (loan repayment). (If ans wered [1] continue with all sections) [2] The CSB is made up of several members of the community (partners or members) and bean seed is grown in on several lots with several promoters. The members of the CSB make decisions about which seed variety to u se (beginning in 2012), what input to use, who should receive seed (loan), and how to secure repayment of grain (loan repayment). (If answered [2] continue with all sections) [3] The CSB is made up of one individual who grows the seed and makes decisio ns, sometimes with the help of INTA, as to who should receive the seed (loan). (If answered [3], please skip to section E and answer the questions on a personal basis. The CSB is the person, so if the question asks, who supports the CSB it is actually as king who supports the individual) 91 C. Information about the Establishment of the Community Seed Bank C.1 When was the first President elected (CSB founded)? C1a. Month___________ C1b. Year___________ C.2 What organizations supported the seed bank when it was first established? [1] Yes [2] No C.2.a INTA ____ C.2.b FAO ____ C.2.c NGO ____ ( if yes, include name ) _____________________________________ C.2.d Other ____ ( if yes, describe support and include name )____________________ __________________________________________________________________ C.3 How many CSB members founded the seed bank and voted for the original president/secretary/promoter? ___________ members C.4 Of the number of fo unding members listed in C.3, how many were men and how many were women? C4a. Number of Men: _____________ C4b. Number of Women: ______________ C.5 During the first year after the CSB was established, indicate which of the following SEED was produc ed by the CSB [1] Yes [2] No C.5.a: Beans ______ C.5.b: Rice ______ C.5.c: Maize _____ C.5.d. Sorghum ______ C.5.e. Other (specify):________ D. Seed Bank Operations D.1 How often does the CSB have regular meetings per month? __________________ times per month D.2 Of the active members, on average how many attend regular meetings? _______________members D.3 Are minutes kept at the meetings? _______ [1] YES [ 2] NO ( if no, go to D5 ) D.4.a If meeting minutes are kept, is someone responsible for recording them? _____ [1] Yes [2] No D.4.b If yes, who is responsible for recording minutes? _______ [1] President [3] Treasurer [2] Secretary [99] Other 92 D.5 [1] Yes [2] No D.5.a Does the CSB have internal by - laws? ___________ D.5.b Is there a written lending policy? ___________ D.5.c Is there a production plan with projected planting area, date and costs?_________ D.5.d Is there a training plan that identifies training needs? ___________ D.5.e Does the CSB have a bank account to deposit cash? ___________ D.5.f Are there financial reco rds and reports available? ___________ ( If yes, please list which financial reports are available ) D.5.f.1 _____________________________________________________ D.5.f.2 _____________________________________________________ D.5.f.3 _____________________________________________________ D.6. How were the CSB members selected? ______________________________________________________________________________ ______________________________________________________________________________ ____ __________________________________________________________________________ _______________________________________________________________________ D.7 Can anyone become a member of this CSB? _____________ [1] Yes [2] No If No, to go s ection E D.8 If yes, to become a new member, must the person looking to join the CSB be: [1] Yes [2] No D.8.a A resident of the community? ______________ D.8.b A BEAN grower? ______________ D.8.c Be recommended by a current member? ______________ D.8.d Have seed production experience? ______________ D.8.e Be a relative of a current member? ______________ D.8.f Other ______________ ( if yes please describe ) _________________________________________ _________________________________ __________________________________________________________________________ __ E. Seed Bank Information (Current situation) E.1. What groups currently support the CSB? [1] Yes [2] No E.1.a INTA ____ E.1.b FAO ____ E.1.c NGO ____ ( if yes, include name ) _____________________________________ E.1.d Other ____ ( if yes, describe support and include name )____________________ __________________________________________________________________ E.2 Does the CS B have access to the following vehicles and are they available for transport of grain/inputs? (select one option) [1] Yes -- [2] No 93 E.2.a Pickup truck ________ E.2.b. Tractor ________ E.2.c Horse, mule or donkey ________ E.2.d Other (describe) ________ ( if yes please name and describe )____________ ________________________________________________________________________ E.3 Does the CSB currently have access to the following facilities? (select one option) [1] Yes - [2] No E.3.a Bodega for equipment _______ E.3.b Area for drying seed/grain _______ E.3.c Threshing machinery ________ E.3.d Automatic dryer ________ E.3.e Pump/backpack sprayer for applying inputs ________ E.3.f Sewing machine for sacks ________ E.4 a. How many silos for gra in or seed storage does the CSB currently have? ________ ( Include silos used by the CSB even if they were not acquired through the project USAID - MSU - INTA ) E.4.b. If the CSB has one or more silos, how many quintals of grain can be stored in ALL these silos? ________ QQ E.4.c Were any of the Silos donated to the CSB? ___________ [1] Yes [2] No ( Include silos used by the CSB even if they were not acquired through the project USAID - MSU - INTA ) E.5 Will this CSB continue in the project USAID - MSU - INTA in Primera 2012? _____ [1] Yes [2] No E.6 In what types of training/capacity building activities have the CSB members participated? [1] Yes [2] No E.6.a CSB organization and formation ___ ___ E.6.b Bean seed production (agronomics, pest management ______ E.6.c. Post - harvest treatment, handling and storage (threshing, drying, cleaning, packaging) ______ E.6.d. Seed bank management (financial, administrative, accounting) ______ E.6.e. Seed marketing ______ 94 F. Seed Production ONLY FROM USAID - MSU - INTA project F.1 How much planting material (registered BEAN seed) was received from INTA in 2011, how much area was planted with this seed and how much was produced and distributed? F101 F102 F103a F103b F104 F105 F106a F106b F107 F108 F109 F110 F111 F112 ID Variety Date Received Seed Received (qq) Area Planted (MZ) Date Harvested Total Production of Seed (qq) Total Distributed as Seed to Farmers (qq) Total Retained for Storage (qq) Total Sold (qq) Total Seed distributed among CSB members (qq) Other (qq) Day Month Day Month 1 2 3 ( The sum of F108, F109, F110, F111and F112 must equal F107) [1] Yes [2] No ( If no, go to F.3 ) F.2.a How much land was used? ______MZ F.2.b How was the promoter compensated?__________ [1] Was not compensated [2] Received equal amount of seed as other CSB members ( go to F.2.d) [3] Received cash (go to F.2.c) [4] Received more seed than other CSB member ( go to F.2.d) [5] Received other form of compensation (specify) _______________________ F.2.c ( If answer for F2b was [3] Received Cash , please answer:) What was the cash amount of compensation? Cordobas ____________ F.2.d ( If answer f , please answer:) How much seed did the promoter receive as his/her compensation? ___________ (lbs) 95 11? ______ [1] Yes [2] No ( If no, go to F.4 ) F.3.a How much land was used? ______MZ F.3.b How was the CSB member compensated?__________ [1] Was not compensated [2] Received equal amount of seed as other CSB members ( go to F.3.d) [3] Received cash (go to F.3.c) [4] Received more seed than other CSB member ( go to F.3.d) [5] Received other form of compensation (specify) _______________________ F.3.c ( If answer for F3b was [3] Received Cash , please answer:) What was the cash amount of compensation? Cordobas ____________ F.3.d ( If answer , please answer:) How much seed did the CSB member receive as his/her compensation? ___________ (lbs) F.4 Was bean seed grown on the rented land in Primera 2011? _____ [1] Yes [2] No ( If no, go to F.5 ) F.4.a How much land was rented? ______MZ F.4.b Was the land owner compensation paid in cash?_______ [1] Yes ( go toF4c ) [2] No ( go to F4d ) F.4.c ( If answer for F4b was Yes , please answer:) What was the cash amount of compensation? Cordobas ____________ F.4.d ( If answer for F4b was No, please answer:) How was the land owner compensated? (detail) _____________________ F.5 Did the CSB receive material inputs for seed production in Primera 2011 from the following organizations? [1] Yes [2] No F.5.a INTA ___________( if yes, list inputs ) ______________________________ F.5.b FAO ___________( if yes, list inputs ) ______________________________ F.5.c NGO ___________( if yes, name NGO and list inputs ) ____________ __________________________________________________________________ F.5.d Other ___________(describe) _______________________________ ___________________ _______________________________________________ F.6 Who provided labor for bean seed production in Primera 2011? [1] Yes [2] No F.6.a CSB Members ___________ F.6.b Hired labor ___________ F.6.c Community members ___________ 96 If the response was yes for F.6.a, please ask the following question: F.6.a.1: How were the CSB members who provided their labor compensated?__________ [1] Were not compensated (worked voluntarily) [2] Compensated in - kind (e.g., a share in seed production) [3] Compensated in cash [4] Received other form of compensation (specify)_________________ If the response was yes for F.6.c, please ask the following question: F.6.c.1: How were the community members who provided their labor compensated? _______ [1] Were not compensated (worked voluntarily) [2] Compensated in - kind (e.g., a share in seed production) [3] Compensated in cash [4] Received other form of compensation (specify)_________________ F.7 Quality of Bean Seed Produced, Germination Rate, Humidity and Observed Physical Purity F701 F702 F703 F704 F705a F705b F705c F706d F706e Name of Variety Germination Test result (% of seed germination) Humidity of Seed after drying (%) Pure Seed (% of acceptable seed from random 100 seeds selected) lumps/divots (Terrones in Spanish) (%) Germinated (Germinado in Spanish) (%) Distinct Characteristics (Contraste in Spanish) (%) Fungus (Hongo en Semilla in Spanish) (%) Other Varietyies (Semilla otras variedades) (%) F.8 Amount of Bean Seed distributed (loaned) in 2011 (and Apante 2012) F801 F802 F803 F804 F805a F805b F805c Season Total amount of seed distributed in pounds (lbs) Total number of farmers (persons) who received seed Total number of farmers (persons) who paid back 2 lbs of grain for every pound of seed received Total pounds of grain received for repayment (lbs) Value of Grain accepted for repayment (Cordobas/lb) Total value of non - grain repayments (Cordobas) Primera 2011 Postrera 2011 Apante 2012 97 F.9. How were the seed (loan) recipients chosen? ______________________________________________________________________________ ______________________________________________________________________________ ___________________________________________________________ ___________________ F.10 In Primera 2011, who made the decision on how to utilize the harvested seed from the CSB? __________ [1] General Assembly of CSB members [2] Board of Directors of CSB [3] INTA [99] Other (describe)__________________________________________ F.11 What is the most popular bean variety seed demanded in this community? ______ [1] INTA Rojo [2] Rojo de Seda [3] INTA Matagalpa [4] INTA Sequia (Fuerte Sequia) [99] Other (name)_____________ _______________ F.12 Was the CSB able to meet the demand in the community for the variety of seed it produced in Primera 2011? _________ [1] Yes [2] No F.13 Were other bean seed varieties that were not offered by the CSB requested by the com munity? _________ [1] Yes [2] No F.14 ( If answered Yes is F13) Please list the other bean seed varieties requested. a.___________________________ b.___________________________ c.___________________________ F.15 What feedback or comments did the CSB receive from the farmers who received seed? ______________________________________________________________________________ ______________________________________________________________________________ ______________________ ______________________________________________________ F.16.a Were labels attached to the bags of seed distributed? _______ [1] Yes [2] No ( If No, go to G.1 ) F.16.b Did the labels include the following: [1] Yes [2] No F.16.b.1 Name of seed variety _________ F.16.b.2 Germination rate _________ F.16.b.3 Weight of bag _________ F.16.b.4 Date of production _________ 98 G.1 Information about the Board of Directors (CSB members) a. Name b. Title/Designation [1] President [7] Vocal [2] Vice President [3] Secretary [8] Member [4] Treasurer [5] Promoter [6] Fiscal [99] Other c. Age [1] Less than 18 [2] 18 - 30 years [3] 31 - 50 years [4] 51 - 65 years [5] Older than 65 d. Gender [1] Male [2] Female e. Years serving as a member in the Seed bank [1] Less than 1 [2] 1 - 3 Years [3] More than 3 f. Education Level [1] No formal [2] Grade 1 - 5 [3] Finished 6 [4] Finished 9 [5] Finished 12 [6] More than 12 i. Name of the Village where this member lives g. Brother/Sister, Father/mother, or son/daughter in this seed bank? [1] Yes [2] No h. Uncle/Aunt, Grandparent, Grandchild or Cousin in this CSB? [1] Yes [2] No 1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. 13. 14. 99 a. Name b. Title/Designation [1] President [7] Vocal [2] Vice President [3] Secretary [8] Member [4] Treasurer [5] Promoter [6] Fiscal [99] Other c. Age [1] Less than 18 [2] 18 - 30 years [3] 31 - 50 years [4] 51 - 65 years [5] Older than 65 d. Gender [1] Male [2] Female e. Years serving as a member in the Seed bank [1] Less than 1 [2] 1 - 3 Years [3] More than 3 f. Education Level [1] No formal [2] Grade 1 - 5 [3] Finished 6 [4] Finished 9 [5] Finished 12 [6] More than 12 i. Name of the Village where this member lives g. Brother/Sister, Father/mother, or son/daughter in this seed bank? [1] Yes [2] No h. Uncle/Aunt, Grandparent, Grandchild or Cousin in this CSB? [1] Yes [2] No 15. 16. 17. 18. 19. 20. 21. 22. 23. 24. 25. 26. 27. 28. 100 APPENDIX B: Descriptive Statistics by INTA Administrative Region Table B.1: Summary statistics of variables used in the duration analysis: Difference across INTA Administrative Regions INTA Administrative Region Variables Centro Norte Centro Sur Las Segovias Pacifico N. Pacifico S. TOTAL # of Observations 37 40 29 28 20 154 Mean Years participation in BTD 1.68 a 1.68 a 2.10 a 2.07 a 3.15 2.02 Failure after Year (% Yes ) 1 48.65 ~ 55.00 ~ 41.38 ~ 35.71 ~ 10.00 ~ 2 40.54 ~ 32.50 ~ 20.69 ~ 21.43 ~ 0.00 ~ 3 5.41 ~ 2.50 ~ 24.14 ~ 42.86 ~ 55.00 ~ 4 5.41 ~ 10.00 ~ 13.79 ~ 0.00 ~ 35.00 ~ CSB Organizational Structure # Years operation to beginning of BTD 0.00 a 0.00 a 0.62 b 0.00 a 0.70 b 0.21 # of CSB members 7.70 a 1.00 15.31 4.00 9.65 a 6.97 CSB or Community had voice in use of seed produced ( % Yes ) 86.49 b 57.50 a 75.86 ab 78.57 ab 70.00 ab 73.38 Number of monthly meetings 1.71 b 0.40 1.14 a 1.14 a 1.20 ab 1.09 % of CSB members attending meetings 78.77 ab 25.00 66.26 a 96.43 b 88.75 ab 66.96 Meeting Minutes Recorded ( % Yes) 32.43 a 22.50 a 82.76 b 50.00 a 50.00 ab 44.81 CSB has written bylaws ( % Yes) 21.62 b 10.00 b 82.76 a 85.71 a 75.00 a 48.70 % of CSB members with Immediate family members in CSB 51.64 b 0.00 33.86 a 32.14 a 33.90 ab 29.03 % of CSB members with Extended family members in CSB 24.97 b 0.00 c 20.69 ab 7.14 ac 23.13 ab 14.20 Community Characteristics Distance to paved road (KM) 21.48 9.75 a 10.86 a 9.91 a 6.58 a 12.40 Travel time to Municipal Seat in private car (minutes) 34.19 a 23.38 a 27.93 a 22.61 a 20.85 a 26.36 PCA of community Level Development - 0.57 a - 0.40 a - 0.03 a 1.14 b 0.29 ab 0.00 Leadership Characteristics President older than 30 ( % Yes) 97.30 a 97.50 a 96.55 a 75.00 b 90.00 ab 92.21 President's Gender ( % Male) 81.08 a 80.00 a 89.66 a 100.00 a 100.00 a 88.31 Promoter's Gender (% Male) 81.08 a 80.00 a 96.55 a 89.29 a 90.00 a 86.36 President is Promoter ( % Yes) 75.68 a 100.00 72.41 a 14.29 55.00 a 67.53 101 Variables Centro Norte Centro Sur Las Segovias Pacifico N. Pacifico S. TOTAL Land Land used for seed production (MZ) 1.06 a 0.99 a 1.13 a 0.68 1.85 1.09 Seed produced on CSB member land ( % Yes) 78.38 a 7.50 82.76 a 71.43 a 85.00 a 60.39 Seed produced on rented land ( % Yes) 18.92 a 25.00 a 27.59 a 3.57 a 30.00 a 20.78 Labor CSB members provided labor ( % Yes) 94.59 a 75.00 a 79.31 a 50.00 100.00 a 79.22 Hired workers provided labor ( % Yes) 8.11 a 25.00 ab 17.24 a 50.00 b 25.00 ab 24.03 Assets/facilities # of Silos 1.35 bc 1.10 b 3.72 a 2.93 a 2.50 ac 2.17 CSB received Silo from BTD Project ( % Yes) 16.22 b 10.00 b 58.62 a 75.00 a 55.00 a 38.31 CSB has access to backpack sprayer ( % Yes) 94.59 b 10.00 44.83 a 42.86 a 100.00 b 54.55 CSB has access to seed/grain drying area ( % Yes) 2.70 a 12.50 a 13.79 a 0.00 a 10.00 a 7.79 CSB has transportation assets (pickup/mule/ horse/ox) ( % Yes) 86.49 ac 95.00 a 48.28 b 96.43 a 65.00 bc 80.52 CSB has animal transportation assets (mule/ horse/ox) ( % Yes) 86.49 ab 77.50 ab 48.28 c 96.43 b 65.00 ac 75.97 Human Capital CSB trained in Formation and Organization ( % Yes) 89.19 a 27.50 89.66 a 100.00 a 80.00 a 74.03 CSB trained in Seed Marketing ( % Yes) 10.81 a 32.50 ab 20.69 a 50.00 b 15.00 a 25.97 CSB trained in Seed Production ( % Yes) 83.78 a 90.00 a 93.10 a 100.00 a 100.00 a 92.21 Output and Efficiency Indicators Yield (QQ/MZ) 12.29 ab 16.42 b 11.20 a 11.68 ab 7.19 a 12.38 Potential Yield (QQ/MZ) 14.73 a 17.20 a 14.15 a 11.90 a 24.25 15.98 % of production distributed to beneficiaries 21.59 45.88 a 47.20 ab 80.90 c 67.35 bc 49.45 # of Beneficiaries per MZ seed production 17.97 a 38.40 b 18.07 a 39.58 b 29.74 ab 28.75 % of Beneficiaries Fully Repaying (2lb per 1lb) 1.25 66.78 b 42.62 a 41.69 a 55.25 ab 40.43 Recovery rate (repaid/seed distributed) 0.07 1.22 a 0.80 a 0.85 a 0.98 a 0.76 CSB supplied variety demanded ( % Yes) 27.03 a 70.00 b 34.48 a 50.00 ab 45.00 ab 46.10 Notes: INTA Administrative Regions that share a letter are not significantly different at the 10% level ~ indicates that a significance tests across groups has not be en performed for these variable 102 APPENDIX C: Interaction Terms in Three Distributional Forms and Interpretation Table C.1 Duration Analysis with Interaction Variable COX (PH) Weibull (PH) Log - Normal (AFT) Variables Hazard Ratio Std. Err. Hazard Ratio Std. Err. Coef. Std. Err. 1=Community had a voice 1.687 ** 0.413 2.054 *** 0.517 1=Meeting minutes recorded 0.392 *** 0.100 0.304 *** 0.082 0.337 *** 0.100 % CSB members with immediate family member in CSB 1.005 0.003 1.008 ** 0.004 - 0.003 ** 0.001 Travel time to city in private car (minutes) 0.990 ** 0.005 0.986 *** 0.005 0.004 ** 0.002 1=President's Age>30 0.322 *** 0.125 0.234 *** 0.094 0.402 *** 0.151 1=President is male 0.267 ** 0.161 0.152 *** 0.095 0.241 0.226 1=CSB has Horse, Mule or Ox 0.316 * 0.203 0.166 *** 0.111 0.287 0.240 1=President male and CSB has animal 3.158 * 2.102 5.405 ** 3.723 - 0.263 0.253 1=seed produced on CSB member land 1.424 0.342 2.017 ** 0.567 - 0.177 * 0.096 1=land rented 1.526 0.449 1=labor from CSB members 1.441 0.394 1.533 0.428 # of Silos 0.873 ** 0.050 0.843 *** 0.050 0.062 *** 0.023 1=access to backpack sprayer 0.143 0.102 1=trained in CSB formation 1.766 0.666 2.025 * 0.805 - 0.212 0.131 1=trained in seed marketing 0.571 0.257 0.452 * 0.212 0.256 0.163 Year 1 Yield (qq/mz) 0.940 *** 0.015 0.926 *** 0.016 0.024 *** 0.006 1=trained in seed marketing*yield 1.078 ** 0.036 1.092 *** 0.037 - 0.022 * 0.011 beneficiaries/mz 1.013 ** 0.007 1.017 ** 0.007 - 0.006 *** 0.002 # of beneficiaries repaying 2x1 lbs 0.347 *** 0.127 0.305 *** 0.114 0.300 ** 0.126 1=Supplied variety demanded in community 0.744 0.150 0.748 0.153 Centro Sur 1.636 1.193 1.940 1.427 - 0.097 0.273 Las Segovias 1.654 0.684 1.874 0.804 - 0.119 0.168 Pacifico Norte 0.649 0.308 0.632 0.310 0.317 * 0.187 Pacifico Sur 0.309 *** 0.136 0.248 *** 0.113 0.615 *** 0.163 1=Parceled Bank 1.663 0.530 1.682 0.542 - 0.107 0.119 1=Individual Bank 2.074 1.348 3.174 * 2.102 - 0.221 0.229 Constant 1.853 1.491 - 0.420 0.317 Parameter p=2.944 0.192 0.023 Log likelihood - 589.330 - 77.620 - 78.592 Restricted LL - 625.129 - 124.207 - 118.378 AIC 1226.660 209.239 205.1839 BIC 1299.547 291.237 278.0708 Variable Significant (at least 10%) 13 18 12 Note: ***, **, and * indicate significance levels of 1%, 5%, and 10% respectively 103 Only in the Weibull model are all three hazard rates included in each of the two interactions are statistically significant. The Weibull model indicates that for two CSBs without access to the animals used for transportation, the male headed CSB has a (1 - 0.152=0.848) 84.8% lower hazard ratio, or failure rate, than the CSB with a female president. Also from the Weibull results, a CSB with a female president and access to animals used for transportation has a (1 - 0.166=0.834) 83.4% lower failure rate than a female headed CSB without access to animals used for transportation. To calculate the difference between male headed CSBs with access to animals and a male headed CSB without access as well as the difference between male and female headed CSBs - 1.884), access to animals (ln[0.166]= - 1. 796) and the interaction term (ln[5.405]=1.687) are obtained. STATA also provides the coefficients by using the nohr command to request that the default display with hazard ratios is not displayed. Comparing two CSBs with animals, [exp( - 1.796+1.687)=0.897 ] the hazard ratio is (1 - 0.897=0.103) 10.3% lower for CSBs with male presidents than those with female presidents. Comparing two CSBs with male presidents, [exp( - 1.884+1.687)=0.821] the hazard ratio is (1 - 0.821=0.179) 17.9% lower for CSBs with animals th an those without. Still using the Weibull model results, a 1 qq/mz increase in yield for a CSB without marketing training, holding all else equal, will lower the failure rate by (1 - 0.926=0.074) 7.4%. After finding the coefficient for yield, (ln[0.926]= - 0. 077) the effect of a 5 qq/mz increase in yield is calculated [exp( - 0.077*5)=0.68] to lower the hazard rate by (1 - 0.68=0.32) 32%, all else equal. For CSBs that experienced crop failure in year 1, (yield was zero) the CSBs with seed marketing training had a (1 - 0.452=0.548) 54.8% lower hazard rate. 104 After calculating the coefficient s for the interaction term (ln[1.092]=0.088), the effect of a 1 mz/qq increase in yield [exp( - 0.077+0.088)=1.011] will increase the failure rate by 1.1% and a 5 qq/mz increase in yie ld will increase the failure rate by 5.7% holding all else equal. These results, without considering their effect at different levels of yield, appear to suggest that training increases failure rates. However, graphical analysis is useful to further expl ain the effect of marketing training. Figure C.1: CSB Hazard Curves with Training Figure C .1 provides the hazard function curves predicted from the Weibull model for CSBs that received training at four levels of yield in year 1, 0 qq/mz (crop failure), 2 qq/mz, 8qq/mz and 16 qq/mz. The curves of the hazard functions are slightly higher at each level as the results above suggest. 105 Figure C.2: CSB Hazard Curves w ithout Training Figure C .2 provides the hazard function curves predicted from the Weibull model for CSBs that did not receive training at the same four levels of yield in year 1. Here, the hazard function curves decrease as yield increases. When the eight total curves from Figure C .1 and Figure C .2 are displayed together, Figure C .3 shows that the curves of the hazard functions for CSBs with training had little variation while the curve of the hazard function for the CSBs without training varied greatly. More importantly, for CSB s with less than 8qq/mz seed production in year 1, the hazard rates were lower for CSBs that received seed marketing training than the CSBs without the training. The data reveals that 37% of the CSBs produced less than 8 qq/mz, 23% produced between 8 and 12 qq/mz, and 40% produced more than 12 qq/mz in year one of the BTD project. Training reduces the variation in hazard functions and is clearly 106 beneficial for CSBs with less than 8 qq/mz production and marginally beneficial for CSBs with production betwee n 8 and 12 qq/mz. Figure C.3: CSB Hazard Curves w ith and w ithout Training The data from the Weibull model might lead the reader to conclude that training has a negative effect on CSBs that produced more than 12 qq/mz . From the Log normal AFT model, it is clear that even with training, an increase in yield by as little as 1 qq/mz delays CSB failure by (0.024 - 0.022=0.002) 0.2%. Thus the models suggest that seed marketing training is able to counteract the effect of va riation in first year yields on the hazard of CSB failure. While the interpretation of the results regarding CSB seed marketing training confirms what is found in the literature, an alternative explanation must be noted. Given the nature of the BTD projec t, survival did not entirely depend on meeting a need (demand) for seed through an economically profitably local enterprise. Higher survival rates can also be due to extension 107 interaction of the trained CSBs despite poor yields in year 1. The friendship or trust developed through the training process might supersede yield (or produc tivity) consideration if the extension worker is asked to make a decision regarding which CSB to exclude from the BTD project in subsequent years. Additionally, the decision to invest the time in training a CSB might be made by a preexisting bias of the e xtension worker in favor of one CSB over another. 108 APPENDIX D: Differentiation of Types of CSBs: With and Without Individual CSBs Table D.1 Weibull Model Duration Analysis without Individual Seed Banks With Individual Banks (Same as Table C.1) Without Individual Banks Variables Haz. Ratio Std. Err. Haz. Ratio Std. Err. 1=Community had a voice 2.054 *** 0.517 1.610 0.558 1=Meeting minutes recorded 0.304 *** 0.082 0.278 *** 0.090 % CSB members with immediate family member in CSB 1.008 ** 0.004 2.439 ** 0.899 Travel time to city in private car (minutes ) 0.986 *** 0.005 0.989 ** 0.005 1=President's Age>30 0.234 *** 0.094 0.224 *** 0.104 1=President is male 0.152 *** 0.095 0.279 0.277 1=CSB has Horse, Mule or Ox 0.166 *** 0.111 0.214 0.222 Interaction: male President and CSB has animal 5.405 ** 3.723 2.542 2.646 2.017 ** 0.567 2.098 ** 0.788 1=Land rented 1.526 0.449 1.433 0.506 1=Labor provided by CSB members 1.533 0.428 2.516 ** 1.006 # of Silos 0.843 *** 0.050 0.828 *** 0.053 1=Trained in CSB formation 2.025 * 0.805 2.096 1.002 1=Trained in seed marketing 0.452 * 0.212 0.483 0.275 Year 1 Yield (qq/mz) 0.926 *** 0.016 0.933 *** 0.020 1=Trained in seed marketing*yield 1.092 *** 0.037 1.073 * 0.046 Beneficiaries/mz 1.017 ** 0.007 1.026 *** 0.008 # of beneficiaries repaying 2x1 lbs 0.305 *** 0.114 0.121 *** 0.058 1=Supplied variety demanded in community 0.748 0.153 0.963 0.220 Centro Sur 1.940 1.427 5.469 6.052 Las Segovias 1.874 0.804 2.626 ** 1.289 Pacifico Norte 0.632 0.310 0.948 0.544 Pacifico Sur 0.248 *** 0.113 0.292 ** 0.144 1=Parceled Bank 1.682 0.542 1.738 0.622 1=Individual Bank 3.174 * 2.102 Constant 1.853 1.491 0.631 0.715 Parameter p=2.944 0.192 p=3.309 0.244 Log likelihood - 77.620 - 47.441 Restricted LL - 124.207 - 88.696 AIC 209.239 146.881 BIC 291.237 219.572 Variable Significant (at least 10%) 18 13 Note: ***, **, and * indicate significance levels of 1%, 5%, and 10% respectively 109 Table D.1 displays the results that show that the sign on the hazard ratios did not change. As a reminder, hazard ratios over 1 indicate a negative relationship between the independent variable and survival, while hazard ratios below 1 indicate a positive relationship betw een the variable and survival. The variables that are no longer significant, beside the individual bank dummy variable, are seed marketing training (although its interaction with yeild remains gender, access to animal transportation, and the interaction of presidents gender with animal transportation. Two variables that were not significant with the full set of observations but are now significant after excluding the individual seed banks are the regional dummy for Las Segovias and the indicator for whether CSB members provided the labor during seed production. CSBs in the Las Segovias region have a significantly higher failure rate than the Centro Norte region an d use of CSB member labor increases the failure rate of CSBs at a 5% confidence level (Table D1 ). Since the parceled and classic CSBs were directly involved in the seed distribution process (unlike the Individual banks in the Centro Sur region where INTA d istributed the seed) the number of communities reached (market size) and the feedback received from clients or seed recipients (quality) are available . Table D.2 gives the results of the Weibull model without Individual CSBs by adding these two variables. 110 Table D.2 Weibull Model Duration Analysis With Quality and Network Variables Variables Haz. Ratio Std. Err. 1=Community had a voice 1.385 0.510 1=Meeting minutes recorded 0.364 *** 0.105 % CSB members with immediate family member in CSB 2.092 * 0.858 Travel time to city in private car (minutes) 0.987 ** 0.006 1=President's Age>30 0.172 *** 0.084 1=President is male 0.618 0.215 1=seed produced on CSB member land 1.623 * 0.457 1=labor from CSB members 3.469 *** 1.565 # of Silos 0.851 ** 0.056 1=CSB has Horse, Mule or Ox 0.416 *** 0.132 1=trained in CSB formation 1.547 0.706 1=trained in seed marketing 1.085 0.394 Year 1 Yield (qq/mz) 0.947 *** 0.018 beneficiaries/mz 1.025 *** 0.009 # of beneficiaries repaying 2x1 lbs 0.102 *** 0.048 1=Seed distributed to 0 Communities (Crop Failure) 0.467 0.338 1=Seed distributed to 2 Communities 0.558 * 0.198 1=Seed distributed to 3 Communities 0.789 0.386 1=Seed distributed to 4 Communities 1.743 0.701 1=Seed distributed to 5 Communities 0.602 0.230 1=Positive feedback from Beneficiaries (Clients) 0.570 * 0.166 Centro Sur 7.039 ** 6.641 Las Segovias 1.506 0.757 Pacifico Norte 0.753 0.443 Pacifico Sur 0.233 *** 0.123 1=Parcelled Bank 1.672 0.551 Constant 1.001 0.880 Parameter p=3.493 0.262 Log likelihood - 42.24 Restricted LL - 88.696 AIC 140.479 BIC 218.761 Variable Significant (at least 10%) 15 Note: ***, **, and * indicate significance levels of 1%, 5%, and 10% respectively Banks that distributed seed to 2 communities had a (1 - 0.558=0.442) 44.2% lower hazard ratio than CSBs that only distributed to one community (Table D.2). There was no statistically 111 significant difference between CSBs that reached more than 2 communities w ith seed compared to CSBs that focused only on one community. The literature indicates that CSBs will be more sustainable if they have a larger market and thus have more consistent annual demand for seed. Table 5.1 indicates that less than half of the CS Bs offered one of the two most popular seed varieties in their communities. While it is possible that the CSB choose to lend to farmers beyond their community of operations, the decision may have been due to the lack of demand for the variety produced in year 1. The results also confirm that quality of seed produced has a positive relationship with sustainability. Holding all else equal, CSBs that received positive feedback from their clients had a (1 - 0.57=0.43) 43% lower failure rate than CSBs that did not receive positive feedback. The positive feedback is also an indicator of satisfaction with the variety, as positive attributes of the seed quality are affected both by the seed produced and attributes associated with the seed variety. The seed mi ght have been free of disease from good agronomical seed production process but also from varietal resistance to the disease. 112 APPENDIX E: Three Options for Removing Heterogeneity Effects One of the other two options is to obtain Huber/White/sandwich esti mators of variance. By specifying this approach, the data is no longer treated as 154 independent observations across freedom in this model, no Wald test is ob tained and the log pseudo - likelihood is the highest compared to the log likelihood ratios of the other two models with heterogeneity removed. The information criterion AIC and BIC, however are the lowest compared to the other two models. A less extreme ap proach is to allow the baseline survivor function of the AFT model (or baseline hazard function for the HP model) to vary by region. This is called a stratified model. One advantage of this model is that we also allow time to accelerate or decelerate with respect to region in addition to the differing baseline survivor functions and as such obtain adjusted coefficients for each region as we had when only controlling for fixed effects. Although the model does produce the best log likelihood and a better AI C than the shared frailty model, the BIC is the worst of the heterogeneity adjusted models. The Huber/White/sandwich estimators of variance, the stratified model and the shared frailty model for removing regional heterogeneity are compared in Table E.1. T he interaction of transportation and gender of president are no longer significant and are dropped in Table E.1 . 113 Table E.1 Log Normal Duration Analysis with Heterogeneity Removed STRATIFIED BY REGION FRAILTY BY REGION VCE CLUSTERS BY REGION Variables Coef. Std. Err. Coef. Std. Err. Coef. Std. Err. 1=Meeting minutes recorded 0.369 *** 0.091 0.295 *** 0.098 0.276 0.202 % CSB members with immediate family member in CSB - 0.325 ** 0.128 - 0.315 ** 0.151 - 0.383 *** 0.125 Travel time to city in private car (minutes) 0.004 ** 0.002 0.003 * 0.002 0.003 ** 0.001 1=President's Age>30 0.326 ** 0.131 0.258 0.165 0.167 0.103 1=seed produced on CSB member land - 0.115 0.093 - 0.100 0.094 - 0.076 0.108 # of Silos 0.055 ** 0.023 0.058 ** 0.023 0.050 *** 0.018 1=access to backpack sprayer 0.139 0.109 0.219 ** 0.102 0.257 * 0.144 1=CSB has Truck, Horse, Mule or Ox 0.120 0.090 0.101 0.100 0.103 0.087 1=trained in CSB formation - 0.288 ** 0.123 - 0.217 0.142 - 0.284 0.188 1=trained in seed marketing 0.251 0.153 0.314 * 0.165 0.310 *** 0.065 Year 1 Yield (qq/mz) 0.022 *** 0.006 0.017 *** 0.007 0.012 ** 0.005 1=trained in seed marketing*yield - 0.023 ** 0.011 - 0.030 *** 0.011 - 0.031 *** 0.004 beneficiaries/mz - 0.006 *** 0.002 - 0.003 0.002 - 0.001 0.002 % of beneficiaries repaying 2x1 lbs 0.392 *** 0.124 0.343 *** 0.130 0.415 ** 0.198 1=Parceled Bank - 0.079 0.100 - 0.082 0.117 - 0.019 0.102 1=Individual Bank - 0.267 0.234 - 0.455 ** 0.184 - 0.522 * 0.284 Centro Sur dummy - 0.159 0.275 Las Segovias dummy - 0.143 0.168 Pacifico Norte dummy 0.250 0.174 Pacifico Sur dummy 0.548 *** 0.140 Constant - 0.115 0.234 - 0.034 0.258 0.114 0.315 Shape Parameters 0.031 0.010 Centro Sur Shape Parameter 0.148 0.175 Las Segovias Shape Parameter 0.113 0.203 Pacifico Norte Shape Parameter 0.075 0.196 Pacifico Sur Shape Parameter - 0.447 ** 0.225 114 Table E.1 STRATIFIED BY REGION FRAILTY BY REGION VCE CLUSTERS BY REGION Constant Shape Parameter - 0.943 *** 0.129 Shared Frailty Parameter 0.095 1.95 0.081 Log likelihood - 75.684 - 91.156 - 92.133 Restricted LL - 116.149 - 113.214 - 118.378 D egree of F reedom 26 19 4 AIC 203.368 220.311 192.266 BIC 282.329 278.014 204.414 Variable Significant (at least 10%) 10 (no t including region variables ) 10 9 Note: ***, **, and * indicate significance levels of 1%, 5%, and 10% respectively 115 APPEND I X F. Results of Duration Analysis Accounting for Fourth Year Table F.1 Comparison of Results Without Fourth Year and Right Censored Data With Year 4 Three Years Three Years w/ Censored 4th Year Variables Coef. Std. Err. Coef. Std. Err. Coef. Std. Err. 1=Meeting minutes recorded 0.300 *** 0.098 0.264 *** 0.091 0.303 *** 0.099 % CSB members with immediate family member in CSB - 0.318 ** 0.151 - 0.333 ** 0.140 - 0.296 * 0.152 Travel time to city in private car (minutes) 0.003 * 0.002 0.003 * 0.002 0.003 * 0.002 1=President's Age>30 0.284 * 0.167 0.266 * 0.156 0.325 * 0.169 1=President is male 0.196 0.234 0.173 0.212 0.190 0.229 1=CSB has Horse, Mule or Ox 0.208 0.249 0.214 0.226 0.225 0.243 Interaction President male and CSB has animal - 0.151 0.261 - 0.148 0.237 - 0.162 0.256 1=seed produced on CSB member land - 0.112 0.094 - 0.122 0.087 - 0.124 0.095 # of Silos 0.056 ** 0.023 0.052 ** 0.021 0.058 ** 0.023 1=access to backpack sprayer 0.221 ** 0.102 0.215 ** 0.093 0.206 ** 0.103 1=trained in CSB formation - 0.216 0.142 - 0.195 0.129 - 0.195 0.145 1=trained in seed marketing 0.300 * 0.167 0.275 * 0.153 0.300 * 0.170 Year 1 Yield (qq/mz) 0.017 *** 0.007 0.014 ** 0.006 0.019 *** 0.007 1=trained in seed marketing*yield - 0.028 *** 0.011 - 0.023 ** 0.010 - 0.027 ** 0.011 beneficiaries/mz - 0.003 0.002 - 0.003 0.002 - 0.004 0.002 % of beneficiaries repaying 2x1 lbs 0.326 ** 0.131 0.244 ** 0.119 0.294 ** 0.136 1=Parceled Bank - 0.066 0.118 - 0.036 0.110 - 0.083 0.120 1=Individual Bank - 0.421 ** 0.184 - 0.373 ** 0.172 - 0.381 ** 0.191 Constant - 0.194 0.327 - 0.133 0.300 - 0.253 0.329 Shape Parameter 0.030 0.028 0.035 Shared Frailty Parameter 0.098 0.102 0.146 p - val=0.082 p - val=0.094 p - val=0.020 Log likelihood - 90.947 - 77.465 - 96.065 Restricted LL - 113.21 - 99.924 - 117.722 116 With Year 4 Three Years Three Years w/ Censored 4 th Year Variables Coef. Std. Err. Coef. Std. Err. Coef. Std. Err. Frailty Distribution Gamma Gamma Gamma AIC 223.894 196.930 234.130 BIC 287.670 260.706 297.906 Variable Significant (at least 10%) 11 11 11 Note: ***, **, and * indicate significance levels of 1%, 5%, and 10% respectively 117 W ORKS CITED 118 WORKS CITED breeding improve seed provision to small - Euphytica , 153 (3), 363 - 372. Almekinders - 5: Eschborn). importance for an improved Euphytica , 78 (3), 207 - 216. Poscos e http://coin.fao.org/coin - static/cms/media/18/13838619740090/protocolo_para_el_manejo_poscosecha_de_la_se milla_de_frijol_web.pdf Accessed December 4, 2015. Araya V., R., & Fonsec Estación Experimental Fabio Baudrit Moreno. http://www.mag.go.cr/acerca_del_mag/p rogramas/pitta - frijol - Protocolo - semillas.pdf Accessed December 4, 2015. Bänziger, M., Sentimela, P.S., & Mwala, M. - based seed Successful Community - Based Seed Production Strategies. Ed. P. S. Sentimela . Mexico, D.F. : CIMMYT . Bentley, J. W., Van Mele, P. A. U. L., & Reece, J. D. 2011. African Seed Enterprises: Sowing the Seeds of Food Security International, Wallingford, UK. Berdegué, J. A., & Fuentealba, R. 2011. Latin America: The state of smallholders in agriculture. In IFAD Conference on New Directions for Smallholder Agriculture (Vol. 24, p. 25). Disea Final R eport . Camagnani, M. 2008. La agricultura familiar en América Latina. Problemas del Desarrollo: Revista Latinoamericana de Economía, Vol. 39, núm. 153, abril - junio 2008 Carter - Farm Productivity: Impact CIAT, FAO, MAF - GoSS, AAH - I, ACTED System Security Assessment, Southe December 2010. http://www.fao.org/docrep/014/ba0032e/ba0032e00.pdf Accessed December 4, 2015. 119 Developing Countries: Institute (UK). No. 65, 114 pp. bean research in Eastern and Centr al Africa, 1992 - Occasional publication series; no. 21 (CIAT) . David, S., Kasozi, S., & Wortmann, C. (1997). An investigation of alternative bean seed marketing channels in Uganda. CIAT. Agriculture and Human Values 21 (4): 387 97. Journal of Sustainable Agriculture , 2 1 (2), 5 - 20. developing countries: A survey Economic development and cultural change , 255 - 298. Greene, W. 2012. Econometric Analysis, Boston: Pearson. Hallensleben Influencing the Success of Failure of Micro - St. Gallen, proved Seed Production and Supply http://oar.icrisat.org/200/1/324_2010_IB_87_SS_Bulletin_VSB.pdf Accessed December 4, 2015. INE. 2008. Estadística Honduras. Tegucigalpa. 2008 http://www.ine.gob.hn/Documentos/EAN%202007 - 2008/TENENCIA.pdf Accessed August 22, 2015. INIDE. 2001. El Tercer Censo Nacional Agropecuario. Managua, Nicaragua. http://www.inide.gob.ni/cenagro/perfiles.htm Access ed August 20, 2015. INIDE. 2012. Informe Final IV Censo Nacional Agropecuario. http://www.inide.gob.ni/Cenagro/INFIVCENAGRO/informefinal.html . Accessed November 12, 2015. FAO - Metodológica Para la Organización de Bancos Comunitarios de Serie: Asistencia Técnica. Guías para Extensionistas Agropecuarios. Rome: Food and Agriculture Organization of the United Nations . 120 Katungi, E., Wozemba, D., & Rubyogo, J. C. 2011. cost benefit analysis of farmer based seed African Crop Science Journal , 19 (4), 409 - 415. of the Dynamics of Varietal Turn Washington D.C. 32 p. - based, non - formal Eastern and Sou thern Africa : Proceedings of a Working Group Meeting, Kampala, Uganda, 10 - 13 October, 1994. Ed. S. David. Network on Bean Research in Africa, Workshop Series No. 32. ClAT, Kampala, Uganda. Resources Institute of the University of Greenwich. Chatham, United Kingdom. 47 p. Disasters , 26 (4) , 343 - 355. Africa: A Conceptual Framework for Creating Coherence Between Practices, Programs, 59. Louwaars, N. Wageningen , The Netherlands: International Agriculture Centre . Evaluación Social de Te rritorios, Ampliación Proyecto de Tecnología Agropecuario y Forestal. http://www.magfor.gob.ni/pr ogramas/pea/salva/Evaluacion%20Social%20de%20Territor ios%20Ampliacion%20PTA%20II.pdf Accessed June 2, 2015 apprenticeship contracts in Australia Journal of Interd isciplinary Economics , 19(4), 379 - 398. Maredia , M.K.; Reyes, B. and DeYoung, D. 2014. Farmer perspective on the use of and demand for seeds of improved bean varieties: Results of beneficiary surveys in Guatemala, Honduras and Nicaragua. Staff Paper No. 2014 - 04, Department of Agricultural, Food and Res ource Economics, Michigan State University. Maredia, M. K., Howard, J. A., Boughton, D., Naseem, A., Wanzala, M. N., & Kajisa, K. 1999. Increasing seed system efficiency in Africa: concepts, strategies and issues 54578). Michigan State University, Department of Agricultural, Food, and Resource Economics. Successful 121 Community - Based Seed Production Strategies. Ed. P. S. Sentimela. Mexico, D.F.: CIMMYT. Journal of Economics and Sustainable Develo pment , 6 (5), 208 - 216 . Quiroz Cortez, M. G., Reyes Vallejos, J. Perez Cruz, M. & Urbina H., A. P. la Agricultura (IICA) Nicaragua http://cenida.una.edu.ni/relectronicos/RENE70I59f.pdf Accessed August 20, 2015. and tools Treadmill with CRS Seed Vouchers an Disasters , 26 (4), 316 - 328. Reyes, B., D. DeYoung, and M. Maredia of the Bean Seed Dissemination Models in Central America as Implemented Under the Bean Technology Dissemination (BTD) Project - 03 Department of Agricultural, Food and Resource Economics, Michigan State Univers ity . Rubyogo, J. C., Sperling, L., Nasirumbi, L., & Kasambala, S. 2007. with and for the marginalized: case of common beans (Ph aseolus vulgaris L.) in East, Proceedings of Farmer First Revisited Conference, Sussex, UK (pp. 12 - 14). innovaciones promovid http://repiica.iica.int/DOCS/B3425E/B3425E.PDF Accessed August 4, 2015 al American Maize - Catholic Relief Services (CRS). . http://static1.1.sqspcdn.com/static/f/752898/20570413/1349796038263/climate - change - maize - beans - full - report.pdf?token=I7CyFPTizHNaiDYKMVX3y9iOfVY%3D Accessed October 9, 2012. Sentimela, P.S., Monyo, E., & Bänziger, M. (2004) Successf ul Community - Based Seed Production Strategies . Mexico, D.F.: CIMMYT. - BIRD/USC Canada 122 Asia/Oxfam/The Development Fund/IFAD/Bioversity International, 14 - 15 June 2012, Pokhara, Nepal. Community Biodiversity Management . Nueva York: Routledge . (pp 109 - 117). S Brief #3. INTEGRATING SEED SYSTEMS. Ag Partner XChange. http:// seedsystem.org/wp - content/uploads/2014/03/Integrating - Seed - Systems - .pdf Access August 14, 2014. The Journal of Development Studies , 44 (4), 586 - 612. Sperling, L. 2008. When Disaster Strikes : A Guide to Assessing Seed System Security. Cali, Colombia: International Center for Tropical Agriculture. based approach t o on - farm conservation and sustainable use of agricultural biodiversity - 104. developed through participatory pl and Gender Analysis. Proc. of the International Seminar on Participatory Research and - 14, 1996), CGIAR System wide Program on PRGA for Technolo gy Development and Institutional Innovation, Cali, Columbia, pp. 155 - 164. Food policy , 22 (5), 433 - 446. Tripp, R., & Rohrbach, D. 2001. Policies for African seed enterprise de velopment. Food policy , 26 (2), 147 - 161. Development Institute, UK. Agriculture in the Sahe l, 1 5. http://www.syngentafoundation.org/db/1/447.pdf. - KERKINACTIE Agrodok - Series (37). http://journeytoforever.org/farm_library/AD3 7.pdf Accessed March 27, 2012. Van Mele, P., Bentley, J., & Guei, R. 2011. African Seed Enterprises: Sowing the Seeds of Food Security CAB International, Wallingford, UK. 123 Wierema, H., Almekinders, C., Keune, L. & Vermeer, R. sina en Centroamérica : los IVO P.O. Box 90153, Tilburg, the Netherlands. World Development , 23(3), 413 - 422. Witcombe - based seed Experimental Agriculture , 46 (04), 425 - 437.