THE ECONOMIC IMPACT OF IMPROVED BEAN VARIETIES AND DETERMINANTS OF MARKET PARTICIPATION: EVIDENCE FROM LATIN AMERICA AND ANGOLA By Byron Alejandro Reyes A DISSERTATION Submitted to Michigan State University in partial fulfillment of the requirements for the degree of DOCTOR OF PHILOSOPHY Agricultural, Food and Resource Economics 2012 ABSTRACT THE ECONOMIC IMPACT OF IMPROVED BEAN VARIETIES AND DETERMINANTS OF MARKET PARTICIPATION: EVIDENCE FROM LATIN AMERICA AND ANGOLA By Byron Alejandro Reyes This dissertation consists of two essays. The first essay studies the economic impact of improved bean varieties (IVs) in Costa Rica, El Salvador, Honduras, Nicaragua, and Ecuador. In these countries, the National Agricultural Research Systems (NARS), in collaboration with private and public institutions, have actively generated and promoted IVs over the past 20 years. There are two types of yield gains derived from the use of IVs: Type I gains in areas where IVs replace traditional varieties, and Type II gains in areas where new IVs replace old IVs. Previous studies have only estimated Type I benefits in Honduras and northern Ecuador. This study estimated the Type II yield gains associated with varietal development of small red and red mottled bean varieties over time in Central America, Honduras, and northern Ecuador, using experimental yield data. Further, it provided estimates of current total adoption rates of IVs in each country, using bean expert opinions. The economic impact of bean IVs was estimated by combining the Type I and Type II yield gains. The results suggest that the Type II yield gains from small red varieties averaged 0.49% per year for Central American countries and 0.56% per year for Honduras. Similarly, the Type II yield gains from red mottled varieties averaged 1.68% per year for Ecuador. Breeders estimated that adoption rates for 2010 ranged from 46% in Honduras to 82% in Nicaragua. Amadeus 77 was the most widely adopted small red IV in Central America and accounted for an estimated 49.7% of the total bean area. Similarly, Portilla, the most widely planted red mottled IV in northern Ecuador and accounted for an estimated 43% of the red mottled bean area in northern Ecuador. Ex post benefit/cost analysis for the period 1991-2015 indicate that returns to investments in bean research have been negative in Costa Rica and positive in all other countries, with a regional net present value of $358 million and a regional IRR of 32%. The surplus per hectare per year was estimated at $74/ha/yr in the region. The second essay studies the factors affecting farmers’ marketing decisions in the rural highlands of Angola, focusing on potatoes, beans, and onions. This essay uses single equation ordinary least squares regressions for analysis of factors affecting production of potatoes, beans, and onions in the central highlands of Angola. Furthermore, it implements double hurdle (DH) regressions to study the factors associated with farmers’ marketing decisions among potato, bean and onion growers, focusing on gender of the household head, asset ownership, and transaction costs, while controlling for potentially endogenous variables. The DH regression results suggest that the factors associated with marketing decisions depend on the crop analyzed and on whether marketing decisions are analyzed conditionally (i.e., probability of selling and, conditional on selling, quantity sold) or unconditionally (i.e., unconditional quantity sold). The results also suggest that boosting sales would be a challenge for the government of Angola, donors, and organizations working with farmers in this region since, due to Angola’s strong currency, overcoming the limiting factors found in this study may require large financial and human resources. Dedicated to Elena, Alexander, Adrián, our baby boy to be named, and to my parents Teresa and Alejandro iv ACKNOWLEDGMENTS I would like to thank my mentor and major professor, Richard Bernsten, for his extremely valuable comments and guidance provided during these past several years. I would also like to thank the feedback and assistance provided by the rest of my examination committee: Mywish Maredia, Cynthia Donovan, Eric Crawford, and James Kelly. I appreciate the assistance provided by the faculty and staff of the Department of Agricultural, Food, and Resource Economics at MSU, especially Songqing Jin, Debbie Conway, Janet Munn, and Nancy Creed. Especial thanks go to my friends and colleagues at MSU, especially Ricardo Hernandez, Alexandra Peralta, Nicky Mason, Wolfgang Pejuan, Sonja Perakis, Tim Komarek, Jake Ricker-Gilbert, and Guilherme Signorini. I am very grateful for the help provided by Juan Carlos Rosas from Zamorano’s bean breeding program in Honduras, and by Eduardo Peralta from National Program on Food Legumes and Andean Grains’ bean breeding program in Ecuador, during the fieldwork data collection and analysis. Furthermore, I’m also grateful for the help provided by key informants during my fieldwork in Honduras, El Salvador, Nicaragua, Costa Rica, and Ecuador. I’m very thankful for the assistance provided by PRORENDA staff in Angola, especially Deodato Guilherme and Sarah Grindle. While conducting this research, I was fortunate to receive the support of the Dry Grain Pulses Collaborative Research Support Program with funding from the United States Agency for International Development (Cooperative Agreement No. EDH-A-00-07-00005-00). Additional funding was provided by the Bill and Melinda Gates Foundation through its World Vision’s ProRenda project, and by the Department of Agricultural, Food, and Resource Economics at MSU. Any opinions and errors are solely my own. v TABLE OF CONTENTS LIST OF TABLES ....................................................................................................................... viii LIST OF FIGURES ...................................................................................................................... xii KEY TO ABBREVIATIONS ...................................................................................................... xiii CHAPTER ONE. INTRODUCTION ..............................................................................................1 1.1 Introduction .............................................................................................................................1 1.2 Objectives ...............................................................................................................................4 CHAPTER TWO. THE ECONOMIC IMPACT OF IMPROVED BEAN VARIETIES IN LATIN AMERICA: A SURPLUS ANALYSIS ..................................................................5 2.1 Introduction .............................................................................................................................5 2.2 Research Gap ..........................................................................................................................9 2.3 Research Questions ...............................................................................................................14 2.4 Conceptual Framework .........................................................................................................14 2.4.1 Research Benefits .......................................................................................................14 2.4.1.1 Estimation of Potential Genetic Yield Gains .............................................................18 2.4.1.2 Total Adoption Rates..................................................................................................22 2.4.2 Research Costs ...........................................................................................................23 2.4.3 Measures of Project Worth .........................................................................................25 2.4.4 Sensitivity Analysis ....................................................................................................27 2.5 Data Description ...................................................................................................................28 2.5.1 Bean Research Programs ............................................................................................29 2.5.2 Bean Varieties Released between 1990-2010 ............................................................35 2.5.3 Research Benefits .......................................................................................................38 2.5.3.1 Estimating Yield Gains ..............................................................................................38 2.5.3.2 Estimating Adoption Rates .........................................................................................42 2.5.4 Research Costs ...........................................................................................................45 2.5.5 Additional Parameters ................................................................................................49 2.6 Results ......................................................................................................................................52 2.6.1 Adoption of Improved Bean Varieties ..........................................................................52 2.6.2 Estimated Rates of Diffusion ........................................................................................55 2.6.3 Bean Yield Gains Estimation Using Experimental Yield Data .....................................57 2.6.3.1 Descriptive Results .....................................................................................................57 2.6.3.2 Econometric Results ...................................................................................................60 2.6.4 Net Present Value and Internal Rate of Return .............................................................65 2.7 Chapter Summary and Policy Recommendations....................................................................69 2.7.1 Chapter Summary ..........................................................................................................69 2.7.2 Policy Recommendations ..............................................................................................71 vi CHAPTER THREE. DETERMINANTS OF MARKET PARTICIPATION DECISIONS: EVIDENCE FROM THE CENTRAL HIGHLANDS OF ANGOLA ...............................74 3.1 Introduction ...........................................................................................................................74 3.2 Research Gap ........................................................................................................................77 3.3 Research Questions ...............................................................................................................82 3.4 Conceptual Framework .........................................................................................................83 3.4.1 Economic Model ........................................................................................................83 3.4.2 Econometric Estimation .............................................................................................87 3.5 Data Used ..............................................................................................................................93 3.6 Results .................................................................................................................................100 3.6.1 Descriptive Statistics ................................................................................................100 3.6.2 Wealth of the Households ........................................................................................109 3.6.3 Households’ Gross Margins .....................................................................................112 3.6.4 OLS Regression Results of Factors Influencing Production ....................................114 3.6.4.1 Potato OLS Results ..................................................................................................115 3.6.4.2 Bean OLS Results ....................................................................................................120 3.6.4.3 Onion OLS Results ...................................................................................................123 3.6.5 Double Hurdle Regression Results of Factors Affecting Marketing Decisions .......126 3.6.5.1 Potato DH Results ....................................................................................................127 3.6.5.2 Bean DH Results ......................................................................................................133 3.6.5.3 Onion DH Results ....................................................................................................139 3.7 Chapter Summary and Policy Recommendations...............................................................145 3.7.1 Chapter Summary .....................................................................................................145 3.7.2 Policy Recommendations .........................................................................................151 APPENDIX ..................................................................................................................................154 REFERENCES ............................................................................................................................195 vii LIST OF TABLES Table 2.1. Red bean varieties included in the red ECAR trial data. Zamorano, Honduras, 1999-2009. .........................................................................................................................40 Table 2.2. Improved bean varieties included in the PRUEBA trial data. PRONALEG-GA / INIAP, Ecuador, 2003-2010. .............................................................................................42 Table 2.3. Bean research investments ($) devoted to the development of red beans (Central America) and red mottled beans (Ecuador). 1990-2010. ...................................................48 Table 2.4. Estimations of adoption rates (%) of improved small reds and red mottled bean varieties for 1996 and 2010................................................................................................53 Table 2.5. Estimated mean yields (kg/ha) and other statistics of red bean varieties based on the experimental trial data. Central America, 1999-2009. .................................................59 Table 2.6. Estimated mean yields (kg/ha) and other statistics of red mottled bean varieties based on the experimental trial data. Ecuador, 2003-2010. ...............................................59 Table 2.7. Linear regression results of the vintage model using experimental yields of small red bean varieties released in Central America. 1999-2009. .............................................61 Table 2.8. Linear regression results of the vintage model using experimental yields of small red bean varieties released in Honduras. 1999-2009. ........................................................63 Table 2.9. Linear regression results of the vintage model using experimental yields of red mottled bean varieties released in Ecuador. 2003-2010. ...................................................64 Table 2.10. Summary of net present value (NPV) and internal rates of return (IRR) estimations of investments on bean research in Central America and Ecuador. 19912015....................................................................................................................................66 Table 3.5.1. Independent variables included in the production and marketing decision regressions. Angola, 2009. .................................................................................................97 Table 3.6.1. Percentage of households growing key crops, per economic status index and gender of household head (HHH). Central Highlands of Angola, 2009. .........................101 Table 3.6.2. Descriptive statistics of the variables used in the Double Hurdle analysis. Central Highlands of Angola, 2009. .............................................................................................103 Table 3.6.3. Scoring factors and means per gender of household head (HHH) and crop grown for asset indicators entering the computation of the first principal component (asset ownership)........................................................................................................................111 viii Table 3.6.4. Average receipts, costs and margins of households selling key crops,1 per economic status index and gender of household head (HHH). Central Highlands of Angola, 2009. ...................................................................................................................113 Table 3.6.5. Linear regression models of factors influencing potato, bean and onion quantity produced (kg). Central Highlands of Angola, 2009. ........................................................116 Table 3.6.6. Double-Hurdle model of factors influencing potato marketing decisions. Central Highlands of Angola, 2009. .............................................................................................128 Table 3.6.7. Double-Hurdle model of factors influencing bean marketing decisions. Central Highlands of Angola, 2009. .............................................................................................134 Table 3.6.8. Double-Hurdle model of factors influencing onion marketing decisions. Central Highlands of Angola, 2009. .............................................................................................140 Table A 2.1. Bean trials planted in Central America and Ecuador. 1999-2010. .........................155 Table A 2.2. Costa Rica: Improved bean varieties released. 1990-2010. ....................................156 Table A 2.3. El Salvador: Improved bean varieties released. 1990-2010....................................158 Table A 2.4. Honduras: Improved bean varieties released. 1990-2010. ......................................159 Table A 2.5. Nicaragua: Improved bean varieties released. 1990-2010. .....................................161 Table A 2.6. Ecuador: Improved bean varieties released. 1990-2010. ........................................162 Table A 2.7. Quantity (MT) of seed of improved bean varieties sold or distributed by government programs in 2010. ........................................................................................164 Table A 2.8. Estimated average yearly salaries ($) of bean breeding programs' permanent staff for Costa Rica, El Salvador, Honduras (DICTA only) and Nicaragua by education level. 2010. ......................................................................................................165 Table A 2.9. Parameters α and β used for the estimation of logistic diffusion curves in the countries of interest. .........................................................................................................166 Table A 2.10. Base scenario: Estimations of total adoption rates (%) of improved bean varieties using a logistic diffusion curve. 1990-2015. .....................................................167 Table A 2.11. Scenario A: Estimations of total adoption rates (%) of improved bean varieties using a logistic diffusion curve. 1990-2015. ....................................................................168 Table A 2.12. Scenario B: Estimations of total adoption rates (%) of improved bean varieties using a logistic diffusion curve. 1990-2015. ....................................................................169 Table A 2.13. Linear regression results of factors influencing experimental yields of small red bean varieties released in Central America. 1999-2009. ...........................................170 ix Table A 2.14. Linear regression results of factors influencing experimental yields of small red bean varieties released in Honduras. 1999-2009. ......................................................171 Table A 2.15. Linear regression results of factors influencing experimental yields of red mottled bean varieties released in Ecuador. 2003-2010. .................................................172 Table A 2.16. Costa Rica: Base scenario Net Present Value (NPV) and Internal Rate of Return (IRR) calculations for improved small red bean varieties. 1991-2015. ...............173 Table A 2.17. El Salvador: Base scenario Net Present Value (NPV) and Internal Rate of Return (IRR) calculations for improved small red bean varieties. 1991-2015. ...............174 Table A 2.18. Honduras: Base scenario Net Present Value (NPV) and Internal Rate of Return (IRR) calculations for improved small red bean varieties. 1991-2015. ...........................175 Table A 2.19. Nicaragua: Base scenario Net Present Value (NPV) and Internal Rate of Return (IRR) calculations for improved small red bean varieties. 1991-2015. ...............176 Table A 2.20. Ecuador: Base scenario Net Present Value (NPV) and Internal Rate of Return (IRR) calculations for improved red mottled bean varieties in northern Ecuador. 1991-2015. .......................................................................................................................177 Table A 2.21. Notes for Table A 2.16 through Table A 2.20. .....................................................178 Table A 3.1. Additional demographic characteristics of farm households (HH). Central Highlands of Angola, 2009. .............................................................................................179 Table A 3.2. Major sources of crop and non-crop household incomes by market participation. Central Highlands of Angola, 2009. ................................................................................181 Table A 3.3. Scoring factors, summary statistics, and per tercile means for asset indicators entering the computation of the first principal component (asset ownership). ................182 Table A 3.4. Means per market participation (seller) for asset indicators entering the computation of the first principal component (asset ownership). ....................................183 Table A 3.5. Potato sellers: Average receipts, costs, margins and percentage of production sold, per economic status index and gender of household head (HHH). .........................184 Table A 3.6. Bean sellers: Average receipts, costs, margins and percentage of production sold, per economic status index and gender of household head (HHH). .........................185 Table A 3.7. Onion sellers: Average receipts, costs, margins and percentage of production sold, per economic status index and gender of household head (HHH). .........................186 Table A 3.8. Descriptive statistics of factors influencing potato, bean and onion production. Central Highlands of Angola, 2009. ................................................................................187 x Table A 3.9. Unconditional average partial effects of factors influencing potato sales. Central Highlands of Angola, 2009. .............................................................................................189 Table A 3.10. Unconditional average partial effects of factors influencing bean sales. Central Highlands of Angola, 2009. .............................................................................................190 Table A 3.11. Unconditional average partial effects of factors influencing onion sales. Central Highlands of Angola, 2009. ................................................................................191 xi LIST OF FIGURES Figure 2.1. National bean yields (kg/ha) in Costa Rica, Ecuador, El Salvador, Honduras, and Nicaragua. FAOSTAT 1990-2009. ......................................................................................8 Figure 2.2. Research-induced supply shift in a small open economy set up. ................................16 Figure 2.3. Flow of breeding materials in the regional and national nurseries in Central America. For interpretation of the references to color in this and all other figures, the reader is referred to the electronic version of this dissertation. .........................................32 Figure 2.4. Base scenario: Total adoption rates of improved bean varieties. 1990-2015. .............56 Figure 3.5.1. Distribution of villages included in the ProRenda 2009 survey. The text in the figure is not meant to be readable but is for visual reference only. ...................................95 Figure 3.6.1. Cumulative distribution of asset index by gender of household head. Central Highlands of Angola, 2009. .............................................................................................112 Figure A 3.1. Cumulative distribution of asset index by potato growers and non-growers. Central Highlands of Angola, 2009. ................................................................................192 Figure A 3.2. Cumulative distribution of asset index by onion growers and non-growers. Central Highlands of Angola, 2009. ................................................................................193 Figure A 3.3. Cumulative distribution of asset index by bean growers and non-growers. Central Highlands of Angola, 2009. ................................................................................194 xii KEY TO ABBREVIATIONS ACDI/VOCA Agricultural Cooperative Development International / Volunteers in Overseas Cooperative Assistance APE Average Partial Effect AY Agricultural Year B/C CRSP Bean Cowpea Collaborative Research Support Program bgm-1 Gene of resistance to BGYMV BGYMV Bean Golden Yellow Mosaic Virus BNF Biologic Nitrogen Fixation BTE Bean Time Equivalents CAFTA Central American Free Trade Agreement CDF Cumulative Distribution Function CENTA Centro Nacional de Tecnología Agropecuaria y Forestal, El Salvador CIAL Local Agricultural Research Committee CIAT International Center for Tropical Agriculture CIDA Canadian International Development Agency COSUDE Swiss Agency for Development and Cooperation COVA Comprobación de Variedades, a national nursery CPI Consumer Price Index DGP CRSP Dry Grain Pulses Collaborative Research Support Program DH Double-Hurdle DICTA Dirección de Ciencia y Tecnología Agropecuaria, Honduras DNA Deoxyribonucleic Acid ECAR Ensayo Centroamericano de Adaptación y Rendimiento, a regional nursery xiii ESPAC Encuesta de Superficie y Producción Agropecuaria Continua, Ecuador FAO Food and Agriculture Organization FC Fixed Costs Fi Filial Generation i; i=1,2,… FO Farmer Organization GARB Gross Annual Research Benefits GDP Gross Domestic Product IDA Angolan Institute for Agrarian Development IMR Inverse Mills Ratio INEC Instituto Nacional de Estadísticas y Censos, Ecuador INIAP Instituto Nacional Autónomo de Investigaciones Agropeciarias, Ecuador INTA Instituto Nicaragüense de Tecnología Agropecuaria, Nicaragua INTA-CR Instituto Nacional de Innovación y Transferencia en Tecnología, Costa Rica IRR Internal Rate of Return IV Improved Variety MINADER Ministério da Agricultura e do Desenvolvimento Rural, the Ministry of Agriculture of Angola MLE Maximum Likelihood Estimator MPLA Popular Movement for the Liberation of Angola MSU Michigan State University MT Metric Tones NARS National Agricultural Research Systems NGO Non-Governmental Organization NPV Net Present Value NSO National Statistical Office, one for each country xiv OLS Ordinary Least Squares PE Partial Effect PIF Programa de Investigaciones en Frijol, Zamorano, Honduras PITTA-Frijol Bean research network of Costa Rica PPB Participatory Plant Breeding PROFRIJOL Programa Cooperativo Regional de Frijol para Centro América, México y El Caribe PRONALEG-GA Programa Nacional de Leguminosas y Granos Andinos, Ecuador PVS Participatory Varietal Selection R&D Research and Development SCARs Sequence Characterized Amplified Regions SIECA Secretariat of Economic Integration of Central America SIMPAH Honduran Agricultural Price Information System TLU Tropical Livestock Units TNR Truncated Normal Regression UNITA National Union for the Total Independence of Angola UPR University of Puerto Rico USAID United States Agency for International Development VC Variable Costs VIDAC Vivero de Adaptación Centroamericana, a regional nursery Zamorano Escuela Agrícola Panamericana, Zamorano xv CHAPTER ONE. INTRODUCTION 1.1 Introduction The goal of any development project is to positively impact people’s life (Maredia 2009). In agriculture, reducing poverty, hunger, and food insecurity is of primary interest. To achieve this, international donors and national governments support a wide range of projects that could assist producers. While many projects focus on generating new technologies (e.g. improved varieties) that offer farmers an opportunity to increase their output, agricultural income, and food security, other projects assist farmers during the production stage by providing technical assistance, access to inputs, etc., or during the post-harvest stage by providing technical assistance regarding storage techniques, better access to markets and information, or assisting farmers to add value to their products. As agricultural research funding becomes increasingly scarce, the demand for impact assessment studies has increased (Alston et al. 2000a) because international donors and national governments must decide where to invest their money and efforts. This increased demand for impact assessment has led to a growing focus and emphasis on a more formal, systematic, and rigorous approach to assessing impacts (Maredia 2009). Although the literature on impact assessment is overwhelming (Feder et al. 1985) and most studies have demonstrated that returns to investments are high (Alston et al. 2000a), impact assessment studies are (and will be) an important and necessary tool to evaluate investments. Similarly, studying the factors associated with farmers’ decisions of whether to engage in marketing activities is important because these studies can provide valuable information about the constraints farmers’ face to participate in markets and increase their agricultural income. 1 Donors and governments can use this information to efficiently allocate scarce resources into constraining areas to boost market participation and positively affect farmers’ welfare. This study focuses on two areas: (1) the economic impact of new bean varieties used by farmers as inputs in their production process in five countries of Latin America, and (2) studying the factors associated with farmers’ marketing decisions in output markets in the Central Highlands of Angola. To estimate the economic impact of new bean varieties in the region, first, experimental yield data were used to econometrically estimate the yield gains associated with the use of new improved bean varieties, compared to old improved bean varieties. This was done because of the time and financial limitations for collecting farm-level data in each country. Furthermore, since the breeding programs in the region continuously release new improved varieties (IVs), farmers can replace old IVs with new IVs and benefit from gains in yields (if any). Thus, the methodology implemented in this study only measures the economic impact derived from these gains. Second, key informants associated with the bean subsector were interviewed to collect information related to adoption rates of improved varieties (IVs), research costs, variety-specific information, etc. Third, additional data were collected from secondary sources. Fourth, to evaluate the economic impact of new bean varieties in Latin America, economic surplus concepts were used. This methodology is commonly used to assess the impact of agricultural projects. Although several studies have evaluated the impact of bean research in the region in the past, there was a need to generate updated information, especially for Costa Rica, El Salvador, and Nicaragua, where no economic impact studies have been carried out to date. To assess the economic impact of bean research, the net present value (NPV) and the internal rate of return (IRR) were estimated for Honduras, El Salvador, Nicaragua, Costa Rica, 2 and Ecuador for a base scenario. Finally, sensitivity analysis was conducted to evaluate the robustness of the results. Since bean research programs in the region have received financial support from donors for several years, the information generated from this study can be used by donors to make decisions of whether financial support should be continued and/or which areas require additional work to augment the economic impact of bean research. Furthermore, government officials can use this information to generate policies targeted at increasing the benefits of bean research to farmers. To study the factors associated with farmers’ marketing decisions in Angola, first, household-level data were used to estimate factors influencing the production of potatoes, beans, and onions by ordinary least squares (OLS). The household-level data were collected by World Visions’ ProRenda project staff in 2009 and included 40 communities across three provinces in the central highlands of Angola. Second, the residuals from these regressions were estimated. Third, for each crop, a double hurdle regression was estimated to determine which factors affect farmers’ marketing decisions, while controlling for self-selection and potential endogeneity problems. Fourth, the average partial effects (APEs) of the variables of interest were estimated via bootstrapping. The APEs were used to measure the unconditional (on market participation) effect of the variables of interest on the total quantity sold. The ProRenda project’s main objective was to increase smallholders’ income through the establishment of competitive value chains of several crops. Because of this, the sampling methodology implemented by this project was out of our control. This caused limitations in our analysis because the research questions included in the present study were not considered during the sampling and questionnaire development. 3 1.2 Objectives The general objective of the study is to generate information that can be used by stakeholders to efficiently allocate resources as to maximize farmers’ welfare from using new improved bean varieties in Latin America and from participating in output markets in Angola. Specifically, the study aims to: 1. Estimate the economic impact derived from smallholder farmers’ use of improved bean varieties released between 1996 and 2010 in five countries of Latin America. 2. Determine the factors that affect smallholder farmers’ marketing decisions in the potato, bean, and onion markets of rural Angola. The document is divided into three chapters. Chapter 2 presents information about the motivation and research questions, conceptual framework, data used, and results of the economic impact of new bean varieties in five countries of Latin America. Chapter 3 presents information about the motivation and research questions, conceptual framework, data used, and results of the factors associated with smallholders’ marketing decisions in rural Angola. 4 CHAPTER TWO. THE ECONOMIC IMPACT OF IMPROVED BEAN VARIETIES IN LATIN AMERICA: A SURPLUS ANALYSIS 2.1 Introduction The most common approach for analyzing welfare effects of agricultural research in a partial-equilibrium framework is the use of economic surplus analysis (Alston et al. 1998). The economic surplus concept can be applied to ex-ante and/or ex-post evaluation of agricultural projects. Norton and Davis (1981) review the major research techniques used to evaluate returns to agricultural research. They found that ex post studies usually use (1) consumer and producer surplus analyses (i.e. average rates of return) and (2) production function analyses (i.e. marginal rates of return) to evaluate returns to agricultural research. Griliches (1958) first used ex-post surplus analysis to estimate the realized social rate of return of private and public investments in research and development of hybrid corn in the US. Since then, hundreds of studies have reported measures of the returns to agricultural research and development (R&D). Alston et al. (2000a) assemble all the available evidence on the returns to investments in agricultural R&D for the period 1953-1999. The fact that they assembled 292 studies reporting a total of 1,886 rates of return estimates demonstrate how wide surplus concepts are used for impact evaluation of agricultural research. Despite its wide use, the surplus approach has been criticized from several perspectives, including: its normativeness (what it should be vs. what is); its potential for measurement error (will surplus measures provide an accurate indicator of changes in social welfare?); the partialwelfare nature of the analysis (it ignores the complex interrelationships with other product and factor markets in the economy); the exclusion from the analysis of (a) externalities and free 5 riders and (b) transaction costs and incomplete risk markets; and its policy irrelevance because (a) of the problems stated previously and (b) important assumptions and variables are usually not explained in the calculations (Alston et al. 1998). However, some of these criticisms (e.g. exclusion of transaction costs and externalities, measurement errors) can be addressed, at least partially, by making refinements to the measure of benefits and costs (Alston et al. 1998). In addition, the policy relevance criticism could be tackled by clearly explaining the implications of the results (Alston et al. 1998). There are several alternatives to using surplus analysis for impact evaluation. Alston et al. (1998) suggest the use of cost-benefit analysis (although this type of analysis is usually complementary to surplus analysis), econometric models (the most commonly used alternative), domestic resource costs models (with the appeal that they provide a simple measure of the social value of inputs used to generate a unit of net output valued at its true social value), and the congruence rule (where resources are invested in relation to the value of output they provide). However, it is very common to find a mix of methods in the literature. For example, Mather et al. (2003) used both econometric and surplus methods to estimate the economic impact of bean research in Honduras. More recently, Mooney (2007) followed a similar approach to estimate the economic impact of bean breeding investments in Ecuador. Both Mather et al. (2003) and Mooney (2007) estimated that investments in bean breeding were profitable, a very common finding across the literature. Despite these criticisms, surplus analysis is a useful methodology because it is common used in all methods for estimating research benefits (Alston et al. 1998). This study uses both econometric and surplus concepts to estimate the economic impact of bean breeding research in four countries of Central America and one in South America. 6 The study focuses on Honduras, El Salvador, Nicaragua, Costa Rica, and northern Ecuador because the National Agricultural Research Systems (NARS) of these countries, in collaboration with private and public institutions, have actively generated and promoted improved bean varieties (IVs) over the past 20 years. 1 Recent aggregate yield data show that yields have been variable over time (Figure 2.1). 2 For the period of 1990-2009, yields averaged 581 kg/ha (CV =0.20) in Costa Rica; 479 kg/ha (CV=0.33) in Ecuador; 3 860 kg/ha (CV=0.11) in El Salvador; 717 kg/ha (CV=0.15) in Honduras; and 713 kg/ha (CV=0.12) in Nicaragua. Therefore, yields varied the most in Ecuador and Costa Rica, and were less variable in El Salvador and Nicaragua. In addition, while yields showed a slightly increasing trend in El Salvador, Nicaragua and Costa Rica, the yield trend is constant in Honduras and decreasing in Ecuador (Figure 2.1). Although most of the variation in yields have been due to weather-related factors (KII, 2010a), the decreasing yield trend in Ecuador is because (1) INIAP’s (the national research institute) food legume breeding program, El Programa Nacional de Leguminosas y Granos Andinos (PRONALEG-GA) primarily focuses on developing bush-type beans targeted for and 4 adopted by farmers in the northern region (KII, 2010a) and (2) FAOSTAT’s dry-bean data does not distinguish between monocropped and intercropped beans, the latter type with lower yields. 1 Although Guatemala was initially considered for this study, due to the fact that no bean varieties were released between 1998 and 2009, it was not possible to estimate bean yield gains from breeding research (since no varieties were released). Thus, this country was excluded from the analysis. 2 CV = coefficient of variation. Estimated by dividing the standard deviation by the mean. 3 4 For Ecuador, FAOSTAT yield data combines monocropped and intercropped beans. Mainly the provinces of Carchi and Imbabura. 7 Figure 2.1. National bean yields (kg/ha) in Costa Rica, Ecuador, El Salvador, Honduras, and Nicaragua. FAOSTAT 1990-2009. 5 The national statistical institute of Ecuador (INEC) reports detailed bean data since 2002. At the national level, INEC and FAOSTAT report the same yield levels for 2002-2009. However, dry bean yields for monocropped beans (estimated from INEC data) averaged 655 kg/ha and dry bean yields for intercropped beans averaged 194 kg/ha. This suggests that intercropped beans may be driving the low average national yields observed in Figure 2.1 for Ecuador. Although it is not clear why yields sharply declined after 2001, it is possible that this may be a medium-term negative effect of the country’s 2000 dollarization, which reduced farmers’ purchase power for inputs. 5 The data is published annually in the Encuesta de Superficie y Producción Agropecuaria Continua, ESPAC. 8 The fact that the yield trend has been constant in some countries does not suggest that bean research has not had a positive effect on (aggregate) production. Morris and Heisey (2003) note that, over time, most successful crop breeding programs generate genetic gains in yields. However, genetic yield gains have two components: (a) increased yield potential, which is observable because yields are higher, and (b) increased biotic and abiotic stress resistance, which is aimed at avoiding losses from stresses (yields may not be higher; instead, losses are averted in the presence of stresses). Therefore, without bean research, it is possible that in these countries, yields could have been much lower. Because of this, it is important to empirically estimate whether improved bean varieties released over time show genetic yield gains and what has been the economic impact of investments in bean research in the past two decades. 2.2 Research Gap Much of the returns-to-research literature has dealt with varietal improvement research (Pardey et al. 2006). In his pioneering work, Griliches (1958) estimated the realized social rate of return on public and private funds invested in hybrid-corn research. For this, he estimated the 6 annual gross social returns and subtracted the cost of producing hybrid seed to obtain an annual flow of net social returns. In addition, he estimated yearly private and public research expenditures. He then brought forward the net returns and expenditures to the year of interest (i.e. 1955) and estimated that at least 700 percent per year was earned (in 1955) on the average dollar invested in hybrid-corn research in the U.S.. Byerlee and Traxler (1995) used total economic surplus concepts to estimate the benefits of wheat improvement research. Although the concepts used were similar to the ones 6 Also called the gross annual research benefits (GARB) by other authors. 9 implemented by Griliches (1958), they estimated IVs’ yield advantage from (a) IVs replacing traditional varieties (TVs) and (b) new IVs replacing old IVs, by using experimental yield data. These authors found that the ex post rate of return of investments in wheat improvement research during the post-green revolution period was above 50%, and projected that the return on future investments would be between 37% and 48%. Pardey et al. (2006) used examples of varietal improvement research in Brazil, focusing on rice, edible beans, and soybeans, to attribute the benefits of varietal improvements to different research institutions. They used the total gross annual research benefits (GARB) approach first implemented by Griliches (1958) for this purpose. Contrary to Griliches (1958), Pardey et al. (2006) used experimental yield data to generate indexes of varietal improvement to estimate the 7 change in yield from the use of new crop varieties. Furthermore, they used the last-cross rule 8 and the geometric rule to attribute the credit (of the benefits) to different research institutions. These authors found that, when the benefits are attributed to different institutions, the benefitcost ratio fall from 78:1 when no distinction is made to 16:1 for the institution of interest (the Brazilian public research corporation Embrapa in their case). In a similar fashion, Maredia et al. (2010) used the GARB approach to attribute the benefits of bean research to different research institutions in Michigan, U.S. Following Pardey et al. (2006), they used experimental yield data to estimate yield gain indexes from the use of IVs and attributed benefits to the different institutions working on bean research in Michigan. These 7 While both Byerlee and Traxler (1995) and Pardey et al. (2006) used experimental yield data to estimate benefits, the former used total economic surplus analysis (instead of GARB). 8 While the last-cross rule attributes all the credit to the breeder institution that produced the variety, the geometric rule attributes the credit to different institutions depending on the share of the genetic material coming from each institution in the variety. 10 authors found that bean research investments made by a public institution (Michigan State University in their case) generated benefit-cost ratios between 0.7:1 and 2.2:1. Marasas et al. (2003) used net present value (NPV), internal rate of return (IRR), and benefit-cost ratios to estimate the economic impact of efforts made by the International Maize and Wheat Improvement Center (CIMMYT) to generate disease-resistant wheat varieties in developing countries. Their study differs from Byerlee and Traxler (1995) in that they estimated the benefits of research in terms of yield losses avoided with the use of IVs, instead of yield gains. These authors found that the ex post IRR of investments made to generate disease-resistant wheat varieties was 41% and that the benefit-cost ratio was 27:1. In the countries of interest, several studies have been carried out to estimate adoption of improved varieties and the economic impact of varietal improvement research. Mather et al. (2003) used econometric methods to estimate adoption rates of bean IVs in two major beanproducing regions of Honduras and the yield loss averted from the use of bean IVs. They estimated that 41-46% of bean farmers had adopted IVs, planting them in 22-37% of the bean 9 area. Additionally, they found that IV adopters gained the equivalent of 7-16% in bean income from the yield loss averted through the use of IVs. These authors then used surplus concepts to estimate the profitability of bean research investments and found that the ex post internal rate of return to bean research was 41.2% (from 1984-2010). Hernández and Elizondo (2006) used descriptive statistics to estimate adoption of IVs in one of the largest bean-producing regions of Costa Rica, the Brunca Region. They estimated that, in 2004, IVs were planted in approximately 70% of the bean area. Although these authors 9 The area planted to IVs varied by season and region. 11 provided detailed information about adoption rates, the study did not include an estimate of the economic benefits derived from adoption of IVs in Costa Rica. Similar to Mather et al. (2003), Mooney (2007) used econometric and surplus analysis to estimate the impact of bean research in northern Ecuador. He estimated that, when diseases are present, IV adopters enjoy 40% higher yields and 20% lower per-unit production costs than nonadopters. Furthermore, he estimated that bean research investments (in red mottled beans) have an ex post internal rate of return of 29% (from 1982-2006). In contrast, CENTA (2004) documented the advantages and disadvantages of one bean IV (CENTA San Andrés) developed by the breeding program of El Salvador. Although it reported detailed information about the characteristics of this variety and its acceptability (by farmers), the study did not include an analysis about the economic benefits from the use of this IV. Similarly, there are no studies that estimate the returns to bean improvement research in Nicaragua. Thus, estimating the economic impact of bean research investments in the countries of interest will provide valuable information that could be used by stakeholders. This study is a step in that direction. Many studies use surplus methods to estimate the economic impact of agricultural research; however, they differ in the way they estimate the benefits of research. For example, while Griliches (1958), Byerlee and Traxler (1995), Pardey et al. (2006), and Maredia et al. (2010) estimate the benefits of agricultural research investments in terms of yield gains, Marasas et al. (2003), Mather et al. (2003) and Mooney (2007) estimate the benefits in terms of yield loss averted by the use of IVs. Within the studies that estimate benefits in terms of yield gains, there are differences in the way these yield gains are estimated. While Griliches (1958) use experts’ estimates, Byerlee 12 and Traxler (1995), Pardey et al. (2006), and Maredia et al. (2010) use experimental data to estimate these gains. Similarly, there are differences in the way yield loss averted is estimated. While Marasas et al. (2003) use experimental data to estimate the yield loss averted by the use of IVs, Mather et al. (2003) and Mooney (2007) use a combination of farm-level data and experimental data to estimate these parameters. This study implements a combination of expert opinions and experimental yield data to estimate the economic benefits from the use of bean IVs. As noted, several studies demonstrate that investments in bean breeding have been profitable. Despite this, it is necessary to generate additional and updated information about the economic benefits of bean research in the countries mentioned above. For countries such as El Salvador, Nicaragua and Costa Rica, this information will be useful for learning whether investments in bean improvement research have been profitable, since no such information currently exists. The contribution of this study is two-fold. First, it provides estimates of adoption rates and yield gains of IVs released over time for four Central American countries and one South American country, using expert opinions and experimental yield data. Second, it estimates the economic impact of bean improvement research in these countries and provides policy recommendations based on these results. Within the countries included in the study, several NARS collaborate to develop bean IVs, mostly using the same genetic materials to develop bean IVs. 10 Therefore, attributing benefits to each research institutions is not pursued in this study. 10 The exception is the Ecuadorian breeding program because it does not collaborate with NARS from the other countries included in this study. Although the International Center for Tropical Agriculture (CIAT) played an important role in the 1980s and 1990s, during the last decade its contribution has drastically decreased. 13 2.3 Research Questions Although the study’s main objective is to generate information about the economic impact of bean research in Latin America, it also attempts to answer the following research questions:  What were the adoption rates of improved varieties in the 2009/2010 agricultural year and which IVs were most widely planted in each country?  What is the cumulative adoption rate of IVs over time?  What are the estimated yield gains from IVs released during the last decade?  What is the economic effect of bean improvement research in the countries of interest?  What policy recommendations can be provided based on these results? 2.4 Conceptual Framework This section presents the economic rationale to estimate research benefits, costs and measure of project worth. Although the methodology is based in surplus concepts, the analysis is tailored to the situation faced in each of the five countries of interest. 2.4.1 Research Benefits The rationale for the use of economic surplus models is straightforward and a large body of literature explains this topic (see for example Alston et al. 1998). Although the researchinduced technical changes in the bean sub-sector could affect different sectors of the economy (e.g. labor markets), it is assumed that these secondary effects are exogenous and were not addressed in the analysis. 14 For each country, a small open economy surplus model was used (Alston et al. 1998). The assumption of an open economy is appropriate because the countries of interest trade (export and import) beans with each other and with other countries in the world. Similarly, the term small is fitting because the bean supply of each country does not influence international prices. While it is common that Central American countries trade beans freely 11 with each other, Ecuador’s main bean-trade partner is Colombia, its northern neighbor. In the small open economy set up (where quantity supplied does not affect world prices), the demand curve is perfectly elastic and all of the benefits accrue to producers because there is no research-induced reduction in price (Alston et al. 1998). Therefore, the change in total surplus (Δ TS) equals the change in producer surplus (Δ PS). Following Maredia and Byerlee (2000), Mather et al. (2003) and Mooney (2007), it is assumed that the supply curve is linear and that its shift (due to technological change) is parallel. One potential problem of this assumption is that the benefits from a parallel shift may be overestimated (almost twice) if the supply shift is indeed pivotal (Alston et al. 1998). However, assuming parallel shifts is appropriate in this context because (a) previous studies have shown that adoption of IVs is scale-neutral (Mather et al. 2003) and (b) the production technology of bean producers is relatively homogeneous (Mooney 2007). Figure 2.2 illustrates the economic benefits derived from the technological changes. Typically, the curves in Figure 2.2 are defined as annual flows (Alston et al. 1998). The original supply curve (before research investments) is represented by S0 and the equilibrium price and quantity under this technology are represented by P0 and Q0, respectively. As farmers adopt the 11 The Central American Free Trade Agreement (CAFTA) allows Central American countries to export and import beans freely. 15 Figure 2.2. Research-induced supply shift in a small open economy set up. new technology (i.e. IVs), the original supply curve shifts outwards to S1 and the new equilibrium quantity shifts from Q0 to Q1. Given that the demand curve (D) is perfectly elastic, the price remains constant at P0. The total benefit from research-induced supply shift is equal to the area beneath the demand curve, D, and between the two supply curves, S0 and S1 (area eabd in Figure 2.2). This benefit is given by the sum of (a) the cost savings on the original quantity produced (area eacd) and (b) the economic surplus from incremental production (area abc) (Alston et al. 1998). 16 Empirically, in the case of a small open-economy, the formula for estimating research benefits from a parallel shift in the supply curve is given by: (2.1) Δ TS = Δ PS = P0 × Q0 × Kt (1 + 0.5 Kt ε) where P0 is the exogenous market price for beans, Q0 is the initial quantity produced before bean research, Kt represents the shift in the supply curve for each year, and ε represents the supply elasticity (Alston et al. 1998). The most critical variable in Equation (2.1) is the supply-shift parameter Kt (Maredia et al. 2010). This parameter can be estimated in different ways. For example, for each IV, Mather et al. (2003) estimated Kt by multiplying the change in net bean income due to yield loss (due to disease pressure) averted from adopting an IV by the adoption rate. Similarly, Mooney (2007) estimated Kt by multiplying the proportional change in unit cost (from IV adoption) by the probability of disease pressure and the cumulative adoption rate. These authors used experimental and farm-level data in their estimations. In contrast, Byerlee and Traxler (1995), Pardey et al. (2006), and Maredia et al. (2010) used experimental yield data to estimate yield gains by generating yield gain indexes and then used this information to estimate the supply-shift parameter Kt. In this study, experimental yield data were used to estimate yield gains from new bean varieties released over time in the countries of interest. Thus, these gains reflect the gains obtained by farmers “with” bean research vs. “without” bean research during the period of evaluation. The supply-shift parameter Kt is represented by: (2.2) Kt = At * kt 17 where At is the share of the bean area planted to improved bean varieties in year t and kt is the research-induced yield advantage of new bean varieties; that is, the yield gains from new IVs over old IVs. The methodology for estimating Kt is detailed below. 2.4.1.1 Estimation of Potential Genetic Yield Gains There are two types of yield gains derived from the use of improved varieties: Type I, which occurs in areas where improved varieties are replacing traditional varieties (i.e. new adopters of IVs), and Type II, which occurs in areas where new improved varieties are replacing old improved varieties (i.e. current adopters replace old IVs with new IVs) (Byerlee and Traxler 1995). Although an ad-hoc estimation of Type I yield gains was conducted, the main focus of the study was on estimating Type II yield gains (explained below). Therefore, benefits from bean research shown in the base scenario may be underestimated. Experimental yield data were used to estimate the Type II yield gains of the commercially successful bean varieties released to date in the countries of interest. The advantage of using experimental yield data is that most variables that influence yields are deliberately held constant; hence, the differences in yields reflect the effect of the variety per se (Pardey et al. 2006). The disadvantage of using experimental data is that experimental yields are usually higher than farmers’ yields. However, Pardey et al. (2006) noted that using experimental yields may be appropriate because farmers’ yields are affected by many factors (e.g. weather, change in relative price of inputs and outputs) and, although experimental treatments (e.g. fertilizer levels) may change over time and among locations, this variability is smaller than the variability of farmers’ yields. Additionally, it is yield gains, not yield levels that are relevant in this study. 18 Following Maredia et al. (2010), for each country, the yields of variety i in location j and year t, Yijt, was estimated by least squares using the following regression model: Yijt    (2.3a) T 1 I 1 J 1 R 1 t 1 i 1 j 1 r 1   t Dt    i Di   j D j    r Dr   t where Dt are the dummy variables for each year, Di are the dummy variables for each variety included in the dataset (i.e. equal to one if Yijt corresponds to yields of variety i; zero otherwise), Dj are dummy variables for each location included in the dataset within each country r, 12 Dr are dummy variables for each country, ut are error terms, and α, βt, γi, δj, and πr are the estimated coefficients. The model in Equation (2.3a) can be estimated with a complete dataset. However, the dataset used in this study was not complete. That is, yield information for every variety for every location within a country and in every year was not available because the trials were not included in the same locations each year (Table A 2). Therefore, following Maredia et al. (2010), the model in Equation (2.3a) was modified to: (2.3b) Yit    T 1 I 1 R 1 t 1 i 1 r 1   t D t    i Di    r Dr   t where now Yit is the yield of variety i (averaged across all locations within a country) in year t. Although averaging across locations does not allow us to estimate the effect of the genotype by environment interaction, each year the breeding programs usually use the same format (i.e. average yields across locations) to report their results; therefore, averaging yields across 12 In Central America, the variety trials are evaluated in several locations across several countries. In Ecuador the trials are only evaluated within the country. 19 locations is plausible. For Ecuador, dummies for countries were not included because the trials are not conducted in other countries. Therefore, for Ecuador, Equation (2.3b) now becomes: Yit    (2.3c) T 1 I 1 t 1 i 1   t D t    i Di   t Since the models in Equations (2.3b and 2.3c) were estimated with an intercept, and to avoid the “dummy” variable trap, one dummy variable for each year, each variety, and each country (except Ecuador) was dropped from the regression. Once the parameters were estimated, the fitted values (Ŷit) for the experimental yields of each variety for every year were computed. Using these (fitted) values provided more accurate estimates of the yield effect because they take into account the year effect on variety i; that is, they adjust the mean upwards or downwards to reflect the fact that variety i may have not been tested in high- or low- yielding years (Maredia et al. 2010). The predicted yields from Equations (2.3b and 2.3c) were used to estimate the effect of a vintage variable Vi on yield gains, using the following simplified “vintage” 13 models (adapted from Maredia et al. 2010): (2.4a) (2.4b) ˆ ln (Yit )    ˆ ln (Yit )    T 1 R 1 t 1 r 1   t D t    r D r   Vi   t for Central America T 1   t D t  Vi   t for Ecuador t 1 where Vi is the year in which variable i was released (e.g. 1996) and ln (Ŷit) is the natural log of the fitted values from Equation (2.3b) for the Central American data and from Equation (2.3c) 13 The term vintage refers to the year when the variety was released. Thus, a vintage model is a model that includes the year of release as an explanatory variable. 20 for the Ecuadorian data. 14 Therefore, the relative (percent) per year yield increase is given by 100 dln (Ŷit)/dVi = 100 λ. Once the per year yield gain (λ) was estimated using Equations (2.4a and 2.4b), it was necessary to reflect the impact of these benefits on farmers who had adopted IVs. For this, the 15 research-induced yield advantage was weighted by the yearly cumulative adoption rate of IVs. Furthermore, the research-induced yield advantage was assumed to grow at a compound rate; i.e. s kt = (1 + λ) , where λ is the yield gains from new bean IVs obtained from Equation (2.4) and s = (t – 1996). Therefore, research-induced Type II yield gains were given by: II s K t = At-1 * [(1 + λ) – 1] (2.5) II where K t measures the benefit from new bean IVs released over time and adopted by farmers who were already adopters in previous time period (At-1). Following Equation (2.1), the total Type II benefits from varietal improvement in country r at time t are given by: (2.6) II II Type II Δ PSrt = Prt × Qrt × K t (1 + 0.5 K t ε) Type I benefits were estimated using available data from previous research conducted in the region. For this (ad-hoc) estimation, Equation (2.5) was modified to: (2.7) K tI   At  Ab  * k I I where At is the adoption rate at time t, Ab is the adoption rate in the base year (i.e. 1996), and k is the yield gain associated with replacing traditional varieties with improved varieties, obtained 14 In practice, 2.4a and 2.4b are the same models, except for Ecuador, where there are no dummy variables for countries where the trials were conducted. 15 Total adoption through time was estimated using the logistic diffusion curve (explained in Section 2.4.1.2 below). 21 from previous research and assumed to be constant through time. Following Equation (2.1), the total Type I benefits from varietal improvement in country r at time t are given by: I I Type I Δ PSrt = Prt × Qrt × K t (1 + 0.5 K t ε) (2.8) Thus, Equation (2.8) assesses the economic benefit for farmers who replace their traditional varieties with improved varieties (i.e. new adopters). Following Byerlee and Traxler (1995), the total benefits from varietal improvement in country r at time t are given by the sum of Type I and Type II benefits; that is, the sum of Equations (2.6) and (2.8). 2.4.1.2 Total Adoption Rates To estimate Equations (2.5) and (2.7), it was necessary to obtain estimations of adoption rates of IVs for each year. For this, a logistic diffusion curve was estimated for each country using total adoption rates of IVs at two points in time (i.e. 1996 and 2010). The logistic diffusion curves for each country r were estimated using the following formula (from Alston et al. 1998, pg. 357-358): Art  (2.9) MAX Ar 1  e ( r   r *t ) where Art is the total adoption rate of IVs in country r and time t (i.e. observed adoption rates), MAX Ar is the maximum adoption rate (adoption ceiling) in country r, and αr and βr are parameters that define the path of the adoption rate of IVs that asymptotically approaches its ceiling. The practicality of this formula is that the curve can be generated with as little as three MAX parameters: Ar , αr, and βr. The expression above can be rearranged and written as: 22  Art ln   A MAX  A rt  r (2.10)      *t r r   From this equation, αr, and βr can easily be estimated because, for each country r, we know MAX Ar and two combinations of Art and t. 16 Although Equation (2.9) does not allow for the possibility of disadoption of the variety, this is not expected to be a problem because all current breeding programs are mature (i.e. constantly releasing new varieties); therefore, it is expected that cumulative adoption is increasing. 17 2.4.2 Research Costs Constructing the time-series data of research costs can be difficult and time consuming (Alston et al. 1998). Several considerations must be taken into account to accurately estimate research costs. First, one must define the duration and scope of the research that will be evaluated (Mooney 2007). Second, it is necessary to develop a clear understanding of the institutional history of the research project (Alston et al. 1998) in order to be able to collect accurate cost data. Alston et al. (1998) indicate that there can be three possible sources of funding: the core funds (usually from the government and used to cover routine or core expenditures such as salaries, consumables, etc.); other government funds (used for non-core activities such as 16 That is, Equation (2.10) needs to be estimated two times using: (1) adoption in the base year and (2) adoption at the time of the study. Then, these two equations can be set equal to each other to obtain αr and βr. 17 As will be explained below, only in Costa Rica, adoption of IVs has decreased over time. This formula is still valid to estimate the decreasing adoption rate in this case. 23 publications, equipment, etc.) that usually are more volatile than the core funds; and donor funds and grants (from public or private institutions). In the countries of interest, most breeding programs receive core funds and some receive both core and donor funds. In this study, both core and donor funds were considered. Additionally, all expenditures incurred prior to the research program were not considered because, as Belli et al. (2001) point out, costs incurred in the past are sunk costs that cannot be avoided; therefore, they should be (and were) ignored from the analysis. Furthermore, extension expenditures that would have been spent regardless of the current research program were also excluded from analysis. However, in two of the countries of interest (Honduras and Ecuador), the research programs provide financial and technical assistance to local agricultural research committees (CIAL), 18 which complement the work of plant breeders through participatory varietal selection and/or participatory plant breeding. For these countries, these costs were included. Once the duration and scope of the research are defined, it is important to disaggregate costs in a way that only the costs of the project are reflected (i.e. costs are not overestimated). For example, a breeder may provide additional services that are not related to the research itself, such as teaching. Once the disaggregation categories have been specified, knowledgeable individuals (such as program leaders) can be asked to provide estimates of total program expenditures and the share of total expenditures devoted to the program of interest (Alston et al. 1998). In this study, this approach was followed. 18 A CIAL is a village-based farmer research group that, among other research activities, conducts varietal selection to develop new varieties (Ashby et al. 2000). 24 Finally, in agricultural research, it is common to observe a lag between commencing a 19 research activity and generating the new technology (i.e. new varieties) (Alston et al. 1998). It is assumed that this lag will be of six years because bean-breeding programs usually take five to seven years to develop and release a new bean variety, and multiply and distribute seed. During this period, only expenses are generated. 2.4.3 Measures of Project Worth After the stream of program benefits and costs are estimated, it is necessary to estimate the returns to research in each country. For this, two economic measures were used: Net Present Value (NPV) and Internal Rate of Return (IRR). These measures are useful because they compress the annual flows of benefits and costs into a summary statistic by aggregating the flows over time, which allows comparison and evaluation of alternative investments (Alston et al. 1998). The NPV is commonly used for ex ante research evaluation; however, in this study it was used to estimate ex post research benefits. NPV estimation combines the stream of program benefits and costs over the period of the research. The decision rule is simple: a program is profitable if NPV > 0. A NPV greater than zero means that the initial investment plus the opportunity cost of capital are recovered (AEC-865 2008). The formula for calculating NPV is: (2.11) T B  Ct NPV   t t t 1 (1  r ) 19 Lags are also observed between generating the new technology and seeing it adopted (Alston et al. 1998). However, the estimation of logistic adoption curves takes this into consideration. 25 where T is the total number of years under consideration, Bt is the calculated (from Equation (2.8)) value for annual research benefits in year t, Ct is the program cost in year t, and r is the discount rate. The numerator in Equation (2.11) is the net benefit. The IRR is commonly used for ex post research evaluation. The decision rule is also simple: a program is profitable if the IRR is greater than the opportunity cost of capital. The IRR is estimated by setting the NPV from Equation (2.11) equal to zero and solving for r; that is: T (2.12) Bt  Ct t t 1 (1  IRR ) 0 Therefore, the IRR is the rate of return that will make the present value of benefits equal to the present value of costs (Alston et al. 1998). The IRR is usually estimated by “trial and error,” although available software easily does this. Although useful and relatively easy to estimate, both NPV and IRR have their advantages and disadvantages (AEC-865 2008). Although NPV reflects the size of the investment (i.e. shows the absolute magnitude of incremental net benefits) and the decision rule is simple (i.e. a project is profitable when NPV > 0), it requires explicit specification of the discount rate and implicitly assumes that profits are reinvested at a rate equal to the chosen discount rate. In contrast, the advantage of IRR is that it does not require knowledge of the exact opportunity cost of capital; however, it does not reflect the size of the investment. Although projects could be ranked by either NPV or IRR, NPV is usually preferred because it considers the size (i.e. magnitude) of the benefits and investments (Alston et al. 1998). In the literature, it is a common practice to estimate both economic measures of project worth. Therefore, in this study both measures were estimated. 26 Finally, it is worth mentioning that, when estimating NPV, the higher the discount rate, the less weight is placed on future benefits. Therefore, one must be careful in deciding what discount rate to use. Often, high discount rates are used when estimating NPV of programs located in developing countries (vs. programs in developed countries) because the risk of investing in developing countries is higher (AEC-865 2008). However, recent literature (Alston et al. 1998; Maredia et al. 2010) suggests using a lower (e.g. 3-5%) real discount rate (adjusted for inflation). Using real discount rates is especially common when evaluating the profitability of medium- to long-term, risk-free projects (Alston et al. 1998; Bazelon and Smetters 2001). In this study, a real discount rate of 4% (average of the range mentioned above) was used. 2.4.4 Sensitivity Analysis Sensitivity analysis is commonly used to assess the robustness of the results (Alston et al. 1998). These authors point out that with this type of analysis, it is important to recognize the fact that the parameters are mutually dependent (e.g. adoption rates likely depend on yields). Therefore, varying each parameter and considering all combinations is not adequate. In this study, NPV and IRR were estimated for a base scenario in which breeders’ estimations of adoption rates, econometric estimations of yield gains, and a 4% real discount rate (among other parameters) were used. To test the robustness of results, the NPV and IRR were also estimated using a 10% discount rate (as was used by Mather et al. 2003 and Mooney 2007). Thus, for the sensitivity analysis, the following parameters from the base scenario were modified as follow: 1. Scenario A: Type II yield gains and 2010 adoption rates were modified to reflect a +10% difference from the base scenario. All other parameters were held constant. 27 2. Scenario B: Type II yield gains and 2010 adoption rates were modified to reflect a -10% difference from the base scenario. All other parameters were held constant. Finally, the minimum Type II yield gains and 2010 adoption rates needed to recover investment (i.e. when NPV=0, or break even values) were estimated separately (e.g. Type II yield gains were changed until NPV=0, while holding all other variables constant). 2.5 Data Description The data used in this study were obtained from different sources, including: experimental trials yield data, expert opinion estimates, secondary sources, and parameters from previous studies conducted in the region. The experimental yield data was obtained from the following bean breeding programs: (1) the Programa de Investigaciones en Frijol (PIF) of Zamorano in Honduras and (2) the Programa Nacional de Leguminosas y Granos Andinos (PRONALEG-GA) in Ecuador. Before explaining the data, it is important to provide general information about the breeding process in the countries of interest. For this, first, the breeding programs of Central America 20 and Ecuador are described. Second, a review of the outcomes of these programs during the past two decades is provided. Finally, the data used to estimate benefits and costs are described. 20 Which include Honduras, Guatemala, El Salvador, Nicaragua and Costa Rica. Although Guatemala is excluded from the analysis, it has collaborated in testing lines in the last decade. 28 2.5.1 Bean Research Programs Central America 21 Profrijol, a regional bean research network established in 1981 by CIAT and supported by the Swiss Agency for Development and Cooperation (COSUDE) was the only bean research network conducting bean research in Central America during the 1980s and 1990s. This network included researchers from the national agricultural research systems (NARS) of Mexico, Guatemala, Honduras, El Salvador, Nicaragua, Costa Rica, Panama, and the Caribbean countries of Haiti, the Dominican Republic, Cuba, and Puerto Rico (Rosas 2010a). During this period, Zamorano, a private university located in Honduras, had little participation in this network because only NARS conducted research to generate new varieties. In 1983, the Bean/Cowpea Collaborative Research Support Program (B/C CRSP) became involved in bean research in the region; however, its impact was small. In 1990, although Zamorano and the University of Puerto Rico (UPR) began participating in Profrijol, they did not generate breeding materials through crosses. 22 In addition, these institutions received only limited funding from Profrijol (Rosas 2010a). In 1996 Zamorano, using funds from the B/C CRSP and Profrijol, was given a mandate to lead efforts to breed small red beans for the region. 23 In 1999, Zamorano also became responsible for breeding small black beans for Central America (mostly for Guatemala). In 2002, COSUDE’s funding to Profrijol ended and CIAT’s participation in the region was drastically 21 The information in this section comes from Rosas (2010a). 22 These institutions were in charge of selecting bean varieties with heat tolerance and biologic nitrogen fixation (BNF) properties only. 23 In the same year, UPR took the lead of the breeding process of Andean beans for Panama and the Caribbean region. 29 reduced (Rosas 2010a). However, given that the network (established with Profrijol) provided a great advantage for testing and disseminating breeding materials, Profrijol members continued to collaborate and the B/C CRSP became the major supporter of this (now informal) network (Rosas 2010a). 24 Since then, Zamorano’s bean program has provided leadership to the region’s bean research network, which currently includes NARS from Guatemala, El Salvador, Honduras, Nicaragua, Costa Rica, Puerto Rico, and Haiti. 25 In 2004, CIAT, in collaboration with this regional network, implemented Agrosalud, a biofortification research project designed to benefit Central American and Caribbean (Cuba and Haiti) countries through the development, promotion, and dissemination of biofortified crops (including beans). By 2007, the Dry Grain Pulses CRSP (follow up to the B/C CRSP) unofficially became the major supporter of the regional network through funds provided to Zamorano (Rosas 2010a). One of the major contributions of Profrijol was the establishment of regional bean nurseries (or trials) in which lines from different breeding programs were put together in nurseries that were distributed to collaborators for testing. These nurseries generated information needed to select materials 26 adapted to a wide range of environments. Currently, Zamorano’s bean program is responsible for supplying breeding material which is included in nurseries throughout the region, including: VIDAC (Central American Adaptation Nursery) and ECAR (Central American Adaptation and Yield Trial). 24 Although the B/C CRSP never provided funds directly to Profrijol, it provided funds to Zamorano’s breeding program. Therefore, it has indirectly supported the bean network because Zamorano’s bean breeding materials are used by NARS throughout the region. 25 This project is coordinated by CIAT and funded by the Canadian International Development Agency (CIDA). 26 The words “materials” and “lines” are used interchangeable in this document. 30 Figure 2.3 illustrates the sequence of steps required to generate a new bean variety in Central America. As the figure illustrates, Zamorano’s bean program makes crosses and puts together the regional nurseries of homogeneous materials that are distributed to collaborators. These nurseries also contain materials from CIAT and UPR (Rosas 2010a). From the regional nurseries, the NARS select materials to include in their own national nurseries. 27 Over the breeding process, while the number of lines decreases in each nursery, the plot size per line increases. These (regional and national) nurseries are used to select material that is used to develop new bean varieties. 27 Sometimes, a particular program (from other country) requests Zamorano to make crosses of specific lines that they want to improve. After the crosses are made, Zamorano sends segregating lines to that particular program for breeding. This case is not illustrated in Figure 2.3. 31 Figure 2.3. Flow of breeding materials in the regional and national nurseries in Central America. For interpretation of the references to color in this and all other figures, the reader is referred to the electronic version of this dissertation. One key step is that, together with the regional nurseries, Zamorano provides a spreadsheet for NARS to collect data from these trials and a copy of the data is returned to Zamorano for further analysis. Each year, approximately 50% of the datasets (i.e. half the number of nurseries) are returned to Zamorano. Zamorano uses the information from the regional nurseries to select the best materials until a variety is released. Without this collaboration, testing lines would be limited and more expensive (Rosas 2010a). 32 Ecuador Ecuador’s national program on food legumes and Andean grains, El Programa Nacional de Leguminosas y Granos Andinos (PRONALEG-GA) is in charge of conducting bean-breeding activities in the country. PRONALEG-GA is located at the Santa Catalina Research Station in Quito, Ecuador and is part of the country’s national agricultural research institute, El Instituto Nacional Autónomo de Investigaciones Agropecuarias (INIAP). PRONALEG-GA consolidated its activities in 1990 and, since 1994 it has supported other experimental stations throughout the country (INIAP 2009). In the 1990s, the PRONALEG-GA program collaborated with Profriza, an Andean bean research network established by CIAT and supported by COSUDE. 28 During this period, PRONALEG-GA depended on CIAT to generate new varieties (i.e. no crosses were made in Ecuador). Although PREDUZA, a Dutch organization provided funding to PRONALEG-GA from 2000-2004 (Mooney 2007), since 2003, the CRSP has been the major external supporter of bean research in Ecuador. Through this collaboration, PRONALEG-GA has been able to make its own bean crosses, which has reduced its germplasm-dependence on other institutions (Peralta 2010). Although several bean varieties were released prior 1990, the varieties that have had the largest impact were developed post 1990. In addition, in 2000 PRONALEG-GA implemented participatory research methods to develop and disseminate new varieties (INIAP 2009). Although farmers are involved in the breeding process, segregating materials are evaluated at the Santa Catalina Research Station. 28 Profriza was established following the Central American model of Profrijol. Both programs were coordinated by CIAT and supported by COSUDE. 33 Although most of its research is done on bush-type beans, the PRONALEG-GA program also conducts research on climbing beans and other legumes. Currently, PRONALEG-GA collaborates with international institutions (e.g. CIAT, Michigan State University), and with local NGOs and local farmers groups (CIALs). Its current collaboration with Michigan State University (MSU) has allowed it to supply black and red mottled bean lines to Rwanda’s bean program, the first time that this program has supplied materials to another continent (KII, 2010a). PRONALEG-GA has three main nurseries for testing advanced lines: (1) Prueba, (2) Comprobación and (3) Producción. These nurseries generally include different market classes and are tested in different locations under farmer conditions. Similar to the Central American case, while the number of lines decrease from one nursery to another, the plot size per line increases. The Producción nursery is the last step before releasing a variety. Participatory Research: Participatory plant breeding (PPB) vs. participatory varietal selection (PVS) Participatory research is a methodology through which farmers are involved in the breeding process by providing them with either early or late generation materials to select from (Ceccarelli et al., 2000). In principle, farmers can be provided with a set of segregating (i.e. >F3) 29 or homogeneous (i.e. >F6) materials to select from and, assisted by scientists, release a variety. There are two major types of participatory research: participatory plant breeding (PPB) and participatory varietal selection (PVS). While PPB is the process where farmers are given a set of segregating materials to select from, PVS is the process where farmers are given a set of homogeneous materials to select from (Ceccarelli et al., 2000). 29 In the breeding literature, Fi is the ith generation after crosses were made. Thus, F3 is the third generation, etc. 34 Within the countries of interest, all countries are currently implementing participatory breeding approaches (KII, 2010a). However, only in Honduras, Nicaragua, Costa Rica and Ecuador, have new bean varieties been released using this approach. While participatory breeding was implemented in 2006 in El Salvador, no bean varieties have yet been released using this methodology (KII 2010a). Furthermore, this methodology is implemented differently across countries. The major differences lie in the degree of farmer participation (i.e. PPB vs. PVS) and the type of group of farmers included in the process. In Costa Rica, Ecuador, El Salvador, and Nicaragua, farmers participate in the breeding process by evaluating and selecting advanced (i.e. >F6) materials. 30 In contrast, in Honduras, farmers evaluate both segregating (i.e. >F3) and 31 advanced (i.e. >F6) materials. Regarding the types of group of farmers participating in plant breeding, in Honduras and Ecuador, the process is implemented in collaboration with farmers organized in CIALs. In contrast, in Costa Rica, El Salvador, and Nicaragua, this process is implemented with farmers organized in cooperatives or associations. The major difference between the CIALs and the cooperatives is that the former are smaller groups of farmers set up to conduct research locally and are usually located in niche (marginal) environments. 2.5.2 Bean Varieties Released between 1990-2010 Between 1990 and 2010, a total of 90 bean varieties (all market classes) were released in the five countries of interest. However, some of the varieties released in Central America were 30 However, in Costa Rica and Ecuador, farmers sometimes evaluate segregating materials at experimental stations. 31 The farmers’ group implements either PPB or PVS. That is, the same group does not implement both methods at the same time. 35 released in several countries, usually with a different name in each country. The varieties released in more than one country are: Dorado and Amadeus 77 (four countries each), Tío Canela 75 (three countries), and Deorho, Carrizalito, DOR 390 and Don Silvio (two countries each). Hence, 78 unique varieties were released in all five countries. From these, most varieties were small reds or reds (46 of 78), followed by red mottled (10 of 78), and blacks (7 of 78). Furthermore, at least 44 of the 78 varieties were developed using (direct or indirect) CRSP funding (Table A 2.2-Table A 2.6). In Costa Rica, as many as 18 improved bean varieties have been released in the last two decades, 56% of them since 2000. The PITTA-Frijol network released all of these varieties. From these varieties, three have Tío Canela 75 (Line ID: MD 3075), a variety released in Honduras in 1996 with resistance to BGYMV, as a parent. Surprisingly, although black beans are most widely consumed in Costa Rica, the last black variety was released in 2000 (UCR 55) and it was not widely adopted (Table A 2.2). In El Salvador, only nine improved varieties have been released in the last two decades, five since 2000. Furthermore, all varieties released over the past 20 years are small red varieties and were released by CENTA, the national center in charge of bean research (Table A 2.3). Five of the nine varieties were developed using indirect funding provided by the CRSP (through germplasm provided by Zamorano), all five in the last decade. In Honduras, as of 2010, 21 bean IVs have been released in the last two decades, 76% in the last decade and 57% (12 out of 21 IVs) were developed using participatory methods (Table A 2.4). From these varieties, four have Amadeus 77 germplasm (Line ID: EAP 9510-77), a variety released in the country in 2003, which is resistant to BGYMV and has a light red seed color, as a parent. Amadeus 77 was developed using Tío Canela 75 as one of its parents. Moreover, three 36 other varieties had Tío Canela 75 as a parent, which brings the total number of varieties with genetic share of Tío Canela 75 to eight. Similar to El Salvador, all varieties released to date are small reds. In addition, 81% of the varieties released over the last 20 years were developed using CRSP funding. Three institutions supported the development of varieties using participatory methods: CIALs, NGOs assisting farmers, and Zamorano. In Nicaragua, 16 bean IVs have been released in the last two decades, less than half were released since 2000 and all varieties were developed using participatory methods (Table A 2.5). Three of these varieties have Tío Canela 75 as one of its parents. In Nicaragua, most varieties released to date are small reds. Although most varieties were released by INTA, one of these varieties (INTA Pueblo Nuevo JM) was released in collaboration with CIPRES, an NGO. Among all of the countries, Ecuador has released the highest number of bean varieties in the past two decades--26 varieties in total. From these, 17 varieties were developed using direct CRSP funding, 16 varieties were released in the last decade, and 12 varieties were developed through participatory methods (Table A 2.6). Although the CRSP is currently the main external supporter of this program, CIAT played an important role in the 1990s, when it was the main supplier of germplasm to the program. 32 In contrast to all other countries, PRONALEG-GA’s efforts have focused on developing varieties of several market classes (Table A 2.6). However, the three main market classes are: red mottle (38% of varieties belong to this market class), yellow (23% of varieties) and white (12%). Furthermore, both INIAP and CIAL groups are credited with the release of most varieties developed through participatory breeding, with the exception of Canario Guarandeño, Libertador, and Canario Siete Colinas, credited only to INIAP. 32 For example, Paragachi Andino, released in 2009, came from a CIAT cross. 37 2.5.3 Research Benefits In order to estimate the value of research benefits, yield gains and adoption rates were needed. The data used to estimate these parameters are explained below. 2.5.3.1 Estimating Yield Gains There are two types of yield gains derived from adopting IVs: Type I, in areas where farmers replace traditional varieties with IVs, and Type II in areas where farmers replace old IVs with new IVs. Type I gains were obtained from previous studies and are detailed in section 2.6.4. To estimate Type II gains from new bean varieties released through time (i.e. Equations 2.3b, 2.3c, 2.4a and 2.4b), experimental yield data were used. The rest of this sub-section explains the Central American and Ecuadorian data used for estimations of Type II yield gains and the market classes analyzed. Central America As explained above, the experimental yield data were obtained from Zamorano’s bean breeding program. These data were used to estimate λ for red beans. 33 There were two nurseries from which data could be drawn: VIDAC and ECAR. The ECAR data were used because it had an adequate number of observations (not as many as the VIDAC), it had three repetitions per line evaluated (vs. no repetitions in the VIDAC), and it was planted in several countries each year (Table A 2.1). Although the ECAR nursery has been implemented since the 1990s, data on red varieties were only available for the period 1999-2009. The ECAR included 14 advanced lines plus one local check (usually a traditional variety) plus one universal check (Dorado, one of the first IVs 33 Although there is a similar nursery for black beans, it was not possible to estimate λ for this market class because none of the few black varieties released were in the dataset (since they were released in 2001 or before, thus they likely were evaluated before 1999). 38 released in the region), for a total of 16 lines per nursery. Each of the (16) lines had three repetitions (Table A 2.1). The data were averaged across repetitions. Although the trial dataset contained information for other countries outside the region of interest, only information from Costa Rica, El Salvador, Guatemala, Honduras, and Nicaragua was used in the analysis since we were interested in the effect of IVs in this region. 34 On average, the ECAR nursery was planted in six locations across four Central American countries each year (Table A 2.1). The data were averaged across locations to obtain one yield observation per country per year. As explained above, this would control for the fact that in some years, the trial was not planted in every location. Thus, the data used reflected average yields at the country level. Although many varieties were released in the last decade in the region, the ECAR dataset did not contain yield information for all varieties. Table 2.1 summarizes which varieties were included in the dataset. As expected, data for varieties released in Honduras were included the most. In contrast, the dataset contained information for only two of the varieties released in Nicaragua, for example. Ecuador Similar to Central America, there were three nurseries from which data could be drawn for the analysis: Prueba, Comprobación and Producción. The Prueba data were used because of similar reasons to the ones stated above: the number of lines was large, there were several repetitions per nursery (although the repetitions were in different locations, not in the same location), and it was the most common nursery for which information was available. 34 Although Guatemala is not included in the study, the trial data from this country was included in the dataset because of its proximity with El Salvador and Honduras and to increase the number of observations. 39 Table 2.1. Red bean varieties included in the red ECAR trial data. Zamorano, Honduras, 1999-2009. Country(ies) where 1 2 Year of release released Variety Name Line ID in trial data 1989, 1990, 1992, ES, HND, GUA, CR, Dorado DOR 364 1992, 1993 NIC 2002, 2002, 2003, ES, NIC, HND, CR Amadeus 77 EAP 9510-77 2003 2003 HND Cedron PTC 9557-10 2003 HND Cayetana 85 PRF 9653-16B-2A 2003, 2004 HND, CR Carrizalito EAP 9510-1 2005 ES CENTA Pipil PRF 9653-16B-3 2007 HND Don Cristobal SRC 1-12-1-8 2007 CR Tongibe BCH 9901-14 2007 HND Cardenal MER 2226-41 2007, 2008 HND, ES Deorho SRC 2-18-1 2008 ES CENTA C.P.C. PPB 11-20-MC 2009 HND Briyo AM IBC 306-95 2009 HND La Majada IBC 301-182 Source: Programa de Investigaciones en Frijol Metadata, Zamorano, Honduras. 1 When more than one year of release, the first year corresponds to the first country listed in the column to the right; the second year corresponds to the second country, etc. 2 CR = Costa Rica; ES = El Salvador; GUA = Guatemala, HND = Honduras; NIC = Nicaragua. This dataset contained information for the period 2003-2010. It included an average of 13 advanced lines plus several checks (Table A 2.1) that were not included in the dataset because these were different for each market class and these varied across locations and years. As with the Central American dataset, the data were averaged across repetitions (when available) and across locations. In contrast to Central America, this nursery was planted in the same location for 1-2 years. This was because, once farmers tested the Prueba trial for one or two seasons, they 40 selected the best lines and advanced them to the next nursery (i.e. Comprobación) and they did not plant the prior nursery until a new breeding cycle began (generally several years later). Similar to the Central American case, although many varieties have been released in the last decade, the dataset did not contain yield information for all varieties. Table 2.2 summarizes which varieties were included in the dataset. As expected, data for red mottled varieties were the most common. Therefore, due to limitations in the number of observations, λ was estimated only for red mottled bean varieties developed by Ecuador’s breeding program. Thus, Equation (2.4) was estimated for (a) small red bean varieties in Central America and (b) red mottled bean varieties in Ecuador. 41 Table 2.2. Improved bean varieties included in the PRUEBA trial data. PRONALEG-GA / INIAP, Ecuador, 2003-2010. Year of Variety 1 release ID Variety Name 1996 I427 Blanco Imbabura 2003 I422 Blanco Belen 2003 I423 Canario 2004 I414 Yunguilla 2004 I424 La Concepcion 2004 I424 La Concepcion 2004 I425 Blanco Fanesquero 2004 I425 Blanco Fanesquero 2005 I420 Canario del Chota 2005 I420 Canario del Chota 2007 I427 Libertador 2007 I428 Canario Guarandeno 2009 I429 Paragachi Andino 2009 I429 Paragachi Andino 2009 I430 Portilla 2009 I430 Portilla 2009 I480 Rocha 2009 2010 2010 2011 I480 I481 I482 I483 Rocha Rojo del Valle Afroandino InTag Market Class White White Yellow Red mottled Purple mottled Purple mottled White White Yellow Yellow Red mottled Yellow Red mottled Red mottled Red mottled Red mottled Yellow Yellow Red mottled Black Purple mottled Line ID in trial data Blanco Imbabura I-Blanco Belen I423 Canario I414 Yunguilla I424 Concepcion MIL UNO ABE4 I-Blanco Fanesquero ACE1 I420 Canario del Chota I-Libertador I-Guarandeno AND 1005 I429 Paragachi Andino I430 Portilla Yunguilla X Mil Uno, S23 ACE1 x (Cocacho x San Antonio) s26 p1 I480 Rocha TP6 A55 (Concepcion x (G916 x Concepcion))-1 Source: INIAP/PRONALEG-GA Metadata, Ecuador. 1 Sometimes two names shown because the variety had two line IDs. Thus, information in the last column refers to the same variety. 2.5.3.2 Estimating Adoption Rates In order to estimate the logistic diffusion curves for IVs for each country, three parameters were needed: (1) the current adoption rates, At, (2) the base year adoption rates, Ab, MAX and (3) the maximum adoption rates, A . 42 (1) Current adoption rates of IVs, At. The current total (i.e. for all IVs) adoption rates were obtained from estimations provided by bean breeders in each country. Breeders generally estimate adoption rates taking into consideration farmers’ re-use of grain as seed and parameters from previous studies. Although Maredia et al. (2010) used bean seed sales data to estimate adoption rates in Michigan, U.S., this approach was not appropriate in the Latin American context since most farmers do not purchase seed. Current adoption rates reflect 2010 levels of adoption of IVs in each country, for which bean breeders were asked to estimate adoption rates of IVs in 2010. In most countries of Central America farmers grow only one market class. Thus, the adoption rates reflect the same market class. However, in Costa Rica and Ecuador, farmers plant several market classes. Therefore, in these two countries, experts were asked to specify the share of adoption of IVs to each market class. In addition, in Ecuador both bush and climbing beans are produced. Since our interest is only on bush beans, climbing bean estimations of adoption were not collected. 35 Furthermore, the analysis only focused on the northern provinces of Carchi and Imbabura since this is where most of PRONALEG-GA’s legume breeding effort is targeted. Thus, adoption rates only refer to adoption of bush beans in northern Ecuador. Although seed distribution/sales data were collected, these data were used only to demonstrate the strength of the seed systems in each country. Seed data suggest that the seed systems may be stronger in Central America than in Ecuador (Table A 2.7). However, in all Central American countries with the exception of Costa Rica, governments have implemented 35 This may underestimate the economic benefit from bean research in Ecuador since a few improved climbing beans were released by PRONALEG-GA. 43 programs that distribute free or subsidized seed to farmers. Therefore, the apparent strong seed system in these countries highly depends on the continuity of the government-funded seed programs. Key informants related to the bean subsector suggested that without these government programs, seed production will most likely drastically decline because (a) most farmers do not purchase seed and (b) the estimated market price of the seed distributed by these programs (if it were sold at full price) is too high (KII 2010a; KII 2010c; KII 2010d). (2) Base year adoption rates of IVs, Ab. To estimate the diffusion curve, the adoption rates for 1996 were obtained from previous research conducted in the region. The year 1996 was used as the base year because, in most countries, many new varieties were released after this year. The adoption rates for 1996 and 2010 were used together to estimate the diffusion curves (i.e. total adoption rates over time) for all countries. MAX (3) Maximum adoption rate of IVs, A . Since most bean programs are mature, it is expected that 2010 adoption rates are approaching the maximum levels of adoption. Thus, it was assumed that the maximum adoption rate is two percentage points above the current adoption rate. 36 Given that many breeders reported high adoption rates, this assumption is reasonable. These three parameters of adoption rates provided the setting to evaluate the economic effect of bean IVs under bean breeders’ estimations of adoption, referred to as the ‘base scenario’ from now on. For Scenarios A and B in the sensitivity analysis, while the adoption rates for 2010 (hence maximum adoption rates) were modified to reflect a +10% difference from the base scenario, the adoption rates for 1996 were held constant (i.e. logistic curve had a pivotal shift). 36 Except for Costa Rica, where adoption of IVs has decreased. 44 2.5.4 Research Costs As mentioned in Section 2.4.2, obtaining cost information is often difficult and requires knowledge about the institutional history of the breeding programs. The history of the breeding programs was explained above. Since Zamorano’s breeding program supplies breeding materials to all NARS in Central America, the costs of generating these materials were imputed to Zamorano. Although this overestimates Zamorano’s costs, it is impossible to attribute these costs to the different programs. Furthermore, costs incurred by donors in their respective countries (e.g. U.S. costs for the DGP CRSP) were not included in these cost estimations. Bean breeding program leaders were asked to identify their 2010 external sources of funding and how much they received from each source. In addition, they were asked to estimate the amount of funding they received during the last ten years from large donors. Funding provided by large donors for several years was easily accessible (e.g. DGP CRSP). However, program leaders found it difficult to estimate their annual core budget (i.e. from their own 37 institution). To estimate the core budget of each program, program leaders were asked how many staff members their programs employed in 2010 and the share of their time devoted to beanrelated activities (to estimate their bean time equivalent, BTE). Furthermore, they were asked to classify their staff by education level and state whether the number of staff has increased/decreased/remained constant over the last decade. Over the past ten years, the number of staff has remained constant across all education levels in Nicaragua and Honduras (Zamorano only). In contrast, the number of staff has decreased across all education levels at DICTA in 37 Except for PRONALEG-GA (Ecuador) and Zamorano (Honduras), where this information was reported without distinction in the total budget and/or was available in the literature (see Mather 2003 and Mooney 2007). 45 Honduras. However, in Costa Rica, while the number of staff holding a Ph.D. degree has remained constant, the number of staff holding a M.Sc., B.S., or F = 0.000 R-squared = 0.9565 Coefficient p-value Variables Year dummy variables (1=Yes): 2000 0.09 ***0.000 2001 0.38 ***0.000 2002 -0.03 0.227 2003 0.36 ***0.000 2004 0.12 ***0.000 2005 -0.10 ***0.006 2006 0.14 ***0.000 2007 0.12 ***0.000 2008 0.34 ***0.000 2009 0.21 ***0.000 Country dummy variables (1=Yes): Costa Rica -0.42 ***0.000 El Salvador -0.33 ***0.000 Guatemala -0.91 ***0.000 Nicaragua -0.31 ***0.000 0.0049 ***0.000 Vintage variable (year of release) Constant -2.11 0.124 *, **, *** indicates the corresponding coefficients are significant at the 10%, 5%, and 1% levels, respectively. Year 1999 and country Honduras were excluded to avoid the dummy trap. Robust standard errors used to estimate p-values because variances are not equal (Prob > Chi2 = 0.0451). Source: Programa de Investigaciones en Frijol Metadata, Zamorano, Honduras. The yield gain results, combined with the information in Table 2.5, indicates that, in Central America, the gain in yield potential averaged roughly 10 kg/ha/year. 51 This number was obtained by multiplying the mean yields of each year by the yield gains, and averaging across years. 51 Is likely that farmers obtained lower yield gains since their yields are generally lower than experimental yields. This fact was accounted for when estimating the economic benefits by using FAOSTAT yields (for the base year) instead of experimental yields. 61 Honduras In Honduras, the breeding program of Zamorano uses yield information from trials conducted in the region (i.e. all countries) to decide which lines to advance to the next stage of breeding. Therefore, trial data from the region were used to estimate yield gains for Honduras. The difference with respect to the estimations discussed above, is that while the estimations for Central America included all IVs released in all Central American countries, the estimations for Honduras only included varieties released in Honduras. The results of the OLS estimation of Equation (2.3b) for Honduras are presented in Table A 2.14. As these results are similar to the ones discussed for Central America as a whole they are not discussed further. The results of the vintage model suggest that the gain in yield potential from varieties released in Honduras from 1989 to 2009 averaged 0.56% (Table 2.8), which is slightly higher than for Central American countries as a region. Therefore, the vintage results suggest that, in Honduras, the gain in yield potential averaged roughly 12 kg/ha/year, which is slightly higher than in Central American countries as a region. 62 Table 2.8. Linear regression results of the vintage model using experimental yields of small red bean varieties released in Honduras. 1999-2009. N = 88 Prob > F = 0.000 Adj. R-squared = 0.9616 Coefficient p-value Variables Year dummy variables (1=Yes): 2000 0.08 ***0.000 2001 0.39 ***0.000 2002 -0.07 ***0.005 2003 0.38 ***0.000 2004 0.13 ***0.000 2005 -0.06 **0.028 2006 0.15 ***0.000 2007 0.11 ***0.000 2008 0.34 ***0.000 2009 0.20 ***0.000 Country dummy variables (1=Yes): Costa Rica -0.46 ***0.000 El Salvador -0.30 ***0.000 Guatemala -0.92 ***0.000 Nicaragua -0.30 ***0.000 0.0056 ***0.000 Vintage variable (year of release) Constant -3.48 **0.012 *, **, *** indicates the corresponding coefficients are significant at the 10%, 5%, and 1% levels, respectively. Year 1999 and country Honduras were excluded to avoid the dummy trap. Source: Programa de Investigaciones en Frijol Metadata, Zamorano, Honduras. Ecuador In Ecuador, the OLS estimation of Equation (2.3c) suggests that yields were significantly higher (at least at the 10% level) in most years compared to 2003. Although most IVs yielded more than Yunguilla (except for Paragachi Andino), none of these differences were statistically significant (Table A 2.15). 63 The results of the estimation of Equation (2.4b) suggest that the gain in yield potential from red mottled varieties released in Ecuador from 2004 to 2010 averaged 1.68% (Table 2.9), which is slightly higher than expected and much higher than the gains found for Central America and Honduras. Therefore, this result together with the information in Table 2.6 suggests that in Ecuador the gain in yield potential averaged roughly 21 kg/ha/year, much higher than in all other countries in this study. Table 2.9. Linear regression results of the vintage model using experimental yields of red mottled bean varieties released in Ecuador. 2003-2010. N = 26 Prob > F = 0.000 R-squared = 0.9491 Coefficient p-value Variables Year dummy variables (1=Yes): 2004 0.58 ***0.000 2005 1.28 ***0.000 2006 0.57 ***0.000 2007 0.87 ***0.000 2008 0.35 ***0.000 2009 1.19 ***0.000 2010 0.81 ***0.000 0.0168 *0.051 Vintage variable (year of release) Constant -27.46 0.106 *, **, *** indicates the corresponding coefficients are significant at the 10%, 5%, and 1% levels, respectively. Year 2003 excluded to avoid the dummy trap. Robust standard errors used to estimate p-values because variances are not equal (Prob > Chi2 = 0.094). Source: INIAP/PRONALEG-GA Metadata, Ecuador. 64 2.6.4 Net Present Value and Internal Rate of Return The estimates of bean yield gains, in combination with estimations of adoption rates, annual bean prices, elasticity of supply, and the annual quantity of bean produced of the market classes of interest were used to estimate the value of benefits realized at the farm level for the period of 1997-2015 for each country of interest. Once this information was obtained, annual research costs and a real discount rate of 4% were used to estimate the net present value and the internal rate of return to bean research investments in each country, assuming a six-year lag period between the start of research and the time when farmers start using the varieties. One more point is worth clarifying before discussing the results. As previously explained, Type II gains were obtained by estimating λ from equations 2.4. In contrast, Type I gains were obtained from the literature. In Honduras, Mather et al. (2003) estimated that adopters gain the equivalent to 7-16% of bean income from yield loss averted from the use of IVs. Thus, for Honduras, the Type I yield gains were assumed to be 11.5%, the average of the values reported by Mather et al. (2003). Since no other studies have reported yield gains at the farm level 52 for other Central American countries, Mather’s values were used for these countries. In Ecuador, Mooney (2007) estimated that adopters enjoy 18.4% lower unit costs when planting IVs in northern Ecuador. Thus, for this country, the Type I yield gains were assumed to be 18.4%. A summary of the NPV and IRR findings is presented in Table 2.10. Results from the base scenario suggest that in all countries except Costa Rica, investments in bean research have been profitable and provided a return well above the assumed opportunity cost of capital because the NPV is positive and the IRR is greater than the discount rate used. When the discount rate 52 CENTA (2004) provides estimations of Type I yield gains for El Salvador. However, its estimations are for test trials in farmers’ fields, which generally are much higher than yields obtained by farmers. Thus, since these were judged to be too high, Mather’s gains were used. 65 was increased to 10%, NPV was also greater than zero for all countries except Costa Rica, suggesting that the results are not greatly affected by the discount rate. The net losses found for Costa Rica are due to the fact that (a) the area planted to beans has decreased since 1996 (and only the red-bean share was included in the estimations) and (b) the adoption rates between 1996 and 2010 have also decreased. Therefore, net losses were expected. Although this is true for small red beans, it is possible that positive gains could be found for black beans because (a) most farmers have adopted the black bean IVs Brunca (released in 1982) and Guaymi (released in 1996) and (b) the area planted to black IVs is more than twice the area planted to red beans (KII 2010a, 2010d). However, estimating the economic impact of black beans for Costa Rica was not possible because only a few varieties have been released recently and available experimental data did not include yield information for these varieties. Table 2.10. Summary of net present value (NPV) and internal rates of return (IRR) estimations of investments on bean research in Central America and Ecuador. 1991-2015. Base NPV($) -2,016,054 77,510,816 58,250,437 214,002,964 10,920,047 Scenario (in constant 2009 US$) Scenario A Scenario B IRR NPV($) IRR NPV($) IRR -5% -1,610,978 -3% n.e. n.e. 40% 93,170,299 43% 62,688,130 37% 34% 73,724,174 37% 43,698,030 31% 42% 254,621,317 45% 175,583,202 39% 37% 13,216,135 39% 8,832,204 35% For 1997-2015 Producer surplus Country per ha per year Costa Rica 26 El Salvador 84 Honduras 63 Nicaragua 73 Ecuador 196 Central American countries 347,748,163 32% 419,904,813 35% 281,969,362 32% 72 All countries 358,668,210 32% 433,120,948 35% 290,801,566 32% 74 Source: Generated by the Author. NOTES: n.e. = not estimated. Scenario A assumes a 10% increase over estimations of Type II yield gains and 2010 adoption rates simultaneously. Scenario B assumes a 10% decrease over estimations of Type II yield gains and 2010 adoption rates simultaneously. Surplus per hectare per year estimated by dividing each year's total surplus (base scenario) by the area planted with IVs. 66 For El Salvador, the NPV ranged from $63 million (Scenario B) to $93 million (Scenario A) and NPV was estimated at $78 million for the base scenario, which represents a surplus of $84 per hectare planted with IVs per year. 53 Similarly, for Honduras, the NPV ranged from $44 million to $74 million and NPV was estimated at $58 million for the base scenario, which represents a surplus of $63 per hectare planted with IVs per year. For Nicaragua, the NPV under the base scenario was estimated at more than $214 million, which represents a surplus of $73 per hectare planted with IVs per year (Table 2.10). There are two reasons for this: (1) the area planted to beans has more than doubled since 1996 and (2) adoption of improved varieties has greatly increased since 1996 due to investments made by donors and the government, especially after hurricane MITCH in 1998. Although the economic benefits are more modest in northern Ecuador than in Central America, the results suggest that investments in bean research have been profitable under all scenarios, with NPV ranging from $9 million to $13 million and IRR ranging from 35% to 39%. The NPV was estimated at $11 million for the base scenario, which represents a surplus of $196 per hectare planted with IVs in northern Ecuador per year, the largest surplus among all countries. As a region (i.e. all countries), investments in bean research were profitable, generating a net present value of more than $358 million, most of which came from Central American countries, particularly Nicaragua. This is due to the fact that Nicaragua is the largest bean producer in the region and the adoption rates in this country were relatively high in 2010. Further, the governments of Nicaragua, Honduras, and El Salvador have implemented (free or 53 See last data-column of Table 2.10. 67 subsidized) seed distribution programs that most likely have contributed to the observed (and large) economic benefits. In contrast, the economic impact was small in Ecuador because the area of impact was relatively small (i.e. concentrated in the north region) and the deficient formal seed system in Ecuador has limited the access to high-quality bean seed. To overcome this limitation, PRONALEG-GA is promoting alternative ways to produce and sell high-quality, low-cost bean seed through seed producers located in villages across different regions. However, PRONALEGGA’s efforts have most likely had limited impact because the amount of seed produced and sold has been relatively small and its limited resources have not allowed to scale up these initiatives. The estimated IRR for the region was 32%, which more than offsets the opportunity cost of capital. Details about estimations of NPV and IRR for the base scenario for each country are presented in Table A 2.16 to Table A 2.20. Although estimations for scenarios A and B were made for all countries, only the estimations for the base scenario are included as appendices since the estimation procedures for scenarios A and B were similar to the procedures described above. Although there were no Type I benefits in Costa Rica because farmers have disadopted IVs through time (thus there are no new adopters), the realized Type II yield gains were not enough to recover investments in bean research. In Costa Rica, the NPV would be zero if the value of λ were 114% higher than estimated. While the benefits derived from Type II gains accounted for less than one-half of the change in total surplus in El Salvador (42%) and Nicaragua (39%), Type II benefits accounted for more than 50% of the change in total surplus in Honduras (61%) and Ecuador (53%). These values were obtained using the change in surplus values in Table A 2.17 through Table A 2.20. 68 Finally, although break-even values (i.e. values that would make NPV=0) of λ and adoption rates were separately estimated using the parameters from the base scenario, the results suggest that for all countries except Costa Rica, even if λ=0 (hence Type II gains = 0), the NPV would be positive. This is because adoption of IVs is assumed to increase (at the base scenario rate) over time, generating enough Type I benefits (from new adopters) to realize positive NPV values. Similarly, it was not possible to estimate break-even values of 2010 adoption rates for these countries since, even if the 2010 adoption rate was only 1% above the 1996 adoption rate (e.g. 26% in El Salvador in 2010 instead of 60%), NPV would still be positive. This was because even with a small increase in adoption of IVs over time, both Type I and Type II benefits are realized and these are large enough to offset the cost of research. In addition, as the 2010 adoption rate decreases (and gets closer to the 1996 adoption rate), the share of the benefits derived from Type I gains become smaller and that of Type II gains become larger. 2.7 Chapter Summary and Policy Recommendations 2.7.1 Chapter Summary There are two types of yield gains derived from the use of improved varieties: Type I gains in areas where improved varieties replace traditional varieties (i.e. new adopters), and Type II gains in areas where new improved varieties replace old improved varieties (i.e. current adopters). This essay estimated the Type II yield gains associated with varietal development of small red and red mottled bean varieties over time in Central America, Honduras, and northern Ecuador separately. Furthermore, it provided estimates of current total adoption rates of IVs in each country. The adoption rates were estimated using bean expert opinions. A base scenario was constructed using the information from econometric regressions, key informants, and secondary 69 sources, and sensitivity analysis was carried out under two additional scenarios: Scenario A where Type II yield gains and 2010 adoption rates were assumed to simultaneously be 10% higher than in the base scenario, and Scenario B where Type II yield gains and 2010 adoption rates were assumed to simultaneously be 10% lower than in the base scenario. OLS regressions using vintage models were conducted to estimate the Type II yield gains from IVs released over time. Results from the vintage model regressions suggest that the gain in yield potential from small red varieties released over time averaged 0.49% per year for Central American countries and 0.56% per year for Honduras. Similarly, the gain in yield potential from red mottled varieties released over time averaged 1.68% per year for Ecuador. Breeders estimated that adoption rates for 2010 ranged from 46% in Honduras to 82% in Nicaragua. Furthermore, breeders estimated that Amadeus 77, the most widely adopted small red IV, accounted for an estimated 49.7% of the total area harvested with beans across the Central American countries (235,028 ha). Similarly, Portilla, the most widely planted red mottled IV in northern Ecuador, accounted for an estimated 43% of the red mottled bean area harvested in northern Ecuador. Ex post benefit/cost analysis of improved small red and red mottled bean varieties for the period 1991-2015 indicate that returns to investments in bean research have been negative in Costa Rica and positive in all other countries. In Costa Rica, negative net gains were observed because both the total bean area and the area planted to small red IVs have decreased over time. Furthermore, small red beans represent a much smaller proportion of total land area planted to beans, compared to black beans. In addition, in contrast to all other Central American countries in this study, Costa Rican farmers do not receive price discounts for dark red beans; instead, they just cannot sell them. Therefore, developing small red varieties that farmers will adopt is more 70 challenging. An analysis to estimate the economic impact of black varieties could not be carried out due to data limitations. Among the countries with positive NPV, under the base scenario, the results indicate a range of IRR values from 34% in Honduras to 42% in Nicaragua. Although this range was different under Scenario A and Scenario B, the countries had the same rank under all scenarios. Furthermore, the estimated NPVs for investments in bean research under the base scenario ranged from $11 million in northern Ecuador to $214 million in Nicaragua. The disparity of benefits is due in part to the limited area of impact and deficient seed system in Ecuador, and the high adoption rates and increased government intervention through seed distribution programs in Nicaragua. Under the base scenario, the NPV was over $347 million among Central American countries and over $358 million across all countries. The regional IRR (i.e. all countries) was estimated at 32%. While the surplus per hectare per year was highest in northern Ecuador ($196/ha/yr) and lowest in Costa Rica ($26/ha/yr), the regional surplus averaged $74/ha/yr. 2.7.2 Policy Recommendations The results of the study support the following policy recommendations:  In Costa Rica, to increase adoption rates, future bean research on small red varieties should give priority to developing varieties that are more acceptable to farmers (i.e. with better market value). Furthermore, since black beans are the most widely produced market class in Costa Rica, increased efforts should be devoted to developing new black varieties.  In Ecuador, additional efforts should be devoted to increasing seed production. As briefly discussed, the seed system in Ecuador is the most deficient among all five countries. In order 71 to increase seed production, financial assistance would likely be required because PRONALEG-GA’s human and financial resources are stretched thin. Financial support from the Government of Ecuador and donors will be key to find alternatives to increase seed production in Ecuador. Although the quantity of seed distributed/sold in Honduras, El Salvador, and Nicaragua is large, the seed provided to farmers is either subsidized or free, which is not sustainable in the long term. Therefore, alternative ways to produce and commercialize low-cost high-quality seed is key to develop a sustainable seed system.  In all countries except Costa Rica, investments in bean research have been profitable. Because of this, donors and governments should continue funding bean research programs in these countries. Although investments in bean research were negative in Costa Rica, this does not imply that funding should be cut in this country. Instead, the bean program of Costa Rica needs to devote efforts to develop better small red varieties and new black bean varieties.  Most research centers in the countries included in the study are highly dependent on external assistance to conduct their activities. A continuous supply of funds from donors is necessary to support varietal development research activities.  The bean programs of Central America collaborate with each other in their breeding activities (through supply of germplasm, data, etc.) This research model has proven successful. Therefore, this model could be implemented in other regions of the world to successfully conduct breeding activities and make a positive economic impact on producers. However, one weakness of the model is that the breeding programs in this region highly depend on Zamorano’s bean program to obtain segregating lines to breed. Thus, if Zamorano’s program in Honduras could no longer provide segregating lines, other bean 72 programs would likely struggle to generate new varieties until finding another source of segregating lines. 73 CHAPTER THREE. DETERMINANTS OF MARKET PARTICIPATION DECISIONS: EVIDENCE FROM THE CENTRAL HIGHLANDS OF ANGOLA 3.1 Introduction Agricultural households can be classified into three categories based on their net position 54 relative to the market: net buyers, net sellers, or autarkic (non-participants). Market participation is both a cause and a consequence of development (Boughton et al. 2007; Barrett 2008). Markets provide households the opportunity to benefit from trade; i.e. they can sell their surpluses and purchase goods and services they need, according to their comparative advantage (Boughton et al. 2007; Barrett 2008). Further, as a household’s income increases, its demand for goods and services also increases, hence stimulating development (Boughton et al. 2007). However, the net position of the households not only depends on market prices; it also depends on households’ access to productive technologies (e.g. improved varieties, inputs, etc). and adequate private and public goods (Barrett 2008) and services. To date, price-based, top-down macro and trade policy interventions have not been enough to stimulate smallholder market participation and agricultural and rural transformation (Barrett 2008). However, understanding the impact of these policies on smallholder farmers’ market participation is important. The fact that market participation is heterogeneous has important implications when studying households’ response to governmental policy interventions and should be considered in policy response estimation (Key et al. 2000). 54 Goetz (1992) called this classification the household trichotomy. 74 Farm households are typically located in environments characterized by a number of market failures 55 (Sadoulet and de Janvry 1995, ch. 6, pg. 9). These authors point out that any market could fail for a particular household when the margin between the low price at which the household could sell a commodity and the high price at which it could buy it is large; hence the household may be better off by being autarkic. This introduces us to the concept of price bands, which are widely described in the literature (De Janvry et al. 1991; Sadoulet and de Janvry 1995; Key et al. 2000). The price band refers to the difference between the effective price paid by buyers and received by sellers (both market participants), which determines the household’s net market position. To boost market participation, one of the government and private sector’s goals could be to target investments at reducing the magnitude of the price band. This magnitude could be affected by transaction costs, the existence of shallow local markets, 56 and price risks and risk aversion (Sadoulet and de Janvry 1995). This paper analyzes the effect of gender of head, transaction costs, and productive assets on household’s marketing decisions, using crosssectional data from three provinces of Angola. Angola ended its 27-year long civil war in 2002. During the war period, each of the two combatant sides controlled the two major commodities of the country: oil and diamonds. While the Popular Movement for the Liberation of Angola (MPLA) monopolized oil exports and revenue, the National Union for the Total Independence of Angola (UNITA) controlled much of 55 De Janvry et al. (1991) demonstrated that market failure was household, not commodity specific. 56 For details see Sadoulet and de Janvry (1995, ch. 6, pg. 9). The idea is that there is a negative covariation between household supply and prices because when the harvest is good and surplus could be traded, the price falls because all other households also have good harvests, widening the price band (the opposite is also true). 75 the diamond wealth of the north and eastern interior (Munslow 1999). Although these two commodities provided significant revenues to the country, 57 during the war, these revenues were used to sustain Angola’s rival armies (Munslow 1999). The war had a large impact in the country’s infrastructure 58 and caused the demise of the rural economy and the subsequent sharp rise of the urban population (World Bank 2007). Land mines and conflict made it dangerous to stay in rural areas and to farm. Other effects were the loss of life of over 1 million people and migration (rural to urban, but also to neighboring countries) (World Bank 2007). Since the Peace Accords, continued rural to urban migration, urbanization, high population growth rate, and increasing household incomes have contributed to an increase in the demand for food in the major cities of the country. For example, the estimated 2005 annual demand for potatoes, onions, carrots, and dry beans in Luanda (the capital city) was a little over 197,000 MT, 61% of which was imported from neighboring countries, especially South Africa (World Vision 2008). Although expenditures in 2003 in energy, agriculture, mining, and transportation were high (10.2% of GDP; US$1.4 billion), by 2005, expenditure in these areas was drastically reduced to only 2.2% of GDP (or US$734 million) (World Bank et al. 2007), suggesting that rural households may still face many limitations to actively participate in markets and reduce import requirements. 57 In 1997, crude oil exports were estimated at over US $4.5 billion, while diamond production in 1998 was estimated to be worth over US $0.5 billion (Munslow 1999). However, oil production was expected to peak in 2010, after which, it is expected to decrease (World Bank et al. 2007) suggesting that the share of other sectors in the economy will grow. 58 It is estimated that US $4 billion will be necessary just to restore the road and bridge network of the country (World Bank 2007). 76 In addition to the country’s transition from war to peace, the country went from a centralized market to a free market (Munslow 1999). However, food aid and assistance programs undermine private sector investments and government control has resulted in a poorly developed trading network (World Bank 2007). Furthermore, Angola has been cut off from agricultural 59 technological advances (e.g. new crop varieties), and increasing farmers’ productivity still remains a challenge because of the disadvantages of Angola’s strong currency and high transportation costs (World Bank 2007). Understanding which factors are associated with farmers’ marketing decisions is important to target government and private donors’ resources, boost crop sales, and increase farmers’ incomes and their food security. Because of this, the study focuses on estimating the determinants of market participation and the quantity of food traded, including the effect of gender, transaction costs and productive asset endowments on marketing behavior, following Boughton et al. (2007), Barrett (2008), and Bellemare and Barrett (2006). Thus, this study will provide the government and private donors the information necessary to better target their assistance and improve smallholder farmers’ livelihoods in rural Angola. 3.2 Research Gap Many studies related to the analysis of market participation by agricultural households have focused on (1) dealing with potential problems of sample selection bias when testing hypotheses about market participation and (2) understanding the role of transaction costs and market failures on households’ marketing decisions. 59 Although Angola enjoys better rainfall than many of its neighbors, crop yields are much lower. 77 Heckman (1979) discussed sample selection bias as a specification error and provided a technique that allowed for the use of simple regression to estimate behavioral functions free of selection bias in the case of a censored sample. Examples of censored samples include data related to market wages, earnings of trainees, or earnings from selling in the market. 60 Sample selection bias arise because (1) individuals may self-select themselves (i.e. they choose whether to participate in the activity) or (2) analysts decide to use only a subset of a random sample obtained from a population (which works in the same way as self-selection). The problem with censored samples is that the functions estimated (e.g. wage or earnings) on selected samples generally do not estimate population (i.e. random sample) parameters. For example, if we are interested in estimating the effect of training on earnings and we only use trainees’ earnings, the parameter of interest (i.e. training) may be biased because it may confound the effect of the probability of receiving the training and the effect of the training itself. 61 The solution proposed by Heckman (1979) to obtain unbiased estimators was simple. 62 First, he demonstrated that the bias that results from using (non-randomly) selected samples could arise from a problem of omitted variables. Second, he proposed that, for the full sample (e.g. trainees and non-trainees), a probit analysis could be used to estimate the probability that an individual may be in the selected sample (e.g. will participate in training). Third, he 60 For example, the earnings of trainees generally do not provide reliable estimates of what nontrainees would have earned had they decided to participate in the training. Gujarati (2003) explains that censored samples are samples in which information on the dependent variable is available only for some observations. 61 A similar problem happens with truncated samples where data is observed only if a certain event is true. 62 See Heckman (1979) for a detailed explanation. 78 demonstrated that by using this probability as a regressor in the equation of interest (e.g. trainees’ earnings) one could obtain unbiased estimators. In his widely cited work, Goetz (1992) modeled the agricultural household’s discrete decision of whether to participate in markets separately from the continuous decision of how much to trade, conditional on market participation; 63 an innovation in market participation analysis at the time. Elaborating on the groundbreaking work of Goetz (1992), Key et al. (2000) studied the effect of proportional and fixed transaction costs on household supply response. They implicitly modeled the household as making the discrete market participation choice simultaneously with the continuous decision of how much to trade. 64 In constructing their agricultural structural household model, they separated the structural supply functions from the production threshold functions. By estimating this model, they were able to separately identify the effect of proportional and fixed transaction costs on supply response, while avoiding the problem of selection bias described by Heckman (1979). As noted, some authors assume households make marketing decisions sequentially, while other assume they make these decisions simultaneously. Bellemare and Barrett (2006) developed a two-stage econometric method that allowed them to test whether rural households in developing countries make market participation and volume decisions simultaneously or sequentially. Using household data from Kenyan and Ethiopian livestock markets, they found evidence in favor of sequential decision making. The major implication of this finding is that households that make sequential marketing decisions are more price-responsive and less vulnerable to trader exploitation. 63 That is, he assumed households make sequential choices: they first decide whether or not to participate in the market; then, conditional on participation, they decide how much to trade. 64 In contrast to Goetz (1992) who assumed households make sequential marketing choices. 79 Although many recent studies have focused on the effect of transaction costs; farmers’ assets and wealth also affect marketing decisions. Boughton et al. (2007) took an asset-based approach to analyze smallholder market participation in Mozambique. They assumed households made sequential marketing decisions 65 and developed a simple structural model of the household’s choice problem, facing two constraints: budget and asset allocation constraints. Fafchamps and Vargas-Hill (2005) analyzed the factors associated with coffee producers’ decision to sell at the market vs. at the farmgate. Although their study did not focus on the decision to participate in the market, 66 it provides insights about why farmers choose different places for their sales. They constructed a simple model of farmers’ form of sale choice, focusing on the relationship between wealth and farmgate sales. They estimated the model with and without access to public transportation. Barrett (2008) provides a detailed literature review about evidence on smallholder market participation in eastern and southern Africa, focusing on staple foodgrains markets. He found that the empirical evidence suggests that most smallholders do not participate as sellers 67 because they face two basic classes of barriers to entry; (1) at the micro-level, households have insufficient access to productive assets, financing, and new production technologies and (2) at the macro-level, especially in remote areas, high transaction costs limit household’s market access, market-level spatial price transmission, and trader competition. Another reason not mentioned by Barrett (2008) is that farmers may not have surpluses they could sell. 65 Similar to Goetz (1992). 66 All coffee producers are sellers because coffee is a cash crop. Therefore, household consumption may be very small, if any. 67 At least not at any significant scale. 80 Markets rarely work perfectly. Market failures are an important consideration for policy analysis because when markets fail, the household’s ability to respond to price incentives is constrained. Household modeling under missing markets is well explained by Sadoulet and de Janvry (1995). De Janvry et al. (1991) analyzed the effect of missing markets on farmers’ supply response. They developed a model of household behavior under various conditions of labor and food market failures and empirically tested their model using simulations. They found that programs directed at reducing the incidence of market failures 68 are very important to increasing the supply elasticity of households--hence increasing household’s response to price incentives. The contribution of this study is as follows. First, it uses a double hurdle approach to control for self-selection bias and provides unconditional effects of variables of interest on the quantity sold. Second, besides focusing on transaction costs, this study also analyzes the effect of productive assets on marketing decisions, following Boughton et al. (2007), gender of household head on marketing decisions. 70 69 and the effect of Third, it provides new empirical results to the rather limited literature on market participation in Africa, especially in Angola, by looking at farmers’ participation in three crop markets in Angola: potatoes, beans and onions. This paper focuses on potatoes, beans, and onions because: 71 (1) these crops, especially potatoes, are very important in the country’s agricultural sector because of their high potential to generate profits to smallholder farmers; (2) there is a strong unmet demand for these crops in 68 Infrastructure investments (which reduce transaction costs), better circulation of information on prices, and access to credit markets (an indirect source of market failure) for example. 69 Bellemare and Barrett (2006) did not explicitly study the effect of productive assets on marketing behavior. 70 Donors are interested in learning about the role of gender on household decisions, especially because after the war, many households are led by females. 71 A detailed explanation is provided in Section 3.5 below. 81 large cities of Angola that currently is satisfied by imports from neighboring countries; and (3) recent private and public investments targeted at improving supply chains in rural Angola are focusing on these crops (World Vision, 2008). Therefore, generating information about the factors affecting smallholder farmers’ marketing decisions will be valuable to target assistance to farmers. 3.3 Research Questions Although the study’s main objective is to generate information about the factors affecting smallholder farmers’ marketing decisions, it also attempts to answer the following research questions:  What are the characteristics of farmers who trade potatoes, beans, and onions in the central highlands of Angola, compared to non-traders?  For each crop, what factors affect production?  For each crop, what factors are influencing farmers’ decision of whether to sell their surpluses?  Conditional on market participation, what factors affect the quantity of food traded by farmers?  What is the unconditional effect of gender, productive assets, and transaction costs on supply of each crop?  What policies could the government and private sector participants implement, based on the empirical evidence, to boost market participation? 82 3.4 Conceptual Framework In this section, first, the economic rationale for analyzing household’s marketing decisions is explained. Then, an econometric framework is presented to empirically estimate the economic model while addressing the econometric challenges of the analysis. 3.4.1 Economic Model To analyze the factors associated with farmer’s marketing decisions, following Boughton et al. 2007 and Barrett 2008, a simple model of household choice is developed. It is assumed that C households will maximize their utility U, by consuming a vector of agricultural commodities, s , for c = 1, 2, 3 crops, and a Hicksian composite of other tradables, x. It earns income from production and possibly sale of any or all crops, and possibly off-farm income, Y, which could be earned or unearned. Crop production is determined by a crop-specific production technology, 72 f C C (A , G), which depends on the flow of inputs (e.g. fertilizer, pesticides, seed, labor) and services provided by privately held quasi-fixed productive assets, represented by the vector A. This function is also affected by the availability of public good and services, G, such as extension services, and farmer associations, because farmers may have access to price information, receive inputs or technical assistance, among other benefits that may affect output. The vector M represents farmer’s choice of whether or not to participate in each market cv cb as a seller, represented by the vector M , or as a buyer, represented by the vector M . The cv vector M 72 cv takes value 1 for every crop c the farmer decides to sell and M = 0 for crops not Due to the nature of the crops, each one could be produced using different technologies. 83 cb sold. Similarly, the vector M cb M = 0 for crops not bought. cv positive if and only if M takes value 1 for every crop c the farmer decides to buy and 73 Net sales of a particular crop, NS cb = 1 and negative if and only if M C º f C (AC, G) - sC, are = 1. However, due to data availability, the focus of this paper is restricted to comparing farmers’ choice as to whether or not to participate in each market as a seller. The parametric market price each household faces, p C cm , is affected by crop-and- C household-specific transaction costs, τ (A, G, Y, Z, NS ). That is, the household faces wide price margins (i.e. a price band) between the low price at which it could sell a crop and the high price at which it could buy that crop (Sadoulet and de Janvry 1995). 74 These transaction costs create a kinked price schedule, which leads some households to self-select out of the market for some crops (de Janvry et al. 1991; Sadoulet and de Janvry 1995; Boughton et al. 2007; Barrett 2008). Following Boughton et al. (2007) and Barrett (2008), transaction costs are assumed to be a function of household’s productive assets, A, access to public good and services, G (e.g. good roads and/or participating in farmer organizations may reduce transaction costs), liquidity from off-farm income, Y, household-specific characteristics, Z, and amount traded, NS. The household’s choice can be represented by the following optimization problem: Max sc , x, Ac , M ci U (sc , x) 73 As mentioned by Boughton et al. (2007) and Barrett (2008), households will not both buy and sell the same crop in this one-period model because of the price wedge created by transaction CV CB costs. Therefore, there exists a complementary slackness condition, M * M = 0, at any optimum. 74 As mentioned above, shallow local markets and price risks and risk aversion also affect the magnitude of the price band (Sadoulet and de Janvry 1995). 84 Subject to the liquidity constraint Y  pxx   p c* M cv  M cb  f c Ac , G  s c   0 3 c 1 And equilibrium conditions for non-tradables 3 A  Ac c 1    f c Ac , G  s c 1  M cb  for c = 1, 2, 3 With each household-specific crop price determined by the household’s net market position:   if M   if M p c*  p cm   c A,G,Y ,Z ,NS c p c*  p cm   c A,G,Y ,Z ,NS c p c*  p a if M CB CV CB = 1 (i.e. net buyer) = 1 (i.e. net seller) CV =M = 0 (i.e. autarkic) Where pa is the autarkic (i.e. non-tradable) shadow price that equates household supply and demand. The second equilibrium condition for non-tradables implies that, if the household does not purchase crop c (i.e. M cb = 0), production must be greater than or equal to the quantity of crop c consumed (may be a net seller) and, if the household does purchase crop c (i.e. M cb = 1), production must be greater than or equal to zero (may produce crop c or not; regardless of which, the household is a net buyer). To find the optimal solution, two steps are necessary. First, the system must be solved for the optimal solution, conditional on the participation regime (i.e. net seller, net buyer, or autarkic). Then, the market participation regime that yields the highest utility level is chosen C C (Key et al. 2000). That is, the optimal choices of {s , A , x} must be replaced into the utility 85 function to obtain the indirect utility function, V. This indirect utility function must be evaluated cv under each feasible combination of M cv cb 75 and M to identify the market participation vectors cb {M , M } that yield the highest level of V (Key et al. 2000; Barrett 2008). Based on the structural model described above, the reduced form of each choice variable cm can be represented as a function of observable (exogenous) variables A, G, Y, Z, p 76 This structural model assumes non-separability x , and p . in household’s production and consumption decisions because the parametric prices are endogenous (because of transaction costs). Because of this, production and consumption behaviors are estimated simultaneously (Sadoulet and de Janvry 1995) in this maximization problem. It is expected that both market participation and quantity traded will be positively affected by asset endowment and access to public goods and services, and negatively affected by transaction costs. In this study, transaction costs include all costs associated with a transaction (e.g. transportation of output to the market, fees paid). Smallholder farmers in rural Angola generally sell their surpluses to itinerant traders at low prices (World Vision 2008). Although this suggests that there may be low barriers to participate in the market, due to high transaction costs (e.g. obtaining price information) farmers’ per unit returns will be small. Therefore, understanding what factors affect smallholder market participation decisions will be useful in designing policies regarding public and private investments oriented to boost market participation by smallholder farmers in rural Angola. 75 There are 27 possible combinations to evaluate for. 76 This implies that production decisions are made as if the household was maximizing profits, while consumption decisions are made as if the household was maximizing utility. For further reading see de Janvry et al. (1991) and Sadoulet and de Janvry (1995). 86 3.4.2 Econometric Estimation As mentioned above, this study attempts to estimate the factors associated with households’ marketing decisions, focusing on households who sell potatoes, beans and/or onions in rural Angola. Given that sales are only observed for a subset of the sampled population because farmers who did not sell any or all of these crops reported zero sales, the function estimated (i.e. quantity traded) on the selected sample may not estimate the population (i.e. random sample) function (Heckman 1979) due to self-selection problems. 77 Therefore, if the parameters were estimated by least squares, they would be biased and inconsistent (Wooldridge 2009). There are at least three alternatives to least squares to estimate unbiased, consistent, and efficient parameters. The first alternative is to estimate the parameters using the standard 78 Heckman sample selection model (two step version ) used by Goetz (1992), Benfica et al. (2006), and Boughton et al. (2007). With Heckman two-step approach, one first estimates a probit model of market participation; then, in the second step, one fits a regression of quantity traded (regression equation below) by least squares, conditional on market participation (Wooldridge 2003). From the probit, one could derive the Inverse Mills Ratio (IMR) and include it as a regressor into the second equation to control for selection bias and obtain unbiased, consistent, and efficient estimators using OLS (for details, see Wooldridge 2003, p. 560-562). A Heckman selection approach would be appropriate in this context because many households reported zero sales. However, the Heckman regression is designed for incidental truncation, where the zeros are unobserved values (e.g. as with wage rate models where the 77 Self-selection arises due to transaction costs, which are reflected in the endogenous market prices faced by farmers. 78 Heckman could also be solved by full maximum likelihood (StataCorp 2009). 87 sample includes unemployed persons) (Ricker-Gilbert et al. 2011). Therefore, a corner solution model is more appropriate in this context because, due to market and agronomic conditions, the zeros in the data reflect farmers’ optimal choice rather than a missing value (as with Heckman). The second and third alternatives to least squares (both corner solution models) are the Tobit estimator proposed by Tobin (1958) and the double hurdle (DH) proposed by Cragg (1971), 79 respectively. Although the Tobit model could be used to model farmers’ marketing decisions, its major drawback is that it requires that the decision to sell a particular crop and the decision about how much of that crop to sell be determined by the same process (i.e. the same 80 variables), which makes it fairly restrictive (Wooldridge 2003 and Ricker-Gilbert et al. 2011). In addition, in a Tobit model, the partial effects of a particular variable, xj, on the probability that the farmer will sell and on the expected value of the quantity traded, conditional on participation, have the same signs (Wooldridge 2008). The DH model is a more flexible alternative than the Tobit because it allows for the possibility that factors influencing the decision to sell a crop are different than factors affecting the decision of how much to sell. Therefore, the DH model proposed by Cragg (1971) is utilized in this paper. In the DH model, 81 the first hurdle estimates the decision of whether or not to participate in the market (i.e. to sell a crop) and, conditional on market participation, the second hurdle estimates the quantity traded (i.e. quantity sold). Conceptually, a simple corner-solution model is where: 79 80 81 He proposed a double-hurdle model that nests the usual Tobit model. For details about the Tobit model, see Wooldridge (2003), pg. 540-546. Also called two-tiered model. 88 yi = s yi * if yi * > 0 yi = 0 if yi * < 0 where yi * = α + Xi β + εi s=1 if participates in the market; s = 0 otherwise In this model, yi is the quantity traded by farmer i, α is the intercept, β is a vector of coefficients, Xi is a vector of explanatory variables, and εi is the error term. The binary variable, s, is used to estimate the maximum likelihood estimator (MLE) of the first hurdle and is assumed to follow a probit model. Therefore, the probability of a farmer choosing not to participate in the market is given by: (3.1) P(s = 0 | x1) = P(y = 0 | x1) = 1 - Φ (x1 γ) where Φ is the standard normal CDF and γ is the vector of coefficients of x1. In the second hurdle, the continuous variable, y (i.e. quantity traded), is assumed to follow a truncated normal distribution. Therefore, the MLE is obtained by fitting a truncated normal regression model (3.2) 82 to the quantity traded (Cragg 1971 and Burke 2009): f (y | x1, x2) = [Φ (x1 γ) (2π) / Φ (x2 β / σ) −1/2 σ −1 2 2 exp { − (y − x2 β) / 2 σ } ] for y > 0 82 The model is called truncated because the distribution of y is truncated at zero to guarantee non-negativity (Cragg 1971). 89 2 where f is the probability density function of positive values of y and σ is the variance of the distribution. Therefore, Cragg’s model integrates Equations (3.1) and (3.2) to obtain: (3.3) f (s, y | x1, x2) = [1 - Φ (x1 γ)] 2 1(s=0) [Φ (x1 γ) (2π) 2 exp { − (y − x2 β) / 2 σ } / Φ (x2 β / σ) ] −1/2 σ −1 1(s=1) where, as mentioned above, s is a binary indicator equal to 1 if y is positive and 0 otherwise. Equation (3.3) demonstrates that the probability of market participation (i.e. y > 0) and the analysis of quantity traded (i.e. value of y), conditional on market participation, could be determined by different factors (the vectors γ and β, respectively) (Burke 2009). The model also puts no restrictions in the variables included in x1 and x2, implying that these could be the same (i.e. x1 = x2) or have several or no variables (i.e. x1 ≠ x2) in common. Furthermore, Equation (3.3) yields the standard Tobit density when γ = β / σ and x1 = x2 (Wooldridge 2002 and Burke 2009). In order to fit Cragg’s model to the data, it is necessary to assume that s and y* are independent, conditional on explanatory variables X (Wooldridge 2008); that is: D(y* | s, X) = D(y* | X) where D is the distribution of the latent variable y*. The above equation implies that the expected value of y conditional on X and s is: E(y | X, s) = s  E(y* | X, s) = s  E(y* | X) From Cragg’s model, one could estimate the same probabilities and expected values as with Tobit. However, Cragg’s model uses an updated functional form (Equation (3.3) above). The probabilities regarding market participation (i.e. whether y is positive) are: 90 (3.4) P(yi = 0 | x1i) = 1 - Φ (x1i γ) i.e. no market participation (3.5) P(yi > 0 | x1i) = Φ (x1i γ) 83 i.e. market participation The expected value of y, conditional on market participation (i.e. y > 0) is: (3.6) E(yi | yi > 0, x2i) = x2i β + σ × λ (x2i β / σ) where λ (x2i β / σ) is the Inverse Mills Ratio (IMR) λ (x2i β / σ) = Φ (x2i β / σ) / Φ (x2i β / σ) where Φ is the standard normal probability distribution function. The “unconditional” 84 expected value y of is: (3.7) E(yi | x1i, x2i) = Φ (x1i γ) [x2i β + σ × λ (x2i β / σ)] From the DH model above, one could estimate the “unconditional” partial effect (PE) of a particular variable, xj, for each observation i. Using these PE, one could estimate the average partial effect (APE) of the variable of interest (i.e. xj) by averaging the PE across all observations in the dataset. However, the standard deviation reported with the APE should not be used as a standard error for inference about the population because it describes only the data. Instead, two alternatives could be used: (a) standard deviations could be re-estimated by bootstrapping or (b) standard errors could be approximated by the delta method (i.e. Taylor expansion around the data mean) (Burke 2009). Burke (2009) provides the Stata programming necessary to make 83 Equations (3.4) to (3.7) were taken from Burke (2009). 84 “Unconditional” refers to market participation (i.e. random population) since all expectations are “conditional” on the explanatory variables. 91 inferences on an “unconditional” APE using methods (a) and (b) above. In this paper, bootstrapping at 500 repetitions was used. Key et al. (2000) showed that while market participation (i.e. household’s decision of whether or not to participate in the market) depends on both fixed and proportional transactions costs, the quantity supplied, conditional on participation, is only affected by proportional transactions costs. The DH model described above allows for different factors to affect the first and second hurdles. However, the variables used as proxies for fixed costs (i.e. distance to market and quality of the road) were included in both the market-participation and the quantitytraded regressions to test whether fixed costs only affect the first hurdle among Angolan farmers. Although the independent variables included in the regressions are explained in the next section, the quantity harvested (included in both hurdles) is worth mentioning here. Quantity harvested is potentially endogenous to the decision of whether or not to participate in the market as a seller and on the decision of how much to sell. For instance, if a farmer produces a crop with the intention of selling his/her surplus, whether he/she participates in the market will depend on how much he/she harvests--i.e. if the quantity harvested is small, he/she might decide not to sell or to sell a smaller amount. Furthermore, market conditions will influence the amount a farmer produces because if the farmer perceives that he/she could sell in the market, he/she may decide to produce more for this purpose. Because of these factors, there may be correlation between the error term in a reduced equation of quantity harvested and the error term of the probability of participation and quantity traded. To deal with this potentially endogenous variable, an OLS regression will be estimated on the quantity produced of each crop. Then, the residuals from these OLS regressions will be estimated and included in both the probit and truncated normal regressions as an additional 92 explanatory variable. This allows determining if the quantity produced is truly endogenous (i.e. if the coefficient of this variable is statistically significant, the quantity produced is endogenous). Although several variables included in the OLS regressions are also included in the DH regressions, the former includes additional variables that are not expected to directly affect marketing participation decisions. 3.5 Data Used Data used in this study come from the cross sectional household- and village-level survey implemented by World Vision’s ProRenda project in Angola in 2009. World Vision, in collaboration with ACDI/VOCA, 85 the Ministry of Agriculture and Rural Development of Angola, and the Angolan NGO HORIZONTE are implementing a four-year project 86 targeted at increasing smallholder-farming families’ annual income from non-perishable crops (World Vision 2008). The ProRenda project attempts to increase smallholder’s incomes, especially those of women, by establishing competitive value chains for potatoes, beans, onions, and other highvalue crops. Michigan State University has been contracted to conduct impact assessment of this project. The baseline survey was implemented from January through April of 2009 and collected data about the latest harvest between September 2007 and December 2008. In Angola, the agricultural year goes from September through May of the following year (MINADER and FAO 2003). Therefore, the data collected refers to the 2007-2008 agricultural year and the first season of the 2008-2009 agricultural year. 85 Agricultural Cooperative Development International/Volunteers in Overseas Cooperative Assistance. 86 The ProRenda Project, which is financed by the Bill and Melinda Gates Foundation. 93 The survey was implemented in three provinces of the central highlands of Angola: Huambo, Bie, and Benguela. These provinces have the most productive lands within the highland region (World Vision 2008) because of good rainfall distribution and environmental conditions; however, yields are usually low (MINADER and FAO 2003). The major crops produced in the highlands are: corn, wheat, rice, potatoes, sweet potatoes, beans, cassava, sugarcane, peanuts, sunflower, sesame, tobacco, and vegetables (MINADER and FAO 2003). The survey included a total of 656 households 87 across 40 villages (Figure 3.5.1). The households were selected using a clustered sampling methodology. This means that the villages were selected first from a listing of all the potential villages within the action area of the ProRenda project. The villages were selected systematically using probability proportional to size for three categories of villages: 1) primary villages for ProRenda project activities; 2) secondary villages for ProRenda project activities and 3) control villages. After selection of the villages in each category, a household listing for each village was used to identify the classification of each household, based on four categories: male head, female head, and participation or no participation in a farmer organization at the time of the listing. Within each 87 Although 656 households were surveyed, only 620 surveys were valid and were used for analysis. 94 Source: ProRenda Survey. Figure 3.5.1. Distribution of villages included in the ProRenda 2009 survey. The text in the figure is not meant to be readable but is for visual reference only. category a random systematic sample of households were selected. 88 In order for the sample estimates to be representative of the population covered by the survey, sampling weights were used. The basic weight for each sampled household is the inverse of its probability of selection (see Reyes et al. 2010 for details). 88 Details about the sampling methodology and weight estimation can be found in Reyes et al. (2010). 95 The household-level survey collected information about household socioeconomic characteristics, productive and non-productive assets, participation in farmer organizations, and production and marketing information of beans, potatoes, onions, carrots, and cabbages. The village-level survey collected information regarding the distance between the village and the main commercial town, availability of public services (e.g. local markets, agricultural extension services) and public transportation, and quality of the road between the village and the main commercial town. The independent variables included in the regressions were classified into five categories: (1) household characteristics, (2) private assets, (3) public assets and quasi-fixed factors, (4) production- and marketing-related variables, and (5) squared and interaction terms (Table 3.5.1). These variables were included because they were theoretically expected to affect production and marketing decisions. A total of 34 independent variables were used in various combinations to estimate the three models proposed in the previous section: (a) linear regression model of quantity produced, (b) probit model of market participation, and (c) truncated normal regression model of quantity traded. Although most variables are self-explanatory, a brief explanation of key variables is provided next. The dependency ratio was estimated by dividing the number of people younger than or equal to 17 by the household size. Having a household member participating in a farmer organization (FO) refers to any member of the household who participated in a FO within the previous 12 months. Adult literacy refers to members older than 17 who can read and write (selfdeclared, not tested). The number of tropical livestock units was estimated using FAO conversion factors for South Africa where, for example, one cattle equals 0.70 livestock units and one sheep equals 0.10 livestock units (FAO 2010). 96 Table 3.5.1. Independent variables included in the production and marketing decision regressions. Angola, 2009. Model where 1 included No. Variable Description of variable Dependent: Quantity produced (kg) 1 Quantity produced of each crop. Market participation (1=yes) 2 Whether or not participates in the market as a seller. Quantity sold (kg) 3 How much was sold. Household (HH) Characteristics: 1 Age of HH head (yr) 1, 2, 3 2 Gender of HH head (1=male) 1, 2, 3 3 Dependency ratio 1, 2, 3 Number of dependants (<17 yr) divided by HH size. 4 HH member is in a farmer organization (1=yes) 1, 2, 3 5 No. adults who can read & write 1, 2, 3 Adults refer to people older than 17 yr of age. 2 6 1, 2, 3 Includes oxen, cattle, goats, sheep, pigs, chicken, and Number of TLU owned rabbits. 7 HH has zinc roof (1=yes) 1 8 Asset Index 2, 3 Estimated using principal component analysis for 9-11 types of assets owned by the HH. Private Assets (1=yes): 9 Own plow 1 Plow used to prepare the soil. 10 Own backpack sprayer 1 Sprayer used to apply pesticides. 11 Own motorcycle 2, 3 12 Own bicycle 2, 3 Public Assets and Quasi-fixed Factors: 13 IDA office in the village (1=yes) 1, 2, 3 IDA is the government's Institute for Agrarian Development. 14 Public market available in the village (1=yes) 1, 2, 3 Farmers could buy/sell food in the public market. 15-21 Seven dummy variables for municipalities (1=yes) 1, 2, 3 Although eight municipalities were surveyed, only seven dummies were included to avoid the dummy variable trap. 97 Table 3.5.1 (cont’d). No. 22 23 24 25 26 27 28 29 30 31 32 33 34 1 2 Variable Distance from village to commercial town (km) Road between village and commercial town in poor condition (1=yes) Production- and Marketing-related Variables: Seed used (kg) Type of plot (1=rainfed plot) Planted intercropped (1=yes) Planted seed of local variety (1=yes) Used fertilizer (1=yes) Used pesticides (1=yes) Reported production costs (Kw/kg) Model where included1 Description of variable 2, 3 2, 3 Poor condition means the road is a clay road, not rehabilitated (i.e. without maintenance). 1 1 1 1 1 1 1 HH reports lower harvest (1=yes) 1 Seller sought price information prior to sales (1=yes) 3 Reported marketing costs (Kw/kg) 3 Quantity produced (kg) Squared terms: Age of HH head squared Dependency ratio squared No. adults who read & write squared No. TLU squared Seed used squared Interaction of 30 * 31 2, 3 Rainfed plots are the most commonly used plots. Only beans could be planted intercropped. Local varieties are usually low-quality seed. Includes expenses on seed, pesticides, labor, transport from field to home Obtained by comparing current year with a normal year. Includes expenses on bags, transportation, load/unload, taxes The predicted residuals are also included in models 2, 3 1 1 1 1 1 1 1 = Ordinary Least Squares for production; 2 = Probit for market participation; 3 = Truncated Normal Regression for quantity sold. TLU = Tropical Livestock Units, calculated using FAO conversion tables. 98 An asset index was estimated as a proxy for household wealth. This index was estimated using principal component analysis. Details about this index are included in Section 3.6.2 below. However, several asset variables were included in the regressions separately from the asset index (e.g. having a home with zinc roof, owning a motorcycle). The quasi-fixed variables included seven dummies for the municipalities where the households were located to control for variations in environment and marketing conditions faced by farmers (at the macro-level). 89 Transaction costs (TC) included the distance between the village and the main commercial town and the quality of the road between these two places. The production-related variables are self-explanatory except for ‘type of plot.’ Angolan farmers in these provinces could cultivate in one (or several) of four possible types of plots: nacas, ombandas, otchumbo, and lavras. 90 Nacas are irrigated lowland areas located close to river deltas, used during the dry season (by exploiting residual moisture), and account for 4% of the cultivated area. Ombandas are medium-level lands with access to gravity-fed irrigation, used in all seasons, and account for 15% of the cultivated area. Otchumbo are small areas close to the homestead, intensively cultivated all year round, and account for 4% of the cultivated area. Finally, lavras are upland areas used for rainfed agriculture and account for 77% of the cultivated area (World Vision 2008). Lavras are the most commonly used types of plots, thus a dummy variable was created to account for whether the crop was produced in this type of plot. Unit production costs were estimated by adding reported expenditures on fertilizers, seed, pesticides, hired labor, and transport from the field to the home and dividing this by the total quantity produced. Family labor contributions were not valued. Similarly, unit marketing costs 89 Although it would have been ideal to include dummy variables for each village, this was not practical because there were 40 villages. 90 These are Portuguese names with no English translation. 99 were estimated by adding farmers’ reported expenditures for bags, sewing of these bags, transportation costs, loading and unloading of the output, and taxes and fees paid at the market and dividing this by the total quantity sold. The squared terms were included to allow for nonlinear relationships between independent and dependent variables. Finally, the land area owned by each household and the area planted of each crop were not included the analysis because the instrument used to collect the household-level information did not include these questions. To control for this potential bias, two variables were included as explanatory variables: the asset index to compensate for the omission of land owned and the amount of seed used to compensate for the omission of area planted. 3.6 Results This section is divided into five subsections. The first subsection describes the sample and provides the socioeconomic characteristics of farm families, focusing on the variables of interest for the double hurdle analysis. For each crop, the results are disaggregated by market participation. The second subsection briefly describes households’ wealth (economic status index). The third subsection discusses household-level receipts and margins for households selling key crops (i.e. potatoes, beans and onions). Subsection four presents the OLS regression results of the quantity produced per crop. The last subsection details the double hurdle regression results. 3.6.1 Descriptive Statistics Beans were planted by the highest number of farmers, followed by potatoes and onions. While almost three out of four farmers planted beans, 55% of farmers planted potatoes, and 46% 100 of farmers planted onions (Table 3.6.1). Further, less than 10% of farmers planted other vegetables, which was expected since farmers in this region mostly depend on maize, bean, and potato production. While 27% of farmers planted both potatoes and onions, 29% of farmers planted beans and onions, and 32% planted potatoes and beans, the most common combination. Regarding sales, approximately 71% of farmers producing potatoes sold part of their harvest, 69% of farmers producing beans were sellers, and 68% of farmers producing onions sold a share of their harvest. In contrast, close to 90% of farmers growing a combination of crops sold at least one of the crops they produced. Table 3.6.1. Percentage of households growing key crops, per economic status index and gender of household head (HHH). Central Highlands of Angola, 2009. 1 1 Economic Status Index by tercile Gender of HHH Lowest Middle Highest Male Female (% of households) 45% a 49% a 64% b 58% 48% ** 77% 73% 68% 69% 75% 33% a 49% b 54% b 51% 37% *** Crop Potatoes Beans Onions 2 Other vegetables Potatoes and onions Potatoes and beans Beans and onions Number of observations 3% 23% 24% 23% 164 a a a 5% 20% 30% 31% a a ab 184 17% 36% 40% 34% 130 b b b Total 55% 71% 46% 11% 30% 35% 32% 4% 22% 26% 23% *** 9% * 27% ** 32% ** 29% 276 250 526 NOTES: All variables are binary (0=NO, 1=YES). Number of observations in Economic Status Index smaller than in last column because of missing values in this variable. Estimates weighted to reflect population (except number of observations). 1 Bonferroni test of difference between means: for Economic Status Index, different letters imply differences are significant at 10%; for Gender of HHH, * significant at 10%; ** significant at 5%; *** significant at 1%. Economic Status Index estimated using principal component analysis. 2 Other vegetables only include carrots and cabbages. Source: ProRenda survey, Angola, 2009. 101 Cultivating potatoes, onions, other vegetables, or a combination of potatoes and onions or 91 potatoes and beans was more common among richer farmers (as classified by their asset index ) and male-headed households. In contrast, most households planted beans--across different economic strata and regardless of the gender of the head (Table 3.6.1). This was expected since, though beans are generally produced for home consumption, they can easily be sold in the markets and are an important source of income for rural households. Further, a higher share of male-headed households planted both beans and onions, compared to female-headed households. Table 3.6.2 reports the descriptive statistics of households in the Central Highlands of Angola, disaggregated by crop and their market participation (i.e. non-sellers vs. sellers). Although most farmers planted beans (Table 3.6.1), they marketed a larger quantity of potatoes than beans or onions. On average, each farmer supplied 200 kg of potatoes, 96 kg of beans, and only 50 kg of onions, which corresponds to roughly 77%, 49%, and 68% of total (i.e. nonsellers’ and sellers’) potato, bean, and onion production, respectively (Table 3.6.2). Furthermore, farmers who did not sell produced less than sellers. While farmers who sold onions were, on average, five years younger than non-sellers (10% significance level, SL), the differences in age between sellers and non-sellers of potato and beans were not statistically significant at the 10% level (Table 3.6.2). Furthermore, as expected, more male- than female-headed households sold their surpluses. Although the proportion of male-headed households participating and not participating in the market (i.e. sellers vs. nonsellers) were similar for beans, among potato and onion producers, the share of male-headed households selling their output was much larger than that of female-headed households (1% SL; Table 3.6.2). 91 Asset index and economic status index are used interchangeable in this section. 102 Table 3.6.2. Descriptive statistics of the variables used in the Double Hurdle analysis. Central Highlands of Angola, 2009. Potato Non-sellers Sellers Demographics Quantity sold (kg) Household Characteristics Age of head (years) Gender of head (% male) Dependency ratio 2 1 Mean S.E. Mean S.E. MT n.a. 200 22.36 -42 3.858 52 0.283 39 78 0.411 0.195 *** 0.50 0.018 0.58 0.022 ** 3 HH member is in FO (% yes) Family members older than 17 4 who are literate No. TLU Bean Non-sellers Sellers 4 0.025 0.054 * 44 70 0.810 0.268 0.53 0.007 10 0.026 1 Mean S.E. MT 96 19.49 -43 68 0.400 0.274 0.56 0.020 1 Mean S.E. n.a. 47 61 5.412 0.345 0.54 0.021 1 Mean S.E. MT 50 5.49 -42 82 0.811 * 0.157 *** 0.55 0.015 0.006 *** 0.7 0.006 8 0.044 * 0.139 0.7 0.085 0.6 0.082 0.9 0.103 *** 0.8 0.126 1.0 0.47 0.121 0.36 0.083 5 0.9 0.198 11 Mean S.E. n.a. Onion Non-sellers Sellers 0.38 0.125 0.44 0.083 0.22 0.039 0.45 0.068 *** 0.04 0.250 11 0.051 14 0.050 0.07 0.239 10 0.026 20 0.064 0.37 0.400 12 0.061 17 0.067 0.22 0.126 7 0.044 26 0.058 6 Modified Asset Index -0.14 0.284 0.37 0.212 *** Owns motorcycle (% yes) 4 0.022 10 0.026 Owns bicycle (% yes) 25 0.155 29 0.061 Public Assets and Quasi-fixed factors 7 IDA office in village (% yes) Public market available in village (% yes) 17 0.052 26 0.040 8 0.078 20 0.028 *** 17 0.084 17 0.024 19 0.052 16 0.052 6 0.045 19 0.023 *** 11 0.035 11 0.045 8 Mean sales price, local market (kw/kg) Percent of HH in following municipalities: Caala Ekunha Bailundo Londuimbali 88.4 5.246 75.1 2.995 *** 23 1 21 35 0.065 0.008 0.062 0.054 11 2 19 15 0.020 ** 0.017 0.053 0.030 *** 64.4 1.140 12 6 46 16 103 0.030 0.035 0.048 0.019 67.8 0.899 ** 5 3 53 28 0.010 ** 0.020 0.040 0.024 ** 83.0 6.155 15 0.6 44 30 0.031 0.007 0.040 0.096 97.8 6.167 6 2 37 23 0.021 0.014 0.069 0.017 ** * Table 3.6.2 (cont’d). Potato Non-sellers Sellers Mean S.E. Demographics Katchiungo 4 0.014 Tchicalachuluanga 7 0.029 Chiguar 9 0.041 Babaera 0.6 0.005 Distance from village to commercial town (km) 10.3 0.990 Road between village and commercial town in poor 9 condition (% yes) 66 0.100 Production and Marketing variables Quantity produced (kg): 30 8.85 In Caala 50 16.02 In Ekunha 66 0.00 In Bailundo 12 1.91 In Londuimbali 30 15.44 In Katchiungo 20 11.34 In Tchicalachuluanga 16 2.66 In Chiguar 37 11.36 In Babaera 12 3.12 Mean 15 2 36 0.2 S.E. 0.021 0.016 0.059 0.002 Bean Non-sellers Sellers Onion Non-sellers Sellers 1 1 1 MT Mean S.E. Mean S.E. MT Mean S.E. Mean S.E. MT ** 11 0.051 3 0.008 *** 2 0.019 9 0.024 ** 6 0.038 2 0.010 ** 2 0.012 4 0.022 *** 4 0.039 6 0.038 6 0.052 20 0.059 ** 0.5 0.004 0.2 0.002 0.1 0.001 0 n.a. -- 11.4 0.908 8.0 1.009 10.0 0.405 81 0.029 ** 80 0.067 73 29.72 *** 59.52 -54.89 -11.47 -26.64 -51.70 -15.89 -61.55 -9.64 -- 53 12.95 26 4.42 26 5.85 41 4.03 149 45.39 35 0.00 23 8.77 12 0.89 9 1.40 143 41 176 137 201 41 71 72 38 25.41 3.47 47.20 16.31 51.86 8.74 9.52 14.59 11.57 63 *** --------- 8.0 0.945 53 0.065 230 359 417 42 172 224 136 312 92 9.7 2.238 0.052 81 0.048 *** 16 24 75 16 17 2 5 6 5 3.25 5.37 0.00 5.85 9.63 0.00 0.45 0.52 0.00 58 103 109 29 102 53 44 50 0 5.66 21.27 26.76 2.49 39.15 48.29 24.96 19.16 n.a. ** --------- Seller sought price information prior to sales (% yes) n.a. Reported marketing costs (Kw/kg) n.a. 0.045 -- n.a. 71 0.008 -- n.a. 63 0.068 -- 2.9 0.249 -- n.a. 3.0 0.703 -- n.a. 3.1 0.435 -- Number of observations 165 89 216 49 125 75 104 Table 3.6.2 (cont’d). 1 MT = test of difference between means: *significant at 10%; **significant at 5%; ***significant at 1%; -- not tested; n.a. = not applicable. 2 Dependency ratio estimated by dividing the number of people 17 years or younger by the household size. A high dependency ratio means more dependants in the household. 3 FO = Farmer organization. 4 5 Literacy refers to people who can read and write (self-declared, not confirmed). TLU = Tropical Livestock Units (estimated using FAO conversion factors). 6 Modified Asset Index excludes owning a motorcycle and owning a bicycle to be able to estimate the effect of these variables on marketing decisions separately from other assets included in the index. 7 IDA = Government's Institute for Agrarian Development. 8 For farmers who sold in local markets, their reported price was averaged per community. Communities with missing prices use average price per the next political division (i.e. town, municipality). This price was imputed to non-sellers. 9 Poor condition means the road is a clay road, not rehabilitated (i.e. without maintenance). Source: ProRenda survey, Angola, 2009. Estimates weighted to reflect population. 105 On average, there were slightly more than one dependent for every two adults in the household (the average dependency ratio was 0.52). Furthermore, potato sellers had more dependents than non-sellers (5% SL; Table 3.6.2). As Table A 3.1 shows, potato sellers had 1.5 children younger than five vs. only one among non-sellers (1% SL). Less than 11% of households had at least one member who participated in a farmer organization (FO) in the year prior to the interview (Table 3.6.2). A higher percent of households selling potatoes and onions, compared to non-sellers, reported having a member participating in a FO (10% SL). In contrast, a higher percent of households producing beans solely for consumption (i.e. non-sellers) reported having a household member participating in a FO (1% SL; Table 3.6.2), which was unexpected. Given that bean non-sellers reported other crops as their major source of crop income and that the large majority of bean sellers reported beans as their major source of crop income (Table A 3.2), it is likely that non-sellers participate in FO related to these other crops, thus explaining this finding. Among bean producers, households selling beans had more literate adults (i.e. older than 17) living at home (1% SL). While the differences in the number of tropical livestock units (TLU) owned between sellers and non-sellers were not statistically significant at the 10% level for potatoes and beans, onion sellers reported more TLU than non-sellers (1% SL; Table 3.6.2). In this study, an asset index, which was estimated using 11 assets, was used as a proxy for household wealth. Details about its estimation and interpretation are included in Section 3.6.2. Households selling potatoes were wealthier than non-sellers (1% SL). While there were no statistically significant differences (at the 10% level) in the asset index between bean and onion sellers vs. non-sellers, it was surprising to find that onion non-sellers had a higher asset index than onion sellers (Table 3.6.2). Since a higher percent of onion non-sellers (20% vs. 2% sellers) 106 reported remittances and other transfers as their major source of non-crop income (Table A 3.2), it is likely that non-sellers invest part of these transfers on improving their home or purchasing assets; thus, explaining this finding. Most of the differences (between sellers and non-sellers) regarding access to public assets and quasi-fixed factors were statistically significant for potato and bean producers (Table 3.6.2). While 19% of bean sellers reported having access to a public market for purchasing food in their villages, only six percent of non-sellers reported having access to this public good (1%SL). Similarly, access to the government’s Institute for Agrarian Development (IDA) office (which provides extension services) was more common among bean sellers (20% vs. 8% non-sellers, 1% SL) (Table 3.6.2). As previously explained, sales prices were collected for farmers who sold at least part of their output. Farmers reported selling their output in different places, including at their farm, their home, local markets, and other markets. To control for (potential) endogeneity problems in market prices, for farmers who reported selling at local markets, the average sale price was estimated for each crop. However, in some villages, none of the farmers who sold their output did so in local markets; thus, the average price could not be estimated. In these cases, the average price of the next political division (i.e. town, municipality) was estimated and used. Although this information is presented in Table 3.6.2, it was excluded from the double hurdle analysis because it was judged to be inaccurate. In villages with no sellers, prices collected at the next political division (e.g. municipality) were imputed to non-sellers in these villages. It is likely that these non-sellers were imputed too high of a price, thus offsetting any positive effect of this variable. 92 92 Furthermore, as one would expect, current prices are likely endogenous. 107 The average distance between the villages and their main commercial town was 9.6 km (Table 3.6.2). Among potato and bean producers, a higher percent of sellers than non-sellers were located in villages farther from the main commercial town (although the differences were not statistically significant); thus, the average distance from their villages to their main commercial town was higher for sellers. Furthermore, a higher share of potato and onion sellers was located in villages with poor quality road between the village and the main commercial town (5% and 1% SL, respectively). Regarding the variables used to control for macro-level environmental and market conditions, potato and onion production was highest in Ekunha municipality among both sellers and non-sellers. In contrast, bean production was highest in Londuimbali (Table 3.6.2). Since Londuimbali accounted for the largest percentage of beans produced, the municipality coefficients from the regression results (of all crops) were compared to this municipality. 93 Furthermore, since no farmers in Babaera municipality sold onions, the binary variable for this municipality was excluded from the double hurdle regressions for this crop. Finally, close to two-thirds of the farmers who sold their output obtained price information before selling their crop and sellers reported an average marketing cost of three Kwanzas 94 per kilogram sold, with the highest marketing costs reported by onion sellers (Table 3.6.2). 93 Although there was one dummy variable for each municipality, the variable for Londuimbali was excluded from the regressions to avoid the dummy variable trap. 94 The exchange rate at the time of the survey was 75 Angolan Kwanzas per US$. 108 3.6.2 Wealth of the Households To analyze wealth across various types of assets, researchers have developed asset indices that take into account the relative importance of the asset in the sample (i.e. giving less weight to assets commonly owned and more weight to assets owned by a few). In this study, following Filmer and Pritchett (2001), McKenzie (2005) and Reyes et al. (2010), principal component analysis was used to estimate an economic status (or asset) index, based on asset ownership. Reyes et al. (2010) details the index estimation process, using the same dataset used for this study; thus, interested readers can refer to these authors for more details on the theory and the construction of the index. In constructing the index, ownership of the following assets was considered: tractors, trucks, cars, plows, carts, backpack sprayers, motorcycles, bicycles, cell phones, radio, and televisions. The index also considered ownership of water storage facilities and a latrine at the homestead, and whether the roof was made of zinc or ‘lusalite,’ both considered improved materials. From the 14 indicators, tractors, trucks, and cars were excluded because no household in the sample owned these items. A high value for the index indicates a higher level of ownership of these assets, implying greater wealth (Reyes et al., 2010). Households were sorted according to their wealth index and the population was split into three groups or terciles (Table A 3.3). These terciles were used to analyze households’ crop margins in the next subsection. The consistency of the index is reflected in the terciles percentages: only 5% of the poorest tercile (lowest 33% of the population) had water storage facilities at home, whereas 19% of the middle tercile and 42% of the highest tercile had this facility at home (Table A 3.3). The mean value of the index (by construction) is zero and its standard deviation is 1.5. The poorest tercile households had an average index of -1.26 while the 109 richest tercile had an average index of 1.72, a difference of 2.98 units (Table A 3.3). One example of a combination of assets that would produce this difference is having a plow (1.11) and a cart (1.87). The wealth index suggests that male-headed households are richer than female-headed ones (Table 3.6.3). Similarly, farmers who grew potatoes and onions are richer than farmers who did not grow these crops (5% and 1% SL, respectively). The differences in the index between bean growers and non-growers were not statistically significant at the 10% level. These results are confirmed by a graphical analysis of the cumulative distribution of the index by gender and crop grown (Figure 3.6.1 and Figure A 3.1 to Figure A 3.3). While potato sellers were wealthier (as per their asset index) than non-sellers (1% SL), farmers who did not sell onions were wealthier than onion sellers (10% SL, Table A 3.4). As explained above, a larger share of onion non-sellers had zinc roof (with a “weight” in the asset index of 0.79, see first data column on Table 3.6.3) and, as Table A 3.4 shows, a higher share of non-sellers owned a motorcycle (with a “weight of 1.71) and a television (“weight” of 1.98, the highest among all assets). Although there were slight differences in the index between bean sellers and non-sellers (sellers had a higher index), these were not statistically significant at the 10% level, suggesting that bean growers have similar wealth, regardless of their market orientation. 110 Table 3.6.3. Scoring factors and means per gender of household head (HHH) and crop grown for asset indicators entering the computation of the first principal component (asset ownership). Percentage of households owning the asset Scoring Gender of HHH Potato grower Bean grower Onion grower Factor / Asset indicators Std. Dev. Male Female MT No Yes MT No Yes MT No Yes MT Own plow 1.11 13% 2% -7% 11% -14% 7% -9% 9% -Own cart 1.87 0% 0% -0% 1% -1% 0% -0% 1% -Own backpack sprayer 1.65 3% 0% -2% 3% -4% 1% -1% 4% -Own motorcycle 1.71 10% 3% -7% 8% -5% 9% -5% 11% -Own bicycle 0.82 30% 4% -18% 25% -30% 18% -20% 24% -Own cell phone 1.27 9% 3% -5% 9% -2% 9% -6% 8% -Have water storage at home 0.66 25% 19% -26% 20% -30% 20% -19% 27% -Have latrine in the house 0.23 95% 70% -87% 87% -85% 88% -87% 88% -Have lusalite or zinc roof 0.79 50% 32% -38% 50% -51% 42% -46% 43% -Own radio 0.71 54% 19% -37% 48% -44% 43% -30% 58% -Own television 1.98 2% 1% -2% 2% -1% 2% -2% 2% -Mean by group Economic Status Index 0.512 -0.625 *** -0.037 0.269 ** 0.270 0.070 -0.081 0.364 *** Number of observations 478 242 236 202 276 130 348 289 189 Notes: Three of the 14 indicators were dropped because they had zero variance. Scoring Factor is the "weight" assigned to each indicator or eigenvector in the linear combination of the variables that constitute the first principal component. The percentage of the covariance explained by the first principal component is 21.06%. The first eigenvalue is 2.32. Data provided in the last eight columns were estimated with weights to reflect population (except number of observations). MT = Bonferroni test of difference between means: * significant at 10%; ** significant at 5%; *** significant at 1%; -- not tested. Source: ProRenda survey, Angola, 2009. 111 Figure 3.6.1. Cumulative distribution of asset index by gender of household head. Central Highlands of Angola, 2009. 3.6.3 Households’ Gross Margins In this study, households’ gross margins equal receipts from sales, which are based only on marketed quantities of potatoes, beans and onions, minus cash expenditures on production and marketing activities. Table 3.6.4 presents households’ margins disaggregated by asset index terciles and gender of the household head. While the richest households and male-headed households reported the highest receipts, these households also reported the highest costs. Regardless of the high costs, households’ margins were higher for households in the middle and highest terciles (10% SL) and for male-headed households (1% SL). 112 1 Table 3.6.4. Average receipts, costs and margins of households selling key crops, per economic status index and gender of household head (HHH). Central Highlands of Angola, 2009. Detail Economic Status Index by tercile Lowest Middle Highest 2 Gender of HHH Male Female 2 Total Receipts (Kw) 5,142 a 9,283 a 19,021 b 14,625 5,103 *** 12,150 Total Costs (Kw) 3,143 a 4,945 a 11,436 b 8,487 3,229 *** 7,121 Margins (Kw) 2,178 a 4,573 ab 8,170 b 6,420 2,337 *** 5,359 Number of observations 119 143 114 241 178 419 NOTES: Kw = Kwanzas. Costs include purchased inputs, hired labor, and reported marketing costs. Variables are at the household level. Number of observations in Economic Status Index smaller than in last column because of missing values in this variable. Estimates weighted to reflect population. 1 Key crops include potato, onion, and bean sales. 2 Bonferroni test of difference between means: for Economic Status Index, different letters imply differences are significant at 10%; for Gender of HHH, * significant at 10%; ** significant at 5%; *** significant at 1%. Source: ProRenda survey, Angola, 2009. Margins were also analyzed in detail for each crop (Table A 3.5 to Table A 3.7). Although the highest receipts were obtained from potato sales, followed by bean and onion sales, farmers who sold potatoes also reported the highest costs. In contrast, farmers selling beans reported the lowest costs. This was expected since, as is common in developing countries where beans are an important staple, bean producers seldom apply purchased inputs (e.g. fertilizer) to their crop (Reyes, 2011). In our sample, while 66% of potato producers and 49% of onion producers applied fertilizers, only 3% of bean producers applied fertilizer (Table A 3.8). Because of this, bean sellers obtained the highest margins. Furthermore, both richest farmers and maleheaded households obtained approximately 2.6 and 3.3 times higher margins from their bean sales (10% and 1% SL) than their counterparts, respectively (Table A 3.6). 113 One additional finding is worth mentioning here. As Table A 3.7 shows, on average, onion sellers in the lowest and highest terciles reported losses, mainly due to the high per unit production costs reported. However, the differences in margins across terciles and gender of household head were not statistically significant across onion sellers. 3.6.4 OLS Regression Results of Factors Influencing Production As explained in the previous section, it was suspected that production could be an endogenous covariate in the double hurdle analysis. Thus, for each crop, linear regression (OLS) estimation was used to determine which factors were affecting production. Then, the residuals of these regressions were included as an additional explanatory variable in the double hurdle analysis and tested for endogeneity. This subsection first discusses the descriptive statistics of factors affecting crop production. Second, for each crop, it presents the results of the OLS regressions on quantity produced. The descriptive results of the factors influencing production are included in Table A 3.8. On average, farmers produced almost 170 kg of potatoes, 90 kg of beans, and 46 kg of onions. Potato producers were slightly younger than both bean and onion producers, had slightly more dependents, more had members of their household participating in a FO, and more had homes with zinc roof (Table A 3.8). In contrast, more female-headed households produced beans and bean producers had more TLUs than potato and onion producers. Furthermore, onion producers had slightly more literate adults in the household (Table A 3.8). Regarding productive assets, a slightly higher share of potato producers owned plows and backpack sprayers (compared to bean and onion producers), which they used to prepare their 114 fields and apply pesticides. Additionally, potato producers had more access to the government’s IDA offices (Table A 3.8). Farmers were asked how much seed they used and whether the seed was a local or imported (usually improved) variety. Potato producers used more seed. However, since the planting rate is higher for potatoes than for beans, the estimated area planted to potatoes was smaller than the area planted to beans (0.015 ha vs. 0.393 ha, respectively). 95 Furthermore, it was more common for potato producers to use improved varieties--almost all bean producers and more than 90% of onion producers reported using local varieties (Table A 3.8). While almost 95% of bean producers planted in rainfed plots (i.e. lavras), less than onehalf of potato producers (44%) and 57% of onion producers planted in this type of field (Table A 3.8). 96 Furthermore, almost 60% of bean producers planted beans as an intercrop, which was expected. Not surprisingly, the use of fertilizer and pesticides was more common among farmers producing potatoes and onions; therefore, these farmers also reported higher production costs (Table A 3.8). Finally, more than 54% of farmers reported lower harvests, compared to a normal year. 3.6.4.1 Potato OLS Results The econometric results of the OLS regression for potatoes are presented in Table 3.6.5. The model had a R-squared of 0.49. The results show that, although male-headed households produced more than female-headed households, these differences were not statistically 95 Potato planting rate = 2,750 kg/ha. Bean planting rate = 60 kg/ha. For onion producers, it was not possible to estimate the area planted using seed data. 96 Although most farmers produced in Lavras, a high percentage of farmers produced potatoes (38%) and onions (25%) in irrigated lowland plots located close to river deltas (i.e. Nacas). 115 Table 3.6.5. Linear regression models of factors influencing potato, bean and onion quantity produced (kg). Central Highlands of Angola, 2009. Potato Bean Onion N = 281 N = 380 N = 162 R-squared = 0.4947 R-squared = 0.6757 R-squared = 0.4417 Coefficient p-value Coefficient p-value Coefficient p-value Independent variables Household (HH) Characteristics Age of HH head (Years) 10.37 0.198 2.05 **0.031 -1.27 0.572 Gender of HH head (1=Male) 36.61 0.198 19.57 0.155 22.87 *0.057 1 Dependency ratio HH member is in farmer organization (1=Yes) 2 No. adults (>17 yr) literate No. of Tropical Livestock Units Home has zinc roof (1=Yes) Productive Assets Ownership (1=Yes) Owns a plow Owns a backpack sprayer Public Assets and Quasi-fixed Factors (1=Yes) IDA office in village Public market in village HH in Caala Municipality HH in Ekunha Municipality HH in Bailundo Municipality HH in Katchiungo Municipality HH in Tchicalachuluanga Municipality HH in Chiguar Municipality HH in Babaera Municipality -265.46 181.33 0.229 0.143 -23.54 16.27 0.609 0.249 -408.42 16.26 *0.069 0.485 13.37 -16.64 68.26 0.739 0.755 **0.032 40.75 33.55 -20.34 **0.018 *0.084 0.109 -37.52 -1.74 65.43 **0.017 0.946 *0.053 186.16 53.22 0.275 0.587 53.10 20.19 ***0.004 0.105 6.33 -60.90 0.818 0.142 52.83 -34.88 101.77 125.13 33.43 63.16 40.11 123.56 -96.35 0.290 0.439 0.115 0.157 0.468 0.434 0.406 **0.020 0.358 23.78 0.62 -24.96 -9.61 1.58 -3.82 10.12 -11.36 -44.83 *0.058 0.894 ***0.001 0.455 0.544 0.762 *0.083 0.277 ***0.010 -46.06 -22.78 -63.57 -76.34 -56.44 -47.08 -112.46 -43.99 -28.10 **0.037 0.214 0.146 *0.085 *0.078 0.377 **0.028 0.351 0.368 116 Table 3.6.5 (cont’d). Potato N = 281 R-squared = 0.4947 Coefficient p-value Bean N = 380 R-squared = 0.6757 Coefficient p-value Onion N = 162 R-squared = 0.4417 Coefficient p-value Independent variables Production-related variables Total seed used (kg) 3.36 **0.012 3.08 ***0.000 -47.85 0.663 Planted in rainfed plot (1=Yes) 25.88 *0.056 3.80 0.563 -18.06 0.406 Planted intercropped (1=Yes) n.a. -3.94 0.531 n.a. Planted local variety (1=Yes) -80.79 0.157 -33.75 **0.012 -22.45 0.594 Used fertilizer (1=Yes) 30.14 **0.027 -6.71 0.604 51.99 ***0.001 Used pesticides (1=Yes) 85.55 **0.013 43.48 ***0.007 49.74 0.351 Reported production costs (Kw/kg) -0.96 0.200 -0.75 **0.014 -0.26 0.191 HH reported lower harvest (1=Yes) -101.04 0.110 -41.67 **0.029 -6.30 0.791 Squared and interaction terms Age squared -0.15 0.113 -0.03 **0.035 -0.02 0.393 Dependency ratio squared 234.79 0.444 30.52 0.654 329.71 *0.052 No. adults literate squared -13.68 0.451 -12.69 **0.016 11.83 **0.015 Tropical Livestock Units squared -5.84 0.600 -9.55 **0.030 2.55 0.507 Total seed used squared -0.003 **0.011 -0.005 ***0.000 129.256 **0.028 Production costs * HH reported lower harvest 0.76 0.348 0.79 **0.016 0.12 0.457 Constant -96.33 0.140 13.04 0.384 296.44 *0.052 Notes: *, **, *** indicates the corresponding coefficients are significant at the 10%, 5%, and 1% levels, respectively. n.a. = not applicable. All municipalities compared to Londuimbali municipality. 1 Dependency ratio estimated by dividing No. members <17 yr by household size. 2 Literacy refers to adults who can read and write. Source: ProRenda survey, Angola, 2009. Estimates weighted to reflect population. 117 significant at the 10% level. Thus, providing technical assistance (related to production) to potato producers will likely generate the same outcome regardless of the gender of the head. However, since female-headed households are poorer, it may be appropriate to devote additional efforts to assist female-headed households. Surprisingly, having members of the household participating in a farmer organization had no statistically significant effect on potato production. Production was positively associated with owning a home made with improved roof materials--farmers who owed a home with a zinc roof produced, on average, 68 kg more potatoes than their counterparts (5% SL). As previously discussed, farmers who produced potatoes were relatively wealthy (see Table 3.6.3); thus, this finding was not a surprise. Although owning productive assets had a positive effect on potato production, the differences between farmers who owned productive assets vs. farmers who did not own these assets were not statistically significant at the 10% level. The variables related to different municipalities were included to control for variations in environmental and market characteristics. Farmers producing potatoes in the Chinguar municipality produced approximately 124 kg more potatoes (5% SL) than farmers in the Londuimbali municipality. The differences in production between all other municipalities and Londuimbali were not statistically significant at the 10% level. Most production-related variables had statistically significant effects on production (Table 3.6.5). Since the dependent variable in these models was production (not yield), it was expected that, as seed use increased, quantity produced would increase (at a decreasing rate). This was true for potatoes (5% SL). Surprisingly, farmers who planted potatoes in rainfed fields obtained higher production (10% SL) than farmers planting in other types of fields. This was because although most farmers (44%) produced potatoes in Lavras or upland rainfed fields, a 118 large share of farmers (38%) produced potatoes in Nacas or irrigated lowlands fields close to river deltas, generally with poorer soil quality compared to Lavras. Furthermore, during the period of analysis, rainfall may have been sufficiently abundant to achieve a good harvest on upland fields. Potato producers who reported obtaining lower harvests (compared to a normal year) stated that the main cause of this was the little or no use of fertilizer (60% of farmers reported this) rather than weather-related problems, thus confirming this finding. Although farmers planting local varieties obtained lower production, the differences between farmers who used local varieties and farmers who used improved varieties were not statistically significant at the 10% level. Since 26% of potato farmers reported using IVs, this suggests that available IVs do not perform better than traditional varieties; thus, efforts to generate better IVs will likely benefit farmers. Potato producers who used fertilizer obtained higher production (5% SL) than farmers who did not apply fertilizer to their fields. Similarly, potato producers who applied pesticides obtained higher production (5% SL) than farmers who did not apply pesticides (Table 3.6.5). The results do not suggest whether farmers did not have access to these inputs or could not afford them (i.e. due to high price). However, given that fertilizer accounted for the largest share of production costs and that more than 65% of farmers applied fertilizer, is likely that most farmers could not afford to purchase the required amounts of fertilizer. Finally, as per unit production cost increased, quantity produced decreased. Similarly, farmers who reported lower harvest during this period (compared to a normal year) obtained lower production. However, the differences in these two variables were not statistically significant at the 10% level (Table 3.6.5). 119 3.6.4.2 Bean OLS Results The econometric results of the OLS regression for beans are presented in Table 3.6.5. The model had a R-squared of 0.68. The results show that age of the head of the household was positively (at a decreasing rate) associated with bean production. Among bean producers, age 97 was positively associated with production until farmers become 34 years old. After this, age was negatively associated with the bean quantity produced. Similar to potato, although male-headed households produced more than female-headed households, the differences were not statistically significant at the 10% level (Table 3.6.5). Thus, providing technical assistance (related to production) to bean producers will likely generate the same outcome regardless of the gender of the head. However, since female-headed households in general and bean growers in particular are poorer, it may be appropriate to devote additional efforts to assist female-headed households and households producing beans. Bean production was positively associated with the number of literate adults in the household (5% SL) and with the number of TLU owned by the household (10% SL). The number of literate adults in the household had a positive effect on production until there were 1.61 literate adults at home, after which its effect becomes negative, suggesting that more educated households may depend less on the bean crop. Bean producers who owned one additional TLU produced, on average, 14.45 kg more beans (Table 3.6.5). 98 Since medium to small animals can easily be sold to relief cash constraints faced by poor bean growers, these farmers are likely using their TLUs to purchase inputs (even in small quantities) for bean 97 Setting the derivative of the bean production model with respect to the age variable equal to zero and solving for the age variable allows us to find this value. 98 This amount was obtained by evaluating the derivative of the bean production model with respect to the TLU variable at one TLU. 120 production. Since this variable was only significant for farmers producing beans (i.e. not for richer farmers producing potatoes or onions), increasing bean farmers’ access to agricultural credit would greatly benefit them. Similarly, bean production was positively associated with owning a productive asset. Farmers who owned a plow produced, on average, 53 kg of beans more than farmers who did not own this implement (Table 3.6.5). Since plows are used to prepare fields for planting, thus allowing the plants to grow better, it was expected that owning a plow would positively affect production. Since less than 1% of farmers producing beans owned a backpack sprayer (see Table A 3.8), it is not surprising that there was no statistically significant effect of this variable on bean production. Although having access to public markets in the villages had no statistically significant effect on production, the presence of the government’s extension office in the village had a statistically significant positive effect on bean production (10% SL; Table 3.6.5). Farmers located in villages with an IDA office obtained, on average, 24 kg more beans than their counterparts, suggesting that extension agents are helping these relatively poor farmers overcome some of their production constraints. Bean producers in the Tchicalachuluanga municipality produced, on average, 10 kg more beans (10% SL) than farmers in Londuimbali. In contrast, farmers in Caala and Babaera municipalities produced fewer beans (1% SL) than farmers in Londuimbali. However, the differences in production between all other municipalities and Londuimbali were not statistically significant at the 10% level. Since bean production is concentrated in Londuimbali, it was expected that production would be the same or lower in other municipalities. 121 As with potatoes, most production-related variables had statistically significant effects on production (Table 3.6.5). Since the dependent variable in this model was production (not yield), it was expected that, as seed use increased, quantity produced would increase (at a decreasing rate). This was true for beans (1% SL). In contrast to potatoes (and onions), planting a local (traditional) bean variety negatively affected production (5% SL). Bean producers who planted local varieties obtained, on average, 34 kg fewer beans than farmers using improved varieties. This result is consistent with the literature. For example, Reyes (2011) found that bean producers in Honduras obtained lower yields when using local varieties versus using improved varieties. Furthermore, since less than two percent of bean growers planted IVs (Table A 3.8), efforts should be devoted at developing and promoting the use of IVs in these regions. Making low cost, high quality seed of IVs available to farmers could also greatly benefit them. As expected, bean production was positively associated with use of pesticides. Bean producers who applied pesticides obtained, on average, 43 kg more beans than farmers who did not apply pesticides. Furthermore, as per unit production cost increased, quantity produced decreased (5% SL). Farmers reported that the largest share of their production costs was due to expenses of purchasing seed (most likely grain), followed by payments for services (e.g. labor). Thus, this result suggests that the seed used by farmers may have poor quality (e.g. low germination rates), which directly affects the quantity produced. This suggests that providing farmers with better (low-cost) seeds may greatly benefit them. Farmers who reported lower harvest during this period (compared to a normal year) obtained lower bean production (5% SL). Farmers who reported harvest losses said that the main reason for these losses was weather related (50%), followed by pest and disease incidence (25%) 122 and lack of fertilizer (21%). This suggests that breeding programs need to develop bean IVs tolerant to abiotic (e.g. droughts) and biotic (e.g. diseases) stresses. The interaction term between per unit production costs and a household reporting lower harvest had a statistically significant positive effect on production (5% SL). However, the overall effect of an increase in production costs would depend on whether farmers reported lower harvest during the period (the opposite is also true). Thus, for farmers who did not report harvest losses during the period (i.e. normal year), as production costs increased, quantity produced decreased. Although for farmers who reported harvest losses during the period, an increase in production costs would positively affect quantity produced, the magnitude of this effect was small--0.04 kg. 99 3.6.4.3 Onion OLS Results The econometric results of the OLS regression for onions are presented in Table 3.6.5. The model had a R-squared of 0.45. The results show that male-headed households produced more onions than female-headed households (10% SL)--male-headed households produced, on average, 23 kg more onions than female-headed ones. Therefore, targeting technical assistance to households led by a woman will greatly benefit them since, in addition to them being poorer than male-headed households, 100 they produce fewer onions. Furthermore, the number of dependents in the household was negatively associated with the quantity produced (10% SL), perhaps because households with more dependents typically 99 This value is found by evaluating the derivative of the bean production model with respect to the production cost variable at ‘reported lower harvest’ = YES. 100 The mean asset index values among onion producers were -0.21 and 0.44 for female- and male-headed households, respectively. 123 have many young children who cannot work in the field. Households with 0.54 dependents (mean dependency rate value) produced approximately 52 kg 101 less. However, the dependency ratio was negatively associated with production until the number of dependents becomes 102 0.62. After this, the dependency ratio was positively associated with onion production. Onion production was negatively associated with the number of literate adults living in the household (5% SL; Table 3.6.5). Since the average number of literate adults was higher among households producing onion, compared to households producing beans or potatoes, this finding was not surprising since more educated households may be less dependent on 103 agricultural outputs, including onions. In addition, onion producers owning a home with improved roof materials produced more onions (10% SL). As previously discussed, farmers who produced potatoes and onions were wealthier than farmers producing beans (see Table 3.6.3); thus, this finding was not a surprise. Onion production was not statistically significantly associated with owning productive assets (Table 3.6.5). This may have been because, in general, farmers plant a smaller area with onions (compared to potatoes and beans), thus reducing the need to own (or rent) a plow to prepare the soil. Furthermore, only a very small percent (4%; Table A 3.8) of onion farmers applied pesticides to their crop, thus making owning a backpack sprayer too expensive to afford. Although having access to public markets in the village had no statistically significant effect on production of onions, the presence of the government’s extension office in the village 101 This value is found by evaluating the derivative of the onion production model with respect to the dependency ratio variable at the mean value. 102 Setting the derivative of the onion production model with respect to the dependency ratio variable equal to zero and solving for the dependency ratio variable allows us to find this value. 103 This variable has a negative effect on onion production until there are 1.59 literate adults at home, then, its effect becomes positive. 124 was negatively associated with onion production (5% SL; Table 3.6.5). The latter suggests that the government’s extension office perhaps provides greater assistance to farmers producing beans (a staple) than to farmers producing onions. Whether to provide greater assistance to farmers producing onions would depend on many factors including the extension agent’s training and available time, the relative importance of onions compared to staple crops, etc. While farmers producing onions in Ekunha, Bailundo, and Tchicalachuluanga produced fewer onions (1% SL, 1% SL, and 5% SL) than farmers in Londuimbali (Table 3.6.5), the differences in production between all other municipalities and Londuimbali were not statistically significant at the 10% level. Regarding the production-related variables, only the use of fertilizer had a statistically significant effect on onion production (Table 3.6.5). Onion producers who used fertilizer obtained, on average, 52 kg more onions (1% SL) than farmers who did not apply fertilizer to their onion crop. Although the results do not suggest whether farmers did not have access to fertilizers or could not afford them (i.e. due to high price), given that the largest share of production costs were due to expenses in fertilizer and that 49% of farmers applied fertilizer, is likely that most farmers could not afford to purchase the required amounts of fertilizer. Although farmers were asked how much onion seed they used, given that this amount is generally reported in grams, farmers had difficulty in estimating/recalling how many grams they used. This may explain the unexpected (albeit not statistically significant) sign of this variable in the regression since is likely that there were errors in measuring this variable. 125 3.6.5 Double Hurdle Regression Results of Factors Affecting Marketing Decisions This subsection presents the double-hurdle regression results for each crop separately. The descriptive statistics for the variables included in these models were already discussed at the beginning of this section (see Table 3.6.2); therefore, this subsection focuses on the double hurdle (DH) regression results. Two additional points are worth discussing here. First, the asset index discussed in section 3.6.2 was re-estimated without two variables: owing a motorcycle and owning a bicycle. This was done to be able to estimate the effect of these variables on marketing decisions separately from other assets included in the index. As expected, the magnitude of the new index was different. 104 Although the results were similar for most comparisons (e.g. male-headed households still were richer than female-headed households) when re-estimated, the differences in the index between bean growers vs. non-growers became statistically significant (10% SL), suggesting that households growing beans were poorer than non-growers. The descriptive statistics discussed at the beginning of the results section refers to this re-estimated index. Second, the coefficient of the OLS regression residuals (estimated from the regressions in subsection 3.6.4) was not statistically significant in any of the two hurdle regressions for potatoes (p-value=0.723 for hurdle 1 and p-value=0.183 for hurdle 2) and beans (p-value=0.272 for hurdle 1 and p-value=0.267 for hurdle 2). Similarly, the onion OLS regression residuals variable was not statistically significant in the first hurdle of the onion regressions (p-value=0.577). These results suggest that quantity produced was not endogenous; therefore, this variable was excluded from both hurdles in the potato and bean regressions and from the first hurdle in the onion 104 The economic status index became: 0.4 and -0.58 (1% SL) for male- and female-headed households, respectively; -0.07 and 0.22 (5% SL) for potato non-growers and growers, respectively; 0.25 and 0.02 (10% SL) for bean non-growers and growers, respectively; and -0.09 and 0.29 (1% SL) for onion non-growers and growers, respectively. 126 regressions. However, the residuals variable was included in the truncated normal regression for onions (second hurdle) since its coefficient was highly significant (p-value=0.002), suggesting that quantity produced was endogenous in this regression. The DH results are presented for each crop separately. Further, for each crop, the discussion is divided into conditional and unconditional (on market participation) results. 3.6.5.1 Potato DH Results Conditional results The double hurdle regression results for potatoes are presented in Table 3.6.6. While maleheaded households were more likely to participate in the market as sellers (5% SL), once the market participation decision has been made, gender of the head had no statistically significant effect on the quantity of potatoes sold. Therefore, targeting assistance to female heads may be necessary to increase their participation in the potato market as sellers, which would benefit them due to increased income from potato sales. In contrast, as the number of literate adults in the household increase, farmers were less likely to sell potatoes (5% SL). However, after the participation decision has been made, having more literate adults in the household had no statistically significant effect on the amount of potatoes sold. As discussed above, the asset index was used as a proxy for household wealth. Wealthier households were more likely to sell potatoes (1% SL); however, conditional on selling potatoes, richer households sold fewer potatoes (5% SL; Table 3.6.6). These results suggest that, although richer households are more likely to participate in the market as sellers, they sell fewer potatoes. 127 Table 3.6.6. Double-Hurdle model of factors influencing potato marketing decisions. Central Highlands of Angola, 2009. HURDLE 1 HURDLE 2 Probability of selling Quantity sold (kg) Truncated Normal Probit Estimator Regression Estimator N = 240 N = 159 Independent variables: the coefficients Pseudo R2 = 0.5006 Prob > Chi2 = 0.000 displayed are the conditional average partial effects (APEs). Coefficient p-value Coefficient p-value Age of HH head (Years) -0.001 0.506 -0.299 0.532 Gender of HH head (1=Male) 0.125 **0.023 21.726 0.139 Dependency ratio 0.179 0.103 14.564 0.655 HH member is in farmer organization (1=Yes) 0.078 0.110 18.430 0.347 No. adults (>17 yr) literate -0.065 **0.024 5.492 0.484 No. of Tropical Livestock Units -0.073 0.114 8.776 0.222 Asset Index 0.051 ***0.009 -10.771 **0.021 Owns motorcycle (1=Yes) -0.043 0.700 17.110 0.689 Owns bicycle (1=Yes) -0.073 0.251 37.889 **0.024 IDA office in village (1=Yes) 0.110 *0.090 44.309 **0.012 Public market in village (1=Yes) -0.185 **0.018 -16.755 0.263 HH in Caala Municipality (1=Yes) -0.162 0.162 73.411 *0.083 HH in Ekunha Municipality (1=Yes) -0.063 0.605 83.047 **0.024 HH in Bailundo Municipality (1=Yes) 0.183 **0.016 -100.143 **0.043 HH in Katchiungo Municipality (1=Yes) 0.201 ***0.004 32.464 0.195 HH in Tchicalachuluanga Municipality (1=Yes) -0.039 0.685 -5.838 0.797 HH in Chiguar Municipality (1=Yes) 0.180 **0.031 -3.190 0.904 HH in Babaera Municipality (1=Yes) -0.086 0.529 -29.850 0.289 Distance from village to sede (km) 0.003 0.336 -1.046 0.414 Road between village and sede in poor condition (1=Yes) -0.016 0.789 -68.381 **0.015 Seller sought price information prior to sales (1=Yes) n.a. -3.418 0.814 Reported marketing costs (Kw/kg) n.a. -0.529 0.577 Total potato production (kg) 0.002 ***0.000 0.543 ***0.000 Notes: *, **, *** indicates the corresponding coefficients are significant at the 10%, 5%, and 1% levels, respectively. Coefficients and p-values obtained using the margins command in Stata. Dependency ratio estimated by dividing No. members <17 yr by household (HH) size. Literacy refers to adults who can read and write. n.a. = not applicable because variable was not included in the regression. Source: ProRenda survey, Angola, 2009. Estimates weighted to reflect population. 128 It was expected that owning vehicles would be positively associated with marketing decisions. While owning a bicycle was not associated with the probability of selling potatoes, owning this type of vehicle (conditional on market participation) was positively associated with the quantity of potatoes sold (5% SL). This was perhaps due to the fact that a bicycle could easily be used to transport potatoes to local markets or other places for sale. Having a government’s extension office in the village had a statistically significant effect on both market participation and quantity traded (10% SL and 5% SL, respectively). In contrast, having access to public markets for purchasing food/selling outputs in the village was a significantly negative factor in market participation (Table 3.6.6), which was unexpected. The main reason for this may be the fact that a higher share of non-sellers reported that public markets were available in their villages (19% vs. 16% sellers). Once the market participation decision has been made, this factor had no statistically significant effect on the quantity of potatoes sold. In contrast to Key et al. (2000) the results suggest that one of the proxies for fixed costs (i.e. road quality) had a statistically significant (5% SL) negative effect on the quantity of potatoes sold. Farmers located in villages with poor road quality between the village and the main commercial town sold fewer potatoes. Although only 33% of farmers reported selling at least one of their outputs in other markets (i.e. outside the village, for whom road quality may be important), these farmers sold more than double the amount sold by farmers selling at home or in the local market (308 kg vs. 145 kg, 1% SL). Thus, investing in improving roads could be an important factor to boost potato sales. Finally, production was a significantly positive factor on the probability of market participation and quantity traded. This was expected since farmers who have greater production 129 have more surpluses they could sell. Although the magnitude on farmers’ market participation decision was small, farmers producing one extra kilogram (above the mean) sold approximately 0.54 kg more potatoes. In summary, these conditional results suggest that, to increase the likelihood that a smallholder farmer in the central highlands of Angola would become a seller, investments are needed to (a) assist female-headed households since this type of households are less likely to sell potatoes, (b) support households with fewer literate adults since adult literacy was negatively associated with the probability of selling potatoes, (c) assist (e.g. production, marketing) poorer potato producers (as classified by the asset index) since these are less likely to sell potatoes, and (d) support farmers with low access to extension services and farmers located in villages without public markets. Conditional on being a seller, the quantity sold would increase if investments are made to (a) assist (e.g. production, marketing) poorer farmers (as per their asset index) and farmers with low access to extension services, and (b) improve the infrastructure, especially the quality of the roads. Further, both the likelihood of a farmer becoming a seller and the conditional quantity sold would be positively affected if farmers increase their production. Therefore, investments are needed to help farmers increase their production (e.g. access to inputs, farm credit, better crop management). Unconditional results The unconditional (on market participation) average partial effects (APE) of all variables are included in Table A 3.9. The APE incorporates the partial effect of both hurdles, which allows making unconditional inferences about the factors affecting the quantity of potatoes sold. Although male-headed households sold more potatoes, the differences between these households and female-headed households were not statistically significant at the 10% level. Thus, the 130 unconditional quantity of potatoes sold was gender neutral. This may be explained by the fact that 41% of female-headed households reported their (male) spouses as the ones responsible for sales (vs. 26% of male-headed households reporting female spouses as responsible for sales). This result suggests that households led by (married) females rely on their (male) spouses for marketing-related decisions; therefore, explaining why the differences in the quantity sold were not statistically significant. In contrast to the conditional quantity of potatoes sold, having a member of the household participating in a farmer organization was positively associated (10% SL) with the unconditional quantity of potatoes sold (Table A 3.9). Thus, promoting participation in these organizations or establishing farmer organizations in villages without them could boost potato sales. Furthermore, the asset index was negatively associated with the unconditional quantity of potatoes sold (10% SL). However, the magnitude of this effect was very small--an increase of one unit in the asset index would reduce the quantity sold by approximately 7 kg. The finding that richer households sell fewer potatoes may be explained by the fact that a lower percent (39%) of richer potato producers reported potatoes as the major source of crop income (compared to 43% of farmers in the poorest tercile) and because a larger percent (30%) of richer potato producers reported services as the main source of non-crop income (compared to 1% of farmers in the poorest tercile). Thus, richer farmers have diverse sources of income, which make them less dependent on potato sales. Owning a bicycle was positively correlated with the unconditional quantity of potatoes sold (1% SL). As previously discussed, a bicycle could easily be used to transport potatoes for sale. Similarly, the presence of an IDA office in the village was positively correlated with the unconditional quantity of potato sold (Table A 3.9). Farmers in villages with IDA offices sold, 131 on average, 43 kg more potatoes than farmers in villages without IDA offices. Thus, providing farmers with extension services could contribute to increase potato sales. Not surprisingly, farmers located in villages with poor road quality between the village and the main commercial town sold fewer potatoes (1% SL). Lastly, farmers producing one extra kilogram (above the mean) sold approximately 0.58 kg more potatoes (1% SL). Therefore, investing in public infrastructure (i.e. improving roads) and helping farmers increase their production could positively affect the unconditional quantity of potatoes sold. In summary, these unconditional results suggest that, to boost the unconditional quantity of potatoes sold by smallholder farmers in the central highlands of Angola, investments are needed to (a) promote farmer participation in organizations and/or establish farmer organizations in villages without them, (b) provide assistance (e.g. production, marketing) to poorer potato producers (as classified by the asset index); however, since this crop requires investments, this assistance cannot focus on farmers who are too poor, (c) support farmers with low access to extension services, (d) improve the infrastructure, especially the quality of the roads, and (e) help farmers increase their potato production, which can be done by making inputs more affordable (e.g. establish credit programs) and/or available. Thus, boosting potato sales would be a challenge for the government of Angola and donors since, due to its strong currency, overcoming these limiting factors would need financial and human resources. 132 3.6.5.2 Bean DH Results Conditional results The double hurdle regression results for beans are presented in Table 3.6.7. In contrast to potato production, the gender of the head had no statistically significant effect on the market participation decision or the quantity sold. Although households with a member participating in a farmer organization were less likely to participate in the bean market (1% SL), once the participation decision has been made, having a member in a farmer organization had no statistically significant effect on the amount of beans sold. Having a member in a farmer organization was expected to have no positive effect on marketing decisions because only a small share (1%) of bean sellers reported having a member of the family participating in a farmer organization (vs. 10% non-sellers). In contrast to potato producers, marketing decisions were not associated with the household’s wealth. This may be explained by the fact that bean producers were the poorest, with an average asset index of 0.07 (vs. 0.269 for potato and 0.364 for onion producers; see Table 3.6.3). Owning a transportation vehicle had a statistically significant effect on marketing decisions. As expected, owning a motorcycle was negatively associated with both the probability of becoming a seller and, conditional on being a seller, the quantity of beans sold. This is because, owning a motorcycle is also an indicative of wealth; hence, farmers who own this type of vehicle may be too rich and less dependent on the bean crop as a source of crop income. This is confirmed by the fact that, while 63% of bean sellers reported the bean crop as their major source of crop income, only 29% of non-sellers reported beans as their major source of crop 133 Table 3.6.7. Double-Hurdle model of factors influencing bean marketing decisions. Central Highlands of Angola, 2009. HURDLE 1 HURDLE 2 Probability of selling Quantity sold (kg) Truncated Normal Probit Estimator Regression Estimator N = 305 N = 206 Independent variables: the coefficients Pseudo R2 = 0.2452 Prob > Chi2 = 0.000 displayed are the conditional average partial effects (APEs). Coefficient p-value Coefficient p-value Age of HH head (Years) -0.002 0.428 -0.435 0.167 Gender of HH head (1=Male) -0.054 0.413 9.242 0.236 Dependency ratio -0.068 0.593 -9.951 0.469 HH member is in farmer organization (1=Yes) -0.409 ***0.000 17.282 0.237 No. adults (>17 yr) literate 0.063 0.326 4.359 0.308 No. of Tropical Livestock Units -0.045 0.491 5.268 0.328 Asset Index -0.004 0.904 1.471 0.684 Owns motorcycle (1=Yes) -0.312 *0.060 -31.898 **0.012 Owns bicycle (1=Yes) -0.114 0.226 26.971 **0.026 IDA office in village (1=Yes) -0.018 0.889 2.216 0.866 Public market in village (1=Yes) 0.115 0.207 -12.830 0.207 HH in Caala Municipality (1=Yes) -0.232 *0.091 -14.464 0.362 HH in Ekunha Municipality (1=Yes) -0.147 0.204 24.203 0.184 HH in Bailundo Municipality (1=Yes) -0.066 0.421 19.040 *0.071 HH in Katchiungo Municipality (1=Yes) -0.076 0.622 -18.010 0.107 HH in Tchicalachuluanga Municipality (1=Yes) -0.115 0.316 -12.267 0.250 HH in Chiguar Municipality (1=Yes) -0.030 0.833 -1.614 0.914 HH in Babaera Municipality (1=Yes) 0.196 **0.037 -41.862 ***0.003 Distance from village to sede (km) 0.011 ***0.009 0.149 0.691 Road between village and sede in poor condition (1=Yes) 0.153 *0.095 -4.850 0.659 Seller sought price information prior to sales (1=Yes) n.a. -12.430 0.200 Reported marketing costs (Kw/kg) n.a. -0.428 0.374 Total bean production (kg) 0.003 ***0.002 0.499 ***0.000 Notes: *, **, *** indicates the corresponding coefficients are significant at the 10%, 5%, and 1% levels, respectively. Coefficients and p-values obtained using the margins command in Stata. Dependency ratio estimated by dividing No. members <17 yr by household (HH) size. Literacy refers to adults who can read and write. n.a. = not applicable because variable was not included in the regression. Source: ProRenda survey, Angola, 2009. Estimates weighted to reflect population. 134 income (see Table A 3.2). Furthermore, farmers selling beans also depended on selling their labor in other farms as a source of non-crop income, which suggest that these farmers may have limited resources (e.g. land, capital) to diversify the crops they plant. In contrast, having a cheaper transportation vehicle (i.e. a bicycle) was positively associated with the amount of beans sold (5% SL). This was no surprise since approximately 15% of farmers in the region owned a bicycle (vs. only 7% owning a motorcycle; see Table A 3.3). Furthermore, a bicycle could easily be used to transport bean surpluses to the place of sale. Having a government’s extension office in the village had no statistically significant effect on marketing decisions (i.e. both hurdles). However, having an IDA office in the village was positively associated with bean production (see Table 3.6.5). These suggest that extension agents may be providing more assistance related to production techniques than to marketing strategies. Both the distance between the village and the main commercial town, and having a poor condition road between the village and the main commercial town oddly had a positive effect on the probability of selling beans. The fact that farmers located farther away were more likely to sell beans highlights that constrained farmers generally produce and sell beans, regardless of how far they are from the main commercial town. This is perhaps because beans could easily be stored and sold at any time after harvest. Several reasons can help explain why poor road quality had a positive effect on becoming a bean seller. First, 76% of farmers reported selling at least one of their outputs at home or in local markets (for whom distance to commercial town and road quality may not be important). Second, although these farmers sold, on average, close to one-half of the amount sold by farmers selling in other markets (77 kg vs. 151 kg, 1% SL), the aggregated volume sold by them was 135 much higher. Third, most farmers selling at home or in local markets reported convenience (58%) and lack of transportation (21%) as the main reasons for selling locally. Fourth, since beans are less perishable and could easily be stored, it is a good crop for farmers to grow when roads are bad since they could sell the crop through time either in local or distant markets (by transporting small quantities over time). Therefore, since bean trade is likely concentrated in local markets, is possible that poor quality roads incentivize participation in local markets. Finally, bean production was a positively associated with the probability of market participation and the conditional quantity traded. As with potatoes, this was expected since farmers who produce more have more surpluses they could sell. Furthermore, although the magnitude on farmers’ market participation decision was small, farmers producing one extra kilogram (above the mean) sold approximately 0.50 kg more beans. In summary, these conditional results suggest that, to increase the likelihood that a smallholder farmer would become a bean seller, investments are needed to (a) assist households with no members participating in farmer organizations since having a member of the family in a FO was negatively associated with selling beans, (b) support farmers with no transportation vehicles, perhaps on alternative ways to market their surpluses, and (c) assist (e.g. production, marketing) farmers located in villages farther away from and with poor road quality to the main commercial town. Conditional on being a seller, the quantity sold would increase if investments were made to assist farmers with no or low-cost transportation vehicles (e.g. marketing alternatives). Since gender of the household head had no statistical effect on both hurdles, any assistance should be targeted to both male- and female-headed households producing beans. Further, both the likelihood of a farmer becoming a seller and the conditional quantity sold 136 would be positively affected if farmers increase their production. Therefore, investments are needed to help farmers increase their production. Unconditional results The unconditional (on market participation) average partial effects of all variables are included in Table A 3.10. Similar to the conditional results, the unconditional APEs suggest that bean sales were gender neutral. This may be explained by the fact that 40% of female-headed households reported their (male) spouses as the ones responsible for sales (vs. 20% of maleheaded households reporting female spouses as responsible for sales). Thus, as with potatoes, households led by (married) females rely on their (male) spouses for marketing-related decisions. In contrast, unconditional bean sales were positively correlated (10% SL) with the number of literate adults living in the household (albeit in a very small magnitude), which suggests that teaching family members to read and write would increase sales perhaps due to the fact that literate adults can make better informed decisions related to marketing activities. Owning transportation vehicles had the same statistical effect on the unconditional sales of beans (Table A 3.10). While owning a motorcycle was negatively associated (1% SL) with the amount of beans sold, owning a bicycle was positively associated (1% SL) with the quantity of beans sold. Furthermore, distance between the village and the main commercial town oddly had a positive effect on the unconditional quantity of beans sold. However, the magnitude of this effect was small. As explained above, this highlights the fact that constrained farmers generally produce and sell beans, regardless of how far they are from the main commercial town. While seeking price information prior to sales was negatively associated with the unconditional quantity of beans sold (1% SL), the magnitude of this effect was small (Table A 3.10). The reason for this could be related to the quality of the information received by farmers, 137 which may have been questionable. Among farmers who sought price information prior to sales, 48% reported receiving this information from a fellow friend (most likely another farmer) and 43% from a trader. It is possible that farmers who sought price information did not trust their source of information or perhaps were expecting a better (higher) price. Thus, after learning about the price, they decided to sell fewer beans. This suggests that investments to establish marketing information systems may be needed to provide farmers with reliable market information so they can make better marketing decisions. 105 Lastly, farmers producing one extra kilogram (above the mean) sold approximately 0.48 kg more beans. Similar to potatoes, this finding suggests that investing in activities targeted at helping farmers increase their bean production could greatly boost bean sales. In summary, these unconditional results suggest that, to boost the unconditional quantity of beans sold by smallholder farmers in the central highlands of Angola, investments are needed to (a) provide assistance (e.g. production, marketing) to younger farmers, (b) teach farmers (and educate children) how to read and write so they can make better informed decisions related to marketing activities, (c) establish marketing information systems to provide farmers with reliable market information, especially about prices, and (d) help farmers increase their bean production, which can be done by providing them with access to education, productive assets, extension services, improved varieties, inputs, and agricultural credit. Thus, as with potatoes, boosting bean sales would be a challenge for the government of Angola and donors. 105 Although there may be another possible explanation related to the existence of shallow markets, proving this is more difficult and outside the scope of this study. 138 3.6.5.3 Onion DH Results Conditional results Table 3.6.8 shows the double hurdle regression results for onions. While older farmers were less likely to sell onions (10% SL), conditional on selling, older farmers sold more onions than younger farmers (5% SL) perhaps due to the fact that older farmers may be better connected with traders (i.e. social capital) and may have more marketing experience. Similar to bean production, the gender of the head had no statistically significant effect on the market participation decision or the quantity sold, suggesting that marketing decisions are gender neutral. Although households with more dependents were less likely to sell onions (10% SL), once the marketing participation decision has been made, households with more dependents sold more onions (5% SL). In contrast to bean production, households with a member participating in a farmer organization were more likely to sell onions; however, this factor had no statistically significant effect on the quantity sold. Thus, establishing farmer organizations in villages without them would contribute to increase market participation. The number of TLUs was positively associated (5% SL) with the probability of selling onions. However, conditional on market participation, this factor was negatively associated (5% SL) with the quantity of onion sold. In contrast, the asset index (proxy for household wealth) was negatively associated (5% SL) with the probability of selling onions. This was expected since non-sellers were richer (as per their asset index; see Table A 3.4) than onion sellers. Furthermore, non-sellers reported having additional sources of non-crop income, including transfers and remittances, and other activities (see Table A 3.2), which may be invested in 139 Table 3.6.8. Double-Hurdle model of factors influencing onion marketing decisions. Central Highlands of Angola, 2009. HURDLE 1 HURDLE 2 Probability of selling Quantity sold (kg) Truncated Normal Probit Estimator Regression Estimator N = 174 N = 103 Independent variables: the coefficients Pseudo R2 = 0.3767 Prob > Chi2 = 0.000 displayed are the conditional average partial effects (APEs). Coefficient p-value Coefficient p-value Age of HH head (Years) -0.005 *0.092 0.286 **0.039 Gender of HH head (1=Male) 0.141 0.132 0.346 0.968 Dependency ratio -0.323 *0.061 16.448 **0.018 HH member is in farmer organization (1=Yes) 0.177 ***0.005 -2.438 0.667 No. adults (>17 yr) literate 0.053 0.164 2.228 0.235 No. of Tropical Livestock Units 0.281 **0.012 -7.777 **0.030 Asset Index -0.087 **0.015 -1.596 0.454 Owns motorcycle (1=Yes) -0.199 0.201 3.766 0.765 Owns bicycle (1=Yes) 0.105 0.257 -0.987 0.881 IDA office in village (1=Yes) -0.024 0.803 3.737 0.702 Public market in village (1=Yes) 0.142 *0.057 1.551 0.804 HH in Caala Municipality (1=Yes) -0.082 0.604 2.857 0.844 HH in Ekunha Municipality (1=Yes) -0.192 0.407 6.849 0.494 HH in Bailundo Municipality (1=Yes) -0.093 0.361 0.776 0.945 HH in Katchiungo Municipality (1=Yes) 0.055 0.693 8.279 0.459 HH in Tchicalachuluanga Municipality (1=Yes) -0.125 0.339 0.269 0.976 HH in Chiguar Municipality (1=Yes) 0.200 *0.082 -0.036 0.996 Distance from village to sede (km) -0.014 **0.011 -0.509 0.310 Road between village and sede in poor condition (1=Yes) 0.253 **0.021 7.166 0.352 Seller sought price information prior to sales (1=Yes) n.a. 2.626 0.244 Reported marketing costs (Kw/kg) n.a. -0.192 0.285 Total onion production (kg) 0.003 **0.044 0.810 ***0.000 Residual from onion production equation n.a. -0.107 **0.025 Notes: *, **, *** indicates the corresponding coefficients are significant at the 10%, 5%, and 1% levels, respectively; p-values obtained via bootstrapping at 500 repetitions in hurdle 2; coefficients in both hurdles along with p-values in hurdle 1 obtained using the margins command in Stata. Dependency ratio estimated by dividing No. members <17 yr by household (HH) size. Literacy refers to adults who can read and write. n.a. = not applicable because variable was not included in the regression. Source: ProRenda survey, Angola, 2009. Estimates weighted to reflect population. 140 improving the home or purchasing household assets, both reflected in the asset index. These results suggest that non-sellers were less dependent on onions since they had diverse sources of income. While having a public market in the village was positively associated (10% SL) with the likelihood of selling onions, once this decision has been made, this factor was not statistically associated with the quantity of onions sold. In contrast, onion producers located farther away from the main commercial town were less likely to sell onions (5% SL). However, once this decision has been made, the distance between the village and the main commercial town had no statistical effect on the quantity of onions sold perhaps due to the small amounts sold by onion sellers (50 kg; see Table 3.6.2). Strangely, having a poor quality road between the community and the main commercial town was positively associated (5% SL) with the probability of selling onions. This may be given by the fact that 81% of farmers selling onions were located in villages with poor road quality between the village and main commercial town (vs. only 53% of non-sellers; see Table 3.6.2). Additionally, since 74% of farmers sold at least one of their harvests either at home or in local markets and since the differences in the quantity sold by farmers selling in local markets compared to farmers selling in other markets were not statistically significant, it is possible that poor road quality incentivized local sales (as also suggested by the effect of having a public market in the village), reflecting sales targeted at meeting local demand. Once the market participation decision has been made, the quality of the road had no statistically significant effect on the amount sold. Finally, onion production was a significantly positive factor on the probability of market participation and the quantity traded. Even after controlling for endogeneity in the second hurdle, 141 quantity produced was a highly statistically significant positive factor in the quantity traded. Although the magnitude on farmers’ market participation decision was small, farmers producing one extra kilogram (above the mean) sold approximately 0.81 kg more onions. In summary, these conditional results suggest that, to increase the likelihood that a smallholder farmer in the central highlands of Angola would become an onion seller, investments are needed to (a) assist (e.g. production, marketing, storage) younger farmers and poorer farmers, (b) promote or establish farmer organizations where farmers could participate in (and perhaps find alternative ways to market their surpluses) and public markets where farmers could trade their surpluses, (c) mitigate cash constraints (e.g. agricultural credit) affecting onion producers since the number of tropical livestock units (proxy for cash availability) was positively associated with being a seller, and (d) assist (e.g. production, marketing, storage) farmers located in villages closer from and with poor road quality to the main commercial town. Conditional on being a seller, the quantity sold would increase if assistance (e.g. production, marketing, storage) is provided to (a) older farmers, (b) households with more dependents, and (c) farmers facing cash constraints (as per the number of tropical livestock units owned). Since gender of the household head had no statistical effect on both hurdles, any assistance should be targeted to both male- and female-headed households producing onions. Further, both the likelihood of a farmer becoming a seller and the conditional quantity sold would be positively affected if farmers increase their onion production. Therefore, as with potatoes and beans, investments are needed to help farmers increase their onion production (e.g. access to inputs, farm credit, better crop management). 142 Unconditional results The unconditional APEs of all variables are included in Table A 3.11. Similar to potato and beans, the quantity of onions sold was gender neutral. Since 34% of female-headed households reported their (male) spouses as the ones responsible for sales and 32% of maleheaded households reported their (female) spouses as responsible for sales, this finding was not surprising. In contrast, the unconditional quantity of onion sold was positively associated with the number of literate adults in the household, suggesting that teaching family members to read and write would increase sales perhaps due to the fact that literate adults can make better informed decisions related to marketing activities. Richer onion producers owning more assets (as indicated by the asset index) would sell fewer onions (1% SL). This was expected since richer farmers may be less dependent on the onion crop as a source of income. Although the presence of an IDA office in the village had no statistically significant (at the 10% level) effect on onion sales, having a public market in the village would positively affect sales (Table A 3.11). Similar to the conditional results, the distance between the village and the main commercial town was negatively associated (5% SL) with the unconditional quantity of onions sold (Table A 3.11). Surprisingly, having a poor quality road between the village and the main commercial town positively affect sales. As explained above, the lack of alternative places for sale and the fact that most transactions occur in local markets may help explaining this finding. Furthermore, since the quantity traded is relatively small, it is perhaps not worth for farmers to travel to other (distant) markets to sell their outputs. 143 Not surprisingly, as per unit marketing costs increased, the unconditional quantity sold decreased (10% SL; Table A 3.11). Since most of the marketing costs were due to transportation costs, this also helps explain why farmers located farther away would sell less. Finally, farmers producing one extra kilogram (above the mean) sold approximately 0.68 kg more onions. As with potatoes and beans, this finding suggests that investing in activities targeted at helping farmers increase their onion production will greatly boost onion sales. In summary, these unconditional results suggest that, to boost the unconditional quantity of onions sold by smallholder farmers in the central highlands of Angola, investments are needed to (a) teach farmers (and educate children) how to read and write so they can make better informed decisions related to marketing activities, (b) provide assistance to poorer onion producers (as classified by the asset index); however, since the onion crop requires investments, this assistance should not focus on farmers who are extremely poor, (c) promote or establish public markets within villages, (d) reduce transaction costs related to marketing activities, especially transportation costs, and (e) help farmers increase their onion production, which can be done by providing them with access to agricultural inputs. 144 3.7 Chapter Summary and Policy Recommendations 3.7.1 Chapter Summary This essay uses single equation ordinary least squares regressions for analysis of factors affecting production of potatoes, beans, and onions in the central highlands of Angola. Furthermore, it implements double hurdle regressions to study the factors associated with marketing decisions among potato, bean and onion growers, focusing on gender of the household head, asset ownership, and transaction costs. The data used in this study came from the cross sectional household- and village-level survey implemented by World Vision’s ProRenda project in 2009, which collected information from 656 households distributed across 40 villages and three provinces in the central highlands of Angola. The results suggest that the quantity produced is exogenous in potato, bean, and onion models for market participation (first hurdle) and in potato and bean models for quantity sold (second hurdle). In contrast, in the onion model for quantity sold, the amount of onion produced is found to be endogenous. The wealth analysis using the asset index determined that male-headed households had more assets than female-headed households and that bean growers had the lowest economic status index among farmers growing potatoes, beans or onions. Furthermore, while potato and onion growers were richer than non-growers, there were no statistically significant differences in wealth between bean growers and non-growers. In contrast, while potato sellers were richer than their counterparts, onion sellers were poorer than non-sellers and there were no statistically significant differences in wealth between bean sellers and non-sellers. The latter finding was no surprise since beans are a staple crop grown both for consumption and sale by farmers across all wealth categories. The assets with more ‘weight’ (i.e., more important) in the estimation of the asset index included owning a television, a cart, a motorcycle, and a backpack sprayer. 145 The OLS regression results suggest that potato production was gender neutral. Furthermore, potato production was positively associated with owning a home with improved roof materials (i.e. zinc roofs). Surprisingly, owning productive assets had no statistically significant effect on potato production. Similarly, although using local varieties negatively affected production, the effect was not statistically significant at the 10% level. Further analysis in this area may be valuable to learn about the effects of using local vs. improved potato varieties. As expected, use of agricultural inputs (i.e., fertilizer and pesticides) positively affected production. The results suggest that, although inputs were available, is likely that farmers could not afford them, which is understandable since Angola’s strong currency makes inputs very expensive. Regarding bean production, the OLS regression results suggest that older farmers produced more beans perhaps because they have better production experience. Similar to potato production, bean production was gender neutral. Furthermore, having more literate adults at home and owning more tropical livestock units (a quick source of cash if needed) were positively associated with bean production. Likewise, owning a productive asset and having a government extension office in the village positively affected production. Not surprisingly, using local varieties had a statistically significant negative effect on bean production. Making available to farmers low-cost, high-quality seed of improved bean varieties would greatly benefit them because (a) farmers who reported lower harvest said that the main reasons for this were weather- and pest and disease-related, which highlights the need to develop/provide IVs with resistance to diseases and tolerance to abiotic factors (e.g. droughts), (b) only a small share of bean producers used IVs, which highlights the need to promote existing bean IVs, and (c) the largest share of production costs was incurred in purchasing seed (most 146 likely grain), which suggest that the quality of the seed used by farmers is low. Furthermore, the use of pesticides was positively associated with the quantity of beans produced. As with potatoes, affording pesticides appears to be the limiting factor. The OLS regression results suggest that, in contrast to potatoes and beans, onion production was higher among male-headed households. Thus, assisting households led by females would contribute to reduce poverty and food insecurity since these households are at disadvantage. Furthermore, having more dependents and more literate adults at home was negatively associated with onion production. In contrast, onion production was positively associated with owning a home with improved roof materials (i.e. zinc roofs). Surprisingly, the presence of a government extension office in the village was negatively associated (5% SL) with onion production, perhaps because extension services may be targeted at farmers producing staple crops (e.g. beans) instead of farmers producing high-value crops (e.g. onions). As expected, use of agricultural inputs (i.e., fertilizer) positively affected production. As with potatoes and beans, it appears that farmers cannot afford purchasing fertilizers. The double hurdle regression results suggest that the factors associated with marketing decisions depend on the crop analyzed and on whether marketing decisions are analyzed conditionally (i.e., probability of selling and, conditional on selling, quantity sold) or unconditionally (i.e., unconditional quantity sold). For potatoes, the conditional results suggest that male-headed households were more likely to sell, and that households with more literate adults and having access to a public market in the village were less likely to sell potatoes. However, once the market participation decision has been made, gender of the head, number of literate adults, or having a public market in the village had no statistically significant effect on the quantity of potatoes sold. 147 In contrast, although richer households (as per their asset index) were more likely to participate in the market as sellers, once this decision has been made, richer households sold fewer potatoes. Furthermore, while owning a bicycle was not associated with the probability of selling potatoes, owning a bicycle was positively associated with the conditional quantity of potatoes sold. Moreover, although the quality of the road between the village and the main commercial town was not associated with the probability of selling potatoes, poor quality road was negatively associated with the amount of potatoes sold. Both having a government’s extension office in the village and the amount of potato produced had statistically significant positive effects on both the likelihood of being a seller and, conditional on selling, the quantity sold. Although the magnitude of the latter on the market participation decision was small, approximately 54% of increased production, conditional on market participation, would be sold. The unconditional analysis suggests that potato sales were gender neutral. Furthermore, having a household member participating in a farmer organization, owning a bicycle, the presence of an IDA office in the village, and quantity produced all positively influenced the unconditional quantity of potatoes sold. For each additional kilogram produced, approximately 58% would end up being sold. In contrast, the asset index and having a poor quality road between the village and the main commercial town both negatively affected the unconditional quantity sold. For beans, the conditional results suggest that marketing decisions were gender neutral. Furthermore, while households with a household member participating in a farmer organization were less likely to sell beans, households located farther away from the main commercial town or with poor road quality between the village and main commercial town oddly were more likely 148 to sell beans. The fact that farmers located farther away or with poor road quality were more likely to sell beans highlights the fact that constrained farmers generally produce and sell beans, regardless of how far they are from the main commercial town or the quality of the road between these two places perhaps because beans are an important source of cash. Once the market participation decision has been made, participating in a farmer organization, distance to main commercial town, or road quality had no statistical effect on the conditional quantity sold. While owning a motorcycle was a significantly negative factor on both the probability of becoming a seller and, conditional on being a seller, the quantity of beans sold, owning a bicycle was positively associated only with the conditional quantity of beans sold. Furthermore, the amount of beans produced positively affected both the likelihood of being a seller and, conditional on selling, the quantity sold. Approximately 50% of increased production, conditional on market participation, would be sold. The unconditional analysis suggests that bean sales also were gender neutral. Furthermore, the number of literate adults living in the household, owning a bicycle, distance to main commercial town, and quantity produced all positively influenced the unconditional quantity of beans sold. For each additional kilogram produced, approximately 48% would end up being sold. In contrast, owning a motorcycle and seeking price information prior to sales both negatively affected the unconditional quantity sold. Lastly, for onions, the conditional results suggest that the likelihood of selling onions was negatively affected by age of the household head and the dependency ratio. However, conditional on selling, these factors positively affected quantity sold. As with beans, marketing decisions were gender neutral. In contrast, households with a household member participating in a farmer organization, with access to a public market in the village, or reporting poor quality 149 road between the village and main commercial town were more likely to sell beans. However, once the market participation decision has been made, participating in a farmer organization, access to a public market in the village, or road quality had no statistical effect on the conditional quantity sold. Richer households (as per their asset index) and households located farther away from the main commercial town were less likely to sell onions. Once the market participation decision has been made, these factors had no statistical effect on the quantity sold. Furthermore, while the probability of selling onions was positively associated with the number of tropical livestock units owned, conditional on selling, quantity sold was negatively associated with the number of TLUs owned. As expected, the amount of onions produced positively affected both the likelihood of being a seller and, conditional on selling, the quantity sold. Approximately 81% of increased production, conditional on market participation, would be sold. The unconditional analysis suggests that onion sales also were gender neutral. Furthermore, the number of literate adults living in the household, having a public market in the village, having a poor quality road between the village and the main commercial town, and quantity produced all positively influenced the unconditional quantity of onions sold. For each additional kilogram produced, approximately 68% would end up being sold. In contrast, the asset index, distance from village to main commercial town, and per unit marketing costs all negatively affected the unconditional quantity of onions sold. 150 3.7.2 Policy Recommendations The results suggest that different policies would be needed to increase the participation of farmers in crop-markets as sellers and the quantity they sell. The specific policies would depend on the crop being analyzed. Thus, for potatoes, the following policy recommendations are proposed: A. To boost the participation of farmers in the potato market as sellers, the government of Angola, donors, and organizations that work with farmers should target their assistance to female-headed households (because they are less likely to sell potatoes), households composed by a high number of illiterate adults, and households without access to local public markets. Furthermore, since potato production requires the use of inputs and since the likelihood of selling potatoes was positively associated with the asset index, a larger impact could be achieved by assisting farmers who are not extremely poor. Additionally, providing extension services would also increase market participation. B. To increase the unconditional quantity of potatoes sold by smallholder farmers in the central highlands of Angola, investments are needed to (1) promote farmer participation in organizations and/or establish farmer organizations in villages without them, (2) provide assistance (e.g. production) to poorer potato producers (as classified by the asset index); however, since this crop requires investments, this assistance should not focus on farmers who are extremely poor, (3) provide farmers extension services related to both production and marketing aspects, and (4) improve the infrastructure, especially the quality of the roads. For beans, the following policy recommendations are proposed: A. To encourage the participation of farmers in the bean market as sellers, the government of Angola, donors, and organizations that work with farmers should target their assistance to 151 both male- and female-headed households, households with no members participating in farmer organizations, and households without transportation vehicles. Furthermore, assistance should be provided to farmers located in distant markets and with poor road access. B. To boost the unconditional quantity of beans sold by smallholder farmers in the central highlands of Angola, investments are needed to (a) provide assistance to younger farmers, (b) teach farmers (and educate children) how to read and write so they can make better informed decisions related to marketing activities, and (c) establish marketing information systems to provide farmers with reliable market information, especially about prices. For onions, the following policy recommendations are proposed: A. To boost the participation of farmers in the onion market as sellers, assistance should target both female- and male-headed households, households led by young heads, households composed by fewer dependents for every adult, households that are asset-constrained, and households located closer to main commercial towns. Furthermore, although it was not possible to separate cause from effect, promoting farmer participation in organizations and/or establish farmer organizations in villages without them would most likely increase market participation. B. To boost the unconditional quantity of onions sold by smallholder farmers in the central highlands of Angola, investments are needed to (a) teach farmers (and educate children) how to read and write so they can make better informed decisions related to marketing activities, (b) provide assistance to poorer onion producers (as classified by the asset index); however, since the onion crop requires investments, this assistance should not focus on farmers who 152 are extremely poor, (c) promote or establish public markets within villages, and (d) reduce transaction costs related to marketing activities, especially transportation costs. Finally, although the quantity sold of each of the three crops was gender neutral, given that female-headed households were poorer than their counterparts, there is a need for special programs targeted at female-headed households to help them move out from poverty. Furthermore, since the quantity sold of each of the three crops would increase if farmers’ outputs were increased, investments should be made to help farmers increase their outputs. For potatoes, this could be achieved by making inputs more affordable and/or available on credit. For beans, providing farmers with access to education, productive assets, extension services, improved varieties, agricultural inputs, and agricultural credit would contribute to increase their production. Finally, for onions, providing farmers with access to agricultural inputs (perhaps through credit) would positively affect their production. Thus, boosting sales would be a challenge for the government of Angola, donors, and organizations working with farmers since, due to its strong currency, overcoming these limiting factors may require large financial and human resources. 153 APPENDIX 154 Table A 2.1. Bean trials planted in Central America and Ecuador. 1999-2010. Year 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 Mean ECAR details 1 Central America: "ECAR trials" # trials included in analysis # countries where planted 2 3 # lines per trial # replications per line Ecuador: "Prueba Trials" 2 # locations where planted Average # lines per trial 3 9 3 9 5 5 3 7 4 6 3 5 3 10 4 8 4 6 3 5 3 n.a. n.a. 7.2 3.5 7 9 9 4 7 6 3 8 6 4 5 n.a. 6.2 16 3 # locations where planted 9 3 16 3 16 3 16 3 16 3 16 3 16 3 16 3 16 3 16 3 16 3 n.a. n.a. 16 3 n.a. n.a. n.a. n.a. 2 2 1 3 11 1 5 1 3 n.a. n.a. n.a. n.a. 17 18 18 8 13 11 12 10 13 Source: Programa de Investigaciones en Frijol Metadata, Zamorano, Honduras. n.a. = not available. 1 ECAR = Ensayo Centroamericano de Adaptacion y Rendimiento. 2 For Central America: sometimes, the same location is used twice per year (i.e. two seasons). Therefore, the number of locations may be different (less) than the number of trials planted. For Ecuador: # locations is the same as the number of trials included in the study. 3 For Central America: from the 16 lines, one is a local check and one is an universal check. Therefore, this nursery contains only 14 advanced lines. For Ecuador: number only includes advanced lines and is the average of the number of lines planted in each location (the number of lines varied per location). 155 Table A 2.2. Costa Rica: Improved bean varieties released. 1990-2010. Year of Variety No. release Name 1 2009 Disquis 2 2009 3 2007 4 2006 5 6 7 8 9 2006 2006 2004 2003 2000 10 2000 11 1996 12 1996 13 1995 14 1995 15 1994 16 1993 Used Developed Ever Market CRSP through widely class funds? PPB? planted? Red YES YES n.a. 1 Genealogy (SEA 15 x MD 2324) F1 x (MD 3075 x G 21212) F1 / MC-6P-MQ-MC-IIC-MC-MC Suru MEB 2232-29 PAN 68 x MD 2324 White Tongibe BCH 9901-14 MD 3075 // SRC 1-1-18 / SRC 1-12-1 Red Changuena MR 13652-39 (BRIBRI x (VAX 1 x RABA 655)) F1 / (NN) Red Q-(NN)-C Curre MPCR 202-26-1 Selection from MPCR 202 Red Gibre MPCR 202-30-2 Selection from MPCR 202 Red Telire EAP 9510-1 MD 3075 / DICTA 105 Small red Cabécar EAP 9510-77 MD 3075 / DICTA 105 Small red UCR 55 NJBC-20601-1- NAB 44 // ROS 24 / G 13689 Black CM(71) Bribri MD 2324 (RAB 310 x XAN 155) x (DOR 391 x Small red Pompadour G) Guaymí MUS 106 XAN 176 x IN 63 Black Maleku RAB 572 (MUS70 x RAO27) x (SEL960 x (RAO29 x Small red (RAB58 x (DOR 164 x IN 199)))) CIAT 95 MUS 181 (XAN 226 x MUS 46) x (G 18252 x (G 13920 Black x (G 13920 x (G 13920 x G2333)))) Chirripo Rojo DOR 489 DOR 367 x (DOR 364 x BAT 1298) Small red UCR 52 DOR 390 (DOR 364 x G 18521) x (DOR 365 x LM Black 30630) UCR 51 DOR 474 DOR 367 x (DOR 364 x BAT 1298) Small red Line ID MR 14215-9 156 YES YES YES YES YES YES n.a. n.a. n.a. YES YES YES YES n.a. YES YES YES YES NO n.a. n.a. n.a. n.a. NO YES YES YES n.a. n.a. NO NO n.a. n.a. n.a. NO n.a. n.a. NO NO NO n.a. n.a. n.a. NO n.a. Table A 2.2 (cont’d). Used Developed Ever Year of Variety Market CRSP through widely 1 Genealogy No. release Name Line ID class funds? PPB? planted? 17 1993 Puricise BAT 76 (G 1741 x G 2045) x (G 4792 x G 5694) Black n.a. NO n.a. 18 1992 UCR 50 DOR 364 BAT 1215 x (RAB 166 x DOR 125) Small red NO NO n.a. Source: CIAT (2001a), INTA and UCR (2005), Hernandez (2010), Martinez (2003), and KII (2010a). n.a. = not available. 1 Same pedigree implies that the lines are sisters, i.e. they come from the same parents. NOTES: Varieties Bribri, Changuena, Curre, and Gibre came from crosses made at the breeding program of Zamorano, which receives CRSP funds. 157 Table A 2.3. El Salvador: Improved bean varieties released. 1990-2010. Year of release Variety Name 1 2008 CENTA Nahuat 2 2008 CENTA C.P.C. 1 3 4 No. 2005 2002 CENTA Pipil CENTA San Andres Genealogy SRC 1-12-1 / MD 3075 Concha Rosada / SRC 1-1-18 / SRC 1-2-12 PRF 9653-16B-3 Bribri / MD 3037 // RS 3 EAP 9510-77 CENTA 2000 x DICTA 105 5 2000 CENTA 2000 MD 3075 Line ID SRC 2-18-1 PPB 11-20 MC DOR 483 x (DOR 391 x Pompadour J) 6 1997 ROJO Salvadoreño 1 DOR 482 DOR 367 x (DOR 364 x LM 30649) 7 1995 CENTA Costeño DOR 585 DOR 364 x SEL 1079 8 1993 DOR 582 DOR 582 n.a. 9 1989 CENTA Cuscatleco DOR 364 BAT1215 x (RAB 166 x DOR 125) Source: CENTA (2005), CIAT (2001a), Martinez (2003), Reyes (2010), and KII (2010a). 1 Same pedigree implies that the lines are sisters, i.e. they come from the same parents. 158 Market class Small red Small red Used Developed Ever CRSP through widely funds? PPB? planted? NO n.a. YES YES NO n.a. Small red Small red YES YES NO NO YES YES Small red YES NO NO Small red NO NO NO Small red Small red Small red NO NO NO NO NO NO YES NO YES n.a. = not available. Table A 2.4. Honduras: Improved bean varieties released. 1990-2010. No. 1 2 3 4 5 6 7 8 9 Year of release Variety Name 2009 Quebradeño 2009 La Majada AF 2009 Briyo AM 2009 Milagrito 2007 Cardenal 2007 Deorho 2007 Victoria 2007 Don Cristóbal 2007 Conan 33 10 2005 Palmichal 1 11 2005 Nueva Esperanza 01 12 2004 Macuzalito 13 2003 Amadeus 77 14 2003 Carrizalito 15 2003 Cedrón 16 2003 Cayetana 85 17 1997 DICTA 113 Line ID IBC307-7 IBC301-182 IBC306-95 F0243 MER 2226-41 SRC 2-18-1 SRS56-3 SRC1-12-1-8 PRF9653-25B-1 PRF 9707-36 DICZA 9801 1 Genealogy TC75//TC75/Cincuenteño Amadeus77//Amadeus77/Paraisito Amadeus77//Amadeus77/Rojo de Seda Mass selection from landrace SRC 1-12-1-47 / Amadeus 77 SRC 1-12-1 / MD 3075 Amadeus77/SEA5 DOR476//XAN155/DOR364 EAP 9503 / RS3 // Bribri / MD 30-37 //// EAP 9503 / RS3 // A429 / K2 /// V8025 / XR 16492 // APN83 / CNC UPR 9356-26 / TC-75 // EAP 9507 / AL12 UPR 9606-2-2 / MD 30-37 PPB 9911-44-513M EAP 9510-77 EAP 9510-1 PTC 9557-10 Concha Rosada // SRC 1-1-18 / SRC 1-121 MD 3075 / DICTA 105 MD 3075 / DICTA 105 EAP 9021 / Bribri // UPR 9356-26 / UPR 9438-129 PRF 9653-16B- EAP 9503 / RS3 // Bribri / MD 30-37 //// 2A EAP 9503 / RS3 // A429 / K2 /// V8025 / XR 16492 // APN83 / CNC DICTA 113 DOR 364 x APN 83 159 Market class Small red Small red Small red Small red Small red Small red Small red Small red Small red Used Developed Ever CRSP through widely funds? PPB? planted? YES YES NO YES YES NO YES YES NO YES YES NO YES NO NO YES NO YES YES YES NO YES YES NO YES YES NO Small red Small red YES YES YES YES NO NO Small red YES YES NO Small red Small red Small red YES YES YES NO NO YES YES YES NO Small red YES YES NO Small red NO NO NO Table A 2.4 (cont’d). Year of No. release Variety Name 18 1997 DICTA 122 19 1996 Tío Canela 75 20 1992 Don Silvio 21 1990 Dorado Line ID DICTA 122 MD 3075 DOR 482 DOR 364 1 Genealogy DOR 364 x APN 83 DOR 483 // DOR 391 / Pompadour J (DOR 367 x (DOR 364 x LM 30649)) (BAT 1215 x (RAB 166 x DOR 125)) Used Developed Ever Market CRSP through widely class funds? PPB? planted? Small red NO NO NO Small red YES NO YES Small red NO NO NO Small red NO NO YES Source: CIAT (2001a), DICTA (1987, 1998), Escoto (2000, 2006), Martel-Lagos (1995), Pejuan (2005), PIF/EAP and DICTA/SAG (2002a, 2002b), PIF/EAP (2003), ASOCIAL Yorito-Sulaco-Victoria et al. (2004), ASOCIALAYO (2005a, 2005b), PIF/EAP and DICTA/SAG (2005a, 2005b), Rosas (2006), and KII (2010a). 1 Same pedigree implies that the lines are sisters, i.e. they come from the same parents. n.a. = not available 160 Table A 2.5. Nicaragua: Improved bean varieties released. 1990-2010. Year of No. release Variety Name Line ID 1 2009 INTA Fuerte Sequia SX 14825-7-1 2 2006 INTA Precoz SRC 2-18 3 2006 INTA Pueblo Nuevo MR 13046-28JM SM4 4 2002 INTA Rojo EAP 9510-77 5 2001 INTA Nueva Guinea DOR 390 Genealogy n.a. Rojo Nacional // Bribri / MD 3075 (VAX 3 x Catrachita) x MD 3075 Red NO YES NO Light red Black YES NO YES YES YES NO Black NO YES NO Small red Small red Small red Small red Small red Small red Small red Small red Dark red Red YES NO NO NO NO NO NO NO NO NO YES YES YES YES YES YES YES YES YES YES YES NO NO NO YES NO NO NO YES NO 1 MD 3075 x DICTA 105 (DOR 364 x G 18521) x (DOR 365 x LM 30 630) 6 2001 INTA Cardenas DOR 500 (DOR 364 x G 18521) x (DOR 365 x IN 100) 7 2001 INTA Canela MD 3075 DOR 483 // DOR 391 / Pompadour J 8 1996 COMPAÑIA RAB 463 G 18244 x MUS 6 9 1994 CNIGB 93 DOR 391 DOR 367 x(DOR 364 x LM 30649) 10 1993 COMPAÑIA 93 PVA 692 G 14013 x(G 13352 x G 21720) 11 1993 DOR 364 DOR 364 BAT 1215 x (RAB 166 x DOR 125) 12 1990 ESTELI 90A CNIGB 1-90 Orgulloso x BAT 1654 13 1990 ESTELI 90B CNIGB 2-90 Orgulloso x BAT 1836 14 1990 ESTELI 150 CNIGB 3-90 Chile Rojo x RAO 36 15 1990 INTA Masatepe DOR 582 n.a. 16 1990 INTA Estelí CM-12214-25 n.a. Source: CIAT (2001a), INTA (2006), Martinez (2003), and KII (2010a). 1 Used Developed Ever CRSP through widely funds? PPB? planted? YES YES n.a. YES YES n.a. Market class Dark red Red Same pedigree implies that the lines are sisters, i.e. they come from the same parents. 161 n.a. = not available Table A 2.6. Ecuador: Improved bean varieties released. 1990-2010. Year of No. release Variety Name 1 2010 Rojo del Valle 2 2010 Afroandino 3 2009 Paragachi Andino 4 2009 Portilla Line ID INIAP-481 INIAP-482 INIAP-429 INIAP-430 5 2009 Rocha INIAP-480 6 2007 Libertador INIAP-427 7 2007 Canario INIAP-428 Guarandeno 8 2005 Canario del Chota INIAP-420 9 2004 Yunguilla INIAP-414 1 Genealogy SEL1308/Red Hawk // JeMa/3/Paragachi Selection of CIAT A55 SUG 26 x CAL 82 INIAP 414 x INIAP 424 Used Developed Ever Market CRSP through widely class funds? PPB? planted? Red Mottle YES YES n.a. Black YES YES n.a. Red Mottle YES YES YES Red Mottle YES YES YES INIAP 420 x (Cocacho x San Antonio), s26 p1 Yellow G 12722 x G 21720 Red Mottle Selection of local variety Yellow YES YES YES YES YES YES YES n.a. n.a. CAP 9 x Canario Bola ICA 24 x ICA 10009 x Mulato Gordo Yellow Red Mottle YES YES YES YES n.a. NO YES YES YES YES YES NO 10 2004 La Concepción INIAP-424 Selection of local variety Mil Uno 11 2004 Blanco Fanesquero 12 2004 Canario Siete Colinas** 13 2003 Blanco Belen 14 2003 Canario 15 2003 Boliche 16 2003 Doralisa 17 1999 Bolívar** 18 1998 Chaupeño INIAP-425 SUG 55 x INIAP 417 Purple Mottled White INIAP-426 TIB 3042 X G 11732 Yellow YES YES n.a. White Yellow Red Kidney Red Mottle Red Cream (Bayo) YES YES YES YES YES NO NO NO NO NO NO NO n.a. n.a. n.a. n.a. n.a. n.a. INIAP-422 INIAP-423 INIAP-473 INIAP-474 INIAP-421 INIAP-419 WAF 82 x INIAP 417 CAP 9 x Canario Bola AFR 298 AFR 722 G 12670 x G 12488 S 24990 x A 197 162 Table A 2.6 (cont’d). Year of No. release Variety Name 19 1996 Blanco Imbabura 20 1996 Je.Ma. 21 1995 Canario 22 1993 Toa** Used Developed Ever CRSP through widely Market class funds? PPB? planted? White NO NO n.a. Red Mottle NO NO n.a. Yellow NO NO n.a. Red Mottle NO NO n.a. 1 Genealogy n.a. G 12722 x G 21720 Selection of G 11780F (L 38 x Cargamanto) x (Mortiño x Diacol Calima) 23 1993 Vilcabamba INIAP-413 ICA 15423 x BAT 1620 Cranberry NO NO n.a. 24 1993 Yunguilla INIAP-414 G 13922 x (G 21721 x G 6474) Red Mottle NO NO n.a. 25 1991 Imbabello INIAP-411 Selection of local variety Cargabello Red Mottle NO NO n.a. 26 1990 Colorado INIAP-472 G 13922 x A 195 Red Kidney NO NO n.a. Source: Mooney (2007) [Who used the following sources: Lepiz (1996); INIAP (1991a, 1991b, 1996a, 1996b, 2004a, 2004b, 2004c, and 2005)]; CIAT (2001a); INIAP (2004a, 2004b, 2004c, 2004d, 2005, 2007a, 2007b, 2009a, 2009b, 2009c, 2010); Peralta et al. (2009), and KII (2010a). Line ID INIAP-417 INIAP-418 INIAP-416 INIAP-412 1 Same pedigree implies that the lines are sisters, i.e. they come from the same parents. n.a. = not available. NOTE: most varieties are developed for the northern region. Yellow varieties are developed for the central region and climbing varieties (denoted by **) are targeted for the southern region. 163 Table A 2.7. Quantity (MT) of seed of improved bean varieties sold or distributed by government programs in 2010. Share (%) of total seed per market class Total bean 1 area (ha) Country (A) Costa Rica 3,944 El Salvador 100,940 Honduras 98,856 Nicaragua 217,518 Ecuador 3,288 Source: KII (2010a, 2010d). Total seed 2 (MT) (B) 286 1,265 1,818 1,717 16 Red (C) 31 100 100 100 n.a. Black (D) 70 n.a. n.a. n.a. n.a. Red mottled (E) n.a. n.a. n.a. n.a. 57 Purple mottled (F) n.a. n.a. n.a. n.a. 22 Yellow (G) n.a. n.a. n.a. n.a. 22 1 In Central America, area refers to red beans only. In Ecuador, area refers to bush-type, red-mottled beans in Carchi and Imbabura only. 2 In Costa Rica, seed data refers to both black and red beans. In El Salvador, Honduras and Nicaragua, seed refers to red beans only (seed of black beans was not distributed/sold). In Ecuador, seed refers to red mottled, purple mottled and yellow beans. n.a. = not applicable 164 Table A 2.8. Estimated average yearly salaries ($) of bean breeding programs' permanent staff for Costa Rica, El Salvador, Honduras (DICTA only) and Nicaragua by education level. 2010. Ph.D. M.Sc. B.S. Technicians < Technician Total 2 3 BTE Salary BTE Salary BTE Salary BTE Salary BTE Salary salaries 1 Country Institution No. (%) ($/yr) No. (%) ($/yr) No. (%) ($/yr) No. (%) ($/yr) No. (%) ($/yr) ($/yr) Costa Rica PITTAFrijol 1 15 4,500 4 44 42,240 3 70 31,500 0 n.a. n.a. 6 80 28,800 107,040 El Salvador CENTA 0 n.a. n.a. 1 100 27,000 2 100 30,000 0 n.a. n.a. 0 n.a. n.a. 57,000 Honduras DICTA 0 n.a. n.a. 0 n.a. n.a. 1 80 12,000 1 50 6,000 0 n.a. n.a. 18,000 Nicaragua INTA 0 n.a. n.a. 3 90 72,900 50 35 183,750 0 n.a. n.a. 0 n.a. n.a. 256,650 Source: The Author, using information provided by KII (2010a) and Mejia (2012, Personal Communication). 1 Zamorano (Honduras) and PRONALEG-GA (Ecuador) were not included because the budget provided by program leaders already includes salaries. PITTA-Frijol includes INTA-CR's and UCR's permanent staff only. 2 BTE = bean time equivalent (i.e. share of staff's time devoted to bean research). 3 Salaries expressed in nominal US$ and only reflect the time devoted to bean-related activities (i.e. monthly salary x 12 x BTE). NOTES: Yearly salaries estimated using the following monthly salaries: Ph.D = $2,750 in Honduras, El Salvador, and Nicaragua and Ph.D. = $2,500 in Costa Rica; M.Sc. = $2,250 in Honduras, El Salvador, and Nicaragua and M.Sc. = $2,000 in Costa Rica; B.S. = $1,250 in Honduras, El Salvador, and Costa Rica and B.S. = $875 (70% of Honduras' salary) in Nicaragua; Technician = $1,000 in all countries listed; and F = 0.000 R-squared = 0.4249 Coefficient p-value Variables Year dummy variables (1=Yes): 2000 234.87 0.367 2001 868.32 ***0.010 2002 34.56 0.903 2003 805.44 **0.028 2004 310.90 0.283 2005 -242.98 0.421 2006 304.78 0.612 2007 156.74 0.645 2008 704.55 *0.072 2009 313.31 0.262 1 Released varieties dummy variables (1=Yes): Carrizalito Amadeus 77 Cedron Cayetana CENTA Pipil Don Cristobal Deorho CENTA C.P.C. Tongibe Cardenal Briyo La Majada Country dummy variables (1=Yes): Costa Rica El Salvador Guatemala Nicaragua Constant 177.13 82.80 -21.85 -2.13 110.34 114.73 203.86 -110.28 161.09 451.34 369.04 354.66 0.583 0.809 0.934 0.995 0.681 0.852 0.451 0.502 0.368 **0.025 0.141 0.134 -944.76 -716.81 -1791.46 -681.14 2155.46 ***0.002 ***0.000 ***0.000 ***0.000 ***0.000 *, **, *** indicates the corresponding coefficients are significant at the 10%, 5%, and 1% levels, respectively. Year 1999, variety Dorado and country Honduras were excluded to avoid the dummy trap. Robust standard errors used to estimate p-values because variances are not equal (Prob > Chi2 = 0.0167). 1 Carrizalito, Amadeus 77 and Deorho were released in more than one country. Source: Programa de Investigaciones en Frijol Metadata, Zamorano, Honduras. 170 Table A 2.14. Linear regression results of factors influencing experimental yields of small red bean varieties released in Honduras. 1999-2009. N = 88 Prob > F = 0.000 R-squared = 0.3986 Coefficient p-value Variables Year dummy variables (1=Yes): 2000 213.83 0.401 2001 877.40 **0.020 2002 -43.75 0.882 2003 876.77 **0.041 2004 266.93 0.377 2005 -158.29 0.619 2006 322.25 0.593 2007 146.81 0.674 2008 694.62 *0.085 2009 303.38 0.281 Released varieties dummy variables (1=Yes): Carrizalito 186.44 0.586 Amadeus 77 91.92 0.786 Cedron 4.15 0.988 Cayetana 8.98 0.979 Don Cristobal 60.88 0.923 Deorho 170.72 0.536 Cardenal 358.86 *0.078 Briyo 369.04 0.154 La Majada 354.66 0.148 Country dummy variables (1=Yes): Costa Rica -1017.94 ***0.005 El Salvador -648.21 ***0.002 Guatemala -1787.99 ***0.000 Nicaragua -640.41 ***0.000 Constant 2128.95 ***0.000 *, **, *** indicates the corresponding coefficients are significant at the 10%, 5%, and 1% levels, respectively. Year 1999, variety Dorado and country Honduras were excluded to avoid the dummy trap. Robust standard errors used to estimate p-values because variances are not equal (Prob > Chi2 = 0.0135). Source: Programa de Investigaciones en Frijol Metadata, Zamorano, Honduras. 171 Table A 2.15. Linear regression results of factors influencing experimental yields of red mottled bean varieties released in Ecuador. 2003-2010. N = 26 Prob > F = 0.001 Adj. R-squared = 0.6984 Coefficient p-value Variables Year dummy variables (1=Yes): 2004 418.75 0.152 2005 1392.00 ***0.000 2006 493.01 *0.070 2007 794.09 ***0.006 2008 230.74 0.364 2009 1262.50 ***0.000 2010 660.47 *0.078 Released varieties dummy variables (1=Yes): INIAP 427 Libertador 75.91 0.758 INIAP 429 Paragachi Andino -58.61 0.733 INIAP 430 Portilla 117.07 0.442 INIAP 481 Rojo del Valle 214.60 0.314 Constant 482.47 **0.037 *, **, *** indicates the corresponding coefficients are significant at the 10%, 5%, and 1% levels, respectively. Year 2003 and variety INIAP 414 Yunguilla were excluded to avoid the dummy trap. Source: INIAP/PRONALEG-GA Metadata, Ecuador. 172 Table A 2.16. Costa Rica: Base scenario Net Present Value (NPV) and Internal Rate of Return (IRR) calculations for improved small red bean varieties. 1991-2015. Area Adoption λ Supply Real Research Kt Change in TS harvested Production rate growth Elasticity Price Costs Net Benefit Year Period (ha) (mt) (%) (%) Type I Type II e ($/mt) Type I Type II ($, real) ($) 1991 -5 17,395 0.86 0 0 158,619 -158,619 1992 -4 15,790 0.86 0 0 153,984 -153,984 1993 -3 14,758 0.85 0 0 149,508 -149,508 1994 -2 14,217 0.85 0 0 145,776 -145,776 1995 -1 14,081 0.85 0 0 141,758 -141,758 1996 Base 8,119 4,063 0.85 0 0 137,692 -137,692 1997 1 11,040 5,524 0.85 0.0049 0.000 0.004 0.7 1,412 0 32,530 134,604 -102,074 1998 2 9,280 4,643 0.85 0.0098 0.000 0.008 0.7 1,212 0 47,026 132,540 -85,513 1999 3 9,063 4,535 0.84 0.0148 0.000 0.012 0.7 1,152 0 65,598 129,676 -64,078 2000 4 7,707 3,856 0.84 0.0197 0.000 0.017 0.7 960 0 62,027 132,136 -70,110 2001 5 5,828 2,916 0.84 0.0247 0.000 0.021 0.7 867 0 52,980 129,454 -76,474 2002 6 5,522 2,763 0.84 0.0298 0.000 0.025 0.7 857 0 59,643 96,711 -37,068 2003 7 5,212 2,608 0.83 0.0348 0.000 0.029 0.7 518 0 39,709 123,918 -84,209 2004 8 4,087 2,045 0.83 0.0399 0.000 0.033 0.7 804 0 55,228 121,616 -66,389 2005 9 4,087 2,045 0.82 0.0450 0.000 0.037 0.7 880 0 67,993 118,514 -50,521 2006 10 3,509 1,756 0.82 0.0501 0.000 0.041 0.7 728 0 53,576 115,666 -62,091 2007 11 3,004 1,503 0.82 0.0552 0.000 0.045 0.7 901 0 62,410 113,295 -50,885 2008 12 2,757 1,379 0.81 0.0604 0.000 0.049 0.7 1,411 0 97,578 109,907 -12,329 2009 13 3,944 1,974 0.81 0.0656 0.000 0.053 0.7 963 0 103,044 86,725 16,319 2010 14 3,460 1,731 0.80 0.0708 0.000 0.057 0.7 1,109 0 111,747 111,747 2011 15 3,460 1,731 0.79 0.0761 0.000 0.061 0.7 1,022 0 110,027 110,027 2012 16 3,460 1,731 0.79 0.0813 0.000 0.065 0.7 1,022 0 116,877 116,877 2013 17 3,460 1,731 0.78 0.0866 0.000 0.068 0.7 1,022 0 123,579 123,579 2014 18 3,460 1,731 0.77 0.0920 0.000 0.072 0.7 1,022 0 130,109 130,109 2015 19 3,460 1,731 0.76 0.0973 0.000 0.075 0.7 1,022 0 136,444 136,444 In constant 1991 US$: NPV = -956,905 In constant 2009 US$: NPV = -2,016,054 IRR = -5% Source: Estimations made by The Author. See Table Notes in Table A 2.21. 173 Table A 2.17. El Salvador: Base scenario Net Present Value (NPV) and Internal Rate of Return (IRR) calculations for improved small red bean varieties. 1991-2015. Area Adoption harvested Production rate Year Period (ha) (mt) (%) 1991 -5 75,097 0.09 1992 -4 76,795 0.12 1993 -3 72,110 0.14 1994 -2 72,042 0.17 1995 -1 58,801 0.21 1996 Base 65,659 51,920 0.25 1997 1 80,495 63,652 0.29 1998 2 75,709 59,866 0.33 1999 3 72,178 57,074 0.37 2000 4 76,659 60,618 0.41 2001 5 82,624 65,335 0.45 2002 6 80,707 63,819 0.48 2003 7 81,362 64,337 0.51 2004 8 84,432 66,764 0.53 2005 9 82,989 65,624 0.55 2006 10 84,758 67,022 0.56 2007 11 91,785 72,579 0.58 2008 12 102,529 81,075 0.59 2009 13 100,940 79,818 0.59 2010 14 92,600 73,224 0.60 2011 15 92,600 73,224 0.60 2012 16 92,600 73,224 0.61 2013 17 92,600 73,224 0.61 2014 18 92,600 73,224 0.61 2015 19 92,600 73,224 0.61 In constant 1991 US$: Kt λ Supply Real growth Elasticity Price (%) Type I Type II e ($/mt) 0.0049 0.0098 0.0148 0.0197 0.0247 0.0298 0.0348 0.0399 0.0450 0.0501 0.0552 0.0604 0.0656 0.0708 0.0761 0.0813 0.0866 0.0920 0.0973 NPV = 0.005 0.001 0.010 0.003 0.014 0.005 0.019 0.007 0.023 0.010 0.026 0.013 0.030 0.017 0.032 0.020 0.034 0.024 0.036 0.028 0.038 0.031 0.039 0.035 0.040 0.038 0.040 0.042 0.041 0.046 0.041 0.049 0.042 0.053 0.042 0.056 0.042 0.060 36,789,922 Source: Estimations made by The Author. See Table Notes in Table A 2.21. 174 0.7 0.7 0.7 0.7 0.7 0.7 0.7 0.7 0.7 0.7 0.7 0.7 0.7 0.7 0.7 0.7 0.7 0.7 0.7 Change in TS Research Costs Net Benefit Type I Type II ($, real) ($) 0 0 196,108 -196,108 0 0 190,377 -190,377 0 0 184,843 -184,843 0 0 180,228 -180,228 0 0 179,485 -179,485 0 0 174,337 -174,337 1,412 427,014 110,126 170,427 366,714 1,212 696,425 207,818 163,864 740,379 1,152 944,756 324,292 160,323 1,108,724 960 1,098,834 431,117 155,110 1,374,841 867 1,303,870 580,788 150,818 1,733,841 857 1,461,830 734,146 109,713 2,086,264 518 996,931 560,563 107,269 1,450,226 804 1,749,924 1,094,180 104,486 2,739,618 880 2,011,168 1,389,285 101,062 3,299,391 728 1,786,251 1,354,845 97,904 3,043,192 901 2,490,189 2,062,075 95,193 4,457,072 1,411 4,483,899 4,032,474 91,673 8,424,700 963 3,083,689 2,997,390 92,000 5,989,079 1,109 3,313,509 3,465,898 6,779,408 1,022 3,096,535 3,471,581 6,568,116 1,022 3,128,228 3,745,398 6,873,627 1,022 3,152,659 4,017,819 7,170,478 1,022 3,171,444 4,289,200 7,460,644 1,022 3,185,860 4,559,919 7,745,779 In constant 2009 US$: NPV = 77,510,816 IRR = 40% Table A 2.18. Honduras: Base scenario Net Present Value (NPV) and Internal Rate of Return (IRR) calculations for improved small red bean varieties. 1991-2015. Area Adoption λ Supply Real Research Kt Change in TS harvested Production rate growth Elasticity Price Costs Net Benefit Year Period (ha) (mt) (%) (%) Type I Type II e ($/mt) Type I Type II ($, real) ($) 1991 -5 104,272 0.20 0 0 258,564 -258,564 1992 -4 67,996 0.23 0 0 211,207 -211,207 1993 -3 79,202 0.25 0 0 197,282 -197,282 1994 -2 111,700 0.27 0 0 171,351 -171,351 1995 -1 64,859 0.29 0 0 125,563 -125,563 1996 Base 79,043 57,718 0.31 0 0 111,262 -111,262 1997 1 78,850 57,577 0.33 0.0056 0.002 0.002 0.7 1,412 180,694 141,196 122,252 199,638 1998 2 74,881 54,679 0.35 0.0112 0.004 0.004 0.7 1,212 285,269 245,432 129,429 401,272 1999 3 106,064 77,449 0.36 0.0169 0.006 0.006 0.7 1,152 555,813 524,885 135,488 945,209 2000 4 114,671 83,734 0.38 0.0226 0.008 0.008 0.7 960 641,461 662,847 164,462 1,139,846 2001 5 72,568 52,990 0.39 0.0283 0.010 0.011 0.7 867 439,057 494,904 168,242 765,718 2002 6 132,661 96,870 0.41 0.0341 0.011 0.013 0.7 857 911,342 1,117,124 155,937 1,872,529 2003 7 99,005 72,295 0.42 0.0399 0.012 0.016 0.7 518 457,807 608,424 160,481 905,751 2004 8 98,347 71,814 0.43 0.0457 0.013 0.019 0.7 804 768,922 1,104,652 156,797 1,716,777 2005 9 111,916 81,722 0.43 0.0515 0.014 0.022 0.7 880 1,026,258 1,589,168 169,698 2,445,728 2006 10 121,600 88,794 0.44 0.0574 0.015 0.025 0.7 728 975,038 1,622,924 164,844 2,433,117 2007 11 133,000 97,118 0.45 0.0634 0.016 0.028 0.7 901 1,382,603 2,467,031 212,451 3,637,183 2008 12 133,000 97,118 0.45 0.0693 0.016 0.031 0.7 1,411 2,246,392 4,285,946 205,016 6,327,322 2009 13 98,856 72,186 0.46 0.0753 0.017 0.034 0.7 963 1,176,031 2,393,291 155,897 3,413,425 2010 14 119,674 87,388 0.46 0.0813 0.017 0.037 0.7 1,109 1,681,314 3,640,993 5,322,306 2011 15 119,674 87,388 0.46 0.0874 0.018 0.040 0.7 1,022 1,583,663 3,641,281 5,224,944 2012 16 119,674 87,388 0.47 0.0935 0.018 0.043 0.7 1,022 1,611,759 3,926,310 5,538,070 2013 17 119,674 87,388 0.47 0.0996 0.018 0.046 0.7 1,022 1,635,459 4,212,450 5,847,909 2014 18 119,674 87,388 0.47 0.1057 0.018 0.050 0.7 1,022 1,655,416 4,499,600 6,155,016 2015 19 119,674 87,388 0.47 0.1119 0.019 0.053 0.7 1,022 1,672,195 4,787,722 6,459,917 In constant 1991 US$: NPV = 27,648,128 In constant 2009 US$: NPV = 58,250,437 IRR = 34% Source: Estimations made by The Author. See Table Notes in Table A 2.21. 175 Table A 2.19. Nicaragua: Base scenario Net Present Value (NPV) and Internal Rate of Return (IRR) calculations for improved small red bean varieties. 1991-2015. Area Adoption harvested Production rate Year Period (ha) (mt) (%) 1991 -5 98,368 0.09 1992 -4 88,139 0.12 1993 -3 100,300 0.15 1994 -2 98,786 0.19 1995 -1 120,677 0.24 1996 Base 104,509 67,653 0.30 1997 1 117,694 76,187 0.36 1998 2 165,025 106,827 0.43 1999 3 180,351 116,748 0.49 2000 4 194,773 126,084 0.55 2001 5 201,338 130,333 0.61 2002 6 218,311 141,321 0.65 2003 7 252,934 163,733 0.69 2004 8 202,692 131,210 0.73 2005 9 236,473 153,078 0.75 2006 10 199,953 129,437 0.78 2007 11 202,613 131,159 0.79 2008 12 209,269 135,468 0.80 2009 13 217,518 140,808 0.81 2010 14 213,165 137,990 0.82 2011 15 213,165 137,990 0.83 2012 16 213,165 137,990 0.83 2013 17 213,165 137,990 0.83 2014 18 213,165 137,990 0.83 2015 19 213,165 137,990 0.84 In constant 1991 US$: Kt λ Supply Real growth Elasticity Price (%) Type I Type II e ($/mt) 0.0049 0.0098 0.0148 0.0197 0.0247 0.0298 0.0348 0.0399 0.0450 0.0501 0.0552 0.0604 0.0656 0.0708 0.0761 0.0813 0.0866 0.0920 0.0973 NPV = 0.007 0.001 0.014 0.004 0.022 0.006 0.029 0.010 0.035 0.014 0.041 0.018 0.045 0.023 0.049 0.028 0.052 0.033 0.055 0.038 0.057 0.043 0.058 0.048 0.059 0.053 0.060 0.058 0.060 0.062 0.061 0.067 0.061 0.072 0.061 0.077 0.062 0.081 101,574,886 Source: Estimations made by The Author. See Table Notes in Table A 2.21. 176 0.7 0.7 0.7 0.7 0.7 0.7 0.7 0.7 0.7 0.7 0.7 0.7 0.7 0.7 0.7 0.7 0.7 0.7 0.7 Change in TS Research Costs Net Benefit Type I Type II ($, real) ($) 0 0 490,362 -490,362 0 0 476,032 -476,032 0 0 462,196 -462,196 0 0 450,657 -450,657 0 0 438,237 -438,237 0 0 425,668 -425,668 1,412 762,195 158,191 416,120 504,266 1,212 1,879,670 460,346 409,738 1,930,278 1,152 2,954,150 847,761 400,884 3,401,027 960 3,517,663 1,173,653 387,847 4,303,468 867 4,016,351 1,544,826 377,116 5,184,060 857 4,999,838 2,197,405 323,952 6,873,290 518 3,910,834 1,946,442 317,584 5,539,692 804 5,282,203 2,951,336 309,346 7,924,193 880 7,173,491 4,462,568 299,208 11,336,851 728 5,249,426 3,608,318 289,858 8,567,886 901 6,814,754 5,140,042 281,831 11,672,965 1,411 11,294,234 9,289,222 271,410 20,312,046 963 8,166,618 7,283,424 271,650 15,178,392 1,109 9,339,858 8,987,250 18,327,108 1,022 8,700,616 8,992,551 17,693,167 1,022 8,765,976 9,692,501 18,458,477 1,022 8,814,450 10,388,869 19,203,319 1,022 8,850,324 11,083,067 19,933,391 1,022 8,876,828 11,776,409 20,653,237 In constant 2009 US$: NPV = 214,002,964 IRR = 42% Table A 2.20. Ecuador: Base scenario Net Present Value (NPV) and Internal Rate of Return (IRR) calculations for improved red mottled bean varieties in northern Ecuador. 1991-2015. Kt Change in TS Area Adoption λ Supply Real Research harvested Production rate growth Elasticity Price Costs Net Benefit Year Period (ha) (mt) (%) (%) Type I Type II e ($/mt) Type I Type II ($, real) ($) 1991 -5 6,983 0.03 0 0 22,827 -22,827 1992 -4 6,970 0.04 0 0 21,763 -21,763 1993 -3 7,185 0.05 0 0 28,098 -28,098 1994 -2 8,599 0.07 0 0 37,864 -37,864 1995 -1 7,759 0.09 0 0 31,751 -31,751 1996 Base 8,489 7,287 0.12 0 0 29,341 -29,341 1997 1 8,534 7,326 0.15 0.0168 0.006 0.002 0.7 600 25,598 8,868 28,903 5,562 1998 2 7,533 6,467 0.19 0.0339 0.012 0.005 0.7 600 48,464 19,965 27,246 41,183 1999 3 7,762 6,663 0.23 0.0513 0.020 0.010 0.7 600 79,158 38,566 23,406 94,318 2000 4 6,427 5,517 0.27 0.0689 0.027 0.016 0.7 600 90,857 52,038 7,480 135,415 2001 5 7,207 6,187 0.31 0.0869 0.035 0.023 0.7 600 130,154 87,051 12,295 204,910 2002 6 8,333 7,153 0.35 0.1051 0.042 0.032 0.7 600 181,461 140,667 13,626 308,501 2003 7 7,996 6,864 0.38 0.1237 0.048 0.043 0.7 600 201,042 179,159 44,755 335,446 2004 8 7,036 6,040 0.41 0.1426 0.053 0.054 0.7 659 216,854 220,264 57,762 379,355 2005 9 8,455 7,258 0.44 0.1618 0.058 0.066 0.7 740 318,152 365,169 57,749 625,572 2006 10 7,353 6,312 0.46 0.1813 0.062 0.079 0.7 753 299,596 385,360 55,944 629,012 2007 11 6,607 5,671 0.47 0.2011 0.065 0.092 0.7 626 234,840 335,885 61,120 509,605 2008 12 6,106 5,241 0.48 0.2213 0.067 0.104 0.7 566 203,029 320,610 131,810 391,828 2009 13 6,085 5,223 0.49 0.2418 0.069 0.117 0.7 839 308,033 533,634 57,375 784,292 2010 14 6,921 5,941 0.50 0.2627 0.070 0.129 0.7 1,039 442,028 835,322 1,277,350 2011 15 6,921 5,941 0.51 0.2839 0.071 0.142 0.7 765 329,964 676,782 1,006,747 2012 16 6,921 5,941 0.51 0.3055 0.072 0.154 0.7 765 333,394 738,908 1,072,302 2013 17 6,921 5,941 0.51 0.3274 0.072 0.167 0.7 765 335,929 801,361 1,137,290 2014 18 6,921 5,941 0.51 0.3497 0.073 0.179 0.7 765 337,796 864,319 1,202,115 2015 19 6,921 5,941 0.52 0.3724 0.073 0.191 0.7 765 339,168 927,959 1,267,126 In constant 1991 US$: NPV = 5,183,118 In constant 2009 US$: NPV = 10,920,047 IRR = 37% Source: Estimations made by The Author. See Table Notes in Table A 2.21. 177 Table A 2.21. Notes for Table A 2.16 through Table A 2.20. Country Notes Costa Rica Area harvested is 25% of total area harvested to reflect only red bean production (i.e. excludes other market classes). Yield for 1996 estimated as the average yields of 1994-1998, using FAOSTAT data. λ=0.0049 (Type II gains); Type I gains=11.5% (from Mather et al. 2003). Price is the average price for Central American countries. El Salvador Area harvested is 97% of total area harvested to reflect only red bean production (i.e. excludes other market classes). Yield for 1996 estimated as the average yields of 1994-1998, using FAOSTAT data. λ=0.0049 (Type II gains); Type I gains=11.5% (from Mather et al. 2003). Price is the average price for Central American countries. Honduras Area harvested is 95% of total area harvested to reflect only red bean production (i.e. excludes other market classes). Yield for 1996 estimated as the average yields of 1994-1998, using FAOSTAT data. λ=0.0056 (Type II gains); Type I gains=11.5% (from Mather et al. 2003). Price is the average price for Central American countries. Nicaragua Area harvested is 87.5% of total area harvested to reflect only red bean production (i.e. excludes other market classes). Yield for 1996 estimated as the average yields of 1994-1998, using FAOSTAT data. λ=0.0049 (Type II gains); Type I gains=11.5% (from Mather et al. 2003). Price is the average price for Central American countries. Ecuador Area harvested is 13.5% of total of dry bean area harvested in the country to reflect only red mottled (i.e. excludes other market classes) bush-bean production in the provinces of Carchi and Imbabura. Yields in northern Ecuador are 35% higher (estimated from ESPAC data) than country-level yields. Thus, FAOSTAT yields data were multiplied by 1.35 to reflect yields in northern Ecuador and yield for 1996 estimated as the average of 19941998. λ=0.0168 (Type II gains); Type I gains=18.4% (from Mooney 2007). For all countries For 2010-2015, area harvested assumed as the average of the previous five years (i.e. 2005-2009). For 2011-2015, price assumed as the average of the previous five years (i.e. 2006-2010). Discount rate=4%. Production estimated by multiplying area harvested in each year times the base year (i.e. 1996) yields. 178 Table A 3.1. Additional demographic characteristics of farm households (HH). Central Highlands of Angola, 2009. Potato Bean Onion Non-sellers Sellers Non-sellers Sellers Non-sellers Sellers Mean S.E. Demographics Marital status of head (%): Married Single Widow Separated Household size No. males >17 yrs No. females >17 yrs No. children <5 yrs No. boys 5-17 yrs No. girls 5-17 yrs 1 Mean S.E. MT 66 2 20 12 4.9 0.8 1.1 1.0 0.9 1.1 0.195 0.016 0.106 0.082 0.281 0.176 0.026 0.118 0.077 0.078 87 1 9 4 5.5 1.0 1.1 1.5 1.1 0.8 0.113 0.012 0.072 0.034 0.109 0.115 0.035 0.088 0.135 0.181 69 0.137 54 23 0.165 11 Mean S.E. *** 1 Mean S.E. MT Mean S.E. 1 Mean S.E. MT 79 1 13 7 5.4 1.0 1.1 1.4 1.1 0.9 0.184 0.012 0.111 0.063 0.478 0.254 0.017 0.189 0.104 0.110 79 2 13 6 5.6 1.0 1.1 1.3 1.2 1.0 0.156 0.019 0.094 0.046 0.334 0.119 0.027 0.124 0.196 0.099 79 2 12 7 5.1 0.9 1.1 0.9 1.3 1.0 0.183 0.016 0.118 0.059 0.253 0.168 0.011 0.173 0.328 0.100 88 1 9 2 5.6 1.2 1.1 1.5 1.0 0.9 0.094 0.014 0.061 0.022 0.126 0.056 *** 0.017 0.118 *** 0.035 0.090 0.135 74 0.164 50 0.230 74 0.174 68 0.127 0.068 6 0.052 26 0.080 26 0.191 14 0.092 0.7 0.169 0.3 0.044 46 0.090 0.5 0.051 ** 0.2 0.020 34 0.022 ** 0.4 0.2 24 0.107 0.035 0.036 0.6 0.073 ** 0.3 0.033 ** 42 0.044 *** 0.5 0.4 39 0.053 0.054 0.036 0.7 0.060 ** 0.2 0.051 * 38 0.049 88 183 101 223 60 *** *** ** ** *** ** 2 If HH member is in FO : Received assistance about 3 production (% yes) Received assistance about 3 marketing (% yes) Family members >17 literate: No. males >17 literate No. females >17 literate Share of all adults (%) Number of observations * 4 179 134 Table A 3.1 (cont’d). 1 2 MT = test of difference between means: * significant at 10%; ** significant at 5%; *** significant at 1%. FO = Farmer 3 4 organization. Asked to member of FO and refers to assistance over past 12 months. Literacy refers to people who can read and write. Source: ProRenda survey, Angola, 2009. Estimates weighted to reflect population. 180 Table A 3.2. Major sources of crop and non-crop household incomes by market participation. Central Highlands of Angola, 2009. Potato Bean Onion Non-sellers Sellers Non-sellers Sellers Non-sellers Sellers Households' (HH) heads 1 declaring Crops below as their major source of crop income (%): Potatoes Corn Beans Onions Other crops Activities below as their major source of non-crop income (%): Commerce Services Farm labor Gifts, retirements, transfers, remittances Handcrafts, processed products None Other activities Number of observations 1 Mean S.E. 2 Mean S.E. MT Mean S.E. 2 Mean S.E. MT Mean S.E. 2 Mean S.E. MT 18 30 33 5 14 0.046 0.046 0.018 0.023 0.026 47 14 19 7 12 0.067 *** 0.039 *** 0.019 ** 0.030 0.035 15 23 29 8 24 0.065 0.058 0.034 0.035 0.083 14 14 63 3 6 0.015 0.067 * 0.086 *** 0.010 * 0.012 *** 11 11 58 9 11 0.037 0.027 0.103 0.066 0.028 31 13 29 15 12 28 3 27 0.120 0.028 0.084 18 23 38 0.060 * 0.079 *** 0.017 * 25 20 21 0.043 0.090 0.114 24 13 32 0.019 0.054 0.039 19 11 22 0.012 0.044 0.087 32 0.040 17 0.068 29 0.021 8 4 28 1 76 0.034 0.025 0.119 0.007 3 10 6 3 186 0.007 ** 0.013 0.027 *** 0.015 2 12 14 7 85 0.017 0.030 0.026 0.036 7 5 15 4 225 0.008 0.022 0.017 0.017 20 8 10 11 47 0.105 0.024 0.104 0.063 2 7 12 1 125 * * * Column sum within sources may not add to 100% due to rounding. 2 MT = test of difference between means: * significant at 10%; ** significant at 5%; *** significant at 1%. Source: ProRenda survey, Angola, 2009. Estimates weighted to reflect population. 181 0.026 *** 0.016 0.026 *** 0.027 0.013 0.003 *** 0.015 0.029 0.010 *** Table A 3.3. Scoring factors, summary statistics, and per tercile means for asset indicators entering the computation of the first principal component (asset ownership). Percentage of households owning the asset Total sample Scoring Asset indicators Factors Own plow 0.326 Own cart 0.121 Own backpack sprayer 0.224 Own motorcycle 0.434 Own bicycle 0.297 Own cell phone 0.299 Have water storage at home 0.261 Have latrine in the house 0.094 Have lusalite or zinc roof 0.393 Own radio 0.347 Own television 0.335 Std. Scor. Fac. Dev. / Std. Dev. 0.292 1.11 0.065 1.87 0.136 1.65 0.254 1.71 0.360 0.82 0.235 1.27 0.398 0.66 0.409 0.23 0.499 0.79 0.485 0.71 0.169 1.98 Mean 0.094 0.004 0.019 0.069 0.153 0.059 0.197 0.789 0.467 0.377 0.029 Economic Status Index Number of observations 0.000 1.522 478 Lowest Middle Highest tercile tercile tercile 0% 4% 23% 0% 0% 1% 0% 0% 6% 0% 0% 23% 0% 5% 58% 0% 0% 21% 5% 19% 42% 71% 93% 94% 0% 46% 79% 2% 55% 62% 0% 0% 6% Mean by tercile -1.259 -0.277 1.719 164 184 130 Notes: Three of the 14 indicators were dropped because they had zero variance. Scoring Factor is the "weight" assigned to each indicator or eigenvector (normalized by its mean and standard deviation) in the linear combination of the variables that constitute the first principal component. The percentage of the covariance explained by the first principal component is 21.06%. The first eigenvalue is 2.32. Data provided in the last three columns were estimated with weights to reflect population (except number of observations). Source: ProRenda survey, Angola, 2009. 182 Table A 3.4. Means per market participation (seller) for asset indicators entering the computation of the first principal component (asset ownership). Percentage of households owning the asset Potato seller Bean seller Onion seller Asset indicators No Yes MT No Yes MT No Yes MT Own plow 16% 9% -4% 9% -7% 11% -Own cart 0% 1% -0% 0% -2% 0% -Own backpack sprayer 0% 4% -0% 1% -0% 4% -Own motorcycle 3% 10% -8% 9% -21% 7% -Own bicycle 20% 27% -17% 19% -22% 25% -Own cell phone 7% 10% -5% 11% -10% 8% -Have water storage at home 19% 21% -23% 19% -44% 20% -Have latrine in the house 81% 90% -86% 89% -80% 91% -Have lusalite or zinc roof 37% 56% -38% 43% -63% 33% -Own radio 35% 54% -48% 40% -47% 63% -Own television 0% 2% -0% 3% -2% 1% -Mean by group Economic Status Index -0.118 0.432 *** -0.086 0.121 0.651 0.224 * Number of observations 97 175 115 232 57 130 Notes: Three of the 14 indicators were dropped because they had zero variance. The percentage of the covariance explained by the first principal component is 21.06%. The first eigenvalue is 2.32. Estimates weighted to reflect population (except last row). MT = Bonferroni test of difference between means: * significant at 10%; ** significant at 5%; *** significant at 1%; -not tested. Source: ProRenda survey, Angola, 2009. 183 Table A 3.5. Potato sellers: Average receipts, costs, margins and percentage of production sold, per economic status index and gender of household head (HHH). 1 Detail Receipts (Kw) Price per kg sold (Kw/kg) 1 Economic Status Index by tercile Gender of HHH Lowest Middle Highest Male Female (mean values) 7,705 a 12,630 ab 17,795 b 15,761 7,287 *** 79 -86 -89 -85 71 -- Total Costs (Kw) Production costs per kg produced (Kw/kg) Marketing costs per kg sold (Kw/kg) 5,544 8,864 Margins (Kw) 2,160 3,766 Share sold (% of total production) 77% 12,796 83% Total 14,087 82 11,379 5,388 ** 10,195 46.5 -- 43.1 -- 61.3 -- 51.9 43.4 -- 50.2 1.8 -- 2.7 -- 2.9 -- 2.4 2.0 -- 2.4 4,999 4,382 1,900 83% 84% 77% 3,892 * 82% Number of observations 39 68 67 127 74 201 NOTES: Kw = Kwanzas. Costs include purchased inputs, hired labor, and reported marketing costs. Variables are at the household level. Number of observations in Economic Status Index smaller than in last column because of missing values in this variable. 1 Bonferroni test of difference between means: for Economic Status Index, different letters imply differences are significant at 10%; for Gender of HHH, * significant at 10%; ** significant at 5%; *** significant at 1%. -- = mean differences not tested. Source: ProRenda survey, Angola, 2009. Estimates weighted to reflect population. 184 Table A 3.6. Bean sellers: Average receipts, costs, margins and percentage of production sold, per economic status index and gender of household head (HHH). 1 Detail Receipts (Kw) Price per kg sold (Kw/kg) Total Costs (Kw) Production costs per kg produced (Kw/kg) Marketing costs per kg sold (Kw/kg) 1 Economic Status Index by tercile Gender of HHH Lowest Middle Highest Male Female (mean values) 3,193 a 4,792 a 11,625 b 8,447 3,039 *** 74 -65 -74 -72 69 -947 987 1,867 1,443 887 Total 6,785 71 1,272 8.8 -- 9.6 -- 9.3 -- 8.2 11.5 -- 9.2 5.0 -- 3.0 -- 0.8 -- 3.2 2.0 -- 2.8 Margins (Kw) 2,247 a 3,805 a 9,758 b 7,004 2,152 *** 5,513 Share sold (% of total production) 60% 64% 57% ** 62% 66% 61% Number of observations 87 79 64 142 113 255 NOTES: Kw = Kwanzas. Costs include purchased inputs, hired labor, and reported marketing costs. Variables are at the household level. Number of observations in Economic Status Index smaller than in last column because of missing values in this variable. 1 Bonferroni test of difference between means: for Economic Status Index, different letters imply differences are significant at 10%; for Gender of HHH, * significant at 10%; ** significant at 5%; *** significant at 1%. -- = mean differences not tested. Source: ProRenda survey, Angola, 2009. Estimates weighted to reflect population. 185 Table A 3.7. Onion sellers: Average receipts, costs, margins and percentage of production sold, per economic status index and gender of household head (HHH). 1 Detail Receipts (Kw) Price per kg sold (Kw/kg) 1 Economic Status Index by tercile Gender of HHH Lowest Middle Highest Male Female (mean values) 1,375 5,135 5,294 5,300 1,883 ** 117 -86 -107 -98 114 -- Total Costs (Kw) Production costs per kg produced (Kw/kg) Marketing costs per kg sold (Kw/kg) 1,671 2,386 Margins (Kw) -296 2,750 Share sold (% of total production) 74% 80% 5,873 Total 4,708 101 4,284 1,096 ** 3,732 95.7 -- 42.7 -- 114.3 -- 90 43 -- 82 2.9 -- 3.6 -- 2.4 -- 2.9 3.8 -- 3.0 -579 1,016 787 976 77% 80% 77% 79% Number of observations 34 50 45 89 52 141 NOTES: Kw = Kwanzas. Costs include purchased inputs, hired labor, and reported marketing costs. Variables are at the household level. Number of observations in Economic Status Index smaller than in last column because of missing values in this variable. 1 Bonferroni test of difference between means: for Economic Status Index, different letters imply differences are significant at 10%; for Gender of HHH, * significant at 10%; ** significant at 5%; *** significant at 1%. -- = mean differences not tested. Source: ProRenda survey, Angola, 2009. Estimates weighted to reflect population. 186 Table A 3.8. Descriptive statistics of factors influencing potato, bean and onion production. Central Highlands of Angola, 2009. Potato Bean Onion N = 281 N = 380 N = 162 Mean Std. Err. Mean Std. Err. Mean Std. Err. Variables Dependent Variable Quantity produced (kg) 169.49 29.120 89.05 17.083 46.28 7.686 Independent Variables Household (HH) Characteristics Age of HH head (Years) 39.71 0.819 43.71 0.396 43.84 0.919 Gender of HH head (% Male) 74.03 0.214 68.75 0.263 76.51 0.208 1 Dependency ratio HH member is in farmer organization (% yes) 0.59 9.37 0.011 0.010 0.54 6.39 0.012 0.034 0.059 0.092 0.048 0.88 0.51 46.58 0.096 0.094 0.077 1.05 0.41 43.50 0.104 0.049 0.040 13.53 4.44 0.027 0.012 11.78 0.75 0.031 0.005 9.40 2.81 0.010 0.021 26.45 17.33 16.62 1.60 15.97 17.11 12.45 187 0.56 2.18 0.80 0.38 56.97 2 No. adults (>17 yr) literate No. of Tropical Livestock Units Home has zinc roof (% yes) Productive Assets Ownership (% yes) Owns a plow Owns a backpack sprayer Public Assets and Quasi-fixed Factors (% yes) IDA office in village Public market in village HH in Caala Municipality HH in Ekunha Municipality HH in Bailundo Municipality HH in Londuimbali Municipality HH in Katchiungo Municipality 0.010 0.046 0.054 0.045 0.021 0.013 0.038 0.038 0.022 19.25 18.97 7.06 3.35 50.19 29.70 3.79 0.039 0.032 0.013 0.023 0.040 0.034 0.004 19.03 12.79 10.45 1.27 34.93 24.06 6.09 0.019 0.043 0.040 0.011 0.049 0.051 0.010 Table A 3.8 (cont’d). Potato N = 281 Mean Std. Err. 2.83 0.018 33.17 0.031 0.25 0.002 Variables HH in Tchicalachuluanga Municipality HH in Chiguar Municipality HH in Babaera Municipality Production-related variables Total seed used (kg) 41.16 Planted in rainfed plot (% yes) 44.03 Planted intercropped (% yes) n.a. Planted local variety (% yes) 73.75 Used fertilizer (% yes) 65.62 Used pesticides (% yes) 10.66 Reported production costs (Kw/kg) 62.54 HH reported lower harvest (% yes) 66.75 Notes: n.a. = not applicable since only beans may be planted intercropped. 1 6.562 0.014 0.046 0.028 0.054 9.583 0.026 Dependency ratio estimated by dividing No. members <17 yr by household size. 2 Literacy refers to adults who can read and write. Source: ProRenda survey, Angola, 2009. Estimates weighted to reflect population. 188 Bean N = 380 Mean Std. Err. 2.36 0.013 3.32 0.022 0.23 0.002 Onion N = 162 Mean Std. Err. 3.52 0.019 19.66 0.044 0.02 0.000 23.59 94.20 59.97 98.10 2.55 0.11 10.93 57.92 0.03 56.57 n.a. 92.90 48.57 3.57 75.66 54.65 3.273 0.016 0.017 0.004 0.006 0.001 2.109 0.020 0.013 0.040 0.031 0.072 0.014 7.912 0.035 Table A 3.9. Unconditional average partial effects of factors influencing potato sales. Central Highlands of Angola, 2009. Quantity sold (kg) Independent variables: the coefficients displayed are the unconditional average partial effects (APEs). Coefficient p-value Age of HH head (Years) -0.309 0.124 Gender of HH head (1=Male) 23.994 0.353 Dependency ratio 20.491 0.702 HH member is in farmer organization (1=Yes) 19.389 *0.091 No. adults (>17 yr) literate 1.822 0.666 No. of Tropical Livestock Units 4.275 0.571 Asset Index -6.980 *0.056 Owns motorcycle (1=Yes) 12.298 0.717 Owns bicycle (1=Yes) 28.198 ***0.000 IDA office in village (1=Yes) 43.242 **0.041 Public market in village (1=Yes) -23.283 0.111 HH in Caala Municipality (1=Yes) 49.531 0.175 HH in Ekunha Municipality (1=Yes) 65.411 **0.042 HH in Bailundo Municipality (1=Yes) -86.079 *0.066 HH in Katchiungo Municipality (1=Yes) 36.192 **0.018 HH in Tchicalachuluanga Municipality (1=Yes) -6.738 0.769 HH in Chiguar Municipality (1=Yes) 4.081 0.805 HH in Babaera Municipality (1=Yes) -29.095 0.378 Distance from village to sede (km) -0.773 0.240 Road between village and sede in poor condition (1=Yes) -59.345 ***0.000 Seller sought price information prior to sales (1=Yes) -2.937 0.761 Reported marketing costs (Kw/kg) -0.454 0.683 Total potato production (kg) 0.577 ***0.000 Notes: *, **, *** indicates the corresponding coefficients are significant at the 10%, 5%, and 1% levels, respectively. Coefficients and p-values obtained via bootstrapping at 500 repetitions. Dependency ratio estimated by dividing No. members <17 yr by household (HH) size. Literacy refers to adults who can read and write. n.a. = not applicable because variable was not included in the regression. Source: ProRenda survey, Angola, 2009. Estimates weighted to reflect population. 189 Table A 3.10. Unconditional average partial effects of factors influencing bean sales. Central Highlands of Angola, 2009. Quantity sold (kg) Independent variables: the coefficients displayed are the unconditional average partial effects (APEs). Coefficient p-value Age of HH head (Years) -0.400 **0.016 Gender of HH head (1=Male) 5.296 0.502 Dependency ratio -10.200 0.188 HH member is in farmer organization (1=Yes) -12.154 0.127 No. adults (>17 yr) literate 5.717 *0.055 No. of Tropical Livestock Units 2.383 0.170 Asset Index 0.971 0.668 Owns motorcycle (1=Yes) -35.318 ***0.008 Owns bicycle (1=Yes) 14.703 ***0.000 IDA office in village (1=Yes) 1.023 0.958 Public market in village (1=Yes) -6.534 0.760 HH in Caala Municipality (1=Yes) -19.680 ***0.000 HH in Ekunha Municipality (1=Yes) 10.260 0.216 HH in Bailundo Municipality (1=Yes) 12.255 ***0.000 HH in Katchiungo Municipality (1=Yes) -16.531 ***0.000 HH in Tchicalachuluanga Municipality (1=Yes) -13.475 0.216 HH in Chiguar Municipality (1=Yes) -2.358 0.910 HH in Babaera Municipality (1=Yes) -32.637 ***0.001 Distance from village to sede (km) 0.538 ***0.003 Road between village and sede in poor condition (1=Yes) 2.300 0.709 Seller sought price information prior to sales (1=Yes) -9.474 ***0.002 Reported marketing costs (Kw/kg) -0.331 0.313 Total bean production (kg) 0.483 ***0.000 Notes: *, **, *** indicates the corresponding coefficients are significant at the 10%, 5%, and 1% levels, respectively. Coefficients and p-values obtained via bootstrapping at 500 repetitions. Dependency ratio estimated by dividing No. members <17 yr by household (HH) size. Literacy refers to adults who can read and write. Source: ProRenda survey, Angola, 2009. Estimates weighted to reflect population. 190 Table A 3.11. Unconditional average partial effects of factors influencing onion sales. Central Highlands of Angola, 2009. Quantity sold (kg) Independent variables: the coefficients displayed are the unconditional average partial effects (APEs). Coefficient p-value Age of HH head (Years) 0.112 0.550 Gender of HH head (1=Male) 3.530 0.211 Dependency ratio 5.954 0.147 HH member is in farmer organization (1=Yes) 0.850 0.851 No. adults (>17 yr) literate 2.795 ***0.000 No. of Tropical Livestock Units -0.201 0.961 Asset Index -2.997 ***0.000 Owns motorcycle (1=Yes) -2.391 0.761 Owns bicycle (1=Yes) 1.157 0.788 IDA office in village (1=Yes) 2.316 0.679 Public market in village (1=Yes) 3.847 *0.065 HH in Caala Municipality (1=Yes) 0.276 0.963 HH in Ekunha Municipality (1=Yes) -0.561 0.852 HH in Bailundo Municipality (1=Yes) -1.466 0.702 HH in Katchiungo Municipality (1=Yes) 7.706 0.105 HH in Tchicalachuluanga Municipality (1=Yes) -2.762 0.410 HH in Chiguar Municipality (1=Yes) 3.635 0.485 Distance from village to sede (km) -0.681 **0.019 Road between village and sede in poor condition (1=Yes) 10.130 ***0.000 Seller sought price information prior to sales (1=Yes) 1.997 0.267 Reported marketing costs (Kw/kg) -0.147 *0.078 Total onion production (kg) 0.675 ***0.000 Residual from onion production equation -0.082 **0.028 Notes: *, **, *** indicates the corresponding coefficients are significant at the 10%, 5%, and 1% levels, respectively. Coefficients and p-values obtained via bootstrapping at 500 repetitions. Dependency ratio estimated by dividing No. members <17 yr by household (HH) size. Literacy refers to adults who can read and write. Source: ProRenda survey, Angola, 2009. Estimates weighted to reflect population. 191 Figure A 3.1. Cumulative distribution of asset index by potato growers and nongrowers. Central Highlands of Angola, 2009. 192 Figure A 3.2. Cumulative distribution of asset index by onion growers and non-growers. Central Highlands of Angola, 2009. 193 Figure A 3.3. Cumulative distribution of asset index by bean growers and non-growers. Central Highlands of Angola, 2009. 194 REFERENCES 195 REFERENCES AEC-865. 2008. Agricultural Benefit-Cost Analysis. Class notes. Michigan State University. Spring 2008. Alston, J. M., C. Chan-Kang, M. Marra, P. G. Pardey, and T. J. Wyatt. 2000a. A Meta-Analysis of Rates of Return to Agricultural R&D: Ex Pede Herculem? IFPRI Research Report 113, Washington, D.C.: International Food Policy Research Institute. 163 p. Alston, J. M., G. W. Norton, and P. G. Pardey. 1998. Science Under Scarcity: Principles and Practice for Agricultural Research Evaluation and Priority Setting. CAB International. Cambridge University Press. 585 p. Ashby, J., A. Braun, T. Gracia, M. Guerrero, L. Hernández, C. Quirós, and J. Roa. 2000. Investing in Farmers as Researchers. Experience with Local Agricultural Research Committees in Latin America. Cali, Colombia, CIAT publication No. 318. 199 p. ASOCIAL Yorito-Sulaco-Victoria, PIF/EAP, FIPAH, and University of Guelph. 2004. Macuzalito: Variedad de frijol rojo desarrollada por y para agricultores de las zonas altas de Yorito mediante métodos de fitomejoramiento participativo. Varietal Information Booklet. ASOCIALAYO, PIF/EAP, PRR; and FIPAH. 2005a. Nueva Esperanza 01: Variedad de frijol rojo desarrollada con agricultores de la ASOCIAL del Lago Yojoa utilizando métodos participativos. Trifolium. __________. 2005b. Palmichal 1: Variedad de frijol rojo desarrollada con agricultores de la ASOCIAL del Lago Yojoa utilizando métodos participativos. Trifolium. Barrett, C. B. 2008. Smallholder market participation: Concepts and evidence from eastern and southern Africa. Food Policy, Vol. 33 (4): 299-317. Bazelon, C. and Smetters, K. 2001. Discounting in the long term. Loyola of Los Angeles Law Review, Vol 35: 277-291. Bellemare, M. and Barrett, C. B. 2006. An Ordered Tobit Model of Market Participation: Evidence from Kenya and Ethiopia. American Journal of Agricultural Economics, Vol. 88 (2): 324-337. Belli, P., J. Anderson, H. Barnum, J. Dixon, and J. Tan. 2001. Economic Analysis of Investment Operations: Analytical Tools and Practical Applications. World Bank Institute Development Series. Washington D.C. 264 p. Benfica, R.; Tschirley, D.; and Boughton, D. 2006. Interlinked Transactions in Cash Cropping Economies: The Determinants of Farmer Participation and Performance in the Zambezi River Valley of Mozambique. Contributed paper presented at the 26th International 196 Association of Agricultural Economists Conference, Gold Coast, Australia, August 1218, 2006. 15 p. Boughton, D.; Mather, D.; Barrett, C. B.; Benfica, R.; Abdula, D.; Tschirley, D.; and Cunguara, B. 2007. Market Participation by Rural Households in a Low-Income Country: An Asset-Based Approach Applied to Mozambique. Faith and Economics, Vol. 50: 64-101. Burke, W. J. 2009. Fitting and interpreting Cragg’s tobit alternative using Stata. The Stata Journal, Vol. 9 (4): 584-592. Byerlee, D. and G. Traxler. 1995. National and International Wheat Improvement Research in the Post-Green Revolution Period: Evolution and Impacts. Amer. J. Agr. Econ., Vol. 77 (2): 268-278. Ceccarelli, S., S. Grando, R. Tutwiler, J. Baha, A. M. Martini, H. Salahieh, A. Goodchild, and M. Michael. 2000. A methodological study on participatory barley breeding. I. Selection Phase. Euphytica, Vol. 111 (2): 91-104. CENTA. 2004. Estudio de caso de la variedad de frijol rojo pequeño “CENTA San Andrés.” Internal Report, San Salvador, El Salvador. 26 p. CENTA. 2005. CENTA Pipil: Variedad de frijol de grano rojo resistente al virus del mosaico dorado amarillo y tolerante al calor. Varietal Information Booklet. CIAT. 2001a. Released Varieties: Latin America. http://webapp.ciat.cgiar.org/beans/ciat_released.htm#america Online available at: __________. 2001b. The impact of the bean research network in Central America: past, present, and future. Online available at: http://webapp.ciat.cgiar.org/beans/impactca.htm. CRSP. 2011. Dry Grain Pulses Collaborative Research Support Program Management Office. Michigan State University. Personal Communication. Cragg, J. 1971. Some Statistical Models for Limited Dependent Variables with Application to the Demand for Durable Goods. Econometrica, Vol. 39 (5): 829-844. Davis, J. S., P. A. Oram, and J. G. Ryan. 1987. Assessment of Agricultural Research Priorities: An International Perspective. ACIAR/IFPRI, Monograph No. 4. 85 p. De Janvry, A.; Fafchamps, M.; and Sadoulet, E. 1991. Peasant Household Behaviour with Missing Markets: Some Paradoxes Explained. The Economic Journal, Vol. 101 (409): 1400-1417. DICTA. 1987. Catrachita: Nueva variedad de frijol rojo precoz y apta para sistemas de monocultivo, relevo y asocio. Trifolium. __________. 1998. Variedad de frijol Dorado. Trifolium. 197 Escoto, D. 2000. Plan Estratégico para la Producción de Frijol: Departamento de Productos Básicos, Secretaría de Agricultura y Ganadería, Dirección de Ciencia y Tecnología Agropecuaria (unpublished). Honduras. __________. 2006. Key Informant Interview, B/C CRSP Survey of PPB project. Danlí, Honduras. Fafchamps, M. and Vargas-Hill, R. 2005. Selling at the Farmgate or Traveling to Market. American Journal of Agricultural Economics, Vol. 87 (3): 717-734. FAO. 2010. Livestock production in tropical Africa with special reference to the tsetse-afferent zone. Online available at: http://www.fao.org/wairdocs/ilri/x5443e/x5443e04.htm FAOSTAT. 2011. Core Production Data. Feder, G., R. E. Just, and D. Zilberman. 1995. Adoption of Agricultural Innovations in Developing Countries: A Survey. Economic Development and Cultural Change, Vol. 33 (2): 255-298. Filmer, D. and L.H. Pritchett. 2001. Estimating Wealth Effects without Expenditure Data-or Tears: An Application to Educational Enrollments in States of India. Demography, Vol 38 (1): 115-132. Goetz, S. 1992. A Selectivity Model of Household Food Marketing Behavior in Sub-Saharan Africa. American Journal of Agricultural Economics, Vol. 74 (2): 444-452. Griliches, Z. 1958. Research Costs and Social Returns: Hybrid Corn and Related Innovations. The Journal of Political Economy, Vol. 66 (5): 419-431. Gujarati, D. 2003. Basic Econometrics. Fourth Edition. McGraw-Hill Publisher. 1002 p. Heckman, J. 1979. Sample Selection Bias as a Specification Error. Econometrica, Vol. 47 (1): 153-161. Hernandez, J. C. 2010. Personal communication. Hernández, J. C. and F. I. Elizondo. 2006. Estudio sobre la adopción de variedades mejoradas de frijol en las principales zonas productoras de frijol de la región Brunca de Costa Rica. Agronomía Mesoamericana, Vol. 17 (3): 357-367. INIAP. 2004a. INIAP 414 "Yunguilla", Variedad mejorada de fréjol arbustivo. Relanzamiento. Plegable No. 253. Programa Nacional de Leguminosas y Granos Andinos. Estación Experimental Santa Catalina. INIAP. Quito, Ecuador. __________. 2004b. INIAP 424 "Concepción", Variedad mejorada de fréjol arbustivo moradomoteado. Plegable No. 254. Programa Nacional de Leguminosas y Granos Andinos. Estación Experimental Santa Catalina. INIAP. Quito, Ecuador. 198 __________. 2004c. INIAP 425 Blanco "Fanesquero", Variedad mejorada de fréjol arbustivo para consumo en grano tierno. Plegable No. 252. Programa Nacional de Leguminosas y Granos Andinos. Estación Experimental Santa Catalina. INIAP. Quito, Ecuador. __________. 2004d. INIAP 426 Canario "Siete Colinas", Variedad mejorada de fréjol voluble. Plegable No. 234. Programa Nacional de Leguminosas y Granos Andinos. Estación Experimental Santa Catalina. INIAP. Quito, Ecuador. __________. 2005. INIAP 420 "Canario del Chota", Variedad mejorada de fréjol arbustivo de color amarillo. Plegable No. 258. Programa Nacional de Leguminosas y Granos Andinos. Estación Experimental Santa Catalina. INIAP. Quito, Ecuador. __________. 2007a. INIAP 427 "Libertador" -rojo moteado-, Variedad mejorada de fréjol arbustivo para la zona de Guaranda, Chimbo y San Miguel de Bolívar. Plegable No. 284. Programa Nacional de Leguminosas y Granos Andinos. Estación Experimental Santa Catalina. INIAP. Quito, Ecuador. __________. 2007b. INIAP 428 "Canario Guarandeño", Variedad mejorada de fréjol arbustivo para la zona de Guaranda, Chimbo y San Miguel de Bolívar. Plegable No. 285. Programa Nacional de Leguminosas y Granos Andinos. Estación Experimental Santa Catalina. INIAP. Quito, Ecuador. __________. 2009a. INIAP 429 Paragachi Andino, Variedad de fréjol arbustivo de grano rojo moteado. Boletín Divulgativo No. 350. Programa Nacional de Leguminosas y Granos Andinos. Estación Experimental Santa Catalina. INIAP. Quito, Ecuador. __________. 2009b. INIAP 430 Portilla, Nueva variedad de fréjol arbustivo de grano rojo moteado. Boletín Divulgativo No. 351. Programa Nacional de Leguminosas y Granos Andinos. Estación Experimental Santa Catalina. INIAP. Quito, Ecuador. __________. 2009c. INIAP 480 Rocha, Nueva variedad de fréjol arbustivo de grano amarillo canario. Boletín Divulgativo No. 352. Programa Nacional de Leguminosas y Granos Andinos. Estación Experimental Santa Catalina. INIAP. Quito, Ecuador. __________. 2010. Base de datos del INIAP. Online available at: http://mail.iniapecuador.gov.ec/isis/search_terms.php?dbinfo=FICHAS&qtype=search&words=FREJOL INIAP. 2009. INIAP: 50 años aportando al país. Quito, Ecuador. 116 p. INTA. 2006. Variedad de frijol rojo de alto rendimiento y valor comercial INTA-PRECOZ. Trifolium. INTA and UCR. 2005. Variedades de frijol de Grano Rojo liberadas en Costa Rica, período 2000-2004. Trifolium. Key, N.; Sadoulet, E.; and de Janvry, A. 2000. Transactions Costs and Agricultural Household Supply Response. American Journal of Agricultural Economics, Vol. 82 (2): 245-259. 199 KII. 2010a. Key Informant Interviews with bean researchers of Costa Rica, Ecuador, El Salvador, Honduras, and Nicaragua. Conducted between June and August of 2010. 18 p. __________. 2010c. Key Informant Interviews with seed producers of Costa Rica, Ecuador, El Salvador, Guatemala, Honduras, and Nicaragua. Conducted between June and August of 2010. 8 p. __________. 2010d. Key Informant Interviews with government officials of Costa Rica, Ecuador, El Salvador, Guatemala, Honduras, and Nicaragua. Conducted between June and August of 2010. 11 p. __________. 2010e. Key Informant Interviews with bean wholesalers of Costa Rica, Ecuador, El Salvador, Guatemala, Honduras, and Nicaragua. Conducted between June and August of 2010. 11 p. Marasas, C. N., M. Smale, and R. P. Singh. 2003. The economic impact of productivity maintenance research: breeding for leaf rust resistance in modern wheat. Agricultural Economics, Vol. 29: 253-263. Maredia, M. K. 2009. Improving the Proof: Evolution of and Emerging Trends in Impact Assessment Methods and Approaches in Agricultural Development. Revised version submitted to Eds. of the Millions Fed Project. October 13, 2009. 43 p. Maredia, M. K. and D. Byerlee. 2000. Efficiency of research investments in the presence of international spillovers: wheat research in developing countries. Agricultural Economics, Vol. 22: 1-16. Maredia, M. K., R. Bernsten, and C. Ragasa. 2010. Returns to public sector plant breeding in the presence of spill-ins and private goods: the case of bean research in Michigan. Agricultural Economics, Vol. 41: 425-442. Martel-Lagos, P. 1995. A Socio-Economic Study of the Honduran Bean Subsector: Production Characteristics, Adoption of Improved Varieties, and Policy Implications. Ph.D. Dissertation. Michigan State University. 209 p. Martínez, L. R. 2003. An Integrated Assessment of the Dry Bean Subsector in Central America. M.S. Thesis. Michigan State University. 138 p. Mather, D. 2003. Essays on the Economic Impact of Disease-Resistant Bean Research in Honduras. Ph.D. Dissertation. Michigan State University. Mather, D. L., R. Bernsten, J. C. Rosas, A. Viana-Ruano, and D. Escoto. 2003. The Economic Impact of Bean Disease Resistance Research in Honduras. Agricultural Economics, Vol. 29: 343-352. McKenzie, D. J. 2005. Measuring Inequality with Asset Indicators. Journal of Population Economics, Vol 18 (2): 229-260. 200 MINADER and FAO. 2003. Revisão do Sector Agrário e da Estratégia de Segurança Alimentar para Definição de Prioridades de Investimentos (TPC/ANG/2907). Angola, Sistemas de Produção Agrícolas. Documento de Trabalho No. 07. 33 p. Mooney, D. F. 2007. The Economic Impact of Disease-Resistant Bean Breeding Research in Northern Ecuador. M.Sc. Thesis, Michigan State University. 153 p. Morris, M. L. and P. W. Heisey. 2003. Estimating the benefits of plant breeding research: methodological issues and practical challenges. Agricultural Economics, Vol. 29: 241252. Munslow, B. 1999. Angola: The Politics of Unsustainable Development. Quarterly, Vol. 20 (3): 551-568. Third World Norton, G. W. and J. S. Davis. 1981. Evaluating Returns to Agricultural Research: A Review. Amer. J. Agr. Econ., Vol. 63 (4): 685-699. Pardey, P. G., J. M. Alston, C. Chan-Kang, E. C. Magalhães, and S. A. Vosti. 2006. International and Institutional R&D Spillovers: Attribution of Benefits among Sources for Brazil’s New Crop Varieties. Amer. J. Agr. Econ., Vol. 88 (1): 104-123. Pejuan, W. 2005. Constraints to and Opportunities for Developing a Bean Seed Production and Marketing System in Honduras. M.S. Thesis. Michigan State University. 125 p. Peralta, E., A. Murillo, and N. Mazón. 2009. Catálogo de variedades mejoradas de fréjol arbustivo para los valles de Chota, Mira e Intag (Imbabura y Carchi), Ecuador. Publicación Miscelanea No. 146. Programa Nacional de Leguminosas y Granos Andinos. Estación Experimental Santa Catalina. INIAP. Quito, Ecuador. 24 p. PIF/EAP. 2003. Variedad mejorada de frijol rojo-pequeño para zonas bajas e intermedias: Tío Canela 75. Varietal Information Booklet. 4 p. PIF/EAP and DICTA/SAG. 2002a. Presentación de la Nueva Variedad de Frijol Rojo Claro Brillante Amadeus-77. Varietal Information Booklet. 12 p. _________. 2002b. Presentación de la Nueva Variedad de Frijol Rojo Brillante Carrizalito. Varietal Information Booklet. 12 p. _________. 2005a. Cardenal: Variedad de frijol rojo claro resistente a enfermedades, tolerante a factores adversos de suelo y clima, de alto potencial de rendimiento. Varietal Information Booklet. _________. 2005b. Deorho: Variedad de frijol rojo claro resistente a enfermedades, de alto potencial de rendimiento y amplia adaptación a ambientes diversos. Varietal Information Booklet. Reyes, C. H. 2010. Personal communication through electronic mail. 201 Reyes, B. 2011. Assessing the socioeconomic impact of participatory plant breeding of beans in two regions of Honduras. M.Sc. Thesis, Michigan State University. 192 p. Reyes, B.; Donovan, C.; Kelly, V. 2010. Baseline Survey Report. Raising the Incomes of Smallholder Farmers in the Central Highlands of Angola: A Model Project for Improving Agricultural Value Chains in Post-Conflict Nations. ProRenda Project. Report submitted to World Vision. 70 p. Ricker-Gilbert, J.; Jayne, T.; and Chirwa, E. 2011. Subsidies and Crowding Out: A DoubleHurdle Model of Fertilizer Demand in Malawi. American Journal of Agricultural Economics. Vol. 93(1): 1-17. Rosas, J. C. 2006. Key Informant Interview, B/C CRSP Survey of PPB project. Tegucigalpa, Honduras. Rosas, J. C. 2010a. Personal communication during a trip made by the author to Honduras in February 2010. Zamorano, Honduras. Sadoulet, E. and de Janvry, A. 1995. Quantitative Development Policy Analysis. The Johns Hopkins University Press; Baltimore and London. 438 p. StataCorp. 2009. Stata: Release 11. Statistical Software. College Station, TX: StataCorp LP. Tobin, J. 1958. Estimation of Relationships for Limited Dependent Variables. Econometrica, Vol. 26 (1): 24-36. Viana, A. 2010. Informal interview during a trip made to Guatemala in 2010. IICA. Wooldridge, J. M. 2002. Econometric Analysis of Cross Section and Panel Data. Cambridge, MA: MIT Press. 752 p. Wooldridge, J. M. 2003. Introductory Econometrics: A Modern Approach. 2nd ed. SouthWestern Publisher. 863 p. Wooldridge, J. M. 2009. Sample selection and missing data. EC 821B class notes. Michigan State University. 108 p. Wooldridge, J. M. 2008. Two-part and hurdle models. EC 821A class notes. Michigan State University. 34 p. World Bank; EC; FAO; UNICEF; UNDP; and WHO. 2007. Angola Public Expenditure Review. Volume II: Sectoral Review. Report No. 39710-AO. 168 p. World Bank. 2007. Angola: Oil, Broad-based Growth, and Equity. World Bank Country Study Series. Report No. 40531. 156 p. 202 World Vision. 2008. PRORENDA. Raising the Incomes of Smallholder Farmers in the Central Highlands of Angola: A Model Project for Improving Agricultural Value Chains in PostConflict Nations. Grant Proposal. 58 p. Zbinden, S. 2005. El impacto de proyectos y programas agropecuaria en America Central apoyados por COSUDE entre 1980 y 2001: Un analisis meta para la estimacion de los efectos agregados de proyectos agropecuaria. Managua, Nicaragua. 36 p. 203