.1. 1.3;?» 1 $3931“ ft, ,. A. a 2 008 This is to certify that the thesis entitled THE ECONOMIC IMPACT OF IMPROVED BEAN VARIETIES IN NORTHERN ECUADOR presented by Daniel F. Mooney has been accepted towards fulfillment of the requirements for the LIBRARY Michigan State University MS. degree in Agricultural Economics —fi%%% Major Professor’s Signature Dec :3. 207 Date MSU is an affinnative-action, equal-opportunity employer PLACE IN RETURN BOX to remove this checkout from your record. TO AVOID FINES return on or before date due. MAY BE RECALLED with earlier due date if requested. DATE DUE DATE DUE DATE DUE APR Mimi E 6/07 p:/CIRC/DateDue.indd-p.1 __.__———— THE ECONOMIC IMPACT OF DISEASE-RESISTANT BEAN BREEDING RESEARCH IN NORTHERN ECUADOR By Daniel F. Mooney A THESIS Submitted to Michigan State University in partial fulfillment of the requirements for the degree of MASTER OF SCIENCE Department of Agricultural Economics 2007 ABSTRACT THE ECONOMIC IMPACT OF DISEASE-RESISTANT BEAN BREEDING RESEARCH IN NORTHERN ECUADOR By Daniel F. Mooney Breeding for disease resistance is one form of productivity-enhancement research that can also potentially contribute to poverty alleviation and sustainability goals. As agricultural research funds become increasingly scarce and investment alternatives multiply, knowledge about the economic impact of such research becomes useful in resource allocation decision making. This thesis evaluates the economic impact of disease-resistant bean varieties (RVs) in northern Ecuador using farm-level survey data fi'om 2006. Regression analysis of the farm-level benefits associated with red mottled RVs reveals that, when high levels of disease pressure are present, adopters enjoy 40% higher yields and 20% lower per-unit production costs than do non-adopters. When high levels of disease pressure are absent, resistant and local varieties perform similarly but RV adopters apply 43-74% less chemical inputs than do non-adopters. Economic impact assessment results under a baseline scenario indicate that red mottled RVs have an ex post internal rate of return (IRR) of 29% and a net present value (NPV) of $1 .29 million USD from 1982 to 2006. Likewise, recently released purple mottled RVs have an estimated ex ante IRR of 34% and a NPV of $536,000 USD from 2004 to 2024. Copyright by DANIEL F. MOONEY 2007 Dedicated to Ignacia, Sebastian, and Marianela iv ACKNOWLEDGEMENTS The completion of this thesis did not happen without incurring many debts of gratitude. I first want to thank my assistantship supervisor and thesis committee chair, Dr. Scott M. Swinton, and the two additional members of my thesis committee, Dr. Eric Crawford and Dr. James Kelly, for their research guidance. Dr. Swinton, in particular, deserves extra thanks for his friendship, professional advice, understanding in times of need, and providing an excellent example of finding balance in life. Thanks are also given to the Bean/Cowpea Collaborative Research Support Project (CRSP) and the Department of Agricultural Economics for providing research and assistantship funding. Related to the CRSP, I’d like to thank both Dr. Irv Widders and former research assistant Ricardo Labarta, in particular, for their advice and research enthusiasm. Of course none of this research would be possible without those who supported the research and field work in Ecuador. So, . . . mil gracias a1 personal de PRONALEG- GA de INIAP quienes hicieron posible este trabajo. En particular, agradezco e1 apoyo de los Ingenieros Cristian Subia y Eduardo Peralta por su apoyo de coleccio'n e ingreso de datos, por su paciencia en responder a misfrecuentes preguntas, y por su amistad. F inalmente, un agradecimiento grande se debe a los productores de los valles del Chota y Mira quienes brindaron su tiempo valioso en participar en Ias encuestas como entrevistados. T uve una experiencia excelente.’ A collective thank you is also owed to my friends and colleagues in Cook Hall— and all others I’ve come to know over the past two years—who made my time in East Lansing enjoyable and who offered their support and encouragement. Each of you deserves an individual thank you; however, I’d inevitability leave out a deserving soul in such an attempt. Instead, I trust that each of you knows who you are. Those in the Department of Agricultural Economics who deserve thanks include Dr. Scott Loveridge, Dr. Eric Crawford (again), Debbie Conway, Nancy Creed, Julia Osborne, and Ann Robinson for their help with administrative issues. I’d also like to thank Dr. Swinton, Dr. Loveridge, Dr. David Schweikhardt, and Dr. Robert Myers for their assistance in the job market. Another thank you is owed to the Department, the College of Agriculture and Natural Resources, and the Graduate School for enabling the presentation of research at the AAEA meetings in Portland. Finally, my last (and of course largest!) thank you is owed to my family. Ignacia and Sebastian provided the necessary encouragement, understanding, and motivation for me to do my best—and they did it all with remarkable patience. Both my immediate and extended families also deserve thanks for the incredible amount of support they provided through times of both joy and sorrow. There is no way to express the size of my gratitude towards them. vi TABLE OF CONTENTS LIST OF TABLES ................................................................................................ LIST OF FIGURES ............................................................................................. KEY TO ACRONYMS ...................................................................................... CHAPTER ONE: INTRODUCTION ............................................................ 1.1 The Economic Impact of Breeding for Disease Resistance... 1.2 Research Motivation ................................................ 1.3 Research Goal and Objectives ..................................... CHAPTER TWO: BEAN PRODUCTION IN NORTHERN ECUADOR... 2. 1 Introduction .......................................................... 2.2 Bean Production Overview ........................................ 2.3 Bean Diseases and Insect Pests .................................... 2.4 Bean Breeding Research and Resistant Varieties ............... 2.5 Chapter Summary ................................................... CHAPTER THREE: FIELD DATA COLLECTION ........................... 3. 1 Introduction ........................................................... 3.2 Survey Design and Implementation .............................. 3.3 Sample Selection Methodology ................................... 3.4 Survey Weights ...................................................... 3.5 Questionnaire Contents ............................................. CHAPTER FOUR: THE DIFFUSION AND ADOPTION OF DISEASE- RESISTANT BEAN VARIETIES IN NORTHERN ECUADOR ............. 4. 1 Introduction .......................................................... 4.2 Conceptual Framework ............................................. 4.2.1 Technology Diffiision ..................................... 4.2.2 Technology Adoption ..................................... 4.3 Quantitative Methods ............................................... 4.3.1 Logistic Diffusion Curve ................................. 4.3.2 Probit Model of Adoption ................................. 4.4 Data Description .................................................... 4.4.1 Diffusion Data ............................................. 4.4.2 Adoption Data ............................................. 4.5 Results and Discussion ............................................. 4.5.1 Estimated Rates of Diffusion ............................. 4.6 4.5.2 Factors Influencing Individual Adoption Decisions... Chapter Summary ................................................... vii X xiii xiv MAI—nu 35 35 36 36 37 40 40 42 43 43 46 49 49 52 54 CHAPTER FIVE: ECONOMETRIC MEASUREMENT OF INCREMENTAL FARM-LEVEL RESEARCH BENEFITS ................... 5. 1 Introduction .......................................................... 5.2 Research Objectives ................................................ 5.3 Conceptual Framework ............................................. 5.4 Conceptual Models ................................................. 5.4.1 Crop Yield Model .......................................... 5.4.2 Input Demand Functions .................................. 5.4.3 Unit Cost Function ......................................... 5.5 Data Description .................................................... 5.6 Empirical Models and Testable Hypotheses ..................... 5.6.1 Crop Yield Model .......................................... 5.6.2 Input Demand Functions .................................. 5.6.3 Unit Cost Function ......................................... 5.7 Results and Discussion ............................................. 5.7.1 Yield and Input Demand Equations ..................... 5.7.2 Unit Cost Equation ......................................... 5.8 Chapter Summary ................................................... CHAPTER SIX: AN ECONOMIC EVALUATION OF BEAN-BREEDING RESEARCH IN NORTHERN ECUADOR ...................................... 6. 1 Introduction .......................................................... 6.2 Conceptual Framework ............................................. 6.2.1 Research Benefits ............................................. 6.2.2 Research Costs ................................................ 6.2.3 Measures of Project Worth ................................. 6.3 Data Description .................................................... 6.3.1 Research Benefits ............................................. 6.3.2 Research Costs ................................................ 6.3.3 Discount Rate ................................................. 6.4 Results and Discussion ............................................. 6.4.1 Ex-Post Impact of Disease-Resistant Red Mottled Varieties .............................................................. 6.4.2 Ex-Ante Impact of Disease-Resistant Purple Mottled Varieties .............................................................. 6.5 Chapter Summary ................................................... REFERENCES ........................................................................ APPENDIX 1: Sample Selection Details ................................. APPENDIX 2: Calculation of Survey Weights ........................... APPENDIX 3: Household-Level Questionnaire........................... viii 56 56 56 57 60 6O 61 62 63 68 69 71 73 74 74 81 84 86 86 87 87 90 91 92 92 94 98 98 98 100 101 103 109 113 116 APPENDIX 4: APPENDIX 5: APPENDIX 6: APPENDIX 7: APPENDIX 8: Village-Level Questionnaire .............................. Description of the Unsatisfied Basic Needs Index. . .. Supplemental Regressions and Regression Diagnostics fiom Chapter Five ......................... Data Elicitation Worksheets for Research Costs ...... NPV and IR Calculation Tables ........................ ix 130 138 139 145 148 Table 2.1 Table 2.2 Table 2.3 Table 2.4 Table 2.5 Table 2.6 Table 3.1 Table 3.2 Table 3.3 Table 3.4 Table 4.1 Table 4.2 Table 4.3 Table 4.4 Table 4.5 LIST OF TABLES Dry Bean Production Costs, Imbabura and Carchi, Ecuador, 2000 and 2004 ........................................... Net Returns to Bean Production, Imbabura and Carchi, Ecuador, 2000 and 2004 ........................................... Plant Diseases Affecting Bush Bean Cultivation in Ecuador.... Insect Pests Affecting Bush Bean Cultivation in Ecuador. . Bean Varieties Released by IN IAP, 1986-2005 ................ Bean-Oriented CIALs in Northern Ecuador ..................... Number of Households Interviewed by Date .................. Number of Households Interviewed by Enumerator ........... Number of Households Interviewed by Canton ................. Number of Households Interviewed by Watershed ............ Bean Varieties Planted by Plot and Land Area, Imbabura and Carchi, Ecuador, 2006 ......................................... Estimated Adoption Rates by Market Class in 2006, Imbabura and Carchi, Ecuador .................................... Estimated Adoption Rates of Disease-Resistant Varieties as a Proportion of Land Area Cultivated, Imbabura and Carchi, Ecuador, 2006 ....................................................... Descriptive Statistics of Variables Included in the Adoption Model for Disease- Resistant Red Mottled Varieties, Imbabura and Carchi, Ecuador, 2006 ............................ Probit Model Results for Factors Influencing the Adoption of Disease-Resistant Red Mottled Varieties, Imbabura and Carchi, Ecuador, 2006 ............................................. 14 16 18 18 20 22 27 27 30 30 44 45 46 47 53 Table 5.1 Summary Statistics of Variables to be Included in the Regression Analysis for Red Mottled Varieties, Imbabura and Carchi, Ecuador, 2006 .......................................... 64 Table 5.2 Sample Means of Explanatory Variables by Adoption Status, Red Mottled Varieties, Imbabura and Carchi, Ecuador, 2006 ....................................................... 65 Table 5.3 SUREG Results for the Quadratic Crop Yield and Linear Input Demand Equations, Imbabura and Carchi, Ecuador, 2006 .................................................................. 77 Table 5.4 SUREG Results for the Log-1.0g Yield and Input Demand Equations, Red Mottled Varieties, Imbabura and Carchi, Ecuador, 2006 ....................................................... 78 Table 5.5 Regression Results for the Log-Log Unit Variable Cost Function, Imbabura and Carchi, Ecuador, 2006 ................. 83 Table 5.6 Summary of Testable Hypotheses and Empirical Results for Farm-Level Regression Models .............................. 84 Table 6.1 Supply Shifi Parameter Values Used in Ex-Post NPV and IR Calculations, Disease-Resistant Red Mottled Varieties, Imbabura and Carchi, Ecuador, 1982-2006.. ......................... 94 Table A. 1 .1 Villages Located Within Targeted Area of Impact by Map Name ............................................................ 111 Table A. 1 .2 List of Villages Selected for Inclusion in the Survey Sample. 112 Table A.2.l Calculation of Survey Weights by Cluster ....................... 114 Table A.2.2 Calculation of Survey Weights by Stratification Level ........ 1 15 Table A6] OLS Regression Results for the Yield, Fungicide Demand, and Insecticide Demand Equations, Imbabura and Carchi, Ecuador, 2006 ....................................................... 140 Table A.6.2 3SLS Regression Results for the Yield, Fungicide Demand, and Insecticide Demand Equations, Imbabura and Carchi, Ecuador, 2006 ....................................................... 141 Table A.6.3 OLS Estimates of the Linear Unit Variable Cost Function, Imbabura and Carchi, Ecuador, 2006 ............................ 142 xi Table A.6.4 Table A.8.1 Table A.8.2 Table A.8.3 Table A.8.4 Table A.8.5 Variance Inflation Factors (VIPs) for the Quadratic Yield Model ......................................................... Baseline Scenario for NPV and IR Calculations, INIAP red mottled varieties, 1982-2006 ......................... Conservative Scenario for NPV and IR Calculations, INIAP red mottled varieties, 1982-2006 ................................. Robust Scenario for NPV and IR Calculations, INIAP red mottled varieties, 1982-2006 .................................. Baseline Scenario w/ All Research Costs for NPV and IR Calculations, INIAP purple mottled varieties, 1988-2024. . Baseline Scenario w/ Purple Mottled Research Costs for NPV and IR Calculations, INIAP purple mottled varieties, 1988-2024 ............................................................ xii 143 149 150 151 152 153 Figure 2.1 Figure 2.2 Figure 2.3 Figure 2.4 Figure 4.1 Figure 4.2 Figure 4.3 Figure 6.1 Figure 6.2 Figure A.l.1 Figure A.5.1 Figure A.6.1 Figure A.7.l Figure A.7.2 LIST OF FIGURES Location of Study Area .................................... Total Bean Area Harvested, Imbabura and Carchi, Ecuador, 1990-2005 ....................................... Average Annual Farm-Level Bean Yields, All Market Classes, Imbabura and Carchi, Ecuador, 1985-2006. Wholesale Prices of Agricultural Commodities, Ibarra Market, 2000-2005 ......................................... The S-Shaped Diffusion Curve ........................... Estimated Diffiision of Improved Red Mottled Beans, Imbabura and Carchi, Ecuador, 1986-2006 ............. Predicted Diffusion of Improved Purple Mottled Beans, Imbabura and Carchi, Ecuador, 2004-2024 ............ Research-Induced Supply Shifi ........................... Sensitivity of NPV to Discount Rate ...................... Cartographic Maps Used in Developing the Area Frame ........................................................ Description of the Unsatisfied Basic Needs Index ....... Scatter Plots of DFBETA Regression Diagnostic Statistics ....................................................... Research Cost Worksheet #1: Operation Costs and External Support Dedicated to Varieties Under Evaluation. . Research Cost Worksheet #2: Human Resource Costs Dedicated to Varieties Under Evaluation ............... xiii 10 13 15 37 51 51 88 99 110 138 144 146 147 3SLS AI B/C CRSP CGIAR CIAL CIAT CIMMY T GPS INIAP INEC IRR m.a.s.l MSU MWD NPV OLS PESAE PPB PRONALEG-GA PVS RUM RV SICA SUREG TEM UBN USAID USD UVC VIF LIST OF ACRONYMS Three Stage Least Squares Active Ingredient Bean/Cowpea Collaborative Research Support Program Consultive Group on International Agricultural Research Comité de Investigacion Agricola Local Centro Internacional de Agricultura Tropical Internacional Maize and Wheat Improvement Center Global Positioning System El Instituto Nacional Autonomo de Investigaciones Agropecuarias E1 Instituto Nacional de Estadistica y Censos Internal Rate of Return Meters above sea level Michigan State University MacKinnon, White, and Davidson Net Present Value Ordinary Least Squares El Programa Especial de Seguridad Alimentaria en el Ecuador Participatory plant breeding Programa Nacional de Leguminosas y Granos Andinos, Participatory Varietal Selection Random Utility Model Resistant Variety Servicio de Informacion y Censo Agropecuario Seemingly Unrelated Regression Treatment Effect Model Unsatisfied Basic Need United States Agency for International Development United States Dollar Unit Variable Costs Variance Inflation Factor xiv CHAPTER ONE: INTRODUCTION 1.1 The Economic Impact of Breeding for Disease-Resistance Post-green revolution research agendas of national and international agricultural research organizations have diversified away from their original focus on enhancing crop productivity to encompass a set of objectives that includes topics such as sustainability and poverty reduction (World Bank 2003; Pingali 2001). As agricultural research funding becomes increasingly scarce and the number of alternative investment possibilities multiplies, national governments and international donor organizations must carefully decide which projects to invest in. Economic impact assessments provide a useful tool for understanding the relative merits of alternative investment opportunities by serving as a yardstick that is comparable across disparate technologies. The most common scholarly approach to economic impact assessment relies on the economic surplus model, which measures the change in consumer and producer surplus associated with the adoption of a new technology adoption and the subsequent research-induced supply shift. By comparing with and without-research scenarios, incremental net benefits derived from the new technology may be determined and then combined with research costs so as to obtain summary statistics such as net present value (N PV) and the internal rate of return (IRR) that are useful in priority setting and resource allocation decision making. Methodology for this class of economic impact assessment is well established in the literature (Alston et al. 1998, Masters et al. 1996). In addition, a set of best practices for impact assessment implementation and data collection also exist (Morris and Heisey 2003, Maredia et al. 2000). Finally, numerous studies highlight the high retums—well over 50% in many cases—to investments in plant breeding over the past half century (Alston et a1. 2000, Evenson 2001 , Evenson and Gollin 2003). Frequently overlooked, however, is the impact that plant breeding research itself can have on non-productivity objectives such as sustainability (be it economic or environmental) or poverty reduction. This is particularly true when the targeted crops are produced by poor households and production systems depend heavily on the use of purchased and potentially hazardous agricultural chemicals. For example new crop varieties endowed with disease resistance (RVs, for resistant varieties) can increase productivity while simultaneously reducing the need for costly and potentially hazardous fungicide inputs. Furthermore, inclusion of small-scale farmers in the research phase of the crop improvement process using participatory techniques may increase RV adoption rates and yield gains among the poor or similarly marginalized groups, many of whom did not benefit from the first generation of green revolution technologies. The appeal of embedding genetic improvements directly into the seed is that additional inputs or changes in management practices are typically not required—farmers already know how to plant and use existing seeds, and seed distribution systems are widespread. Impact analysis methodologies for such RVs differ from those for traditional high yielding varieties without disease resistance characteristics. First, RVs provide two distinct avenues to achieving productivity increases (Smale et a1. 1998; Morris and Heisey 2003). One avenue is through productivity maintenance, where research benefits are not derived from yield gains per se but rather from the avoidance of yield losses associated with disease pressure. The other avenue is through traditional yield gains that RVs may exhibit over currently planted varieties in instances where disease pressure is absent. Second, the embedded resistance traits provide damage abatement services that can substitute for pesticide inputs (Lichtenberg and Zilberman 1986). Mather (2005) showed this to be true for disease-resistant bean varieties in Honduras. This potentially leads to a reduction in both production costs and in the quantity of pesticide active ingredient released into the environment. Given these distinctions, breeding for disease resistance is described in this thesis as a type of productivity-enhancing research rather than yield-enhancing or productivity- maintenance research, both of which are commonly used in the economic impact assessment literature. The term productivity-maintenance is reserved strictly for genetic research that is conducted to keep a previously-bred trait (e. g. disease resistance) viable in the face of genetic adaptation by diseases and their vectors. Similarly, the term yield- enhancing is reserved strictly for genetic research aimed at increasing yields without consideration to resistance characteristics. So defined, the term productivity-enhancing research is thus preferred as it encompasses both other types of research as special cases and therefore does not limit a priori the possibility of either as a source of productivity gain. Few economic impact assessments are devoted to disease resistance research. Those that do exist indicate high rates of return. Morris et a1. (1994) found an ex post IRR of 80% for Nepal’s national wheat RV research program during a 30-year period from 1960-1990. This same study also estimated an ex ante IRR of 49% for the period 1990- 2020. Smale et a1. (1998) assessed research on breeding for disease resistance in Mexican wheat varieties developed by the International Maize and Wheat Improvement Center (CIMMYT) for a 20-year period from 1970-1990 and estimated the IRR to lie between 13% and 40%. In a multiple-country impact assessment of CIMMYT’s disease-resistant genetic research in wheat, Marasas et a1. (2003) found an IR of 41% with an NPV of $5.36 billion USD over the period 1973-1990. Mather et aL (2003) found a rate of return to disease-resistant genetic bean research in Honduras close to 40% for a 16-year period from 1984 to 2000. In each case, the returns to disease-resistant genetic research remained well above the assumed opportunity cost of capital (between 5-10% in each study) which suggests that similar research should be considered by national governments and international donors as an attractive fiinding possibility. 1.2 Research Motivation The economic impact of similar disease-resistant research in northern Ecuador is largely unknown. Since 1982, Ecuador’s national agricultural research institution El Instituto Nacional Auto'nomo de Investigaciones A gropecuarias (INIAP) has developed and released a series of disease-resistant bush bean (Phaseolous vulgaris L.) varieties in the northern sierra provinces of Imbabura and Carchi. The release of these RVs was supported in part by external fiinding fi'om both the Centro Internacional de Agricultura Tropical (CIAT) and the Bean/Cowpea Collaborative Research Support Program (B/C CRSP). The motivation behind this bean improvement effort stems from the central role of bean production within the area. First, beans represent one of the principal crops in the region in terms of total area planted and number of farmers. Second, bean production provides farm households with a source of both income and household nutrition through sales and consumption, respectively. Finally, the northern Andean region serves as a source of bean genetic biodiversity whose values extend beyond national borders. Since INIAP’s release of the first RVs, descriptive statistics obtained from a series of household surveys within the region indicate increasing trends in average yields and in the land area planted to beans (Arévalo 1985, Peralta et al. 1991, Lépiz et aL 1995, and Peralta et al. 2001). In addition, high indices of poverty and the significant health risks posed by agricultural chemicals are well documented throughout the region (SIISE 2001, Crissman et a1. 1998). Nevertheless, no research has been conducted on the farm- level impacts of RV bean adoption or on the estimated return to bean research expenditures. In 2005, this knowledge gap lead INIAP to a request an economic impact assessment of the disease-resistant bean breeding research program during a joint meeting between INIAP and B/C CRSP researchers. The research presented here serves the dual purpose of fulfilling this request while also expanding the economic impact assessment literature, with the principal contribution being the use of treatment effect regression models to determine the farm-level benefits associated with RV adoption and then incorporating these findings into the traditional economic impact assessment methodology. 1.3 Research Goal and Objectives The goal of this thesis is to evaluate the economic impact of disease-resistant bean breeding research and outreach in Northern Ecuador. While RVs from various bean market classes exist, this analysis will focus on RVs developed for the red mottled and purple mottled market classes, which represent over 80% of the total land area cultivated to beans in northern Ecuador. To accomplish this, three specific objectives are identified. They are: 1. Determine the rate of diffusion of red mottled and purple mottled RVs across time, and identify factors that influence farmers’ adoption decisions (Chapter 4). 2. Statistically estimate the farm-level benefits associated with the adoption of RV red mottled varieties, including impacts on yield, input use and the unit cost of production using a set of treatment effect regression models (Chapter 5). 3. Calculate estimates of a) the ex post economic impact of bean breeding research on red mottled varieties from 1982 to 2006, and b) the ex ante economic impact of bean breeding research on purple mottled varieties from 2000 to 2024, along with appropriate sensitivity analyses on key parameters (Chapter 6). Before discussing each of these objectives in turn, however, the subsequent two chapters provide needed background information on bean production in Ecuador (Chapter 2) and on the field survey methodology used in data collection (Chapter 3). CHAPTER TWO: BEAN PRODUCTION IN NORTHERN ECUADOR 2.1 Introduction The common bush bean (Phaseolus vulgaris L.) represents one of Ecuador’s principal agricultural crops. In particular, the Mira and Chota river valleys—located along the shared border of Ecuador’s two northern sierra provinces of Imbabura and Carchi—stand out as the focal point of production (Figure 2.1). This chapter provides an overview of production trends, crop management practices, principal insect pests and plant diseases, and local breeding efforts that surround the bush bean. The information presented here was obtained in collaboration with bean breeders fi'om Ecuador’s national program on food legumes and Andean grains, El Programa Nacional de Leguminosas y Granos Andinos (PRONALEG-GA), located at the Santa Catalina Experimental Research Station in Quito, Ecuador. PRONALEG-GA pertains to Ecuador’s national agricultural research institute, El Instituto Nacional Autonomo de Investigaciones Agropecuarias (INIAP). The information gathering process involved two visits by the thesis author to the Santa Catalina station in Quito and survey site in northern Ecuador. In addition, a number of previous studies on bush bean cultivation in the provinces of Imbabura and Carchi also aided in analyzing changes in production practices over time (Arévalo 1985, Peralta et al. 1991 , Lépiz et al. 1995, Peralta et al. 2001, and Subia et al. 2004). Figure 2.1: j «fig. m-‘ . .T'l rf‘ I Location of Study Area :Egh \":1’\\ (I 'I . I *. "lip“,v- /" . ,I .0 ‘. $3 1‘ ’[I ~. eaten. Ii Imbab , ' , n " t _ chm/sue, a” ” _" t "‘ . . ‘ r as”: *- IKH Afar/U "" -- OOH-Q \ jE,-}fk¢-\ I VJ \I‘ E“ ,1 ( Ecu dor ..’ 'I- (I vii. g .0" [Jr I) T? Y,“ safi‘\ 1 '1i ‘5‘ -'-.'r. A | 1" 0' u” A I ‘ i. , . mg“ fi.lf.|r.n I,b~l I 2.2 Bean Production Overview The majority of bean production in northern Ecuador occurs during two main planting cycles. The first cycle runs from January through April, while the second cycle runs from September through December (Vasquez et al. 1992). Each of these cycles corresponds to a period of increased precipitation. Between cycles, from June to August, the region faces an extended dry season. Precipitation totals vary greatly throughout the region. Communities located near the valley’s center receive an annual average precipitation of 480 mm whereas communities located 10 kilometers from the valley’s center receive a much higher annual average precipitation of 630-795 mm (Rodriguez- Jaramillo 1994). In 2001, the provinces of Carchi and Imbabura together accounted for 40% of all national dry bush bean production (INEC 2001). In Carchi, beans rank as the second most important agricultural crop in terms of both land area cultivated (7,700 hectares) and number of farm households (4,200) after the potato. In Imbabura, beans rank as the second most important crop in terms of land area (4,600 hectares) afier maize, and as the third most important crop in terms of number of farms (2,500) after both maize and potatoes. The total land area dedicated to bean cultivation increased steadily from 1990 to 2000 (Figure 2.2)'. The period fi‘om 2000 to 2004, however, registered a slight decline. Interestingly, this decline directly follows the period of economic instability and eventual dollarization of the Ecuadorian economy in 1999-2000. The estimated area dedicated to beans in 2005 shows a large upward tick to over 30,000 hectares. In 2005, the total bean harvest in Imbabura and Carchi was approximately 24 thousand metric tons. Bean market classes are primarily defined by seed coat color, degree of mottling (i.e. spotting), and whether they are harvested dry (seco) or as fresh pods (en tiemo). The size, shape and texture of the beans are also important, but help to define quality rather than market class. The two largest market classes for dry beans in northern Ecuador are the red mottled and purple mottled classes—both of which have a high demand in regional markets due to their popularity in Colombia. Many additional market classes are also I This figure uses estimated yield data from SICA (2007b) and assumes that Carchi and Imbabura together account for 40% of total national ch'y bean production following INEC (2001). Figure 2.2: Total Bean Area Harvested, Imbabura and Carchi, Ecuador, 1990-2005 35.000 30.000 25.000 20.000 Area Harvested (ha) 15,000 10,000 1990 1992 1994 1996 1998 2000 2002 2004 Ye ar Source: SICA (2007a) cultivated, but are primarily sold locally or are grown for home consumption. These include white, yellow, solid red, solid black, cream, and pink mottled market classes. Land preparation is standardized throughout much of the region and generally involves two steps. The first step, called Ia rastra, consists of harrowing of the bean field using either animal traction equipment or a tractor. The second step, called la surcada, uses animal traction to form topsoil into firrrows that wind throughout the bean field so as to increase the efficiency of gravity-fed irrigation practices. Following la surcada, a process called arreglo de guachos is ofien undertaken using manual labor to put the finishing touches on the firrrows. Occasionally, the above two steps are preceded by an initial plowing, called la arada. The combination of land preparation practices used by a particular farmer depends largely on previously planted crops and/or if the land had been previously fallow. In 2005, Subia et al. (2004) found that 100% of producers prepared 10 their land using the surcada, 85% used the rastra, and only 40% used the arada. These numbers match those presented by Arévalo (1985), suggesting that land preparation practices have changed little over the past two decades. Typical production inputs consist of seed, pesticides, fertilizers, and manual labor. Seed is obtained either through local markets, from other farmers, or is retained fi'om the previous harvest. INIAP recommends a seed planting density of approximately 90 kg (or 2 quintales) per hectare (Vasquez et al. 1992), although a recent survey of 27 bean farmers reported an actual mean of 68 kg ha'l (Subia et a1. 2004). Pesticides are traditionally applied on a prophylactic calendar spray basis, without regard to observed infestations. Farmers typically use at least one insecticide and one fiingicide in each application (Peralta et al. 1991). It is not uncommon, however, for farmers to mix multiple insecticides and firngicides with different commercial names for use in the same application. A reduction in farmers’ reliance on pesticides over the past two decades is evident. In 1985, 90% of all farmers relied on 3 or more pesticide applications per production cycle (Arévalo 1985). In 1991, 83% of all farmers relied on 3 or more pesticide applications per cycle, with an overall average of 3.2 applications per cycle (Peralta et al. 1991). Survey results from the present study (2006) indicate only 70% of farmers relied on 3 or more pesticide applications, with an overall average of 2.9 applications per cycle. An even larger reduction appears to have occurred in villages receiving INIAP extension services. Subia et al. (2004) reported that only 34% of farmers from a set of 9 villages previously receiving INIAP extension intervention related to bean ll production used 3 or more pesticide applications per cycle with an overall average of 1.9 applications per cycle. Fertilizers used in bean production include both foliar fertilizers and soil fertilizers. F oliar fertilizers are frequently included in pesticide applications. Survey results indicate that 84% of all pesticide applications included a foliar fertilizer. Again, it is not uncommon for farmers to include more than one foliar fertilizer in the same application. It is also common for farmers to use a foliar fertilizer that is inappropriate for the stage of plant grth at the time of application. Soil fertilizers, on the other hand, are applied at low rates. Subia et al. (2004) reported only 19% of bean producers apply soil fertilizers, with 18-46-00 NPK being the most common. Labor input is used primarily for four tasks: weeding, el aporque or “mounding” (a cultural practice of covering the plant base with topsoil), irrigation management, and harvesting. Weeding and mounding are often undertaken simultaneously in a process called la pala where farmers use shovels to clear soil sediment out of the irrigation fin-rows while simultaneously managing weed infestations. On rare occasions, herbicides or animal traction are substituted for manual weeding. Irrigation management requires the most labor days, often accounting for the majority of the overall labor requirement (Arévalo, 1985; Subia et al. 2007). In addition, the number of required irrigations per production cycle depends almost exclusively on precipitation levels—and as a result varies greatly from year to year. Land tenure arrangements commonly found among bean producers vary. Data collected for the present study found just under two-thirds of producers (81 of 132) owned their own land. Exactly one-third (44 of 132) entered a sharecropping agreement 12 Figure 2.3: Average Annual Farm-Level Bean Yields, All Market Classes, Imbabura and Carchi, Ecuador, 1985-2006 1500 1372 1300 7‘” I" 13.1.?” A // £1100 [,1’ 5 //// g 900 —————— 860x/ $ 786 ———————— ' I 700 attestsrastt Year Source: Arc'valo (I985), Lépiz et al. (1995). and Peralta et al. (2001). 2006 data is from the present study where production inputs and harvests are shared between the farmer and land owner. Only a few farmers rented land (5 of 132). Average per hectare bean yields appear to have almost doubled over the past two decades—from 785 kg ha'1 in 1985 to over 1300 kg ha'1 in 2006 (Figure 2.3). Yield data is typically reported by farmers as the number of quintales harvested per quintal of seed input, where one quintal is equal to 100 pounds (or, equivalently, 45.4 kilograms). All data reported here, however, is converted into kilograms per hectare. The data presented in Figure 2.3 reflect descriptive statistics obtained fi'om previous farm-level surveys in the Provinces of Imbabura and Carchi. The reported figures utilize different sample sizes and sample selection methods. In general, they include varieties from all market classes and are not representative of the entire area of impact. Data from the present survey (2006) indicate per hectare yields for red mottled varieties to be slightly above average and per hectare yields for purple mottled varieties to be slightly below average. 13 Table 2.1: Dy Bean Production Costs, Imbabura and Carchi, Ecuador, 2000 and 2004 (n=19) 2000 2004 Cost * $/ha % of total $/ha % of total cost cost Land Preparation 55 12 87 18 Pesticides 1 l 7 26 70 15 Other Purchased Inputs 129 28 107 22 Manual labor (hired and household) 157 34 220 45 Total Cost 458 484 ‘Cost data from 2000 was inflated to 2004 prices using the Ecuador producer price index Source: Subia et a1. (2007) Panel data for a sample of 19 bean farmers collected in 2000 and 2004 reported an increase of 5.5% in variable per hectare production costs from 2000 to 2005 (Table 2.1). Analysis of specific cost categories reveals a decrease in total expenditures on pesticides and other purchased inputs, and an increase in total expenditures on land preparation and labor. Labor costs tend to vary widely fi'om year to year, however, since irrigation practices depend on the quantity of rainfall and are labor-intensive. The increase in land preparation costs is also most likely driven by the labor-intensive nature of land preparation practices. The majority of beans harvested are sold for cash to market intermediaries, either at the farm gate or in the local markets. Results fi'om the present survey (2006) indicate that the average percentage of bean harvest sold in the market is 87%. A breakdown of percentages sold by province show 91% of the harvest was sold in Imbabura and 84% was sold in Carchi. These finding are higher but proportionally consistent with Lépiz et al. (1996), who found that 89% of total bean production was sold in Carchi and only 65% of in Imbabura. The local markets of Ibarra and Pimampiro generally serve as the first stop for these beans before they are exported to Colombia. In 1998, dry bean 14 — Beans (Red Mottled) Figure 2.4: Wholesale Prices of Agricultural -.-..Tm.° Commodities, Ibarra Market, 2000-2005 - "maze (Ham dry) ------- Yellow split pea 0.60 0.50 - 0.40 - Jan-00 Source: SICA (2007b) consumption in Colombia exceeded 145,000 metric tonsz, with the cities of Medellin, Bogota, Cali, and Barranquilla accounting for over 70% total demand. Dry bean imports from Ecuador during the same year totaled 11,500 metric tons, or about 8% of total dry beans consumed (CCI 2000). Market prices for beans fluctuate greatly fi'om year to year (Figure 2.4). Between 2000 and 2005, two periods of high prices and two periods of low prices are evident— even after adjusting the price series for both seasonal trends and inflation. These fluctuations contrast with the prices of tomatoes, maize, and yellow split peas. Both tomato prices and split pea prices show an abrupt increase and decrease in price in 2002, but otherwise remain relatively constant compared to beans. Maize prices remain very constant throughout the entire 5 year period. The difference in price fluctuations likely reflects the fact that tomatoes, maize, and split peas are oriented towards the domestic market whereas red mottle beans are produced primarily for export. 2 Total dry bean consumption in Colombia is estimated as total dry bean production plus imports and minus exports. 15 Table 2.2: Net Returns to Bean Production, Imbabura and Carchi, Ecuador, 2000 and 2004 (n=19L Year: 2000 2004 Total Cost ($USD, nominal) 348 484 Yield (kg ha") 1350 1 166 Price ($/kg) (unadjusted) 0.6 0.97 Total Revenue 810 1131 Net Return / ha 462 647 Benefit-Cost Ratio 1 .33 1.34 Source: Subia et al. (2007) This price fluctuation affects net returns to bean production. Panel data from a small sample of 19 farmers in 2000 and 2004 show average revenues per hectare of $810 in 2000 and $1,131 in 2004 (Table 2.2). While yields were lower in 2004, producers received a much higher price, and average revenues are actually higher. Net returns equaled $462/ha in 2000 and $647/ha in 2004. In both years, the ratio of total revenue to total costs per hectare was approximately 1.33. 2.3 Bean Diseases and Insect Pests Plant diseases and insect pests represent two important production constraints. In 1995, Lépiz et al. (1996) reported that 55% of all respondents listed insect pests as a principal production constraint and 40% listed plant diseases. Farmers also mentioned inadequate soil and seed, and weed infestation as other production constraints, but none was listed by more than 15% of respondents. Bean rust and anthracnose are the most widely reported plant diseases. This is expected since bean production in northern Ecuador occurs at altitudes over 1000 meters above sea level (m.a.s.l.) where both diseases be more prevalent. Both Lépiz et al. (1996) and Peralta et al. (1991) report bean rust as the single most prominent disease in northern Ecuador, affecting 70% and 85% of all farmers in each study, respectively. Anthracnose l6 was reported to affect 35% and 25% of all farmers, respectively. Results fi’om the present survey (2006) are consistent with these earlier findings: 70% of farmers reported the presence of bean rust and 62% reported the presence of anthracnose. In addition, 89% of farmers reported at least one of the two diseases and 43% reported both. Other common diseases include powdery mildew, angular leaf spot, bacterial blights, web blights, and root rots (Table 2.3). Successful disease management requires using proper cultural practices and disease recognition ability. Bean rust, powdery mildew, angular leaf spot, and web blight are transmitted through decaying plant materials left from previous harvests in the same or adjacent fields. Root rot disease spores can remain in the soil for many months after harvest. Crop rotations can help to minimize the recurrence of these diseases. Anthracnose and bacterial blights, on the other hand, are transmitted through bean seed, and farmers confi'onting these diseases must obtain seed from a secure source. Proper seed management and renovation can help to minimize the recurrence of these diseases. Nevertheless, these cropping practices are not standardized throughout the region. An additional confounding factor confronting plant disease management is farmer inability to properly identify plant diseases. Lépiz et al. (1995) reported that farmers often refer to all plant diseases collectively as “la lancha,” a local term used by farmers. Whitefly is the most commonly reported insect pest. Peralta et al. (1991) reported an incidence of whitefly of 95%. Other insect pests of minor importance include cutworms, red spider mites, leaf miners, leaf rollers, leaf hoppers, nematodes, aphids, Chrysomelid beetles, l7 Table 2.3: Plant diseases affecting bush bean cultivation in Ecuador Common Name (English) Common Name (Spanish) Scientific Name(s) Bean rust Roya Uromyces appendiculatus Aritliracnose— — _-- _ muAritracr-rosisi _ "— I f S“ _ Colletotrichum lindemuthianum _ Powderymildew Cieriiza/Oidium Ervsiphe polygoni * ’ ' 0 * AEghTar leaf spot ManchaAriguIar _ Phaeoisariopsis griseola Baa—Erich? " Bic—1060510 I“ " 'Xanthomonas emigre '— Web blight Mustia 7 W ' Thanatephorus cucumeris Root to? ' PudricioinRadicularv’ _ thEdetBttla—s; I ''''' F usarium sp. Sclerotium sp. Source: Arévalo (1985); Vésquez et al. (1992), Subia et al. (2004) Table 2.4: Insect pests affecting bush bean cultivation in Ecuador Common Name (English) Common Name (Spanish) Scientific Name(s) Whitefly Mosca blanca Trialeurodes vaporariorum Qfin3{_ "I Minadores de la_h(;j:a__——Lifirionn);a—htlidohr;n:i.s— Hemichalepus sp. Phyllonorycter sp Cut worm Trozadorli Phyllophaga oilisol-eta descortezador A grotis ipsilon Spodopterafrugiperda. tat roller Baiieiradbrl Epihotia £0." " " ._. __ - _- _- _. . evened? . __ _ __ _. _. ._ _ Red spider mite Arafiita roja Tetranychus sp. NeflttEtB‘de—m A -- " __- Nematodo del 0000 'Mefidegiigsfii " ‘ _- F [51‘ seeder “futile (reed; w wedged Itraemeri. ’ ‘ ’ 741131111?“w ”Pill-guillafl 7‘ ” 2135i; Qp. -_ m' ” ' 7 disease-tie beetle ‘Pindam " ‘ ‘ hCeroto‘m‘a s07; 0‘ ”0 .___ '_ fl -1- flfla“ - fl _ Diabrotica balteata _ j 7 Caterpillar Pega pega Omiodes indicata. Mole-cricket Grillo topo 7 Cry/[us sp. Thrip” I Trips ' Thrips palmi’ #fi. _c._-_ _- -__. _Er£'1t1{'3i_el_@§9~_ _ Bean weevil Gorgojo Acanthoscelides obtectus Source: Vésquez et al. (1992) and Are’valo (1985) 18 mole crickets, and thrips (Table 2.4). Cut worm infestations increase susceptibility to plant diseases. Red spider mite infestations increase under dry conditions when planting is delayed (Vasquez et al. 1992). The bean weevil does not affect bean cultivation itself, rather it poses as a pest during storage. 2.4 Bean Breeding Research and Resistant Varieties A network of national and international organizations leads the bean improvement effort in northern Ecuador. This effort is led primarily by PRONALEG-GA of INIAP, which is the arm of Ecuador’s national agricultural research center responsible for the genetic improvement of leguminous crops. International organizations involved in the bean breeding process include the Bean/Cowpea Collaborative Research Support Project (B/C CRSP) funded by the United States Agency for International Development (USAID) and the Centro Internacional de Agricultura Tropical (CIAT) in Cali, Colombia, which is a member of the Consultative Group on International Agricultural Research (CGIAR). In addition, two non-govemmental organizations, the Programa Especial de Seguridad Alimentaria en el Ecuador (PESAE) and the Corporacio'n Randi- Randi have also contributed to the bean improvement process through their extension services. The Ecuadorian Ministry of Agriculture does not cmrently provide agricultural extension services related to bean production within the area of impact targeted by PRONALEG-GA’S bean breeding efforts. PRONALEG-GA began their bean improvement efforts by releasing a set of improved bush bean cultivars in northern Ecuador during the 19805 and 19905. Between 19 Table 2.5 Bean Varieties Released by INIAP, 1986-2005 Varie Name Year of Market Resistance Resistance to fiZZuZIIZZL Yield ty Release Class to Bean Rust Anthracnose p (kg ha") (meters) Paragachi" 1986 Red Mottle Susceptible Susceptible 1800-2500 1200— 2000 Cargabello 1987 Red Mottle Susceptible Tolerant 1600-2500 1500 (INIAP-404) Imbabello 1991 Red Mottle Intermediate Susceptible 1500-2200 1500- (INIAP-41 l) 2900 Je.Ma. 1996 Red Mottle Resistant Resistant 1800-2500 1200- (INIAP-418) 2300 La Concepcion” 2004 Purple Intermediate Susceptible 1400-2400 700- (IN IAP-424) Mottled 1800 Yunguilla 1993 Red Mottled Intermediate Resistant 1400-2400 500- (INIAP-4l4) 2004 2000 Canario del Chota 2005 Yellow Intermediate Susceptible 1400-2400 1200- (INIAP-420) 2200 Blanco F anesquero 2005 White Intermediate Resistant 1400-2400 1090- (IN lAP-425) . 2000 Source: Lépiz (1996); INIAP (1991a, 1991b, 1996a, 1996b, 2004a, 2004b, 2004c, and 2005) * The variety Paragachi is resistant to root rots “The variety La Concepcion was formerly known as the local variety Mil Uno 1987 and 1996, a total of four improved varieties were released, including Cargabello (INIAP-404), Imbabello (INIAP-411), Yunguilla (1993), and Je.Ma. (INIAP-418) (Table 2.5). These varieties all pertain to the red mottled market class and are adapted to production environments between 1500 and 2500 m.a.s.l. They also have varying degrees of resistance to bean rust and anthracnose, with Yunguilla and Je.Ma. resistant or intermediately resistant to both. These cultivars were bred at CIAT, with INIAP’s role limited to testing and evaluation and maintenance breeding. The varieties released were selected by plant breeders with little or no input from farmers. The variety Paragachi, also originally developed by the CIAT, is a cross between varieties of both Andean and Central American descent (INIAP 1996b). It was developed 20 using traditional breeding methods and widely tested by INIAP scientists in northern Ecuador. It is well adapted to the production environment in northern Ecuador and is resistant to root rots. It is also the most widely planted improved cultivar in spite of its susceptibility to both bean rust and anthracnose (Table 2.5). Since its introduction in Ecuador by CIAT, PRONALEG-GA bean breeders have undertaken maintenance breeding on the Paragachi variety. In 2002, a renewed bean improvement effort began and was led by bean breeders from both PRONALEG-GA and the B/C CRSP at Michigan State University. In contrast to earlier traditional bean breeding and testing efforts, this more recent effort relied on participatory plant breeding (PPB) methods to select varieties to be released. PPB involves the close collaboration between researchers and farmers to bring about plant genetic improvements within a species. Instead of playing a passive role of technology recipients as with traditional breeding programs, in PPB farmers are treated as partners in research (Vemooy 2003). In Ecuador, the particular PPB method utilized is called participatory varietal selection (PVS), in which farmers select the final variety to be released from a set of fixed-lines chosen by PRONALEG-GA breeders (Ernest 2004). The PVS process generally consists of four steps: i) identification of the farmer’s varietal needs, ii) a search for suitable genetic materials, iii) farmer experimentation with potential varieties in their own fields under their own crop management practices, and iv) selection of the preferred varieties based on their own selection criteria (Vemooy 2003). To implement the PVS process, PRONALEG-GA researchers established four local agricultural research committees, or CIALs to use their Spanish acronym (Comité 21 de Investigacio'n Agricola Local). As the name implies, CIALs are community-based organizations that implement farrner-run agricultural experiments. CIAL leaders are elected by community members, research topics are collectively identified by the community at large, and each manages a small fund to offset the costs and risks associated with their experiments. Most CIALs are focused on increasing agricultural productivity, however multiple objectives such as agro-biodiversity conservation, increased experimentation, protection of farmer’s breeding rights, or increasing the availability of improved seed, or the targeting of marginalized groups such as women, poor, landless, or historically underrepresented ethnic groups, may also be included (CIAT 2001). The CIAL concept originated at CIAT in Colombia. The four bean-oriented CIALs in northern Ecuador are located in the corrrrnunities of La Concepcion, Santa Lucia, El Tambo, and San Clemente (Table 2.6). Each of these CIALs emphasize PVS processes and are targeted towards communities with high indices of poverty and malnutrition (Mazon and Peralta 2005). PRONALEG-GA breeders carry out three trials together with CIAL members before a final variety is selected for release. First, a test plot is planted with trial varieties without the use of external inputs. Second, a Table 2.6: Bean-Oriented CIALs in Northern Ecuador Community Valley Name Date Founded La Concepcion Mira Cuenca del Rio Mira June, 24 2002 Santa Lucia Mira Nueva Esperanza July, 19 2003 El Tambo Chota E1 Progreso del Tambo January, 27 2004 San Clemente Chota La Esperanza de San Clemente Smtember 8, 2004 22 confirmation plot is planted with the best performing varieties from the test plot to verify the initial results. Finally, a production plot is planted and managed according to INIAP crop management recommendations. By 2004, two CIALs had completed the investigative cycle and released IN lAP- bred disease resistant varieties. The varieties La Concepcion (INIAP-424) and Yunguilla (INIAP-414) were released through the CIAL based in La Concepcion. The variety Canario del Chota (INIAP-420) was released through the CIAL in El Tambo. An additional variety, Blanco F anesquero, was developed using participatory methods but in a non-CIAL affiliated community. The variety La Concepcion belongs to the purple mottled market class, Yunguilla to the red mottled market class, Canario del Chota to the yellow market class, and Blanco Fanesquero to the white market class. F armer-selected criteria used to rank the trial varieties included yield, disease resistance, time to harvest, plant uniformity, and tolerance to drought (Mazon and Peralta 2005). All four of these varieties possess intermediate resistance to bean rust. The varieties Yunguilla and Blanco Fanesquero are also resistant to anthracnose, whereas the varieties Concepcion and Canario del Chota are susceptible. 2.5 Chapter Summary Bean production is of central importance to farmers in Ecuador’s northern provinces of Imbabura and Carchi. Over the past two decades, both the total area planted to beans and farm-level bean yields have increased. These increases correspond to the release of disease resistant bean varieties by INIAP. A first set of improved varieties was released during the late 1980’s and early 1990’s, and included four varieties fi'om the red 23 mottled market class. A second set of improved varieties was released beginning in 2004 and included one variety each fi‘om the purple mottled, red mottled, yellow, and white market classes. The adoption of these improved varieties is expected to reduce the unit cost of production through increased yields and reduced firngicide input requirements, with the latter due to in-bred resistance to the major plant diseases bean rust and anthracnose. The remainder of this thesis seeks to determine the economic impact of these improved INIAP varieties. 24 CHAPTER THREE: FIELD DATA COLLECTION 3.1 Introduction This chapter describes the field data collection process—including survey design and implementation, sample selection methodology, survey weighting, and a description of questionnaire contents. The survey design and questionnaire development were developed by an interdisciplinary team of agricultural economists funded by the B/C CRSP at Michigan State University (including the author) and bean breeders fi‘om Ecuador’s national agricultural research institution, INIAP3. Survey planning details were finalized during a week-long joint meeting at INIAP’s Santa Catalina experimental field station in Quito in August, 20064. 3.2 Survey Design and Implementation The survey design serves the dual purpose of I) obtaining regional estimates of varietal adoption by market class, and 2) analyzing the economic impact of improved INIAP varieties. The population of interest is defined as the set of all bean farmers who planted at least one parcel of mono-cropped bush beans (Phaseolus vulgaris L.) during the first production cycle of 2006, and whose farmstead is located 1) within the provinces of Imbabura or Carchi, and 2) within the predefined altitudinal range of 1200 to 2400 3 More specifically, INIAP’S bean breeders belong to INIAP’s Programa Nacional de Leguminosas y Granos Andinos (PRONALEG-GA). 4 All discussion occurred in Spanish. 25 meters above sea level (m.a.s.l .). The majority of this population lives within the Mira and Chota river valleys, which is INIAP’s targeted area of impact for bean research. In total, 132 farmers from 30 communities were intervieweds. Data collection occurred from October through December of 2006 (Table 3.1), with over 70% of the interviews being conducted in November. Two questionnaires were used to collect data with one targeting the village- level information and the other targeting household-level information (contents described below). Each questionnaire underwent two rounds of pre-testing—first in the community of Peruche, located outside of the targeted area of influence, and second in the community of La Concepcion, which was later included in the survey6. All questions pertained to the first production cycle of 2006 (January through March). A team of INIAP investigators served as enumerators. To prepare for this task, enumerators received a half day of training to review the study’s main objectives, research hypotheses, and questionnaire contents. Training also involved sessions on plot measurement and the use of global positioning system (GPS) devices to record latitude, longitude, and altitude coordinates of bean plots. Enumerators generally worked in teams of two, and spent one whole day in each community. In the morning, enumerators conducted the village-level and household interviews. In the aftemoon, they recorded plot measurements and obtained confirmation of the variety planted by observing either saved seed or plants in fields planted with saved seed. 5 Sample selection details are provided in Section 3.3 6 In the case of La Concepcion, the two farmers who participated in the pre-testing session were excluded as possible survey participants during the actual data collection. 26 Table 3.1 Number of Households Interviewed by Date Number of Percent of Total Interview Dates (2006) Interviews Interviews (%) October 2-7 7 5 October 16-21 12 9 October 23-28 7 5 November 6-11 34 26 November 13-18 29 22 November 20-25 23 17 November 27-30 12 9 December 18-23 8 6 Total 132 100 Table 3.2 Number of Households Interviewed by Enumerator Enumerators) $3333.? Z3333??? A 57 43 B 52 39 C 6 5 D 6 5 E 4 2 F 4 3 G 3 3 Total 132 100 At the initiation of each interview, enumerators presented participants with a declaration of consent which solicited their willingness to participate and guaranteed confidentiality and anonymity to their responses7. Village-level questionnaires lasted approximately 20 minutes each while household questionnaires lasted approximately 30 minutes each. The two principal enumerators (denoted A and B) conducted over 80% of all interviews (Table 3.2). The enumerator denoted E consisted of a team of three INIAP 7 The Declaration of Consent forms used in the survey are included in Appendices 3 and 4 along with questionnaire contents. Approval for involving human subjects in the survey was granted on September 15, 2006 by the Institutional Review Board (IRB) under application number X04-142. 27 researchers. The thesis author participated as an enumerator for 6 interviews during a one-week visit to Ecuador in November of 2006. 3.3 Sample Selection Methodology A clustered, double-stratified sample design was implemented following Deaton (1997). Clusters are defined as villages. Village-level clustering provides two important practical advantages over the use of a purely random sample. First, it is cost-effective given the rugged topography of Ecuador’s northern Andean region. Travel from village to village saves time and resources as compared to visiting dispersed households selected at random. Second, it facilitates repeat visits to collect absent information or clarify confusing data. One disadvantage to the clustering method, however, is that it reduces the precision of population parameter estimates. This is because similarities exist among farmers from the same village. For example, they face similar prices, production environments, and other fixed factors such as transportation or infrastructure. As a result, clustered samples are slightly less representative of the overall population than a completely random sample. This weakness is minimized by interviewing a handfiil of farmers fi'om many villages, as opposed to interviewing many farmers from only a few villages. Stratification provides the advantage of specifically targeting sub-groups of particular interest that are relatively rare in the population as a whole, such farmers who belong to a CIAL. A first stratification involved the division of village clusters into three groupings based on the level of prior extension intervention by INIAP directly related to bean production. All villages with prior intervention were automatically included in the 28 survey—including 4 communities with a CIAL and 8 additional communities without a CIAL but with previous INIAP intervention related to bean production. An additional 18 villages were randomly selected fiom an area fi'ame compiled by the survey design teams. The second stratification occurred only within the four CIAL villages and involved a division of farmers based on CIAL membership. Use of a straightforward random sample would most likely not include enough CIAL farmers and their inclusion must be guaranteed through other means. In total, 132 farm households were interviewed from 30 village clusters. In each of the 4 CIAL villages, a total of 7 farmers were interviewed, including three CIAL members and four non-members. In each of the 8 non-CIAL villages with a previous INIAP extension intervention and in each of the 18 villages without a previous INIAP extension intervention 4 farmers were interviewed at random. To ensure randomness within clusters, enumerators developed a list of bean farmers with community leaders. In cases where this was not possible, separate barrios (neighborhoods) were identified within each village and one farmer from each barrio was selected who had no relation to either the community leader or other survey participants. This sample selection process resulted in four stratification levels and the following breakdown in number of interviews: 1) Lev_e_l_l_: l2 CIAL members from CIAL communities 2) L_eg—:_l_2; 16 non-CIAL members from CIAL communities 3) Level 3: 32 non-CIAL members from non-CIAL communities with previous INIAP intervention 8 For details on development of the area frame see Appendix 1. 29 4) Lgelfi 72 non-CIAL members from non-CIAL communities without previous INIAP intervention The process also resulted in a spatial variation of farmers interviewed across both cantones (local political divisions) and watersheds. As for cantones, the largest number of farmers interviewed lived in Mira (50 total interviews), whereas the smallest number of farmers interviewed lived in Ibarra (23 total interviews) (Table 3.3). As for principal watersheds, the majority of farmers interviewed lived in the Chota watershed (74 total interviews) and Mira watershed (54 total interviews) (Table 3.4). A total of 4 farmers were interviewed who did not live within either the Mira or Chota watersheds. Table 3.3: Number of Households Interviewed by Canton C ant 0. n Number of Percent of Total Interviews Interviews Ibarra 23 l 7 Pimamp iro 28 2 l Bolivar 3 l 24 Mira 50 38 Total 132 100 Table 3.4 : Number of Households Interviewed by Watershed Percent of Watershed N“’"”‘?’ of Total Intervzews . Intervzews Mira 54 41 Chota 74 56 Other 4 3 Total 132 100 30 3.4 Survey Weights Since each cluster and stratification level is representative of an unequal number of farm households, the survey design described above results in unequal selection probabilities. In order to estimate descriptive statistics (such as adoption rates) that are representative of the target population and avoid sample bias, each observation must be appropriately weighted. Two sample weights are calculated: the first corrects for the difference in selection probabilities within clusters, while the second corrects for the difference in selection probabilities within stratification levels. To determine the first sample weight suppose that each stratification level S (where S = 1 to 4) represents a separate population, and that the total population within each strata is given by M. We can define the probability of sample selection, 7r,-, for household i from village-level cluster c as, "c It, :— 3.1 , Nsc ( ) where nc is the sample size chosen for cluster c and NSC indicates the total population of cluster c within stratification level S (Deaton 1997). Next we define a sample weight w,- for each household that is equal to the inverse of its sample selection probability multiplied by the number of draws into the sample nc. This is because households with a high probability of sample selection (such as CIAL members) represent only a small fraction of households in the overall population and should receive a lower survey weight—and vice versa for households with a small probability of selection. Each household’s survey weight is then given by, w.- = (new. 1" (3.2) 31 and is approximately equal to the number of households in the population that are represented by the sample household i (Deaton 1997) 9. In addition, the probability weighted mean for each stratification level v,- can be easily estimated as the proportion: (k = w ,w2,...,wn), (3.3) where k represents all clusters within the stratification level of interest. Using the above formulas, we obtain descriptive statistics that are representative only within a given stratification level. To obtain an estimate that is representative of the target population, the second sample weight is used to correct for the difference in selection probabilities within stratification levels. The weighting process here is identical to the process explained above, with the only difference being that the probability of sample selection, 1a, is defined 35, NS 71,- - -1V- 9 (3.4) or the ratio of total households within stratification level, N,, to the total number of households in the overall population, N"). 3.5 Questionnaire Contents Data collection involved the use of two survey instruments—a village-level questionnaire and a household-level questionnaire. The village-level questionnaire was designed to be a formal way of recording similarities and differences between 9 See Table A.2.l of Appendix 2 for sample selection and sample weight calculations by cluster. '0 See Table A.2.2 of Appendix 2 for sample selection and sample weight calculations by stratum. 32 communitiesl I . In total, 30 village-level questionnaires were executed with an exact one- to-one correspondence to the 30 village-level clusters previously discussed. Upon arriving at a selected village for data collection, the enumerator’s first task involved seeking out a community leader in order to present the village-level questionnaire and obtain a list of all bean farmers within the community. Questions focused mainly on demographic data (including the aggregate number of farrnsteads producing beans), availability of public services, existing community organizations, and support fiom outside agencies other than INIAP (either governmental or non-govemmental). Specific to bean production, the questionnaire solicited data on factors that are expected to change between communities but not necessarily within communities—such as wages for agricultural labor, transportation costs to the point of sale, primary input and output markets, and market access. The household-level survey was designed to obtain economic and agronomic data on bean production for use in estimating both adoption rates by market class and the economic impact of improved varieties bred by INIAP”. Whole farm data was collected on the bean varieties planted, land areas, and related yields. More detailed parcel-level production data was also collected with respect to the largest plot of beans planted by the household during the first production cycle of 2006. During each interview, enumerators obtained a surface area measurement of the plot indicated and also received confirmation of the variety planted whenever possible through the inspection of saved seed or a bean field planted with saved seed. Specific production data collected includes varietal choice ” See Appendix 4 for a copy of the village-level questionnaire. '2 See Appendix 3 for a copy of the household-level questionnaire 33 and desirable varietal traits, pest and disease pressure, pesticide use and other crop management practices, and harvest, yield and price data. In addition to bean production data, the household survey also elicited data on farm and household characteristics, such as household size and demographics, primary occupations, age and education of the household head, and poverty levels. Poverty is measured as the number of unsatisfied basic needs (UBNs) for each household”. Questions were also included to measure the impact of pesticide use on human health for use in a cost of illness model, however only 7 of 132 survey respondents indicated having suffered an acute pesticide poisoning episode during the first bean production cycle of 2006. This is most likely due to the fact that the toxicity of pesticides used in bean production appears to be low compared to those used in the production of potatoes and horticultural crops”. '3 The UBN index was chosen since it is one official indicator of poverty used by the Ecuadorian government. See Appendix 5 for a description of this index. " Subia et a1. (2007) provides background information on pesticides used in the production of bush beans and other rotation crops in the Mira and Chota valleys. 34 CHAPTER FOUR: THE DIFFUSION AND ADOPTION OF DISEASE-RESISTANT BEAN VARIETIES IN NORTHERN ECUADOR 4.1 Introduction The economic impact of agricultural research outputs, such as improved crop varieties, is highly dependent on the total land area on which they are cultivated. The adoption of agricultural research outputs, however, often occurs in some areas or among certain populations, but not in or among others. Consequently, knowledge about the extent of diffusion and what factors impel or constrain a farmer’s technology adoption decision is desirable for an economic impact assessment. The objectives of this chapter are 1) to estimate the rate of diffusion of improved varieties across time, and 2) to determine the factors influencing individual farmers’ adoption decisions. For the first objective, both red mottled and purple mottled market classes are considered. For the second, analysis will focus only on red mottled varieties. The terms adoption and diflusion are frequently used in economic literature, and slight variations in their definitions are common. In this chapter, the term adoption refers to an individual’s discrete decision whether to use a given technology at a specific point in time. The term diffusion refers to the level of cumulative adoption across time and space, and is measured as the proportion of total bean land cultivated with the new variety. 35 4.2 Conceptual Framework 4. 2. 1 Technology Diflusion The period of diffusion between when a technology is initially released and when it reaches its maximum cumulative adoption rate typically follows an S-shaped pattern similar to that shown in Figure 4.1. The exact shape of the diffusion curve for particular technologies varies, but each can be characterized by their adoption ceiling and rate of diffusion. The adoption ceiling ( max) expresses the maximum cumulative adoption as a proportion and ranges in value from 0 to l. The rate of adoption (given by the slope of the diffusion curve) expresses the speed with which diffusion occurs and determines how soon the cumulative adoption rate approaches Am”. The observation of an S-shaped curve was first remarked a half century ago by both sociologists and economists in studies on the diffusion of hybrid corn in the central United States (Rogers 1962, Griliches 1957). Sociological explanations of this pattern generally emphasize the role of both awareness and attitude in influencing the diffusion rates of new technologies. Rogers’ (1962) basic observation is that diffusion depends on the flow of information between adopters and potential adopters. These flows are central to generating awareness and allow farmers to formulate their attitudes towards a technology’s performance. Information costs, however, typically increase with distance from the center of diffusion and the decision to adopt for many farmers often does not become optimal until later in time. Economic studies of technology diffusion, on the other hand, typically emphasize the role of incentives and capacity. Incentive is linked to profitability. New technologies 36 Figure 4.1: The S-Shaped Diffusion Curve oooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooo Adoption Ceiling (Am) Cumulative Adoption Time (t) therefore are expected to be adopted at higher rates wherever they are more profitable. Griliches (1957) found that, indeed, both the adoption ceiling and the rate of adoption parameters for hybrid corn varieties were higher in counties where the profit differential between new and old technologies was also higher. Capacity is generally linked to certain characteristics of the farmstead, individual adopters, and economic institutions. For example, farm size, management capacity, risk preferences, wealth, and land tenure have been recognized as crucial to the widespread cumulative adoption of certain technologies (Feder et al. 1985, Sunding and Zilberman 2001). While in some cases these characteristics indicate the capacity for adoption, in others the lack of certain characteristics among non-adopters indicates an adoption constraint. The S-shaped diffusion curve can be split into three conceptual phases: the early adoption phase, the take-off phase, and the saturation phase (Sunding and Zilberman 37 2001). Low levels of aggregate adoption and high rates of marginal adoption characterize the early adoption phase. Farmers adopting in this phase are known as innovators, and often have higher levels of educational attainment, financial resources, social status, and willingness to accept risk. Awareness and attitude play their most important role in this early stage. The takeoff phase, which begins just before the inflection point, represents a period of rapid adoption within a short period of time. Many farmers choose to adopt in this phase only after they observe the technology as profitable. They are characterized as less willing to accept risk than early adopters and may face certain adoption constraints. Finally, the saturation phase occurs as the adoption rate slows and the diffusion curve reaches its adoption ceiling. In the case of Ecuador, the diffusion process for improved bean varieties is assumed to follow a similar S-shaped pattern since the population of bean farmers is relatively large and heterogeneous with respect to their access to information, farm and household characteristics, and production environments. It is also important to note that since the red mottled RVs were released at different points in time, some degree of varietal replacement likely occurred due to farmers switching from older to newer RVs (Maredia et al. 2000). However, the focus of this study is on the cumulative adoption across all red mottled RVs, which as a group are assumed to exhibit a continually increasing rate of diffusion over the period 1986-2006. 4. 2. 2 Technology Adoption For individual farmers, the decision whether or not to adopt a new technology poses as an economic dilemma. Economic theory suggests that farmers seek to maximize 38 their utility subject to various production constraints. For discrete cases of technology adoption, we accordingly expect to observe the adoption of a new technology whenever the utility to be received from adoption exceeds that fi'om non- adoption. The random utility model (RUM) developed by McFadden (1973) provides a useful fi'amework to formalize the discussion of discrete technology adoption decisions. The RUM states that farmers base their adoption decision for a new technology, v, on the unobservable utility function, U; = U —U (4.1) A N ’ * a a a I a where U v represents a latent variable equal to the difference between the utility received with adoption, U A , and the utility received without adoption, U N . Assuming that . . . . . '1' farmers act to maxrmrze utility, we expect adoption to occur whenever U v takes on a positive value, such that (4.2) A =1 whenever ULZO A =0 whenever UL(X;fl) ' 1, (4.8) where A again represents individual adoption decisions by farmers, and is equal to 1 if a farmer adopts and 0 otherwise. Second, the log likelihood function is obtained by taking the natural log of Equation (4.8) and summing across all i to obtain: In 1. = 2(Ai 1n[(X,6)] + (1 — Ai)ln[1— (0(me . (4.9) Probit estimation is then completed by choosing the B so as to maximize the log likelihood fimction given in Equation (4.9) for the given dataset (Myers 2006). 42 To facilitate the interpretation of probit regression results, the marginal effects will be reported alongside normal probit coefficients. Marginal effects show the expected change in the probability of adoption given a one unit change in a particular explanatory variable Xj, holding all other explanatory variables fixed. In the case of binary explanatory variables, marginal effect estimates report the expected change in the probability of adoption given that the value changes from 0 to 1. All marginal effect estimates will be reported at the data means. 4.4 Data Description The data used to estimate both the logistic diffusion curve and the probit model for discrete adoption decisions comes fi‘om the 2006 field survey carried out by the Bean/Cowpea Collaborative Research Support Project (B/C CRSP) and INIAP in northern Ecuador (as discussed in Chapter 3). 4. 4. 1 Difi’usion Data Survey results obtained from the varietal adoption component of the household questionnaire indicate a wide diversity of bean varieties cultivated with respect to both market class and varietal classification (Table 4.1). The red mottled bean market class is the best represented, with 11 different varieties appearing in the sample, including the two most widely observed varieties Paragachi and Injerto. Additional varieties recorded belong to the purple mottled, yellow, solid black, solid red, and pink mottled market classes. 43 Table 4.1: Bean Varieties Planted by Plot and Land Area, Imbabura and Carchi, Ecuador, 2006 Q = 187) Variety Number Percent of Area Percent Variety Name Market Class Grouping of Plots Plots (%) Planted of A rea (n) (ha) (%) Paragachi Red mottled RV-l9908 36 19.3 44.3 19.1 Injerto Red mottled Local 30 16.0 48.0 20.7 Calima Negro Purple mottled Local 21 11.2 29.0 12.5 Selva Red mottled Local 14 7.5 24.5 10.6 Canario del Chota Yellow RV-20005 12 6.4 10.9 4.7 Calima Rojo Red mottled Local 11 5.9 22.8 9.8 Concepcion Purple mottled RV-20005 11 5.9 10.8 4.7 Negro Solid black Local 8 4.3 5.4 2.3 Capuli Solid red Local 8 4.3 9.0 3.9 Je.Ma. Red mottled RV-l9903 7 3.8 7.3 3.1 Cargabello Red mottled RV-19905 5 2.7 4.8 2.1 Imbabello Red mottled RV-l 9905 3 1.6 2.8 1.2 Yunguilla Red mottled RV-20005 2 1.1 1.8 < 1.0 Uribe o Magola Pink mottled Local 2 1.1 1.3 < 1.0 835* Purple mottled RV-20005 2 1.1 3.5 1.5 S23“ Red mottled RV-20005 2 1.1 1.0 < 1.0 Matahambre Yellow Local 2 1.1 1.0 < 1.0 Blanco Fanesquero White RV-ZOOOs 2 1.1 1.0 < 1.0 Blanco de Leche White Local 2 1.1 < 1.0 < 1.0 Toa Red mottled Other I 0.5 < 1.0 < 1.0 S26* Yellow RV-ZOOOs 1 0.5 < 1.0 < 1.0 Rojo Solid red Local 1 0.5 < 1.0 < 1.0 Radical Solid red Local 1 0.5 < 1.0 < 1.0 Mixturiado Various Local 1 0.5 < 1.0 < 1.0 Blanco Belén White Local 1 0.5 < 1.0 < 1.0 Algarrobo Red mottled Local 1 0.5 < 1.0 < 1.0 Total 187 100.0 229.2 100.0 Source: B/C CRSP and INIAP farm-level varietal adoption survey, northern Ecuador, 2006 * - Indicates varieties currently under participatory evaluation All RVs released by INIAP during the 1990’s (grouping “RV-1990s”'7) are represented in the sample, and all belong to the red mottled market class. The majority of local varieties also belong to the red mottled market class. This will allow for an ex-post comparison between INIAP red mottled RVs and local red mottled varieties for the period 1986-2006. All four recently released INIAP purple mottled RVs released through ‘7 Exceptions are the variety Paragachi released in 1996 and the variety Cargabello released in 1987. 44 participatory breeding techniques (grouped as “RV-20005”) are also represented in the sample. Of these, however, only the purple mottled market class provides enough observations of both resistant and local varieties to permit a comparison. Given the recent introduction of the purple mottled RV “La Concepcion” in 2004, we predict ex-ante diffusion rates for the 21-year period 2004-2024. Together, the red mottled and purple mottled market classes account for an estimated 80% of all land cultivated to beans in the Imbabura and Carchi provinces (Table 4.2). Of these two classes, the red mottled varieties dominate with an adoption rate Table 4.2: Estimated Adoption Rates by Market Class in 2006, Imbabura and Carchi, Ecuador Market Estimated Land Area Estimated Class 0 Number of F arm {/0} (ha) Households Red Mottled 68.4% 21,090 2,893 Purple Mottled 12.3% 3,784 519 Yellow 6.0% 1,837 252 Solid Black 5.8% 1,793 246 Solid Red 5.8% 1,790 246 White < 1.0% < 300 < 50 Pink Mottled < 1.0% < 300 < 50 Other/mixed < 1.0% < 300 < 50 Total 100.0% 30,816 4,227 Notes: Estimated land area proportions are calculated using the survey weighting method described in Section 3.4. Estimated land area is from 2005 (SICA 2007); Estimated total number of households is from 2001 (INEC 2001). 45 Table 4.3: Estimated Adoption Rates of Disease-Resistant Varieties as a Proportion of Land Area Cultivated, Imbabura and Carchi, Ecuador, 2006 Variety Market Class Growing Red Mottled Pumle Mottled Local 54% 84% RV-l990s 45% --- RV-ZOOOS < 1% 10% Other < 1% 6% Total 100% 100% of 66% and an estimated 21,000 hectares planted. Purple mottled varieties account for just over 12% of all land cultivated and an estimated 3,700 hectares of production. Yellow, solid black, and solid red varieties are also found on over 5% of cultivated land each. White and pink mottled varieties are planted on less than 1% of all land. RV adoption rates vary for the red mottled and purple mottled market classes (Table 4.3). Within the red mottled market class, 44.9% of all land was cultivated to RVs—which, again, pertain exclusively to the set of INIAP varieties released during the 19903. Within the purple mottled class, 9.82% of all land was cultivated with an improved variety—which, in this case, refers to the variety “La Concepcion” released by INIAP in 2004 under a participatory varietal selection process. It is important to note that the red mottled RVs are in the saturation phase of their diffusion process, while the purple mottled RV is still in the early adoption phase. 4. 4. 2 Adoption Data Specific variables to be included in the adoption analysis of red mottled RVs are shown in Table 4.4. Price variables include the price of bean seed, the market price 46 Table 4.4: Descriptive Statistics of Variables Included in the Adoption Model for Disease- Resistant Red Mottled Varieties, Imbabura and Carchi, Ecuador, 2006 (n=82) Variable Description Mean Std. Dev. Minimum Maximum Dependent Variable Adoption of an improved variety (1=Yes) 0.49 0.50 0 1 Price Variables Bean Seed Price ($/kg) 0.94 0.30 0.40 1.76 Bean Sale Price ($/kg) 0.64 0.18 0.28 1.10 Cost of transport to market ($/ 100 lb. sack) 0.39 0.42 0.00 1.50 Family and Household Characteristics Age of household head (years) 47.9 13.6 21 76 Education of household had (years) 5.1 2.5 0 16 Number of working household members 2.6 1.7 0 9 Attended a pest management seminar (1=Yes) 0.30 0.46 0 1 Received remittances (1=Yes) 0.20 0.40 0 1 Proportion of Harvest Sold 0.89 0.19 0.00 1.00 Unsatisfied Basic Needs Index (# UBN) 0.7 0.9 0.0 3.0 F arm Characteristics Agricultural land owned (ha) 3.18 2.49 0.25 12.00 Agricultural land planted to beans (ha) 1.93 1.79 0.25 10.00 Received credit for bean production (1=Yes) 0.20 0.40 0 1 Sharecropped land (1=Yes) 0.48 0.50 0 1 Fixed Factors Altitude (m) 2024 329 1306 2583 Time of transport to market (hours) 0.49 0.60 0 3 Located in the Chota Valley (1=Yes) 0.67 0.47 0 1 Village w/out prior INIAP intervention (1=Yes) 0.57 0.50 0 l Source: B/C CRSP and INIAP field survey, Ecuador, 2006 received at the time of harvest, and an additional variable measuring the cost to transport to market. This last variable is included to capture the effective prices paid by farmers (Sadoulet and de Janvry 1995). Household characteristics include the age and education of the household head, the number of working adults, a dummy variable indicating whether the household received financial remittances, a dummy variable indicating whether the household head attended a pest management seminar, the proportion of harvest sold, and the unsatisfied basic needs (UBN) index. Age, education, and attendance at a pest management seminar serve to proxy human capital and management 47 ability. The number of working adults is a proxy for labor availability. The remittance variable is used to proxy financial capital. The proportion of harvest sold indicates the degree of subsistence. The UBN poverty index is included as a proxy for wealth”. Farm characteristics include the total agricultural land area owned by the household, the total area planted to beans, whether the household received credit for bean production during the 2006 cycle, and whether the household utilizes a sharecropping arrangement. Land serves as an indicator of wealth, total area planted to beans serves as an indicator of specialization in bean production, and access to credit is a proxy for financial capital. Sharecropping is a management practice known to reduce input provision. Fixed factors include altitude, the time required to travel to market, whether the household is located in the Chota Valley, and the level of extension intervention by INIAP. Altitude serves as a proxy for both disease pressure and precipitation. The required time to travel to market captures access to information, under the assumption that information costs increase with distance. A binary variable for farmers living within the Chota river valley is included so as to control for unobserved differences in bean production between the two zones. Mean difference tests of the summary statistics between farmers living in the Mira and Chota river valleys reveal that, on average, farmers in the Chota valley are younger, have smaller plot sizes, sell a much higher share of total production, had higher costs of transporting their product to the market, and also received higher bean prices. Finally, the dummy variable for farmers living in villages without previous extension intervention by INIAP is included to control for awareness of improved varieties. ‘8 See Appendix 5 for details on the construction of the Unsatisfied Basic Needs (UBN) Index. 48 4.5 Results and Discussion 4. 5 . 1 Estimated Rates of Difi‘usion A logistic function is estimated for RV adoption within both the red mottled and purple mottled market classes. In the case of red mottled varieties, an ex-post diffusion curve is estimated for the years 1986 through 2006. In the case of purple mottled varieties, an ex-ante diffusion curve is estimated for the years 2004 through 2024. The adoption ceiling, or maximum cumulative adoption rate, for the set of red mottled RVs is 45%, equal to the 2006 adoption rate shown in Table 4.2. Two observations support this assumption. First, the diffusion of red mottled RVs is assumed to be in the saturation phase of adoption. A 10-year window exists between the release of the last of these varieties (Je.Ma. in 1996) and data collection and it is likely that non- adopters are classified as such by choice rather than by a lack of awareness. Second, INIAP began releasing a new set of RVs in 2004. A process of varietal replacement will likely occur and result in a decrease in the cumulative adoption level of red mottled RVs released during the 19905. Two data points are used to estimate values for the two logistic function parameters a and B. First, the 2006 adoption rate of 45% for the red mottled market class is used. Second, we assume that in 1986, the first year a red mottled RV was released, the adoption rate was near 1%. Using this data, the estimated parameter values are a = -795 and B = 0.398 and the logistic fimction is as presented in Figure 4.2. The adoption ceiling for purple mottled RVs is also assumed to be 45%. In lieu of better data on which to base the estimate, this assumption is made using the history of red 49 mottled RVs. As more data becomes available in the future, it may be necessary to change this assumption. Two data points are also used to estimate values the value of a and B. In 2006, we found an adoption rate for RVs from the purple mottled market class of 10%. Second, we assume that in 2004, the first year of release, the adoption rate was near 1%. Using this data, the estimated parameter values are 01 = -25200 and B = 1.25 and the diffusion curve is as presented in Figure 4.3. A comparison of Figures 4.2 and 4.3 shows faster predicted rate of adoption for purple mottled RVs than for red mottled RVs. The method used to estimate the diffusion rate of red mottled varieties, however, most likely overestimates the time needed for the cumulative adoption rate to reach Ama" in 2006—at the time of the survey. Thus, this assumption provides an underestimate of the cumulative adoption rate in previous years and Figure 4.2 then provides a lower bound estimate of diffusion rates for the red mottled varieties. It may not be the case however that the cumulative adoption of red mottled varieties reached Am“ as quickly as predicted for the purple mottled varieties. The diffusion of purple mottled varieties may enjoy a faster rate of adoption and possibly even a higher adoption ceiling than red mottled varieties. First, the popularity of red mottled RVs may lead to faster adoption of newer generations of RVs. Second, the participatory nature of INIAP’s bean improvement program and/0r improved communication networks and other similar factors may accelerate awareness compared to two decades ago when the previous RVs were released. 50 Figure 4.2: Estimated Diffusion of Improved Red Motfled Beans, Provinces of Imbabura and Carchi, Ecuador, 1986-2006 0.50 0.40 / c .. s ‘ / 3 <3 0.30 '0 5 / < a 0 O 5.2 020 3 5i. / g 0.10 / 0.00 *1 F l I M r I I I I I l I l T—T 1 1 l 1 I ’\ Qt '\ ’b ‘3 ’\ 0.) '\ 'b ’\ .999 .99 .99 .9. s9 .99 (ea (or ,eo ,c .99 ,6» Year Figure 4.3: Expected Diffusion of Improved Purple Mottle Beans, Imbabura and Carchi, Ecuador, 2004-2024 0.50 § 0.40 ’ c a 1: '3 4 [ .3 5 0.30 < a O 0 5 u- 5 g 0.20 3 s a 3 s 2 0.10 e 0.00 M I l l 1 1 l I I l I l I T T j r T 1 *1 I '19 09° 080 ’L 1» '» '19 99 119 '19 51 4. 5.2 Factors Influencing Individual Adoption Decisions Moving from analysis of cumulative adoption rates to analysis of individual adoption decisions, probit results of the factors influencing adoption decisions for red mottled RVs indicates a high percentage (71%) of correctly predicted observed adoption decisions (Table 4.5). Across variable categories, price variables and fixed factor variables play the most significant role in influencing farmers’ RV adoption decisions. Statistically significant price variables include the price of bean seed (significant at a .01 level) and the cost of transport to market (significant at a .05 level). Other things equal, an increase in the price of bean seed by one dollar per kilogram reduces the probability of adoption by 77%. Similarly, an increase in the cost of transportation of a 100 pound sack of beans by one dollar decreases the probability of adoption by 63%. While these marginal effects appear large, a one-dollar per kilogram increase is also large relative to the sample means of bean seed price and transport costs of $0.94/kg and $0.39/kg, respectively. Statistically significant fixed factor variables include the dummy variable for farmers living in the Chota watershed (significant at a .01 level), altitude (significant at a .05 level), and the time of transport to market (also significant at a .05 level). A farmer living in the Chota watershed is 58% more likely to adopt a red mottled RV than a farmer living in the Mira watershed. An increase in the altitude by 100 meters reduces the probability of adoption by 6%. Contrary to expectations, an increase in travel time by one hour increases the probability of adoption by 41 %. Finally, the household characteristic dummy variable indicating that a family received remittances is negative and significant 52 at a .10 level. This implies that, other things equal, a family receiving remittances is 29% less likely to adopt RVs. The coefficients on the price variables and altitude have the expected sign. The negative sign on the remittances variable indicates that households with access to financial capital are less likely to adopt RVs. This may indicate that wealthier households Table 4.5: Probit Model Results for Factors Influencing the Adoption of Disease- Resistant Red Mottled Varieties, Imbabura and Carchi, Ecuador, 2006 (N=82) Probit Marginal Explanatory Variables Coefficient Effect P-Value Price of seed ($/kg) -1.9473 -0.7698 ***0.010 Price of beans ($/kg) -0.0961 -0.0380 0.924 Cost of transport to market (S/qq) -l .5848 -0.6264 “0.031 Age (years) 0.0008 0.0003 0.954 Education (years) -0.0816 -0.0322 0.268 Pest management seminar (1=Yes) 0.0319 0.0126 0.941 Number of working adults -0.0308 -0.0122 0.764 Received remittances (1=Yes) -0.8057 -0.2921 *0.086 Percent of harvest sold -1.l344 -0.4484 0.314 Poverty measure (# of UBN) -0.1703 -0.0673 0.445 Agricultural land (ha) 0.2053 0.081 1 0.179 Land planted to bean (ha) -0.2118 -0.0837 0.311 Received credit (1=Yes) 0.1677 0.0666 0.759 Partidario (1=Yes) 0.4273 0.1679 0.326 Altitude (100 meters) -0.0015 -0.0601 “0.028 Time of transport to market (hours) 1.0258 0.4055 "0.025 Chota valley (1=Yes) 1.7596 0.5773 ***0.002 Village w/out prior INIAP intervention 0.2326 0.0915 0.564 Log likelihood -39.26 LR Chi2 35.10 Prob > chi2 0.00 % Correctly Predicted 71% "* = Significant at a=0.01 ** = Significant at a=0.05 "' = Significant at a=0.10 53 invest in areas other than bean production, or that receipt of remittances indicates households with limited production resources. The negative sign on travel time, however, is contrary to expectations. Since the cost of transport to market is also included in this analysis, the positive sign may indicate that better transportation infi'astructure leads to increased ad0ption. The reasoning here is that a longer trip along a route with improved infrastructure would cost the same as a short trip along a route with unimproved infiastructure. Since improved infrastructure lowers communications costs, the negative sign may indeed be in accordance with diffusion theories. 4.6 Chapter Summary The diffusion analysis provides data on the estimated cumulative adoption rate of red mottled RVs from 1986-2006 and predicted cumulative adoption rates for recently release purple mottled RVs for the period 2004-2024. Comparison of the diffusion rates between the red mottled and purple mottled RVs shows a faster rate of adoption for purple mottled varieties. This finding is partially by construction, because red mottled estimates represent a lower bound on estimates for purple mottled varieties. The diffusion data presented here will be used as key parameters in the economic surplus analysis in Chapter 6. The adoption analysis indicated that price variables and fixed factor variables influence RV adoption decisions the most. For the most part, results followed expectations. Low seed prices and low transportation costs (which reflect market access) acted as incentives to adoption. As for the fixed factors, altitude (which is correlated with 54 rainfall and disease pressure) acted as a constraint to adoption, while unobserved factors present among producers located in the Chota valley increased the probability of adoption. 55 CHAPTER FIVE: ECONOMETRIC MEASUREMENT OF INCREMENTAL FARM-LEVEL BENEFITS 5.1 Introduction Disease-resistant varieties (RVs) have the potential to provide an array of farm- level benefits to adopters such as higher yields and lower input requirements, which can jointly reduce unit costs of production”. Ecuador’s national agricultural research institute, the Instituto Nacional Autonomo de Investigaciones Agropecuarias (INIAP) began releasing a series of improved red mottled bean varieties during the 1980’s in the two northern provinces of Imbabura and Carchi as part of it bean improvement program. Each of these varieties is resistant to a specific plant disease that is prevalent in the area as reported in Table 2.5. The objective of this chapter is to estimate the farm-level impact of RV adoption on bean yield, pesticide use, and the unit cost of production. Empirical measures of these impacts are obtained using multiple regression treatment effect models. This knowledge is important for agricultural researchers within Ecuador, as well as to other decision makers who are seeking information about estimating the economic impact of disease- resistant varietal technologies. In addition, these measures can be used as parameters in an overall economic impact assessment of INIAP’s efforts to breed for disease resistance. 5.2 Research Objectives Breeding for disease-resistance seeks to enhance productivity through strict yield- gains, yield-losses avoided (Morris and Heisey, 2003), or by embedding damage '9 In this chapter, unit cost is defined as the cost per unit of output (i.e. $/kg). 56 abatement services into improved seed which can then substitute for chemical fungicide inputs (Lichtenberg and Zilberman 1986). In this chapter, three hypotheses concerning RV adoption are tested using multiple regression techniques: 0 H 11 Farmers who plant RVs have higher yields per hectare compared to farmers who plant local varieties when faced with disease pressure. 0 H2: Farmers who plant RVs apply less pesticide active ingredients per hectare compared to farmers who plant local varieties. This hypothesis is tested for both fungicides and insecticides. 0 H3: Farmers who plant RVs have a lower unit variable cost of production than do farmers who plant local varieties”. 5.3 Conceptual Framework The empirical estimation of these hypotheses requires the appropriate identification of the with- and without-research scenarios, also known as the counterfactual (Morris and Heisey 2003). However, we are only able to observe one of these outcomes for each farmer. For adopters, the challenge lies in determining what their production outcomes would have been without RVs. For non-adopters, we face the opposite challenge of determining what their outcomes would have been with RVs. In practice, a number of approaches to formulating the counterfactual are possible. One possibility is to compare production outcomes for the same farmers both before and after adoption. This requires that panel data, but most adoption surveys (including this one) collect data for a single year and thereby exclude this possibility. In 2° Unit variable cost (UVC) is defined as those costs expected to vary upon the adoption of a resistant variety. A more formal definition is provided in Section 5.5. 57 addition, the before-and-after approach assumes that the counterfactual scenario remains constant, which is rarely the case. A second possible approach is to compare the production outcomes of adopters to those of non-adopters. This approach is common within the growing body of literature on the economic impact of input-saving biotechnologies. A number of studies compare the yield and input-saving impacts using only a simple mean comparison approach (Bennett et al. 2004, Brookes 2005, Qairn and Traxler 2005). This may lead to biased conclusions, however, since the observed differences in production outcomes between adopters and non-adopters is generally not a result of the improved variety alone. Many secondary factors such as plot characteristics, crop management techniques, and even certain household characteristics (such as management ability) are also often correlated with observed crop yields, input use, and unit costs. The failure to control for these factors leaves open the possibility of attributing observed differences in farm-level benefits to the improved technology rather than its true source. To avoid these criticisms, this paper employs the use of multiple regression treatment eflect models (TEMs), which allow for production outcomes between adopters and non-adopters to be compared conditional upon a set of explanatory covariates (Wooldridge 2002). This methodology draws from the literature on program evaluation whose focus is on identifying the impact of economic or social programs, called treatments, on program participants, called the treated. In our case, the treatment of interest is the adoption of an INIAP RV and the impacts of interest are the effect of this RV adoption on yield, input-savings, and the unit cost of production. An empirical 58 measure of each impact is given by the treatment effect, which is the partial effect of the binary treatment variable on a dependent variable. Treatment effect models provide unbiased results provided that two assumptions hold (Wooldridge 2002). The first is called the stable unit treatment value assumption. It states that the treatment of one observation does not affect another’s outcome. Given that the technology of interest is embedded in self-pollinating bean seed, this assumption clearly holds”. The second is called the ignorability of treatment. It implies that selection of those who receive treatment does not occur based on a set of observable characteristics. In cases where selection on observables does occur, it can be controlled for by including those characteristics that partially determine selection into the regression analysis as covariates”. Applications of TEMs to agriculture are limited. Godtland et al. (2004) uses a TEM to isolate the impact of participation in farmer field schools on farmer knowledge about integrated pest management practices. More specific to input-saving agricultural technology, F emandez-Comejo et al. (2002) used a TEM in their study of on-farm impacts of herbicide-tolerant soybean varieties on yield and herbicide demand. Qaim and de Janvry (2005) use a treatment effect model (although not stated as such) to estimate the farm-level economic impact of Bt cotton adoption yield and pesticide use. More recently, Gardner and Nelson (2007) use a TEM to estimate labor savings in US. agriculture that have resulted from the adoption of genetically modified crops. 2' An example of when this assumption may not hold is with knowledge-intensive technologies, such as integrated pest management, where diffusion is likely to occur between the treated and un-treated. 22 The selection on observables can be examined using a mean comparison test of descriptive statistics between the treated and tm-treated (Godtland et al. 2004). In this case, differences between adopters and non-adopters are explored in Section 5.5. 59 A further issue in using treatment effect models to compare production outcomes is an assumption that no self-selection occurs among farmers. Previous economic impact assessments of disease-resistant bean breeding did control for this possibility (Mather 2005). This correction was necessary given that the farm-level survey included areas both with and without disease pressure. Given individual farmers’ knowledge of disease infestation probabilities in their own fields, adoption occurred only in areas with high disease-pressure. Thus, a comparison of adopters to non-adopters would not provide a good estimate of what the production outcome of adopters would have been in absence of the RVs. Data fi'om the B/C CRSP household survey in Ecuador however includes only farmers living in disease-prone areas and will not result in a self-selection process similar to that described in by Mather (2005). 5.4 Conceptual Models Before presenting empirical tests for each of the three stated hypotheses, this section introduces the three conceptual models to be used in evaluating the farm-level impact of improved varieties: a crop yield model, a pesticide demand model, and a unit cost model. For the pesticide demand model, two separate functions are specified, one for fiingicides and the other for insecticides. 5. 4.1 Crop Yield Model In order to form a priori expectations about the farm-level impact of resistant varieties on crop yields, it is important to understand the economic motivation behind their development. The breeding of crop varieties for disease resistance is generally 60 undertaken in order to maintain current yield levels in the face of disease pressure (Smale et a1. 1998; Morris et al. 1994). Thus, our expectation regarding the impact of RV adoption on yield is positive when compared to susceptible varieties in the presence of disease pressure. When disease pressure is absent, we expect there to be no yield differential between the resistant and susceptible varieties. A crop yield model is used to test this hypothesis. The crop yield model is stated as a production function where Y represents the maximum obtainable yield for a set of fixed inputs, X, that is conditioned by certain covariates such that, Y = f(X|A,D,L,F,Z) (5.1) where A represents the RV adoption decision, D represents pest and disease pressure, L plot characteristics, F farm household characteristics, and Z represents community-level fixed factors. The Z variables are included to control for unobserved variables that differ between communities but remain constant within communities. 5. 4. 2 Input Demand Functions In addition to yield gains, the disease resistance embodied in the improved seed provides damage abatement services that can be substituted for fungicide inputs in a biological production function (Lichtenberg and Zilberman 1986, Mather 2005). This suggests that RV adoption should reduce the quantity of fungicides used. Since farmers in northern Ecuador typically apply firngicides and insecticides jointly in a single application, there may also be an indirect impact of RV adoption on insecticide use. Two input demand functions are developed to test these hypotheses, one for fungicides and the other for insecticides. These models state the quantity of pesticide 61 active ingredient (AI) demanded, X, as a firnction of price variables, P, and conditioned by a set of covariates such that, X = f(P|A,D,L,F,Z) (5.2) where A, D, L, F, and Z are identical to those described in the yield model. 5. 4.3 Unit Cost Function An economic measure of the farm-level impact of improved varieties can be obtained using a unit cost fimction (Alston et al. 1998). This approach is advantageous, since it reflects the economic benefits of both yield-enhancing and input-saving components of a new technology. While the yield and pesticide demand firnctions measure changes in the agronomic quantities of yield, seed, and pesticides associated with the adoption, they do not reflect differences in the prices paid for these products between adopters and non-adopters. A farmer may obtain a higher yield upon adopting an improved variety, but if his or her variable production costs also increase then the final economic benefits realized through adoption may not be as large as expected. Assuming farmers act to maximize profits, the unit cost function describes the average cost of producing one unit of product (C) for a given set of input prices (W) and level of output (Q) and covariates such that, C = f(W,Q|A,D,L,F,Z) (5.3) where A, D, L, F, and Z are as previously defined. 62 5.5 Data Description Data comes from the 2006 B/C CRSP and IN IAP field survey conducted in the provinces of Imbabura and Carchi in northern Ecuador in 2006 (see chapter 3). A total of 132 farmers fi'om 30 communities were surveyed, with results indicating that 82 farmers planted a red mottled variety on their largest bean plot whereas 23 planted a purple mottled variety. The other 27 farmers planted varieties from either the yellow, white, solid red, solid black, or pink mottled market class. In addition, 12 farmers planting a red mottled variety could not provide complete information on specific pesticides used. For the purposes of the regression analysis to follow, only those using a red mottled variety with complete data on all pesticide applications are considered. A description of variables to be included in the regression analysis and their units of measurement, along with descriptive statistics are provided in Table 5.1. Bean yield (Y) is measured in terms of kilograms of beans produced per hectare and serves as the dependent variable for the yield impact model. The per-hectare yields reported here appear much higher than that presented in Figure 2.3. This is due to the restriction of analysis to red mottled varieties”. Production inputs (X) include the quantity of insecticide, fungicide and foliar fertilizer active ingredient (AI) used per hectare and seed input. The same insecticide and fungicide values serve as dependent variables in the two input demand functions“. 23 It is important to note that two of the top three observations with respect to per-hectare yields (i.e. > 3000 kg/ha) pertain to farmers who planted relatively small plots (i.e. <0.25 ha). The conversion of these figures to ones representative of a per-hectare basis results in higher per hectare yield values than might be normally expected. This helps explain the egregiously large maximum values for per hectare yields. 2" Fungicides, insecticides and foliar fertilizers used in northern Ecuador come in liquid and powder forms resulting in various units of measurement A conversion to kilograms of AI assumes 1 co = 1 ml = lg. 63 Table 5.1: Summary Statistics of Variables to be Included in the Regression Analysis for Red Mottled Varieties, Imbabura and Carchi, EcuadorL2006 Qi=73) Variable Category: Mean Std Dev Min Max Dependent Variables: Yield (kg/ha) 1526 906 239 4915 Fungicide Al (kg/ha) 15.8 34.0 0.1 187.5 lnsecticida AI (kg/ha) 6.1 14.0 0.1 96.1 Unit variable cost ($/kg) 0.16 0.13 0.03 0.86 Treatment Eflect Variables: Adopted improved variety (1=yes) 0.60 0.49 0 1 High disease pressure (1=yes) 0.29 0.46 0 1 High pest pressure (1=yes) 0.23 0.43 0 1 Production Inputs: Fungicide Al (kg/ha) 15.8 34.0 0.1 187.5 Insecticida Al (kg/ha) 6.1 14.0 0.1 96.1 Foliar fertilizer AI (kg/ha) 12.2 25.4 0 153.39 Seed (kg/ha) 112.8 44.7 53.1 294.9 Plot Characteristics: Plot size (ha) 1.04 1.00 0.23 7.00 Altitude (m.a.s.l.) 2034 323 1306 2583 Loam soil (1=yes) 0.42 0.50 0 l Irrigated plot (1=yes) 0.95 0.23 0 1 Plot prev. cropped w/ beans (1=ye5) 0.30 0.46 0 l Sharecropped plot (1=yes) 0.40 0.49 0 l Rented plot (1=yes) 0.08 0.28 0 1 Household Variables: Age (years) 46.4 13.8 21 76 Attended pest man. seminar (1=ye5) 0.33 0.47 0 l Symptom-based pest man. (1=yes) 0.16 0.37 0 1 Poor household (1 or more UBN) (1=yes) 0.42 0.50 0 1 Price Variables: Market price for beans ($/kg) 0.64 0.18 0.28 1.10 Cost of transport to point of sale ($/qcD 0.34 0.36 0.00 1.00 Avg. price of fungicide ($/kg A1) 0.75 0.37 0.30 2.00 Avg. price of insecticide ($/kg A1) 1.35 0.37 0.80 3.40 Seed price ($/kg) 0.89 0.27 0.40 1.76 Community-Level Variables: Chota valley (1=yes) 0.62 0.49 0 1 Prev. extension intervention (1=yes) 0.45 0.50 0 l Source: 2006 B/C CRSP and INIAP field survey 64 Table 5.2: Sample Means of Explanatory Variables by Adoption Status, Red Mottled Varieties, Imbabura and Carchi, Ecuador, 2006 (n=73) Variable C ategorv: A dopters (n =42) Non-A dopter (n =3 11 P- Value Mean Std Dev Mean Std Dev a/ Dependent Variables: Yield (kg/ha) 1554 776 1372 1060 0.36 Fungicide Al (kg/ha) 11.3 16.0 17.9 46.1 0.16 Insecticide Al (kg/ha) 3.60 3.79 7.61 19.91 0.34 Unit variable cost ($/kg) 0.15 0.13 0.17 0.12 0.53 Treatment Eflect Variables: High disease pressure (1=yes) 0.26 0.44 0.40 0.50 0.17 High pest pressure (1=ye5) 0.24 0.43 0.20 0.41 0.66 Production Inputs: Fungicide AI (kg/ha) l 1.3 16.0 17.9 46.1 0.35 Insecticide Al (kg/ha) 3.60 3.79 7.61 19.91 0.16 Foliar fertilizer A1 (kg/ha) 0.08 0.11 0.08 0.14 0.34 Seed (kg/ha) 112.09 26.95 114.67 63.41 0.80 Plot Characteristics: Plot size (ha) 1.00 1.16 1.05 0.74 0.81 Altitude (m.a.s.l.) 1986 307 2091 351 0.15 Loam soil (1=yes) 0.36 0.48 0.63 0.49 "*0.01 Irrigated plot (1=yes) 0.94 0.24 0.91 0.28 0.65 Plot prev. cropped w/ beans (1=yes) 0.34 0.48 0.23 0.43 0.27 Sharecropped plot (1=yes) 0.40 0.49 0.34 0.48 0.60 Rented plot (1=yes) 0.57 0.53 0.58 0.49 0.93 Household Variables: Age (years) 45.0 12.8 50.0 14.1 *0.10 Attended pest man. seminar (1=yes) 0.34 0.48 0.32 0.47 0.80 Symptom-based pest management (1=yes) 0.12 0.33 0.26 0.44 0.11 Poor household (at least 1 UBN) (1=yes) 0.44 0.50 0.43 0.50 0.92 Price Variables: Market price for beans ($/kg) 0.66 0.19 0.59 0.13 I""‘0.05 Cost of transport to point of sale (S/QCI) 0.38 0.36 0.34 0.42 0.60 Avg. price of fungicide ($/kg A1) 0.69 0.30 1.00 1.04 ‘0.07 Avg. price of insecticide ($/kg A1) 1.37 0.43 1.31 0.27 0.54 Seed price (S/kg) 0.84 0.22 1.06 0.35 "*0.00 Community-Level Variables: Chota valley (1=yes) 0.76 0.43 0.40 0.50 "*0.00 Prev. extension intervention (1=yes) 0.40 0.49 0.51 0.51 0.30 a/ P—value is for a mean-difference t-test between adopters and non-adopters assuming equal variances "* = significant at a 1% level; ** = significant at a 5% level; * = significant at a 10% level Source: 2006 B/C CRSP and INIAP field survey 65 Variables used to measure the treatment effect of improved varieties include binary indicators representing the RV adoption decision (A) and disease and pest pressure (D). For each model, A=I indicates adoption occurred and A =0 indicates non-adoption. Survey results indicate that 44 of 73 (60%) farmers planting a red mottled variety had adopted an RV. To corroborate that farmer’s correctly identified the variety planted, enumerators obtained confirmation through either viewing saved seed or visiting a bean plot planted from saved seed. To measure D, two binary indicators are used to compare plot-level pest and disease pressure experienced by farmers when compared to an average year. In both cases, D=1 indicates higher than average disease (pest) pressure and D=0 indicates either normal or lower levels of disease (pest) pressure. Data collection for the construction of D included a question in the survey that asked farmers to relate the pest pressure they observed in the first production cycle in 2006 to the average level of pest pressure observed in previous years. Explanatory variables included as physical plot characteristics (L) include plot size, altitude, binary indicators for a loam soil texture and access to irrigation, and another two binary variables indicating whether the plot was either Sharecropped or rented versus the base case of ownership, and finally, whether the plot was previously cropped to beans (bean diseases often survive in the soil and thus affect yields). To analyze potential violations of the selection on observables assumption, Table 5.2 presents p—values for an equality of mean t-test between adopters and non-adopters. For plot characteristics, results indicate the only significant difference between adopters and 66 non-adopters is that a greater proportion of non-adopters who have plots with a loam soil texture versus either clay or sandy soils. Price variables (P) include the market price of beans, the cost of transport to market, the average price of fungicides and insecticide, and the price of bean seed. The cost of transport to point of sale helps capture the effective price received by producers (Sadoulet and de Janvry 1995). T-test results indicate a difference in the mean prices received by adopters and non-adopters both in the market price and seed price. Adopters received higher market prices and paid lower seed prices. A similar difference was found for the price of fimgicide inputs, with non-adopters paying higher prices (Table 5.2). Household characteristics (F) include the age of the household head, whether the household head had previously attended a pest management seminar, whether the household follows a symptom-based pest management strategy compared to the base case scenarios of relying on a calendar-spray pest management strategy, and whether the household is categorized as poor, according to the UBN index”. The UBN poverty index is used to proxy wealth, with “poor” households categorized as such if they have one or more UBN. Finally, community-level variables (Z) include two binary variables indicating, first, whether the community is located in the Chota Valley versus the base case of living in the Mira Valley, and second, whether the community received previous extension intervention by INIAP related to bean production versus having received no previous extension intervention. Of all the household characteristics, t-test results indicate a difference between adopters and non-adopters only for the mean age of farmers, with adopters being significantly younger than non—adopters (Table 5.2). 25 UBN index is discussed in Chapter 2 and in Appendix 5. 67 For the unit cost model, the dependent variable is defined as total production costs that are expected to vary upon adoption of an improved variety divided by total output. For each individual farmer, i, unit cost values are constructed using the following formula, UVC. =w 1 t Y: (5.4) where x represents the set of production inputs j that are expected to vary between adopters and non-adopters, w represents input prices, and Y' represents total output. In this case, three production inputs are expected to vary depending on a farmer’s choice of varietal technology: seed inputs, fungicide use, and insecticide use. Production inputs that are not expected to vary between adopters and non-adopters, such as [and preparation, fertilizers, and labor inputs”, are assumed to be fixed. This definition of unit variable cost (UVC) is used for the remainder of this paper. 5.6 Empirical Models and Testable Hypotheses Empirical treatment effect models define two possible outcomes for each observation: the outcome without treatment, yo, and the outcome with treatment, y]. Together with the farmers’ adoption decision, A, the observed outcome for each observation is, y = (1 - A)y0 + A(y1) (5.5) 26 While RV adoption may impact labor costs for pesticide applications, a mean difference test did not reveal a statistically significant difference in the number of applications between adopters and non-adopters (p=0.42). Thus, this model accounts for changes in the quantity of pesticide active ingredient applied but not for a reduction in the number of total applications. 68 Allowing yo and y, to become random variables and assuming error terms have a mean of zero, the outcome can be modeled as a function of treatment status and set of covariates, X, such that, E(y|A,X) =a0 +alA + xp (5.6) where a; is the treatment effect of adoption on the outcome y. In the cases where the impact of adoption on a production outcome is assumed to also be dependent on disease pressure, the appropriate covariate and interaction term with adoption can be included. 5.6.1 Crop Yield Model The yield model is specified empirically as a quadratic function, following Qaim and de Janvry (2005) and Femandez-Cornejo et al. (2002), as well as a log—log function, based on a model specification test”. Using the notation provided in the data description above, it follows that, 27 The quadratic specification was selected based on a review of economic literature on farm-level impact analyses of varietal technologies with input-saving characteristics. However, a MacKinnon, White, and Davidson (MWD) test (Gujarati 2003, p. 280) rejected the null hypothesis of a linear model (p=0.04) but failed to reject the alternative hypothesis of a log-log model (p=0.38). Empirical results for both models will be presented in Section 5.7.1. 69 — a: Y—ao +alA+a2D1+a3A Dl +oz4D2 + 2 (5.7) +X,B1 +X ,82 +L7+F¢+ Z€+8 where, Y = Per hectare bean yield, A = Adoption of improved crop variety (binary; 1=Yes), D1 = Above average disease pressure (binary; 1=Yes), D2 = Above average pest pressure (binary; 1=Yes), A *D, = Interaction term between A and D1 (binary; A=1, D,=1), a = Parameter coefficients on the intercept, adoption, and disease pressure variables, X, L, F, and Z = Matrices of regression covariates as previously defined, fl, 7, (D, and t9 = Vectors of coefficients on X, L, M, F, and Z, and 8 = A normally distributed error term The treatment effect of adoption on bean yield can now be elicited through interpretation of the appropriate parameter coefficients. The parameter coefficients on the binary variables representing RV adoption, disease pressure, and pest pressure serve as differential intercept coefficients such that, a] = differential intercept effect of adoption on average bean yield without disease pressure (12 = differential intercept effect of disease pressure on average bean yield without adoption (1. + a2 + a3 = differential intercept effect of adoption on average bean yield when disease pressure is present, and (14 = differential intercept effect of insect pest pressure on average bean yield without adoption. 70 The hypothesis (H1) that farmers planting an RV obtain superior yields to those planting susceptible varieties when disease pressure is present can be tested as, H a+a +a <0 10 ' l 2 3 " (5.8) Hla:a'1+ar2+ar3 >0 Rejection of the null hypothesis (H10) in favor of the alternative (H 1 a) will confirm a positive RV yield impact in the presence of disease pressure. That is, the productivity- maintenance aspect would be confirmed. Conversely, a failure to reject H .0 would suggest that no yield differential exists between resistant and local varieties. Empirically, this hypothesis is tested by evaluating the sum of those coefficients in H10 deemed significant at conventional levels. In the case that only a] is significant and positive, then there would be evidence suggesting that RV varities provide a yield-enhancing advantage over susceptible varieties. 5. 6. 2 Input Demand Functions The two input demand firnctions, one for firngicides and the other for insecticides, are specified using a linear firnctional form as follows”, X. =6 +6A+6 D+6 A*D+ +Pw+Ly+F¢+Z6+e; (j=l,2) where, 28 Similar to the yield equation, a linear input demand specification was selected based on a literature review of impact analyses of varietal technologies. However a MacKinnon, White, and Davidson (MWD) test also rejected the null hypothesis of a linear model for both the insecticides (p=0.01) and fungicides (p=0.00) but failed to reject the alternative hypothesis of a log-log model for both insecticides (p=0.54) and fungicides (p=0.70). Empirical results for both models will be presented in Section 5.7.1. 71 = Denotes either fungicides (i=1) or insecticides (i=2), = Denotes either disease pressure (when j=1) or pest pressure (when j=2), = Parameter coefficient for the intercept and binary variables representing adoption and disease/pest pressure, = Vector of input and output prices, = Vector of parameter coefficients associated with P, 8 ”a one: and all other variables are as previously defined. The input demand function for fungicides includes a binary variable for disease pressure and an interaction term between D1 and adoption. The demand function for insecticides, however, includes only a binary variable for pest pressure. Similar to the yield model, the treatment effect of RV adoption on input use is analyzed using the appropriate parameter coefficients. The parameter coefficients 6 1 and 6 2 represent differential intercept coefficients for RV adoption and disease (or pest) pressure. The parameter coefficient on 5 3 is the differential intercept with both adoption and disease pressure. The hypothesis (H2) that farmers who plant RVs reduce the quantity of pesticides applied versus those who plant susceptible varieties can be tested as, Hj 26+§ +5 20 20 l 2 3 (5.10) I . ' Hza.6l+62+53<0 As stated, rejection of the null hypothesis (H {a ) in favor of the alternative (H 5.0 ) will confirm that RV adoption leads to a decrease in the quantity of fungicides (for j=1) or insecticides (for j=2) applied when disease (pest) pressure is present. Note that if only 6 1 is significant and negative, then RV adoption leads to a decrease in pesticide use regardless of disease pressure levels. Alternatively, a failure to reject H2o suggests that the adoption of an improved variety does not result in significant input-savings. 72 Empirically, this hypothesis is tested by evaluating the coefficient 6 1 at conventional levels of significance using a one-tailed t-test. 5. 6.3 Unit Cost Function The unit cost function is specified using a log-log functional form”. Following the notation previously provided, the unit cost firnction can be stated empirically as, 1n(UVC) = 770 +771A +712D1+n3A *Dl +174D2 + + 1n(W )1 + 1n(Y )7: + ln(L)y+ ln(F)¢ + 1n(Z )0 + e (5.11) where, n = Parameter coefficients on the intercept term and binary variables, representing adoption and disease/pest pressure, W = Vector of input prices, ,1 = Vector of parameter coefficients on W, Y = Bean yield (level of output variable), 7: = Parameter coefficient on the yield (output) variable, and all other characters are as previously defined. As in the yield model, :7 1, 172, and :74, represent differential intercept terms on RV adoption, disease pressure, and pest pressure, respectively. The coefficient on the interaction term between adoption and disease pressure, n3, is the differential intercept term for the unit cost of bean production with adoption in the presence of disease pressure. Hypothesis three concerns the treatment effect of RV adoption on unit variable costs (UVC). The null hypothesis (H3) that farmers who plant RVs varieties have lower 29 The log-log specification was selected based on a MacKinnon, White, and Davidson (MWD) test of the unit cost function (Gujarati 2003, p. 280). The MWD test rejected the null hypothesis of a linear model (p=0.00) but failed to reject the alternative hypothesis of a log-log firnctional form (p=0.62). 73 unit costs than those who plant susceptible varieties can be tested for both the firngicide and insecticide models as, H30 2771+772 +773 20 H . 0. (5.12) 361"71+"2+’73< Rejection of the null hypothesis (H30) in favor of the alternative (H33) will confirm that RV adoption leads to a decrease in UVC when disease pressure is present. Alternatively, a failure to reject H30 would suggest that the RV adoption does not significantly reduce UVC. Empirically, this hypothesis is tested by evaluating statistically significant coefficients in H3 with a one-tailed t-test. Finally, it is important to note that if the coefficient on m alone is statistically significant and negative, then RV adoption is expected to reduce unit variable costs regardless of the presence of disease pressure. 5.7 Results and Discussion This section reports information on model specification, empirical results, and regression diagnostics for the crop yield, input demand, and unit cost equations. A discussion of key findings is also included. 5. 7.1 Yield and Input Demand Equations The yield and input demand equations are estimated as a system using seemingly unrelated regression (SUREG)3°. Given that the system estimates two demand functions for similar products, correlation among the errors terms is expected and this approach 3° Fernandez-Cornejo et al. (2002) also estimated yield and input demand treatment effect models as a SUREG system but with a profit fimction. In this case, however, a unit cost function is estimated separately as opposed to a profit function so as to obtain an empirical measure of the farm-level economic impact that is consistent with the economic surplus fiamework used in economic impact analyses of agricultural technologies (i.e. the reduction in variable cost per unit of output). 74 provides efficient and unbiased standard errors compared to OLS3 '. The crop yield model is specified as quadratic production function and the two input demand firnctions are linear. All three models are regressed through the origin to preserve a maintained hypothesis of non-negative yields and input demands. Each model also shows a high degree of overall statistical significance (Chi2 statistics significant at p< 0.00 for all models) and statistical fit (R2 of 0.79 or higher for all models). Coefficients on continuous explanatory variables are interpreted as the expected marginal change in the dependent variable for a one unit increase in a given explanatory variable. The coefficients on binary explanatory variables are interpreted as the expected change in the dependent variable for a change in the binary variable from 0 to l. The null hypothesis for the yield model (H 1 o) is rejected, as shown by the magnitudes and significance levels of the adoption and disease pressure treatment effect variables (Table 5.3). The high disease pressure variable and the interaction term between RV adoption and high disease pressure are significant at the 10% and 1% levels, respectively. Other things equal, higher than average disease pressure decreases yields by 486 kg ha", while the use of an RV in the presence of disease pressure increases yields by 1350 kg ha]. The RV adoption variable alone is not significantly different fi'om zero at conventional levels, and is thus treated as zero in testing H1. Together, these coefficients sum to 658 kg ha", suggesting a positive impact of resistant varieties on bean 3 ' The same yield and input models were also estimated using two alternative estimation procedures: first using ordinary least squares (OLS) with heteroskedasticity-robust standard errors, and second using three- stage least squares (3SLS). Both SUREG and 3SLS proved superior to OLS in terms of efficiency. The potential advantage of 3SLS is that it controls for the endogeneity of fungicide and insecticide inputs. However, SUREG and 3SLS estimates reported identical significance levels and magnitudes for all treatment effect variables. Thus, SUREG was chosen for presentation here since the additional benefit to be gained by using 3SLS appears minimal and doesn’t justify the additional computational complexity required by BSLS. See Tables A61 and A62 in Appendix 6 for empirical OLS and 3SLS regression results. 75 yields with disease pressure is present”. The hypothesis that farmers who plant RVs obtain higher yields than those who plant susceptible varieties holds when disease pressure is present. To provide a reference figure in analyzing the relative magnitude of these effects, the mean yield across the sample of 72 farmers is equal to 1526 kg ha'l suggesting the yield-losses avoided from RV adoption equal to 43%. Additional variables in the yield model that are statistically significant include altitude and the two community-level variables. Altitude has a positive effect on yield. This is expected since altitude in northern Ecuador is positively correlated with both humidity and disease pressure. Since the model does control for disease pressure but not rainfall, the coefficient on altitude likely reflects increased precipitation levels at higher altitudes. It is also important to note the presence of multicollinearity among the four production input variables and their squared counterparts”. While the influence of individual inputs cannot be identified, as a group they offer much stronger explanatory power. Given that the coefficients on these variables are not central to the hypotheses of interest, however, the ambiguity created by this multicollinearity is acceptable and no further corrective action is taken. 32 DFBETA regression diagnostic statistics for the disease pressure and the interaction between disease pressure and RV adoption were plotted (Figure A6] of Appendix 6). Only one outlier was determined to have an exceptionally large impact on coefficient magnitudes, with its exclusion decreasing disease pressure by 0.5] standard deviations and increasing the interaction term by 0.58 standard deviations. This implies that the initial results provided a conservative estimate. In addition, observations with the highest individual per-hectare yields do not have a statistically significant influence on the coefficient magnitudes. 33 Variance inflation factors (VIFs) for all production inputs are greater than 10, and an F-test of their joint significance reported a Chi2 value of 282. These measures suggest the presence of multicollinearity. See Table A.6.4 for empirical VIF test results. 76 Table 5.3: SUREG Results for the Quadratic Crop Yield and Linear Input Demand Equations, Red Mottled VarietiesI Imbabura & CarchiI EcuadorI 2006 (n=73) Yield Fungicide Demand Insecticide Demand Explanatory Variables by Category: (kL/ ha) (kg AI / ha) (kg AI / ha) Treatment E fleet Variables: Adopted improved variety (1=yes) -222 (0.34) -1 1.7 (0.02)“ -4.72 (0.01)**"‘ High disease pressure (1=yes) -486 (0.09)* -2.70 (0.60) Adopted x Disease pressure (1=yes) 1350 (0.00)*** 7.34 (0.31) High pest pressure (1=yes) 65.3 (0.76) 2.98 (0.04)“ Production Inputs: Fungicide A1 (kg ha") 5.73 (0.79) Fungicide AI squared -0.080 (0.63) Insecticide A1 (kg ha") 28.8 (0.54) Insecticide AI squared 0.167 (0.65) Foliar fertilizer AI (kg ha" ) -279 (0.87) Foliar fertilizer AI squared -0.012 (0.95) Bean seed (kg ha") -0.980 (0.92) 0.545 (0.00)*** 0.219 (0.00)*** Bean seed squared 0.029 (0.46) Plot Characteristics: Plot size (ha) 25.6 (0.76) -2.03 (0.35) -0.222 (0.80) Altitude (m.a.s.l.) 0.508 (0.03)“ 0.013 (0.02)" 0.003 (0.20) Loam soil (1=yes) 59.6 (0.73) -1.49 (0.73) -0.500 (0.78) Irrigated plot (1=yes) -377 (0.32) -12.6 (0.19) -4.26 (0.27) Plot prev. cropped w/ beans (1=yes) -202 (0.28) 3.74 (0.44) 1.83 (0.36) Sharecropped plot (1=yes) 137 (0.44) -9.81 (0.04)" -4.79 (0.02)" Rented plot (1=yes) -465 (0.18) -13.1 (0.15) 0.045 (0.99) Household Variables: Age of HH (years) -1.29 (0.84) -0.716 (0.00)"* -0.246 (0.00)**"‘ Attended pest man. seminar (1=yes) 124 (0.47) -0.301 (0.95) -3.49 (0.07)"‘ Symptom based pest man. (1=yes) 5.99 (0.28) 7.06 (0.00)*" Poor household (1=yes) -3.01 (0.49) 1.06 (0.55) Price Variables: Market price for beans ($/kg) -25.3 (0.06)* -7.23 (0.19) Cost of transport ($/qq) -2.26 (0.73) -4.48 (0.09)* Avg. price of firngicide ($/kg A1) -1.39 (0.76) Avg. price of insecticide ($/kg AI) -0.847 (0.62) Community-Level Variables: Chota valley (1=yes) 523 (0.00)"* 9.58 (0.08)* 3.52 (0.1 1) Prev. extension intervention (1=yes) 321 (0.04)“ -2.76 (0.54) -1.31 (0.46) R2 0.88 0.79 0.80 Chi2 587 296 299 P>Chiz 0.00 0.00 0.00 Notes: p—values in parentheses; data is from the 2006 B/C CRSP and INIAP farm-level survey *** significant at 1%; ** significant at 5%; * significant at 10% 77 Table 5.4: SUREG Results for the Log-Log Yield and Input Demand Equations, Red Mottled Varieties, Imbabura and Carchi Ecuador 2006 (n=73) # Log onield Log Bgflicide Log (13f :gfisidde Explanatory Variables by Category: lnfig / ha) 1n(kg Al / ha) 1n(kfll/ ha) Treatment Eflect Variables: Adopted improved variety (1=yes) -0. 147 (0.40) 0.230 (0.42) -0.1 12 (0.60) High disease pressure (1=yes) -0.059 (0.80) 0.568 (0.04)" Adopted x Disease pressure (1=yes) 0.606 (0.04)“ -0.411 (0.27) High pest pressure (1=yes) 0.009 (0.95) 0296 (0.07)* Production Inputs: Log of firngicide A1 (kg ha") -0.084 (0.44) Log of insecticide A1 (kg ha") 0.269 (006)“ Log of foliar fertilizer A1 (kg ha") 0.034 (0.37) Log of bean seed (kg ha") 0.581 (0-00)*** 0.009 (0.00)*** 0.007 (0.01)" Plot Characteristics: Log of plot size (ha) 0.160 (0.25) -l .37 (0.00)""' -1.25 (0.00)*** Log of altitude (m.a.s.l.) 0.660 (0.00)”* -0-013 (0.95) 43.085 (0.67) Loam soil (1=yes) 0.077 (0.58) -0. 146 (0.56) -0.371 (0.08)‘ Irrigated plot (1=yes) -0.364 (0.27) 0.928 (0.14) 0.701 (0.19) Plot prev. cropped w/ beans (1=yes) -0.064 (0.66) 0.240 (0.40) 0.328 (0.17) Sharecropped plot (1=yes) 0.090 (0.53) 0021 (0.94) -0-l57 (0.50) Rented plot (1=yes) -0.585 (0.03)" 0189 (0.71) 0503 (0.23) Household Variables: Log of age of HH (years) -0.183 (0.38) 0039 (0.82) 0.136 (0.68) Attended pest man. seminar (1=yes) -0.028 (0.84) '0-141 (0.59) '0-023 (0.90) Symptom based pest man. (1=yes) -0.138 (0.67) 0.097 (0.72) Poor household (1=yes) -0-423 (0.10)* '0230 (0.28) Price Variables: Log of market price for beans (S/kg) 0.631 (0.21) 0.604 (0.16) Log of cost of transport ($/qq) 0.055 (0.24) 0-023 (0.90) Log avg. price of firng. ($/kg AI) -0-124 (0.50) Log avg. price of insect. (S/kg AI) -0. 161 (0.60) Community-Level Variables: Chota valley (1=yes) 0.434 (0.00)*** -0.385 (0.25) -0. 180 (0.52) Prev. extension intervention (1=yes) 0.288 (0.03)“ -0.211 (0.40) -0.251 (0.22) R2 0.99 0.81 0.76 Chi?- 15408 34 22 P>Chi2 0.00 0.00 0.00 Notes: p-values parentheses; data is from the 2006 B/C CRSP and INIAP farm-level survey at" significant at 1%; ** significant at 5%; * significant at 10% 78 The null hypotheses for the fungicide and insecticide input demand functions (H20) are rejected at the 5% and 1% levels, respectively (Table 5.3), when disease or insect pest pressure is not present. Other things equal, RV adoption leads to a decrease in the quantity of fungicide active ingredient applied by of 11.7 kg ha'1 and in insecticide active ingredient by 4.7 kg ha'l (shown by (11). Again, to compare the relative size of these effects, the mean quantities of pesticides applied across the sample of 72 farmers is equal to 15.8 kg ha'I for fungicides and 6.1 kg ha'I for insecticides. These decreases represent a 74% decrease in fungicide and a 40% decrease in insecticide use. The coefficients on the disease/pest pressure variable are statistically significant only for the insecticide model but large in magnitude in both models. This suggests that while RV adopters reduce their input application rates compared to non-adopters when no disease pressure is present, the difference decreases as disease/pest pressure increase. For example, RV adopters are expected, on average, to apply only 1.7 kg ha’1 (28%) less insecticide than non-adopters when disease pressure is present. Additional variables in the input demand functions that are statistically significant across both models include planting density (kg ha’l of bean seed), whether the plot is Sharecropped, and farmer age. An increase in planting density by 1 kg ha'l increases the demand for fungicides by 0.55 kg ha'1 and for insecticides by 0.22 kg ha". A sharecropping production arrangement results in a decrease in the average amount of pesticides demanded by 9.8 kg ha'I for fungicides and 4.8 kg ha'1 for insecticides. This is consistent with the economic theory of sharecropping, which suggests that such output sharing agreements reduce the incentive to provide inputs since the lower expectation of income leads to a lower marginal value product for each unit of input. The age variable 79 indicates that older farmers use lower quantities of pesticides, on average, than do younger farmers. Similar to the yield model, fungicide demand is influenced by altitude and whether the farmer resides in the Chota valley. Against expectations, the market price of beans is negatively correlated with fimgicide demand. An increase in the market price of beans is expected to decrease, on average, farmers’ use of fungicides. In the insecticide model, however, the cost of transport variable that proxies market transaction costs does agree with intuition, showing that the demand for insecticides decreases as its prices increase. The log-log specification of the same system of equations is given in Table 5.4. Given that MWD test results rejected the linear models in favor of the log-log model as discussed in Section 5.6, it is important to analyze the robustness of findings with respect to this separate firnctional form. H1 and H2 can be tested in a similar fashion as previously discussed; however the log-log model parameter coefficients now describe the proportional change in the dependent variable for a one-percent change in the value of the explanatory variable. Here, we reject H1 since the parameter coefficient is statistically significant. Other things equal, farmers who use an RV obtain yields 55% higher than do farmers who plant a local variety. When disease pressure is present, this finding is consistent with the linear specification although slightly larger since the parameter coefficient on disease pressure is not significant. Unlike the linear specification, we cannot reject H2 for either the firngicide or insecticide demand model. Since H2 is not robust across model specifications, the conclusion that farmers who adopt an RV apply less pesticide active ingredient than do farmers who plant local varieties should be interpreted with less certainty than H1. 80 5. 7.2 Unit Cost Equation The unit cost filnction is reported using a log-log model specification and has a high overall statistical significance (F=10.31) and statistical fit (R2 = 0.77)“. Given the log-linear specification, coefficient estimates of continuous explanatory variables, Xj, are interpreted as elasticities, which give the expected percentage change in UVC for a 1% increase in X,-, holding all other factors fixed. For binary explanatory variables, the interpretation of log-log coefficient estimates changes slightly. They are interpreted as the expected proportional change in UVC for a change in value fi'om 0 to l (Wooldridge 2000, p. 218). Hence, multiplying the coefficient by 100 will provide an approximation of the expected percentage change in UVC. The null hypothesis (H30) for the unit cost model is rejected, as shown by the magnitude and significance level of the adoption and disease pressure treatment effect variables (Table 5.5). The coefficient on the interaction term between high disease pressure and RV adoption is significant at the 10% level. The coefficients on the individual disease pressure and adoption variables are not significant, however, and are assumed to be zero for the purpose of testing H3. RV adoption is expected to decrease UVC by an average of 40%. Thus, the hypothesis that farmers who plant improved varieties have a lower unit variable cost of production than do farmers who plant susceptible varieties holds. It is important to note that this decrease refers only to those costs that are expected to vary upon the adoption of an RV (i.e. pesticide and seed costs), and not to all variable costs associated with the production of beans in northern Ecuador. Data provided in Peralta et al. (2001) are used to estimate the overall decrease in variable 34 See Table A53 in Appendix 6 for regression results fi'om the linear unit cost model. 81 production costs from RV adoption. Cost of production estimates show that pesticides and seed costs, together, to represent 46% of total variable production costs. A 40% reduction in these costs due to RV adoption translates into an overall decrease in variable production costs of approximately 18.4%. Apart from the hypothesis of interest, price, output, and community-level variables are also significant. An increase in the prices of both firngicide and bean seed have a positive effect on unit variable costs (UVCs). A one percent increase in the price of fungicides is expected to increase UVCs by .25%. Likewise, a similar increase in the price of bean seed will increase UVCs by .88%. Both output (kg ha") and plot size have a negative effect on UVCs, with a 1% increase leading to an average decrease in UVC by .67% or .26%, respectively. Communities where IN IAP has previously intervened are expected to have UVCs 25% lower than the base case of no previous intervention. Finally, altitude and renting are both significant determinants of UVC. A 1% marginal increase in altitude is expected to lower UVC by .51%, whereas farmers who rent plots as opposed to own them or enter sharecropping agreements are expected to have 38% higher UVC. 82 Table 5.5: Regression Results for the Log-Log Unit Variable Cost Function, Imbabura & Carchi, Ecuador, 2006 (n=73) Log of Unit Variable Cost Erplanatory Variables by Category: ($/kg) Treatment E fleet Variables: Adopted improved variety (1=yes) 0.160 (0.20) High disease pressure (1=yes) 0.106 (0.48) Adopted x Disease pressure (1=yes) -0.396 (0.07)‘ High pest pressure (1=yes) -0.064 (0.54) Price Variables: Log of avg. filngicide price ($/kg A1) 0.255 (0.02)" Log of avg. insecticide price (S/kg Al) 0.008 (0.97) Log bean seed price (S/kg) 0.882 (0.00)*** Output Variable: Log of bean yield (kg ha") -0.676 (0.00)*** Plot Characteristics: Log of plot size (ha) -0.261 (0.00)"* Log of altitude (m.a.s.l.) -0.513 (0.08)"I Loam soil (1=yes) 0.030 (0.75) Irrigated plot (1=yes) 0.072 (0.73) Sharecropped plot (1=yes) -0. 140 (0.15) Rented plot (1=yes) 0.376 (0.04)" Household Variables: Log of age of HH (years) -0.069 (0.66) Attended pest man. seminar (1=yes) 0.078 (0.43) Community-Level Variables: Chota valley (1=yes) -0.015 (0.89) Prev. extension intervention (1=yes) -0.249 (0.01)*" Constant 7.11 (0.@*** R2 0.77 F (k,c#) 10.31 Prob>F 0.00 p-values in parentheses "‘ significant at 10%; "”" significant at 5%; "*significant at 1% 5.8 Chapter Summary This chapter analyzed three farm- level hypotheses regarding the adoption of improved bean varieties released by INIAP in northern Ecuador. These hypotheses are that farmers who plant an improved variety 1) obtain higher yields, 2) utilize fewer pesticide inputs, and 3) experience a lower unit cost of production than do farmers who plant local varieties. Three average treatment effect models were developed to test each of these hypotheses using multiple regression techniques: a crop yield model, an input demand model, and a unit cost function. Regression analysis rejected the null hypotheses for the yield, input demand, and unit variable cost models, as the results summarized in Table 5.6 suggest. Other things equal farmers who adopt an RV obtain higher yields by about 40% when compared to the sample mean of 1590 kg ha". Likewise, those adopting an RV apply 70% less firngicides and 43% less insecticide than do non-adopters when compared to the sample means and with or without disease pressure. These percentages are likely much lower when disease Table 5.6: Summary of Testable Motheses and Empirical Results for F arm-Level Regression Models Model Null Hypothesis Result Conclusion On average, RV adopters have 40% higher yields than do non-adopters when disease Yield H10: a] + a; + (13 S 0 Reject H10 pressure is present. This confirms the. Model productrvrty-malntenance characteristics of INIAP RV beans. On average, RV adopters apply 74% less Input Use . fungicide (and 43% less insecticide) than do Model H ; 5 + 5 + 5 2 0 Reject H20 non-adopters when disease (insect) pressure is 20 1 2 3 absent. This suggests that farmers may perceive RVs as a substitute for chemical inputs. On average, RV adopters have 18% lower unit Unit Cost , costs of production than do non-adopters when Function H 30: '11 + ‘12 + '13 Z 0 R816“ H30 disease pressure is present. This confirms that RV adoption shifis producers’ marginal cost curves downwards. 84 or pest pressure is indeed present. These two impacts lead to an average reduction in unit variable costs of 40% (again, for only those costs that vary upon RV adoption). It is important to note that findings for the yield model are robust across linear and log-log model specifications. The findings for the input use model, however, are significant only for the linear specification and should be interpreted with less confidence. Now that both the diffusion rates and farm-level benefits of the improved varieties have been identified, attention may be turned to estimating the economic impact of bean-breeding research. 85 CHAPTER SIX: AN ECONOMIC EVALUATION OF BEAN-BREEDING RESEARCH IN NORTHERN ECUADOR 6.1 Introduction This chapter introduces the economic surplus approach commonly used in measuring the aggregate economic benefits of agricultural research and calculates two well-known measures of project worth——the net present value (NPV) and the internal rate of return (IRR)—for disease-resistant bean breeding research in northern Ecuador. Discussion is focused primarily on the research-induced supply shift. Basic prior knowledge of supply, demand, and economic surplus concepts is assumed. Three scenarios are presented in order to determine a range of NPV and IR values: a baseline scenario which bases parameter values on the best estimates available, a conservative scenario which utilizes lower-bound parameter values, and a robust scenario which uses parameter values at the upper-bound of possible values. For red mottled resistant varieties (RVs), the ex post NPV and IR values are estimated for the period 1982—2006. Since the first RV (Paragachi) was released in 1986, this provides a 21-year benefit stream with which to assess their economic impact. The period of time from 1982-1986 therefore represents a four-year lag period in which research costs were incurred without any research benefits. For purple mottled varieties, ex ante NPV and IR values are estimated for the period 1998-2024. Given that the first purple mottled RV (La Concepcion) was released in 2004 though the CIAL network, this period was chosen to allow for a 21-year benefit stream similar to that of red mottled RVs for consistency across estimates. The difference 86 in the overall length of the time period considered is due to the longer, six-year lag period between the first research expenditures in 1998 and release of the first variety in 2004. When appropriate, a comparison of methodologies and parameter values fi'om economic impact assessments of similar technologies will be discussed (Mather et a1. 2003, Johnson et al. 2003, Boys et al. 2007, Oehmke and Crawford 1996). 6.2 Conceptual Framework 6. 2. 1 Research Benefits The estimation of economic benefits derived from agricultural research is generally undertaken as an exercise in partial-equilibrium analysis. The conceptual underpinnings of this approach, also known as the economic surplus model, are fairly straightforward and a large body of literature on the topic exists (see for example Alston et a1. 1998 or Masters et a1. 1996). While research-induced technological change in the bean sub-sector can also affect other sectors of the economy such as food prices, labor markets, or returns to different factors of production, all secondary effects are considered exogenous and not addressed here. The model presented here is that of a small open economy. The term open reflects the export-oriented nature of bean production in northern Ecuador, while the term small refers to the supply share of Ecuador’s dry bean production with respect to its primary export market, Colombia. In 1998”, Ecuador exported 11,500 metric tons of dry beans to Colombia, or about 8% of Colombia’s total dry bean consumption that year (2000). A small increase in dry bean production in Ecuador, therefore, is not expected to affect market price. This implies a perfectly elastic demand curve. Such an assumption is ofien 3’ 1998 is the only year for which data was available. 87 appropriate for impact analyses of agricultural technologies in a small, developing country context (Alston et al. 1998, p. 226). The magnitude of economic benefits derived fiom agricultural technology is determined by the size of the downward shift in the supply curve. Graphically, this is akin to the reduction in unit production costs found in Chapter 5 (Figure 6.1). The supply curve under the original technology is labeled S 1, and the initial equilibrium occurs at the point (P), Q;). As cumulative adoption increases, the supply curve shifts outward fi'om S 1 to S; and the equilibrium quantity increases fi‘om Q; to Q2. The equilibrium price remains unchanged at P1 given the perfectly elastic demand curve (due to the small open economy assumption). A parallel shift of the supply curve, as diagrammed, is appropriate when 1) the production technology of bean producers is fairly homogenous, and 2) technology is scale neutral, as crop varietal technologies often are. When valid, these assumptions ensure that linear approximations of supply schedules will provide a good basis for estimating research benefits (Alston et a1. 1998). Figure 6.1: Research-Induced Supply Shift $/kg / SI / SI Pl : b D K i e a d . l l : l 3 01 Q2 m 88 In the small open economy fi'amework, all research benefits accrue to producers as shown by eabd. This results in the special case where total economic surplus equals producer surphls. There is no impact on the economic welfare of domestic consumers or importing countries. The portion of the benefit derived from incremental output (holding inputs fixed) is the area abc. The portion of benefit derived fiom reduced production costs (holding output constant) is the area eacd. An empirical measure of annual research benefits in a small open economy for ex post analyses can be estimated by the equation, ATS, = PlQth (1 + 0.5K (a) (6.1) where P; represents the exogenous market price for beans, Qt represents the initial before-research production level, 8 is the supply elasticity of demand, and K represents the vertical shift in the supply curve (Alston et a1. 1998, p. 227). The time-path for K is determined by three factors: i) AUC, the proportional farm—level change in unit costs from RV adoption, ii) P(D), the probability of disease pressure”, and iii) A the cumulative adoption rate at time t. This time path is calculated 35, K, = AUC x P(D) x At. (6.2) Specific data and parameter values to be used are presented in Section 6.3.1. Note that in this instance, only A varies with time whereas AUC and P(D)remain constant for the period of analysis. Due to the lag period between the release of an improved variety and ’6 K depends on disease pressure in this instance because results fiom the unit cost model in Chapter 5 indicated a reduction in proportional unit costs only in the presence of disease pressure. 89 the maximum cumulative adoption level, this time path must be determined before calculating Equation (6.1). 6. 2.2 Research Costs The appropriate identification of research costs begins with a statement of the duration and scope of the research project to be evaluated. All expenditures prior to the research project are considered sunk costs and not included, since these resources would have been spent with or without the current agricultural research program (Masters et al. 1996). Extension expenditures incurred during the period of analysis that would have been spent regardless of the current research program are also not included. Some extension programs, however, are indeed specifically designed to complement the work of plant breeders, such as the local agricultural research organizations, or CIALs, formed in northern Ecuador. In this case, CIAL extension costs are included in the ex-ante analysis of purple mottled varieties presented in Section 6.4.4 since their primary focus to assist plant breeders through participatory varietal selection. Once the time period and scope of the research project are determined, care must be taken in disaggregating research expenditure data so as to reflect only the technology of interest. When operating costs or scientist salaries are not broken down into the level of detail required to estimate research costs, knowledgeable individuals (such as program directors) can be asked to provide estimates of total expenditures and the share of this expenditure devoted to the program or technology of interest (Alston et al. 1998). This is the approach followed here. Details on specific data collected are included in Section 6.3.2. 90 6. 2.3 Measures of Project Worth In this paper, two economic measures are used to evaluate the stream of benefits and costs associated with the development of the improved bean cultivars, namely the net present value (N PV) and the internal rate of return (IRR). These measures are advantageous since they summarize information about the economic returns to research investments into simple summary statistics. The NPV calculation combines the flow of research benefits and costs over the period 1982-2006 into a single value in constant 2006 dollars ($USD). Positive values of NPV indicate that a project is profitable, while negative values indicate that a project is not profitable. Profitable in this case implies that all investments plus the opportunity cost of capital are recovered and the project is therefore worth investing in3 7.It is calculated by first taking the difference in research benefits and research costs for each time period, t, and then discounting these sums to a single base period using a discount rate, r, using a present value formula, NPV, = Z _T t=l (1+r) (6.3) where T is equal to the number of time periods (years) under consideration. The value of research benefits minus research costs in any given year is referred to here as the net benefit. The IR indicates the value of discount rate, r, from Equation (6.3) for which the NPV is equal to zero: ’7 In the case of a constrained investment budget, the NPV decision rule becomes more complicated. In this case we are only evaluating one alternative, so the present decision rule suffices. 91 0 = —, t=l (1+ IRR)’ (6.4) or, equivalently, the discount rate that makes the present value of benefits equal to the present value of costs. The general decision rule for a profitable project is that the IRR be greater than the opportunity cost of capital. Both measures have their strengths and weaknesses. The NPV provides an excellent measure of total net benefits received, but does not allow for a comparison between alternatives based on the rate of return. Nevertheless, a measure of the NPV is particularly useful in ex ante evaluations where priority-setting among alternative projects is still occurring. The IR, on the other hand, does allow us to rank programs based on the rate of return but does not provide the analyst with information about the scale of benefits or the monetary value of the overall program A measure of IR is typically useful in ex post studies where information is sought on the actual rate of returns received. The outcome of both NPV and IR calculations depends partly on the discount rate, r. In the NPV calculation, the higher the discount rate, the less weight is placed on future benefits. In the IRR calculation, the discount rate is not directly used, but serves as a benchmark to compare IR and determine relative profitability of investments. 6.3 Data Description 6. 3. I Research Benefits Determination of annual research benefits and NPV and IR summary statistics depends on the values used in calculating the time-path of K, and total surplus, given in Equations (6.2) and (6.1), respectively. Due to uncertainty surrounding some of these 92 values, three scenarios are presented: a baseline scenario, a conservative scenario, and a robust scenario. The baseline scenario utilizes the best estimates available for each parameter value and represents the most likely NPV and IR. For the conservative scenario, parameter values are chosen so as to provide a lower-bound estimate of NPV and IR. Likewise, for the robust scenario, parameter values are chosen so as to provide upper-bound estimates of NPV and IR. This scenario-based approach to sensitivity analysis provides a range of possible NPV and IR estimates. While one can perform sensitivity analysis on individual parameters, it is important to recognize that most parameters are mutually dependent and will be expected to co-vary to some degree. Treating individual parameters as independent and considering all cases from high to low could lead to misleading interpretations of NPV and IR values (Alston et a1. 1998). Parameter values used in calculating the NPV and IR for each scenario are listed in Table 6.1. The change in unit cost parameter, AUC, is assumed equal to 18.4% for the baseline scenario. This is determined by multiplying the percentage reduction in unit variable costs (UVC) from Chapter 5 (found to be 40%) by the proportion of total production costs represented by filngicide, insecticide, and seed costs, which is approximately 46% (Peralta et al. 2001). Values for the conservative and robust scenarios are determined by adjusting the baseline value by 5% in either direction, resulting in AUC values of 13% and 23% respectively. For P(D), no data exist that describe the distribution of farmers who report above average disease pressure. Nor do reliable data exist that describe the relationship between 93 Table 6.1: Supply Shift Parameter Values Used in Ex-Post NPV and IRR Calculations, Disease- Resistant Red Mottled Varieties, Imbabura and Carchi, Ecuador, 1982-2006 Parameter Values by Scenario Symbol Description “Conservative ” “Baseline ” “Robust " AUC I’mpfmma' Change 0.166 0.184 0.202 m unit costs P (D) Incidence of disease 042 pressure Cumulative At adoption rates: 1986 1.0% 1.0% 1.0% 1987 1.5% 2.4% 3.3% 1988 2.2% 6.0% 9.8% 1989 3.1% 12.7% 22.3% 1990 4.5% 19.7% 34.8% 1991 6.4% 23.9% 41.5% 1992 8.9% 26.4% 44.0% 1993 12.1% 28.4% 44.7% 1994 16.0% 30.4% 44.9% 1995 20.3% 32.6% 44.9% 1996 24.7% 34.9% 44.9% 1997 29.0% 37.0% 45.0% 1998 32.8% 38.9% 45.0% 1999 36.1% 40.5% 45.0% 2000 38.6% 41.8% 45.0% 2001 40.5% 42.7% 45.0% 2002 41.9% 43.4% 45.0% 2003 42.8% 43.9% 45.0% 2004 43.5% 44.3% 45.0% 2005 44.0% 44.5% 45.0% 2006 44.3% 44.7% 45.0% severity of disease pressure and impact on yields and unit costs. Values reported in Chapter 5 report only average statistics. The baseline P(D) parameter value, therefore, is calculated as the average between the proportion of farmers surveyed who reported both bean rust and anthracnose (0.43) and those who reported at least one of these diseases (0.89), resulting in a value of 0.64. Parameter values for the conservative and robust scenarios are taken as the high and low values of P(D) used in calculating the baseline estimate. 94 Cumulative adoption rates, A,, for red mottled RVs are unknown. However, they are expected to fall between the conservative estimate for red mottled RV diffusion (previously shown in Figure 4.2) and the estimated diffilsion path of the recently released purple mottled RVs (previously shown in Figure 4.3). Baseline values for Al in Table 6.1 are then defined as the average cumulative adoption rate between these two figures. For the conservative scenario, values for A are obtained from Figure 4.2 since it represents a lower-bound of expected diffusion rates. Likewise for the robust scenario, values for A. are obtained from Figure 4.3 since the recently released purple mottled RVs are expected to diffuse at least as fast as the earlier red mottled RVs and should provide an appropriate upper bound. To complete the calculation of total research benefits (ATS), data on the supply elasticity, market price, and total production are also needed. The supply elasticity parameter, a, is assumed equal to 0.7 for all three scenarios. No primary research on supply elasticities for semi-subsistence crops exists for Ecuador. The value is thus assumed identical to that used by Mather et al. (2003) for the supply elasticity of export- oriented bean production in Honduras. This value lies at the mid-point of supply elasticities for developing country agriculture from 0.2 to 1.2 suggested by Masters et al. (1996). Wholesale price data for the red mottled bean variety is available fiom 2000 to 2005 for the market in Ibarra (SICA 2007b). The price used in this analysis is assumed to be $600/mt, which is determined by calculating the average wholesale price over the same period (in constant 2005 dollars) and inflating to adjust the wholesale prices for an assumed 20% retail mark-up. Data on total annual dry bean production in the provinces of Carchi and Imbabura is available from 1990 to 2005 (SICA 2007a). For the period 95 from 1980-1989, annual production totals are calculated as the average of annual totals fiom 1990-1995. For 2006, the production data from 2005 are used. Together, Imbabura and Carchi are assumed to be responsible for 40% of national dry bean production. Of this, red mottled varieties are assumed to represent 68.4% of the total for Imbabura and Carchi and purple mottled varieties are assumed to represent 12.3% (Table 4.2). However, as suggested by Figure 2.4, the market price is historically volatile. The sensitivity of IR and NPV findings from the baseline scenario to changes in price and quantity produced will be examined in Section 6.4.1. 6. 3.2 Research Costs The calculation of total research costs relied on records of PRONALEG-GA’s total operating budget for bean research for the period fiom 1982 to 2004. Input from PRONALEG-GA senior staff’8 allowed for a decomposition of these figures (as a share of total expenditure) into two components: total operating expenditures and human resource costs (e. g. plant-breeders and support stafl). Within each of these categories, expenses were again decomposed based on the proportion of resources devoted only to those varieties being evaluated”. For the period corresponding to research investments in red mottled RVs, from 1982 to 1998, 60% of PRONALEG-GA’s total operating expenditure is allocated to the varieties under evaluation. During the same period, 60-80% of bean breeder salaries and 40-80% of technical assistant salaries are allocated to the red mottled varieties under evaluation. ’8 Ing. Eduardo Peralta, Director of PRONALEG-GA, e-mail message to author, September 15, 2007. ’9 The worksheets and values used are included in Appendix 7. 96 For the period corresponding to research investments in purple mottled RVs, from 1998 to 2004, two research cost scenarios are considered. The recently released purple mottled RV (La Concepcion) represents one of four varieties released through the local agricultural research committees (CIALs) network during this same timeframe. The first scenario uses a similar research cost allocation to that of the red mottled varieties for PRONALEG-GA operating expenditures and employee salaries, but is then multiplied again by 0.25, or the proportion of research costs devoted only to the purple mottled RV itself. The second scenario considers the fill] set of research costs for all four varieties. This second scenario is intended to analyze whether research benefits derived fi'om the purple mottled RV alone cover research costs for the entire participatory breeding program devoted to disease-resistant varieties from 1998 to 2004. Data collected on external support indicated three organizations provided financial support to PRONALEG-GA’S bean breeding efforts. The first is CIAT, which provided supplemental firnding for the development of improved red mottled varieties during 1990-1999. The second is PREDUZA, a Dutch organization that provided funding from 2000 to 2004. The third is the B/C CRSP, which began providing funding in 2003. In all three cases, the share of total fimding assumed to be devoted to the development of the improved varieties under consideration is 60%, identical to the assumption made earlier regarding PRONALEG-GA’S operating expenditure. Previous B/C CRSP funding and CIAT research expenses incurred outside of Ecuador are considered sunk costs which did not influence INIAP’s breeding-efforts for the varieties under evaluation. 97 6. 3.3 Discount Rate A final important matter is the selection of an appropriate discount rate to use in NPV calculations. Gittinger (1982) indicates a range of discount rates, r, between 0.08 and 0.15 that are appropriate for developing country contexts. Higher discount rates, however, are commonly used by governments and/or international donors to evaluate the short-term profitability of different investments. Since no primary research identifying a discount rate specific to Ecuador was identified, the specific discount rate used here, r = 0.10, is set equal to that used by both Mather et al. (2003) and Smale et a1. (1998) who also evaluated the economic impact of disease-resistant breeding research on agricultural crops. To deal with this uncertainty, the sensitivity of the NPV provided in the baseline scenario to a range of discount rates will be examined in the following section. 6.4 Results and Discussion 6. 4. I Ex-Post Impact of Disease-Resistant Red Mottled Varieties Results from the baseline scenario indicate an IR of 29% and an NPV $1.29 million USD of for bean-breeding research in northern Ecuador related to red mottled RVs for the period 1982-2006 (calculation provided in Table A81 in Appendix 8). Results from the conservative and robust scenarios provide an IR range of 13-46% and an NPV range of $43,000 USD to $2.74 million USD around the baseline figures (Tables A82 and A83). This implies that research investments in red mottled RVs have been profitable, at minimum, and most likely provide a return well above the assumed opportunity cost of capital (r = 0.10). Certainly the baseline estimate of 31% would compete well with alternative investment opportunities in Ecuador. In addition, the NPV 98 values refer to the profit earned by research investments above that which would have been earned by investing the money elsewhere at a rate of return of 10%. A further consideration is that the period of evaluation extends from 1982 to 2006, with no assumption regarding a continuation of benefits. While some decline in total land area planted to improved red mottle varieties may occur due to varietal replacement, the stream of positive research benefits should continue for a number of years and add to this total. In addition to sensitivity analysis on the supply shift parameters, the impact to both price and quantity and the choice of discount rate are also considered. Using the baseline scenario results, price and quantity were jointly increased and decreased 10%. This resulted in a range of IR values from 26-33% and in NPV values of $903,000 USD and $1.57 million USD. Given uncertainty of the true opportunity cost of capital in Ecuador, Figure 6.2 presents estimated NPVs graphically for a range of discount rates Figure 6.2: Sensitivity of NPV to Discount Rate $3,000 ‘1' $2,500 \ $2.000 \ \ $1.000 \ \ $0 - 4. Net Present Val ($1000 8 USD ‘5500 r i W m i 1 a 0.05 0.10 0.15 0.20 0.25 0.30 0.35 Discount Rate (r) 99 between 5% and 35%. At discount rate of 5%, results in an NPV over $2.5 million USD whereas discount rates above 29% the NPV becomes negative (as suggested by the IR) and alternative investments become more attractive. 6. 4. 2 Ex-Ante Impact of Disease-Resistant Purple Mottled Varieties In calculating the ex ante impact analysis of purple mottled RVs, parameter values from red mottled baseline scenario were used to calculate two IR and NPV estimates based on the share of research costs considered. The first estimate considers only the proportion of research costs directed towards the development of purple mottled RVs. Results indicate an IR of 34% and an NPV of $536,000 USD for the period 1998- 2024 (calculations provided in Table A.8.5 of Appendix 8). While the IR is comparable to that found for red mottled varieties, the much smaller NPV figure indicates the much smaller magnitude of research costs and benefits associated with purple mottled RVs. As with red mottled varieties, sensitivity analysis was carried out around the estimated baseline ex ante IR and NPV values for purple mottled varieties. The price and quantity values were jointly increased and decreased by 10%. This resulted in a range of IR values from 31% to 37% and of NPV values fi'om $419,000 USD to $665,000 USD. The second estimate includes all research costs related to the development of the four varieties released through the CIAL network. Results here show an IR of 17% and an NPV of $295,000 USD (calculations provided in Table A.8.4 of Appendix 8). Thus, in theory, the additional three varieties developed and released through the CIAL networks 100 could produce zero net benefits and the total research expenditure would remain profitable, albeit at a much lower rate of return. 6.5 Chapter Summary This chapter estimated the stream of annual research benefits derived fiom investments in disease resistant bean breeding research on red mottled and purple mottled varieties in northern Ecuador and their associated measures of project worth. For each variety class, this entailed the specification of three possible scenarios, namely a conservative, baseline, and robust scenario. Ex post scenario analysis for disease resistant red mottled varieties for the period 1982-2006 indicated a range of IR values fi'om 13-46%, with a baseline estimate of 46%. The estimated NPVs for this investment ranged from $43,600 USD to $2.74 million USD and a baseline estimate of $1 .29 million USD. The benefit stream is likely to continue for a number of years and add to this total. This indicates that the investment has been profitable and brought returns higher than the assumed opportunity cost of capital (r=0.10). These results are not very sensitive to a 1. 0% change in both the price received and quantity produced. Ex ante scenario analysis for disease resistant purple mottled varieties for the period 1998-2024 indicate a baseline IRR of 34% and an NPV of $536,000 when only those costs devoted exclusively to purple mottled varieties are considered. These values decrease to 17% and $295,000 USD, respectively, when all costs devoted to release of RVs through the CIAL network. In either case, the research investment on purple mottled varieties is expected to be profitable and bring returns higher than the assumed opportunity cost of capital. The small NPV values relative to those associated with red 10] mottled RVs is due to the fact that purple mottled market class represents a much smaller proportion of total land area planted to beans. Like the red mottled varieties, these results are not very sensitive to a 10% change in both the price received and quantity produced. IRR estimates for both the red mottled and purple mottled RVs approach the IRR of 40% reported by Mather et a1. (2003) for investments in bean-breeding research in Honduras and exceed the IRR of 13% reported by Boys et a1. (2007) for investments in cowpea technology in Senegal. 102 REFERENCES Alston, J ., G. Norton, and P. Pardey. 1998. Science Under Scarcity: Principles and Practice for Agricultural Research Evaluation and Priority Setting. Wallingford, UK: CAB International. Alston, J ., M. Marra, P. Pardey, and T.Wyatt. 2000. Research returns redux: a meta- analysis of the returns to agricultural R&D. The Australian Journal of Agricultural and Resource Economics 44(2): 185—215. Arévalo, V. 1985. Sistemas de Produccion y Tecnologia del Fréjol Arbustivo en Pimampiro. Documento de Trabajo 85.SE.4, Proyecto INIAP-Comell. Departamento de Planificacion y Economia Agricola. Instituto Nacional de Investigaciones Agropecuarias. Quito, Ecuador. December, 1985. Bennett, R., Y. Ismael, U. Kambhampati, and S. Morse. 2004. Economic Impact of Genetically Modified Cotton in India. AgBioForum 7296-100. Boys, K, M. Faye, J. Fulton, and J. Lowenberg-DeBoer. 2007. The economic impact of cowpea research in Senegal: an ex-post analysis with disadoption. Agricultural Economics 36:363-375. Brookes, G. 2005. The Farm-Level Impact of Herbicide-Tolerant Soybeans in Romania. AgBioForum 8:235-241. CCI. 2000. Inteligencia de Mercados: Frijol. Perfil de Producto No. 8. Corporacion Colombia Internacional (CCI). Bogota, Colombia. April-June. CIAT. 2001. The CIALs at a Glance. Centro Internacional de Agricultura Tropical (CIAT). Cali, Colombia. http://www.ciat.cgiar.orglipra/inglglancehtm (accessed May 16, 2006). Crissman, C., J. Antle, and S. Capalbo, eds. 1998. Economic, Environmental and Health Tradeofifs in Agriculture: Pesticides and the Sustainability of Andean Potato Production. Norwell, MA: Kluwer Academic Publishers. Deaton, A. 1997. The Analysis of Household Surveys: A Microeconometric Approach to Development Policy. The World Bank: Washington, DC. Ernest, E. 2004. Developing improved bush bean varieties in Ecuadorian market classes using farmer participatory crop improvement methods and marker assisted selection of an anthracnose resistance gene. MS. Thesis. Plant Breeding and Genetics Program, Department of Crop and Soil Sciences. Michigan State University, East Lansing, MI. 103 Evenson, R., and D. Gollin, eds. 2003. Crop variety improvement and its efl'ect on productivity: the impact of international agricultural research. Wallingford, UK: CABI Publishing. Feder, G., R. Just, and D. Zilberman. 1985. Adoption of Agricultural Innovations in Developing Countries: A Survey. Economic Development and Cultural Change. 33(2):255-298. F emandez-Comejo, J ., C. Klotz-Ingram, and S. Jones. 2002. Farm-Level Effects of Adopting Herbicide-Tolerant Soybeans in the USA. Journal of A gricultural and Applied Economics. 34: 149-163. Gardner, 1., and C. Nelson. 2007. Genetically Modified Crops and Labor Savings in. US Crop Production. Paper presented at the Annual Meeting of the Southern Agricultural Economics Association. Mobile, AL, February 4-7. Gittinger, J. 1982. Economic Analysis of Agricultural Projects. Baltimore: The Johns Hopkins University Press. Second Edition. Godtland, E., E. Sadoulet, A. de Janvry, and O. Ortiz. 2004. The Impact of Farmer Field Schools on Knowledge and Productivity: A Study of Potato Farmers in the Peruvian Andes. Economic Development and Cultural Change. 53:63-92. Griliches, Z. 1957. Hybrid Corn: An Exploration in the Economics of Technological Change. Econometrica. 25:501-522. Gujarati, D. 2003. Basic Econometrics. New York, NY: McGraw-Hill Higher Education. Fourth edition. INEC. 2001. III Censo Nacional Agropecuario. E1 Instituto Nacional de Estadistica y Censos (INEC). Quito, Ecuador. INIAP 1991a. INIAP-404 Cargabello Seleccionado: Variedad de Fréjol Arbustivo Cargabello Seleccionado. Plegable No. 96. Instituto Nacional Autonomo de Investigaciones Agropecuarias, Estacion Experimental Santa Catalina, Programa Nacional de Leguminosas. Quito, Ecuador. February. INIAP 1991b. INIAP-411 Imbabello: Variedad de Fréjol Arbustivo. Plegable No. 122. Instituto Nacional Autonomo de Investigaciones Agropecuarias, Estacion Experimental Santa Catalina, Programa Nacional de Leguminosas. Quito, Ecuador. November. INIAP 1996a. INIAP-418 Je.Ma.: Variedad Mejorada de Fréjol Arbustivo para la Cuenca del Rio Chota. Plegable No. 160. Instituto Nacional Autonomo de Investigaciones Agropecuarias, Estacion Experimental Santa Catalina, Programa Nacional de Leguminosas. Quito, Ecuador. 104 INIAP 1996b. Paragachi: Variedad Local de Fréjol Arbustivo. Plegable miscelanea. Instituto Nacional Autonomo de Investigaciones Agropecuarias, Estacion Experimental Santa Catalina, Programa Nacional de Leguminosas. Quito, Ecuador. INIAP 2004a. INIAP-414 Yunguilla: Variedad Mejorada de Fréjol Arbustivo. Plegable No. 253. Instituto Nacional Autonomo de Investigaciones Agropecuarias, Estacion Experimental Santa Catalina, Programa Nacional de Leguminosas. Quito, Ecuador. INIAP 2004b. INIAP-424 Concepcion: Variedad Mejorada de Fréjol Arbustivo Morado- Moteado. Plegable No. 254. Instituto Nacional Autonomo de Investigaciones Agropecuarias, Estacion Experimental Santa Catalina, Programa Nacional de Leguminosas. Quito, Ecuador. INIAP 2004c. INIAP-425 Blanco Fanesquero: Variedad Mejorada de Fréjol Arbustivo para Consumo en Grano Tierno. Plegable No. 252. Instituto Nacional Autonomo de Investigaciones Agropecuarias, Estacion Experimental Santa Catalina, Programa Nacional de Leguminosas. Quito, Ecuador. INIAP 2005. INIAP-420 Canario del Chota: Variedad mejorada de Fréjol Arbustivo de Color Armarillo. Plegable No. 258. Instituto Nacional Autonomo de Investigaciones Agropecuarias, Estacion Experimental Santa Catalina, Programa Nacional de Leguminosas. Quito, Ecuador. Johnson, N., D. Pachico, and C. Wortmann. 2003. The Impact of CIAT’s Genetic Improvement Research on Beans. Chapter 12 in R. Evenson and D. Gollin, eds. Crop variety improvement and its eflect on productivity: the impact of international agricultural research. Wallingford, UK: CAB International. Lépiz, R., E. Peralta, L. Minchala, and R. Jimenez. 1995. Diagnostico Agro socioeconomico del Cultivo de Fréjol en la Sierra Ecuatoriana. Documento de Trabajo INIAP-PROFRIZA No. 1. Proyecto Frijol de la Zona Andina (PROFRIZA). Quito, Ecuador. Lépiz, R. 1996. Variedades de Frijol Liberadas por los Programas Nacionales de Bolivia, Peru, y Ecuador: Periodo 1988-1995. Proyecto Frijol de la Zona Andina (PROFRIZA). Quito, Ecuador. Lichtenberg, E. and D. Zilberman. 1986. The econometrics of damage control: why specification matters. American Journal of A gricultural Economics 68(2): 26 l - 273. 105 Marasas, C., M. Smale, and R. Singh. 2003. The Economic Impact of Productivity Maintenance Research: Breeding for Leaf Rust Resistance in Modern Wheat. Agricultural Economics 29(3):253-263. Maredia, M., D. Byerlee and J. Anderson. 2000. Ex Post Evaluation of Economic Impacts of Agricultural Research Programs: A Tour of Good Practice. In The Future of Impact Assessment in the C GIAR: Needs, Constraints and Options. Rome, Italy: CGIAR Technical Advisory Committee Secretariat, F AO. Masters, W., B. Coulibaly, D. Sanogo, M. Sidibé, and A. Williams. 1996. The Economic Impact of Agricultural Research: A Practical Guide. Unpublished manuscript. West Lafayette, Indiana: Department of Agricultural Economics Purdue University. Mather, D., R. Bernsten, J. Rosas, A. Viana Ruano, and D. Escoto. 2003. The economic impact of bean disease resistant research in Honduras. Agricultural Economics 29:343-352. Mather, D. 2005. Essays on the Economic Impact Of Disease-Resistant Bean Research In Honduras. Ph.D. Dissertation, Department of Agricultural Economics, Michigan State University, East Lansing, MI. Mazon, Nelson, and Eduardo Peralta. 2005. E1 Comité de Investigacién Agricola Local (CIAL), una Estrategia para el Mejoramiento de la Produccion Agricola en Comunidades Rurales: La Experiencia del CIAL “Cuenca del Rio Mira.” Publicacion Miscelanea No. 131. Programa Nacional de Leguminosas y Granos Andinos (PRONALEG-GA), Estacion Experimental Santa Catalina, Instituto Nacional Autonomo de Invesigaciones Agropecuarias (INIAP). Quito, Ecuador. McFadden, D. 1973. Conditional logit analysis of qualitative choice behavior. In P. Zarembka, ed. Frontiers in econometrics New York, NY: Academic Press, pp 105-142. Morris, M., and P. Heisey. 2003. Estimating the benefits of plant breeding research: methodological issues and practical challenges. Agricultural Economics 29(3):241—252. Morris, M., H. Dubin, and T. Pokhrel. 1994. Returns to wheat breeding research in Nepal. Agricultural Economics 10(3):269-282. Myers, R. 2006. Qualitative Response Models. Class Notes from ABC 835: Introductory Econometrics. Department of Agricultural Economics, Michigan State University, East Lansing, MI. Oehmke, J., and E. Crawford. 1996. The Impact of Agricultural Technology in Sub- Saharan Africa. Journal of African Economies. 5(2):271-292 106 Peralta, E. 1991. Subproyecto 6: Control Integrado de Plagas. Informe Tecnico. Proyecto Frijol de la Zona Andina (PROF RIZA). Instituto Nacional Auténomo de Investigaciones Agropecuarias, Estacion Experimental Santa Catalina, Programa Nacional de Leguminosas. Quito, Ecuador. Peralta, E., V. Barrera, J. Unda, M. Guala, and M. Tacan. 2001. Estudio de la produccion, poscosecha, mercadeo y consumo de fre'jol arbustivo en el valle del Chota. Publicacién Miscelanea. Instituto Nacional Autonomo de Investigaciones Agropecuarias, Estacion Experimental Santa Catalina, Programa Nacional de Leguminosas. Quito, Ecuador. Pingali, P. 2001. Milestones in Impact Assessment Research in the CGIAR, 1970-1999. Mexico, D.F.: Standing Panel on Impact Assessment, Technical Advisory Committee of the Consultative Group on International Agricultural Research. Qaim, M., and A. de Janvry. 2005. Bt cotton and pesticide use in Argentina: economic and environmental effects. Environment and Development Economics 10:179- 200. Qaim, M., and G. Traxler. 2005. Roundup Ready soybeans in Argentina: farm level and aggregate welfare effects. Agricultural Economics 32:73-86. Rodriguez-Jaramillo, Lourdes. 1994. T enencia de la T ierra en los Valles del Chota y Salinas. F ondo Ecuatoriano Populorum Progresso, Quito, Ecuador. Rogers, E. M. 1962. Diffusion of Innovations. New York, Free Press of Glencoe. Sadoulet, E. and A. de Janvry. 1995. Quantitative Development Policy Analysis. The Johns Hopkins University Press: Baltimore. SICA. 2007a. Superficie mensual cosechada de fréjol seco 1990—2001. Servicio de Informacion y Censo Agropecuario (SICA). Ministerio de Agricultura, Ganaderia, Acuacultura y Pesca del Ecuador. http://www.sica.gov.ec/cadenas/fiejol/indexhtml (accessed March 22, 2007). SICA. 2007b. Precios mayoristas del fi'éjol tipo rojo moteado en el mercado de Ibarra 2000-2005. Servicio de Informacion y Censo Agropecuario (SICA), Ministerio de Agricultura, Ganaderia, Acuacultura y Pesca del Ecuador. http://www.sica.gov.ec/cadenas/frejol/index.html (accessed January 11, 2007). SIISE. 2001. Indicadores Sociales. CD-ROM. Sistema Integrado de Indicadores Sociales de Ecuador (SIISE), Ministerio de Coordinacién de Desarrollo Social, Gobiemo Nacional del Ecuador, Quito, Ecuador. 107 Smale, M., R. Singh, K. Sayre, P. Pingali, S. Rajaram, and H. Dubin. 1998. Estimating the economic impact of breeding nonspecific resistance to leaf rust in modern bread wheats. Plant Disease 82: 1055-1061. STATA. 2006. Regression with STATA: Regression Diagnostics. Chapter 2 in STATA Web Books, UCLA Academic Tecnology Services. http://www.ats.ucla.edu/stat/stata/webbooks/reg/chapterZ/statareg2.htm (accessed November 11, 2006). Subia C., P. Estrella, J. Ocafia, and E. Peralta. 2004. Monitoreo del uso de pesticidas en el control de plagas y enfermedades que afectan a1 cultivo de fi'e’jol en los valles de Chota y Mira de las provincias de Imbabura y Carchi, Ecuador, 2004. Informe de Trabajo. Instituto Nacional Auténomo de Investigaciones Agropecuarias, Estacion Experimental Santa Catalina, Programa Nacional de Leguminosas. Quito, Ecuador. Subia, C., E. Peralta, E. Falconi, D. Mooney, and S. Swinton. 2007. El cultivo de fie'jol arbustivo en los valles de Chota y Mira: Sistemas de produccion y uso de plaguicidas. Publicacion miscelanea No. 138. Instituto Nacional Autonomo de Investigaciones Agropecuarias, Estacion Experimental Santa Catalina, Programa Nacional de Leguminosas. Quito, Ecuador. Sunding, D., and D. Zilberman. 2001. The Agricultural Innovation Process: Research and Technology Adoption in a Changing Agricultural Industry. In B. Gardner and G.C. Rausser, eds. Handbook of Agricultural and Resource Economics, vol. 1, part 1. Amsterdam: North Holland, pp 207-261. Vasquez, J ., E. Peralta, J. Pinzon, and R. Lépiz. 1992. E1 Fréjol Arbustivo en Imbabura: Sugerencias para su Cultivo. Publicacién Miscelanea No. 57. Instituto Nacional Autonomo de Investigaciones Agropecuarias, Estacion Experimental Santa Catalina, Programa Nacional de Leguminosas. Quito, Ecuador. Vernooy, Ronnie. 2003. Seeds that give: participatory plant breeding. Ottawa, Canada: International Development Research Centre. Wooldridge, J. 2000. Introductory Econometrics: A Modern Approach. Cincinnati, OH: South-Westem College. Wooldridge, J. 2002. Chapter 18: Estimating Average Treatment Effects, in Econometric Analysis of Cross Section and Panel Data. Cambridge, MA: The MIT Press. World Bank. 2003. The CGIAR at 31: An Independent Meta-Evaluation of the Consultative Group on International Agricultural Research. Volume 1: Overview Report. Operations Evaluation Department. 108 APPENDIX ONE: VILLAGE-LEVEL SAMPLE SELECTION PROCESS 109 Figure A.1.1 = name map * = Indicates the number of communities on the map located from 1200 and 2600 m.a.s.1 L = Indicates that map corresponds to targeted area, but lies below 1200 masl. H = Indicates that map corresponds to targeted area, but lies above 2600 m.a.s.1. NC = Indicates that map corresponds to targeted area but does not contain villages NB = Indicates that map corresponds to targeted area but does not contain rural villages OT = Indicates that the entire map lies outside of the targeted area of influence Source: Cartas Topogréficas, Escala 1:50.000. E1 Instituto Geografico Militar, Quito, Ecuador 110 Table A.l.1 Villa es Located Within Tar eted Area of Im act b Ma Name . Community Names Map Name Community Names Ma Name La Carolina (11) La Carolina Bolivar (6) "*El Tambo Sn. Juan de Lachas Los Andes Sn. Fco. de Tablas Puntales Bajo Tablas Sn. Joaquin Luz de América Guaranton Chorrera de Tablas Cuesaca Guadal Tumbabiro (1) Salinas Naranjito Sn. Vic. de Pusir (11) "Sn. Vicente de Pusir El Corazon Mascarilla *El Naranjal *Chota Chiquito (Alor) El Rosal Tababuela Estac. Carchi (15) Guadrabamba “Turnbatl'l Potrerillos Pusir Chiquito Mascarillas Chota *Sta. Marianita Yacucaspi H. El Refugio Yacuscapi *Chirimoyal (Espadilla) "Chamanal Playa de Ambuqui “"‘Sta. Lucia Ambugui Sn. Francisco Carpuela (16) Cabras Cuajara La Cruz Hato Chamanal Puntales H. La Loma Sn. Joaquin Tarabita Sn. Fco. De Chutan Sn. Guillermo Sn. Fco. De Villacis Estacion Carchi El Izal H. Cabuyal ‘Cunquer La Concepcion (8) Empedradillo Pusir *El Milagro *Carpuela ‘Convalecencia Piquiucho "*La Concepcion El Rosal H. Sta. Ana El Juncal Santa Luisa ‘Sn. Fco. De Caldera Juan Moutalvo "Chalguayacu Santa Ana *Caldera Mira (10) *Cabuyal Pimampiro (13) "San Rafael "Piquer Pueblo Nuevo Huaquer Yunguilla Pueblo Viejo Pimampiro “Mira San Antonio Las Parcelas I"Santa Cecilia (San José) “Pisquer ‘Pugarpuela ‘La Portada El Tejar Lorna Sn. Juan Alto San Juan Yascén “Los Arboles (Yucatan) Ibarra (3) "‘“San Clemente El Inca ‘H. Piman La Mesa *Yucatan ‘Buenos Aires Notes: Targeted area of impact pertains to the cantons of Mira and Bolivar in the province of Carchi and the cantons of Ibarra and Pimampiro in the Province of lrnbabura; 94 total villages. ”"‘ Indicates a CIAL village; automatically included in survey (4) ” Indicates a village with previous INIAP extension intervention; automatically included in survey (10) “ Indicates a village without previous extension intervention and chosen at random (18) 111 Table A.l.2 List of villa es selected for inclusion in the survey sample No. Mapa Comunidades Canton Provincia 1 La Concepcion ***La Concemion Mira Carchi 2 Estacion Carchi ***Sta. Lucia Mira Carchi 3 Ibarra ***San Clemente Ibarra Imbabura 4 Bolivar ***El Tambo Bolivar Carchi 5 Estacion Carchi **Chamanal Mira Carchi 6 Mira **Piquer Mira Carchi 7 Mira **Mira Mira Carchi 8 Mira **Pisquer Mira Carchi 9 Sn. Vic. De Pusir **Sn. Vicente de Pusir Bolivar Carchi 10 Sn. Vic. De Pusir **Tumbatl'r Bolivar Carchi 1 1 Carpuela **Chalguayacl’l Pimampiro Imbabura 12 Pimampiro “Pimampiro (no beans) Pimampiro Imbabura l3 Pimampiro **Los Arboles Pimampiro Imbabura l4 Pimampiro **El Inca (Pat. Viejo) Pimampiro Imbabura 15 La Carolina *Naranjal Mira Carchi 16 Sn. Vic. De Pusir *Chirimoyalfllspadilla) Ibarra Imbabura l7 Sn. Vic. De Pusir *Chota Chiquita (Alor) Bolivar Carchi 1 8 Carpuela *Carpuela Ibarra Imbabura 19 La Concepcion *El Milagro Mira Carchi 20 Pimampiro *Santa Cecilia (San Jose) Pimampiro Imbabura 21 Mira *Cabuyal Mira Carchi 22 Estacion Carchi *Sta. Marianita Yacucaspi Ibarra Imbabura 23 Carpuela *EI Rosal (Cunquer) Bolivar Carchi 24 Estacion Carchi *Piman Ibarra Imbabura 25 Mira *La Portada Mira Carchi 26 Pimampiro *Buenos Aires Pimampiro Imbabura 27 Carpuela *Sn Fco. Chutan (Caldera) Bolivar Carchi 28 Pimampiro *Puggrpuela Pimampiro Imbabura 29 Pimampiro *San Rafael Bolivar Carchi 30 La Concepcion *Sta. Ana (Convalecencia) Mira Carchi R1 Sn. Vic. De Pusir Tababuela Ibarra Imbabura R2 La Carolina La Carolina Ibarra Imbabura R3 Ibarra *Yucatan Pimampiro Imbabura R4 Carpuela Puntales Bolivar Carchi R5 La Carolina El Rosal Mira Carchi R6 Pimampiro San Antonio Pimampiro Imbabura R7 La Congrpcion Santa Luisa Mira Carchi R8 Bolivar Puntales Bajo Bolivar Carchi R9 Carpuela Sn. Fco. De Villacis Bolivar Carchi R10 Estacién Carchi H. La Loma Mira Carchi R1 1 Bolivar Guaranton Bolivar Carchi R12 Estacion Carchi Cuajara Ibarra Imbabura R13 Bolivar Cuesaca Bolivar Carchi R14 Mira Loma San Juan Alto Mira Carchi R15 Pimampiro La Mesa Pimampiro Imbabura Notes: Villages denoted ** or *** were automatically included in the sample. **"' = CIAL village ** = Non-CIAL village with previous INIAP extension intervention * = Non-CIAL village without previous INIAP extension intervention Rl-RIS = Reserve communities (chosen at random) 112 APPENDIX TWO: CALCULATION OF SURVEY WEIGHTS 113 Table A.2.l Calculation of Survey Weights by Cluster Number of Level of INIAP 8 Sample Sampling Survey No. Cluster (Village) Interventionb ean c Size Probability Weight Producers . (S) (N g) (’7) (71?) (W1) 5( 1 La Concepcion 1 3 3 1.00 0.33 2 Santa Lucia 1 4 3 0.75 0.44 3 San Clemente 1 4 3 0.75 0.44 4 El Tambo l 4 3 0.75 0.44 5 La Concepcion“ 2 100 4 0.04 25.00 6 Santa Lucia” 2 32 4 0.13 8.00 7 San Clemente" 2 29 4 0.14 1.79 8 El Tamboa 2 90 4 0.04 22.50 9 Chamanal 3 60 4 0.07 15.00 10 Piquer 3 27 4 0.15 6.75 11 Mira 3 48 4 0.08 12.00 12 Pisquer 3 86 4 0.05 21.50 13 San Vincente 3 130 4 0.03 32.50 14 Tumbatu 3 1 10 4 0.04 27.50 1 5 Chalguayacu 3 200 4 0.02 50.00 16 Patio Viejo 4 70 4 0.06 17.50 17 Los Arboles 3 30 4 0.13 7.50 18 San José 4 30 4 0.13 7.50 19 Naranjal 4 23 4 0.17 5.75 20 Espadilla 4 44 4 0.09 l 1.00 21 Alor 4 60 4 0.07 15.00 22 Carpuela 4 300 4 0.01 75.00 23 El Milagro 4 13 4 0.31 3.25 24 Yucatan 4 l 0 4 0.40 .625 25 Cabuyal 4 30 4 0.13 7.50 26 Piman 4 44 4 0.09 1 1.00 27 Caldera 4 100 4 0.04 6.25 28 Imbiola y Sta. Marianita 4 16 4 0.25 4.00 29 La Portada 4 50 4 0.08 12.50 30 Buenos Aires 4 42 4 0.10 10.50 31 Cunquer 4 60 4 0.07 15.00 32 Pugarpuela 4 20 4 0.20 1.25 33 San Rafael 4 160 4 0.03 40.00 34 Convalecencia 4 23 4 0.17 5.75 Totals 2078 132 A — Villages are repeated here since farmers from multiple strata live within the same cluster b — Level of INIAP intervention is categorized into four stratification levels: 1) CIAL members from a CIAL village, 2) non-CIAL members from a CIAL village, 3) villages with previous INIAP intervention, and 4) villages without prior INIAP intervention c — All data was obtained from the village-level survey, except for San Clemente, Yucatan, Caldera and Pugarpuela for which data was estimated by INIAP researchers 114 Table A.2.2 Calculation of Survey Wews by Stratification Level Number of Sampled Level of INIA P Number of Began Households within Sampling Survey Intervention“ Producers each cluster Probability Weight ”0' (S) (W 02.) (7a) (ta) 1 1 15 12 0.800 1.25 2 2 251 16 0.064 15.69 3 3 953 32 0.034 29.78 4 4 3008c 72 .024 41.78 Totals 4227 132 a — Level of INIAP intervention is categorized into four stratification levels: 1) CIAL members fiom a CIAL village, 2) non-CIAL members from a CIAL village, 3) non-CIAL members from a village with previous INIAP intervention, and 4) non-CIAL members from a village with no prior INIAP intervention b — Data for stratification level 1-3 were obtained fi'om the community level survey, data for stratification levels 4 and the total number of producers was obtained from Ecuador’s 2001 Agricultural Census c — The value for N4 was obtained by subtracting N1, N2, and N3 from N and is therefore treated as a residual. This presents the possibility of biasing the influence of observations from communities without previous INIAP intervention. If the number of bean farmers has increased, then the weight assigned to N4 will be biased downward. If the number of bean farmers has decreased, then the weight assigned to N4 will be biased upward. 115 APPENDIX THREE: HOUSEHOLD-LEVEL QUESTIONAIRE 116 @in p EVALUACION DE IMPACTO DE NUEVAS VARIEDADES DE FREJOL EN EL NORTE DE ECUADOR —-CUESTIONARIO ACERCA DEL HOGAR-- Declaracién de Consentimiento Nosotros estamos conduciendo un estudio de irnpacto de variedades de fréjol resistentes a enfermedades liberadas en el Norte de Ecuador. Este estudio es realizado por el INIAP (Instituto Nacional Autonomo de Investigaciones Agropecuarias) en colaboracidn con Michigan State University (Universidad Estatal de Michigan). Me gustaria observar su campo de fréjol y hacerle algunas preguntas sobre su produccion. Su participacion es voluntaria. Si usted no desea participar en esta encuesta o desea suspender su participacion, usted no sera penalizado de ninguna manera. Esta encuesta consistira en una sola visita que tomara aproximadamente 45 minutos. Usted tiene plena libertad para no responder a las preguntas que le haga. Sin embargo, yo 1e alentaria a participar y a responder las preguntas porque sus respuestas nos ayudarén a mejorar los métodos de la produccion del fréjol. Toda la informacion que nos proporcione sera confidencial, lo cual implica que nadie mas tendra acceso a sus respuestas y su identidad permanecera protegida en cualquier publicacion relacionada con la informacion que nos proporcione. Su privacidad sera protegida a1 maximo de acuerdo a lo que es permitido por ley. Si usted tiene cualquier pregunta sobre este estudio, por favor comuniquese con el Profesor Scott M. Swinton en el departamento de Economia Agricola, Michigan State University, 304 Agricultural Hall, East Lansing MI 48824, Estados Unidos de América. Teléfono (1-517-3537218) y correo electronico swintons@msu.edu. Si usted tiene preguntas o dudas respecto a sus derechos como participante del estudio o esta insatisfecho en cualquier momento con cualquier aspecto del estudio, usted puede comunicarse con el doctor Peter Vasilenko, Ph.D., Jefe del Comité Universitario para Investigaciones que Involucren Aspectos Humanos (UCRIHS), teléfono (1-517-432- 4503), correo electronico irb@msu.edu, correo 202 Olds Hall, East Lansing MI 48824- 1047, Estados Unidos de América. Usted indica su acuerdo voluntario de participar en esta investigacion en siguiendo con la entrevista. 117 Gill 0 EVALUAC'ION DEL IMPACTO DE LAS NUEVAS VARIEDADES DE FREJOL ARBUSTIVO EN EL NORTE DEL ECUADOR DMD/COMIC”? --CUESTIONARIO ACERCA DEL HOGAR-- A. Datos Basicos **Datos para llenar antes que empieza la entrevista** A1. Numero de la encuesta A2. Fecha de A3. Iniciales del del hogar: Entrevista: _/ /_ entrevistador: dia / mes / afio A4. Comunidad: (marca una) [ 1 ] La Concepcion [ 9 ] San Vicente [ 17 ] Puntales [ 25 ] La Portada [ 2 ] Santa Lucia [ 10 ] Tumbatu [ 18 ] Carpuela [ 26 ] Yunguilla [ 3 ] San Clemente [ 11 ] Chalguayacu [ 19] E1 Milagro [27 ] Cunquer [4 ] El Tambo [ 12 ] Pimampiro [ 20] Santa Rosa [ 28 ] Pugarpuela [ 5 ] Chamanal [ l3 ] Los Arboles [ 21 ] Cabuyal [ 29] San Rafael [ 6 ] Piquer [ 14 ] El Inca [ 22 ] Sta. Marianita [ 30 ] Santa Ana [ 7] Mira [ 15 ] Naranjal [ 23 ] Caldera [ 8 ] Pisquer [ 16 ] Espadilla [ 24 ] lrnbiola A5. Canton: (marca una) Imbabura: Carchi: [ 11 ] Ibarra [21 ]Bolivar [ 12 ] Pimampiro [ 22 ] Mira A6. Valle (marca una) [ 1 ]Mira [ 2 ] Chota [ 3 ] Salinas A7. Datos GPS del hogar (OBSERVACION) a. Latitud ° . ' (grados y minutos N) b. Longitud ° . ' (grados y minutos W) c. Altitud (metros s.n.m.) 118 B. Datos del Hogar **Antes que discutamos sobre la produccion de fréjol, quisiera saber un poco acerca del hogar" Bl. Nombre y apellidos del entrevistado(a): Parentesco del entrevistado respecto a1 jefe del hogar: (marca una) [ l ] Jefe del hogar [ 2 ] Esposa/o [3 ] Hijo/a [ 99 ] Otro, Especifique: B2. Jefe del Hogar 3. (Que edad tiene? b. LCuantos afios de educacidn? c. Jefe(a) del hogar es: [0] Hombre [ 1 ] Mujer d. LCual es su actividad principal? (marca una) [ 1 ] Agricultor [6] Comerciante [ 2 ] Jomalero [ 7 ] Construccion/Albaflil [ 3 ] Ama de casa [ 8 ] Duefio de negocio [ 4 ] Empleada domestica [ 9 ] Jubilado [ 5 ] Empleado/a en oficina/tienda [ 99] Otro, Especifique: e. (A que' organizaciones pertenece? 1. LAgricultores? no/ 51' ii. gRegantes? no/ si iii. LCIAL‘? no/si iv. gMujeres? no/ 51 v. LCooperativa? no/ si B3. LQuién toma decisiones sobre el cultivo de fre'jol? (marca una) [ l ] Jefe del hogar crasa a pregunta B4) [ 2 ] Esposa/o (sigue con el orden dc preguntas) [ 3 ] Hi jo (sigue con el orden dc preguntas) [ 99] Otro, Especifique: a. gQué edad tiene? b. LCuantos afios de educacion? c. Esta persona es: [0] Hombre [ 1 ] Mujer d. gCual es su actividad principal? (marca una) [ 1 ] Agricultor [ 6 ] Comerciante [ 2 ] Jomalero [ 7] Construccién/Albafiil [ 3 ] Arna de casa [ 8 ] Dueflo de propio negocio [ 4 ] Empleado/a domestico/a [ 9 ] Jubilado [ 5 ] Empleado/a en oficina/tienda [ 99] Otro, Especifique: 119 e. (A qué organizaciones pertenece? i. LAgricultores? no/si ii. LRegantes? no/ si iii. LCIAL? no / si iv. gMujeres? no / si v. LCooperativa? no / 51 B4. LCuantos familiares vivian en el hogar en enero de 2006? a. (De ellos, cuantos trabajan? b. gDe ellos, cuantos dependen de los que trabajan? BS. LHay algun nifio de entre 6 y 12 afios que no asista a la escuela? [0] No [ 1 ] Si B6. LTiene familiares que le apoyaron economicamente y que vivian fuera de la casa en enero de 2006? [0 ] No [ l ] Si B7. Vivienda a. (gTiene electricidad instalada? [0 ] No [ 1 ] Si b. LDispone de agua en casa? [0 ] No [ 1 ] Si c. LDispone de servicio higiénico? {(1); 2.0 l (1. Material predominante de paredes exteriores [ O ] No permanente (plastico, cafla, carton, lata, otros) [ 1 ] Permanente (adobe, bloque, ladrillo, cemento, otros) e. Material predominante del piso [ 0 ] Tierra [ l ] Otro material (cemento, madera, mixto, etc.) f. Numero de cuartos (sin contar cocina, bafio): C. Produccion del Fréjol **Ahora vamos a hablar acerca de su produccion del fi'éjol en el primer ciclo 2006'” C 1. LAproximadamente, cual es la superficie total de tierra que usted cultivo? a. Superficie total: b. Unidad de medida: 120 C2. gAproximadamente, cuanto de esa superficie sembro con fréjol en el primer ciclo 2006? a. Superficie total: b. Unidad de medida: C3. LCuales variedades sembro Ud. en el primer ciclo de 2006? Variedad Cantidad de Semilla Superficie Rendimiento (nombre que to da e1 agricultor) Sembrado (qq) aproximada (Ha.) (qq total) 1. 2. 3. 4. C4. (JDonde suele vender e1 fréjol seco? [ 0 ] La misma localidad que no sea consorcio (pasar a pregunta C4) [ 1 ] Consorico [ 5 1 Quito [ 2 ] Ibarra [ 6] Colombia [ 3 ] Pimampiro [ 7] E1 Juncal [4 ] Tulcan [99 ] Otro, Especifique: a LCuanto cuesta transportar un quintal de fi'éjol seco hasta alla? $ b. LCuanto tiempo de viaje demora en transporte particular? horas C5. LRecibio Ud. algl'm crédito para el cultivo de fréjol durante el primer ciclo de 2006? [0 ] No [ 1 ] Si C6. LCdmo decide cuando aplicar plaguicidas en el cultivo de fiéjol? [ 0 ] No aplica plaguicidas en la produccion dc fréjol (pasa a pregunta C15, pg 6) [ 1 ] Aplica plaguicidas por costumbre (0 sea, en intervalos de tiempos fijos) [ 2 ] Aplica plaguicidas cuando aparece algun sintoma dc daflo [ 3 ] Aplica plaguicidas cuando lo recomienda un técnico [ 99] Otro: Especifique C7. gQuién 1e indica qué plaguicidas aplicar? a. LVendedor? Si / No b. LTécnicos (no vendedores) Si / No c. LTecnico de consorcio? Si / No (1. (familiares? Si / No e. (,Vecino? Si / No f. (,Propia experiencia? Si / No g. (,Radio? Si / No h. LOtro? Si / No, Especifique: C8. gQué tipo de equipo de fumigar es el que mas usa y cual es su capacidad? [ 1 ] Bomba de mochila Capacidad: [ 97 ] Sin repuesta [ 2 ] Bomba de motor Capacidad: [ 98 ] No sabe [ 3 ] Bomba de motor (manguera) Capacidad: [ 99 ] Otro, Especifique: 121 C9. (Durante el primer ciclo de produccion de fi'e'jol de 2006, quién aplico los plaguicidas? [ 1 ] (,Miembros del hogar?* Si / No [ 2 ] (,Partidario? Si / No [ 3 ] (Jomalero? Si / No (Si responden que NO pasa a C10) C10. LPor que' se contrata a un jomalero? (marca uno) [ 1 ] Miembros del hogar no saben como aplicar plaguicidas [ 97 ] Sin repuesta [ 2 ] No alcanza el tiempo o familiares disponibles [ 98 ] No sabe [ 3 ] Prefiere proteger los miembros de la familia [ 99 ] Otro, Especifique: de exposicion a los plaguicidas * Si ningtin miembro del hogar aplico' plaguicidas (vea Ia respuesta a pregunta C 9a), pasar a la pregunta C16. Si algtin miembro del hogar si glicé plaguicidas, seguir con el order: dc preguntas. C11. LDurante el primer ciclo de produccion de fréjol de 2006, paso alguna vez que después de aplicar plaguicidas su piel quedaba mojada de producto? 0 No 1 i 1 a C12. (En el primer ciclo de produccion de 2006, comic 0 bebié (refrescos) en el campo durante la aplicacion en el cultivo de fi'éjol? [0]No [ 1 ] Si C13. (Durante el primer ciclo de produccion de 2006, firmaba mientras fumigo e1 cultivo de fre'jol con plaguicidas? [0]No [ l ] Si C14. LCuél es el color de etiqueta de plaguicida de mayor peligrosidad? (Sin leer los colores, marca una respuesta) [ 1 ] Amarillo [97] Sin Respuesta [2]Azu1 [98]Nosabe [ 3 ] Rojo [ 99] Otro, Especifique: [4] Verde C15. (,Qué ropa y equipo protector uso para la aplicacion de plaguicidas? Ropa o Equipo Si / No a LMéscara protectora? Si / no b LGafas protectoras? si / no c. (,Guantes de caucho? si / no d. LProtector de la espalda? si / no e. LCamisa de manga larga? si / no f. LPantalones protectores? si / no g. LBotas? si / no h. 1,0tro?, Especifique: _ si / no C16. 5H3 asistido a algl'm seminario sobre el manejo de plagas o la aplicacion de plaguicidas en los ultimos dos afios? [0]No [ l ] Si, LNl'lmero de dias?: 122 D. Sintomas de Intoxicaciones Agudas D1. {Antes del afio 2006, ha experimentado un miembro del hogar algunos de los siguientes Sintomas dentro de pocas horas después de aplicar plaguicidas? *Ninguno X a. LIrritacion de la piel? si /no g. (JDolor de cabeza? si / no b. gVision nublada? sl' / no h. (,Diarrea? si /no c. Llrritacion de los ojos? si /no i. (,Dolores musculares? si /no (1. LNausea y vomito? si /no j. LDificultades en respirar? si /no e. gMolestias estomacales? si / no k. gAngustia — Desesperacion? si / no f. gMareo? si / no 1. gDesmayo? si /no m. (,Otro?, Especifique: D2. (Durante el primer ciclo de 2006, en el cultivo de fréjol ha experimentado un miembro del hogar algunos de los siguientes Sintomas dentro de pocas horas después de aplicar plaguicidas? *Ninguno X a. (,Irritacion de la piel? Si /no g. (,Dolor dc cabeza? si / no b. LVision nublada? si /no h. LDiarrea? si /no c. Llrritacion de los ojos? si /no i. (Dolores musculares? si /no (1. LNausea y vomito? si /no j. LDificultades en respirar? si /no e. LMolestias estomacales? si / no k. LAngustia — Desesperacion? si /no f. LMareo? si / no 1. LDesmayo? si / no m. LOtro?, Especifique: * Si respondieron Ninguno en D2, pasar a pregunta E Si respondieron que S_I en alguna opcién de D2, seguir con las preguntas . . . Ahora me gustaria preguntarles acerca del sintoma que experimenté después de fumigar elfréjol y sobre los gastos del remedio . . . D3. (gCuanto gastaron en medicinas (sin receta médica)? $ D4. (gNumero de dias incapacitado en la casa? dias D5. LBusco atencion médica? (marca uno) [ 0 ] = No (pasa a pregunta D6) [ 1 ] = Si (siga can preguntas abajo) a. LNl'rmero de visitas a1 doctor 0 clinica? b. gCuantos adultos acompailaron a1 enfermo? c. LCosto de consulta por visita? $ (1. (,Costo de medicacion comprada con receta médica? S e. LCosto de transporte de ida y vuelta por visita? $ f. LTiempo dedicado a visitar la clinica por visita? horas En casa de haber sido internado: g. (,Nl'lmero de dias hospitalizado? dias h. gCosto total de estadia en la clinica 11 hospital? $ D6. LHabia otros gastos relacionados con la intoxicacion? (Total) $ *Especifique: 123 F ICHA PARCELA **Me gustaria visitar la parcela ma’s grande defrefiol sembrado par usted en el primer ciclo de 2006 y preguntarles rm poco rmis sobre la variedad usada y su produccio'n. Adema's, me gustaria ver semillas de reserva o visitar urr late que esté actualmente con la misma semilla del primer ciclo 2006 ** E. Parcela mas Grande Sembrada dc Fréjol - Primer Ciclo de 2006 (Nata: Si se presentan ma’s que rma variedad sembrada en la parcela, las siguientes preguntas solo debe referirse a la variedad con rmis superficie y no todas las variedades) E1. C6mo identifica a la parcela: E2. {Tenencia de la parcela primer ciclo 2006? (marca uno) [ l ] Es dueflo de la parcela [ 2 ] Arrendo la parcela [ 3 ] Fue partidario E3. Pendiente estirnada de la parcela: (marca una) [ 1 ] Plano [ 2 ] Poco inclinado [ 3 ] Ladera E4. Clasificacion del suelo que mejor describa la parcela (marca una) [ 1 ] Arcilloso [ 2 ] Arenoso [ 3 ] Limo (Franco) E5. Presencia de piedras: (marca una) [ 0] Sin piedras [ l ] Pocas [ 2 ] Muchas E6. gDispone de agua dc riego? [0]No [ 1 ] Si E7. (,Que' cultivo estaba sembrado en esta parcela antes del fréjol del primer ciclo 2006? [ 0] Ninguno [ 1 ] Fréjol [ 97 ] Sin respuesta [ 2 ] Leguminosa no fréjol (arveja, otros) [ 98 ] No sabe [ 3 ] Cereal 0 graminea (maiz, cafla, pastos) [ 99 ] Otro, Especifique: [ 4 ] Hortaliza (tomate, pimiento, aji, otros) [ 5 ] Frutales (Limon, aguacate, otros) [ 6] Raiz o tubérculo 124 F. Variedades del Fréjol **Las siguientes preguntas se refieren a la variedad sembrada en el primer ciclo de 2006* * F1. Variedad dc fréjol sembrada: a. Nombre Dado por el Productor: [ l ]Concepcion [ 10]Capuli [ 18 ] Rojo [2] Paragachi [ 11 ] Negro [ 19] Margarita [ 3 ] Yunguilla [ 12 ] Calima rojo [ 20 ] Uribe [ 4 ] Blanco fanesquero [ l3 ] Calima negro [ 21 ] Matahambre [ 5 ] Je.Ma. [ 14 ] Canario Pallatanga [ 22 ] Toa [ 6] Canario del Chota [ 15 ] Cargabello [ 23 ] Torta [ 7 ] Imbabello [ l6 ] Blanco de leche [ 24 ] Panamito [ 8 ] Selva [ l7 ] Campeén [ 25 ] Magola / Magolita [ 9 ] Injerto [ 99 ] Otro, Especifique: b. Color de grano (marca una) [ 1 ] Rojo moteado [4] Amarillo [ 97 ] Sin respuesta [ 2 ] Morado moteado [ 5 ] Rojo solido [ 98 ] No sabe [ 3 ] Blanco [ 6 ] Negro [ 99 ] Otto *Especifique: c. Comprobacion: (marca una) (Nombre luego de ver Ia plants/semilla — dado por Encuestador) [ 0 ] No se puede comprobar [ 1 ] Concepcion [ 10 ] Capuli [ 18 ] Rojo [2] Paragachi [ 11 ] Negro [ 19] Margarita [ 3 ] Yunguilla [ 12 ] Calima rojo [20] Uribe [ 4 ] Blanca fanesquero [ l3 ] Calima negro [ 21 ] Matahambre [5 ] Je.Ma. [ 14 ] Canario Pallatanga [22 ] Toa [ 6 ] Canario del Chota [ 15 ] Cargabello [ 23 ] Torta [ 7 ] Imbabello [ 16 ] Blanco de leche [ 24] Panamito [ 8] Selva [ 17] Campeon [25 ] Magola / Magolita [ 9] Injerto [ 99] Otro, Especifique: F2. gDe dondc obtuvo la semilla que sembro? (marca una) [ 1 ] Guardo de la cosecha anterior [ 2 ] Agricultor de la comunidad [ 3 ] Agricultor de otra comunidad ‘Cual: [ 97 ] Sin respuesta [4 ] Mercado *Cual: [ 98] No sabe [ 5 ] Organismo gubemamental, *Cual: [ 99 ] Otro, ‘Especifique: [ 6 ] Organismo no-gubernamental, ‘Cual: F3. gPorque decidio obtener su semilla de ese lugar/persona/etc.? F4. {Hace cuantos afios que siembra esta variedad? 125 F5. LConoce e1 origen de ésta variedad? (marca una) [ 1 ] Agricultor, *De donde: [ 97 ] Sin respuesta [ 2 ] Mercado *Cual: [ 98 ] No sabe [ 3 ] Organismo gubemamental, *Cual: [ 99] Otro, *Especifique: [4 ] Organismo no-gubernamental, *Cual: F6. (,Por qué prefiere la variedad actual sobre otras variedades? (marcar hasta 3 y el orden) Orden (l a 3): Orden (1 a 3): [ 1 ] Precio bajo de la semilla [ 6 ] Resistencia a enfermedades [ 2 ] Alto rendirniento _ [ 3 ] Se vende a mejor precio [ 4 ] Requiere menos insumos [ 5 ] Calidad del producto [ 7 ] Resistencia a plagas [ 8 ] No hubo otras opciones [ 9 ] Autoconsumo [ 99 ] Otro, : Especifique: —._..._.. F7. (En el primer ciclo 2006, qué cantidad de semilla sembro en ésta parcela? a. Cantidad: qq b. Precio: $ / qq c. En que mes? F8. LCual the el rendirniento total de la parcela? a. fréjol seco: qq b. en tiemo: bultos (grano seco estimado qq) F9. (381 hubiera contratado jomales sin uso de maquinaria, con cua'ntos jomales hubiera cosechado y trillado? F10. gCuanto de la cosecha total vendio? a. fréjol seco: qq b. en tiemo: bultos F11. (3A qué precio vendio la cosecha? a. fréjol seco: $ /qq b. en tiemo: $ /bu1tos 0. En que mes? F12. bCémo compara e1 dafio de enfermedades durante el primer ciclo de 2006 con afios anteriores? [ 1 ]mayor [97 ] sin respuesta [2]menor [98]nosabe [ 3 ] lo misrno F13. (,Cuales de estas enfermedades se presentaron en esta parcela durante el primer ciclo de 2006, aun si no las reconoce por nombre? (Deja que el entrevistado identifica cuales usando las fotos) l. si/no 5.si/no 9. si/no 2. si/no 6. si/no 10.si/no 3. si/no 7. si/no 11.si/no 4. si/no 8. si/no 126 F14. LCuales de estas enfermedades reconoce por nombre, al'm si no se presentaron en su fmca? (Ensefla las fotos de las enfermedades a1 entrevistado y marcar con X solamente las enfermedades correctamente identificadas) [0] NINGUNA [ 1 ] Antracnosis [5] Pudricion radicular [ 9 ] Moho blanco [2] Roya [6] Mustia [10] Mancha anillada [ 3 ] Mancha angular [ 7] Virus [ 11 ] Afiublo de halo [4] Bacteriosis [8] Oidiooceniza F15. LCémo compara e1 dafio de plagas durante el primer ciclo 2006 con afios anteriores? [ l ] mayor [97 ] sin respuesta [2]menor [98]nosabe [ 3 ] lo mismo F16. gCuales plagas se presentaron durante el primer ciclo de 2006? (marca todas que aplican) Mosca blanca(palomi11a) [ 6 ] Pinda/mariquita/diabrotica [ l l [ 2 ] Trips [ 7 ] Saltador dc hojas [ 97 ] Sin respuesta [ 3 ] Trozadores [ 8 ] Minadores [ 98 ] No sabe [4] Arafia roja [ 9] Enrollador [ 99] Otro *Especifique: [ 5 ] Grillo topo F17. LQue’ variedad de semilla sembraba antes que la variedad actual? a. Nombre Dado por el Productor: [ 1 ] Concepcion [ 10 ] Capuli [ 18 ] Rojo [ 2] Paragachi [ 11 ] Negro [ 19 ] Margarita [ 3 ] Yunguilla [ 12 ] Calima rojo [ 20] Uribe [4 ] Blanco fanesquero [ 13 ] Calima negro [ 21 ] Matahambre [ 5 ] Je.Ma. [ 14] Canario Pallatanga [22 ] Toa [ 6 ] Canario del Chota [ 15 ] Cargabello [ 23 ] Torta [ 7 ] Imbabello [ 16 ] Blanco de leche [ 24] Panamito [ 8 ] Selva [ 17 ] Campeon [ 25 ] Magola / Magolita [ 9] Injerto [ 99] Otro, Especifique b. Color de Grano (marca una) [ 1 ] Rojo moteado [4 ] Amarillo [ 97] Sin respuesta [ 2 ] Morado moteado [ 5 ] Rojo solido [ 98 ] No sabe [3] Blanco [6]Negro [99]Otro *Especifique: G. Uso de plaguicidas G1. gCuantas veces aplico fertilizantes quimicos a1 suelo? a. LCuantos quintales aplico en total? G2. LCuantas veces aplico herbicidas? G3. LCuantas fumigaciones realizo? 127 G4. Plaguicidas Usados Por Fumigacion (incluye insecticidas, firngicidas y abonos foliares) Nombre comercial Unidad (marcar unidad) Aplicacion #1: (Numero de tanques: ) Nombre comercial Unidad (marcar unidad) Aplicacion #2: (Numero de tanques: ) Aplicacion #3: (Numero de tanques: ) Aplicacion #4: (Numero de tanques: ) 128 C osto/unidad Dosis usada ($) (por tanque 20 l) C osto/unidad Dosis usada ($) (por tanque 200 I) Aplicacion #5: mumero de tanques: ) Aplicacion #6: (Numero de tanques: ) H. Observaciones de la Parcela H1. Datos GPS de la parcela: (Desde el centro de la parcela) a. Latitud . ' (grados y minutos N) b. Longitud ° . (grados y minutos W) c. Altitud (metros s.n.m.) H2. Croquis de la parcela (para medir superficie) 129 APPENDIX FOUR: COMMUNITY-LEVEL QUESTIONAIRE 130 C5? mine ESTUDIO DE IMPACTO DE VARIEDADES DE FREJOL RESISTENTES A ENFERMEDADES LIBERADAS EN EL NORTE DE ECUADOR --CUESTIONARIO ACERCA DE LA COMUNIDAD-- Declaracion de Consentimiento Nosotros estamos conduciendo un estudio de impacto de variedades de fréjol resistentes a enfermedades liberadas en el norte de Ecuador. Este estudio es realizado por el INIAP (Instituto Nacional Autonomo de Investigaciones Agropecuarias) en colaboracion con Michigan State University (Universidad Estatal de Michigan). Me gustaria hacerle algunas preguntas sobre la comunidad y su agricultura. Su participacion es voluntaria. Si usted no desea participar en esta encuesta o desea suspender su participacion, usted no sera penalizado de ninguna manera. Esta encuesta consistira en una sola visita que tomara aproximadamente 30 a 40 minutos. Usted tiene plena libertad para no responder a las preguntas que le haga. Sin embargo, yo 1e alentaria a participar y a responder las preguntas porque sus respuestas nos ayudaran a mejorar los me'todos de la produccion del fréjol. Toda la informacion que nos proporcione sera confidencial, lo cual implica que nadie mas tendra acceso a sus respuestas y su identidad permanecera protegida en cualquier publicacion relacionada con la informacion que nos proporcione. Su privacidad sera protegida a1 maximo de acuerdo a lo que es permitido por ley. Si usted tiene cualquier pregunta sobre este estudio, por favor comuniquese con el Profesor Scott M. Swinton en el departamento de Economia Agricola, Michigan State University, 304 Agricultural Hall, East Lansing MI 48824, Estados Unidos de América. Teléfono (1—517-3537218) y correo electronico swintons@msu.edu. Si usted tiene preguntas o dudas respecto a sus derechos como participante del estudio o esta insatisfecho en cualquier momento con cualquier aspecto del estudio, usted puede comunicarse con el doctor Peter Vasilenko, Ph.D., Jefe del Comité Universitario para Investigaciones que Involucren Aspectos Humanos (UCRIHS), teléfono (1-517-432- 4503), correo electronico irb@msu.edu, correo 202 Olds Hall, East Lansing MI 48824- 1047, Estados Unidos de América. Usted indica su acuerdo voluntario de participar en esta investigacion en siguiendo con la entrevista. 131 @in p ESTUDIO DE lMPACTO DE VARIEDADES DE FREJOL RESISTENTES A ENFERMEDADES LIBERADAS EN EL NORTE DE ECUADOR --CUESTIONARIO ACERCA DE LA COMUNIDAD-- A. Informacion General: A1. Numero de la encuesta A2. F echa de Entrevista: comunitaria: _/ / dia / mes / afio A4. Comunidad: (marca una) [ 1 ] La Concepcion [ 9 ] San Vicente [ 17 ] Puntales [ 2 ] Santa Lucia [ 10 ] Tumbatu [ 18 ] Carpuela [ 3 ] San Clemente [ 11 ] Chalguayacu [ 19] El Milagro [ 4 ] El Tambo [ 12 ] Pimampiro [ 20] Santa Rosa [ 5 ] Chamanal [ 13 ] Los Arboles [ 21 ] Cabuyal [ 6 ] Piquer [ 14] E1 Inca [ 22 ] Sta. Marianita [7] Mira [ 15] Naranjal [23 ] Caldera [ 8 ] Pisquer [ 16 ] Espadilla [ 24 ] Imbiola A5. Canton: (marca una) Imbabura: Carchi: [11]Ibarra [21]BoIivar [ 12] Pimampiro [22 ] Mira A6. Valle (marca una) [ 1 ] Mira [ 2 ] Chota [ 3 ] Salinas 132 A3. Iniciales del entrevistador: _ [ 25 ] La Portada [ 26 ] Yunguilla [ 27 ] Cunquer [ 28 ] Pugarpuela [ 29 ] San Rafael [ 30 ] Santa Ana B. Servicios Pl'lblicos e Infraestructura Transporte Bl. gQué tipo de transporte publico llega a la comunidad? (Menciona) [ 0 ] Ninguna gCuantos minutos hay que caminar para encontrar transporte publico? [ 1 ] Bus de ruta [ 2 ] Carnioneta (carrera) [ 99 ] Otro, Especifique: Educacion B2. LCuél es el nivel de educacion mas alta que se puede completar dentro de la comunidad? [ 0] Ninguna [ 1 ] Primaria / Escuela [ 2 ] Secundaria / Colegio Salud B3. gQué facilidades de salud publica y privada se encuentran en la comunidad? [ 0 ] Ninguna [ l ] Sub-Centro de salud publico [ 2 ] Centro de salud publico [ 99 ] Otro, Especifique: B4. LDonde esta 1a facilidad de salud con un doctor permanente mas cercana? [ l ] Misma comunidad (Pase a pregunta B6) [ 2 ] Comunidad vecina [ 3 ] Jefatura parroquial [4 ] Ciudad secrmdaria (Mira, Pimampiro, etc.) [ 5 ] Ciudad principal (Ibarra / Tulcan) [ 99] Otro, Especifique: a. Qué tiempo en vehiculo horas b. Costo de viaje (ida y vuelta) : $ Infraestructura B6. LQué tipo de servicio de telefonia tiene la comunidad? (Marcar todas que aplican) [ 0 ] Ninguno [ 1 ] Convencional [ 2] Celular C. Informacion Agricola Informacion demogra'fica Cl. LCuantas personas habitan en la comunidad? C2. LCuantas familias hay dentro de la comunidad? C3. LCuantas familias se dedican a la agricultura en la comunidad? C4. (JDe ellos (familias agricolas), cuantos siembran el fiéjol? 133 C5. LCuanto se paga el jomal agricola? (sin almuerzo) C6. LExiste alguna diferencia en el costo del jornal, segl'm 1a labor agricola(pa1a, aplicacion, etc.)? [0 ] No [ l ] Si, Razon: Cultivos principales C7. LCuales son los cultivos principales en la comunidad? a. Fréjol no/si b. Maiz no/ si c. Tomate no/ 51 d. Cafla no/ si e. Frutales no/si Especifique: f. Raicesotubérculos no/ si Especifique: g. Otrasleguminosas no/ 51 Especifique: h. Otros ccreales no/ si Especifique: i. Otro no/si Especifique: Insumos C8. LDonde se compran los insumos agricolas? (marca todos que aplican) a. Almacén agropecuaria dentro de la comunidad no / 51 b. Comerciantes arnbulantes que entran a la comunidad no/ si e. Alrnacenes fuera de la comunidad no / si d. Cooperativa / Consorcio no / si e. Otro no / si Especifique: C9. LExisten lotes de produccion exclusivamente para hacer semilla dc fréjol dentro de la comunidad? [ 0 ] No (pasa a pregunta C10) [ 1 ] Si (siga con la pregunta C9.a) a. LES e1 productor? [ l ] Individual [ 2 ] Empresa privada [ 3 ] Grupo comunitario *Especifique cual: C10. LCuales variedades estan cultivando? (marca todos que aplican) Nombre Dado por el Productor: [ 1 ]Concepcién [9] Capuli [ 17 ] Rojo [2 ] Paragacbi/ Injerto [ 10 ] Negro [ 18 ] Margarita [3 ] Yunguilla [ 11 ]Calima rojo [ l9] Uribe [ 4] Blanco fanesquero [ 12 ] Calima negro [ 20] Matahambre [5 ] Je.Ma. [ 13 ] Canario Pallatanga [21 ]Toa [ 6] Canario del Chota [ 14 ] Cargabello [ 22 ] Torta [ 7] Imbabello [ 15 ] Blanco de leche [ 23 ] Panamito [ 8] Selva [ 16 ] Campeon [ 24 ] Magola / Magolita [ 99 ] Otro, Especifique: 134 Observaciones: Fuentes de A gua para la produccion agricola C11. Dispone de agua de riego? [0]No [ 1 ] Si meses/ai‘lo frecuencia de turno promedio (cada cuantos dias le corresponde el turno) % Area de la comunidad bajo riego Observaciones: D. Acceso al Mercado Exterior D1. A qué distancia se encuentra e1 mercado de fiéjol mas importante para la comunidad? a. En minutos: b. (,Cual es el costo de transportar un quintal de fréjol ha este mercado? D2. Mencionen los canales diferentes para vender su cosecha de fréjol y el orden de importancia a la comunidad: [ 1 ] Venta directa a1 mercado mayorista Orden( ) [ 2 ] Venta al interrnediario en la finea Orden ( ) [ 3 ] Consorcio o Cooperativa Orden ( ) [ 99 ] Otro, Especifique: Orden ( ) Observaciones: 135 E. Organizaciones Comunitarias g'Cudles organizaciones comunitarias realizan actividades relacionados con la produccio'n delfrejol? Organizacion #1: Tipo de organizacion: [ 0] Ninguna [ l ] CIAL [ 2 ] Cooperativa [ 3 ] Asociacion de productores [ 4 ] Grupo de semilleristas [ 5 ] Grupo de Mujeres [ 99 ] Otro, Especifique: gCon qué se relacionan sus actividades principales? Semilla Si / No Credito Si / No Plaguicidas Si / No Comercializacion Si / No Investigacion Si / No Otro, Si / No Especifique: Observaciones: Organizacion #2: Tipo de organizacion: [ 0] Ninguna [ 1 ] CIAL [ 2 ] Cooperativa [ 3 ] Asociacion de productores [ 4 ] Grupo de semilleristas [ 5 ] Grupo de Mujeres [ 99 ] Otro, Especifique: aCon qué se relacionan sus actividades principales? Semilla Si / No Crédito Si / No Plaguicidas Si / No Comercializacion Si / No Investigacion Si /No Otro, Si / No Especifique: Observaciones: 136 F. Apoyo de Organismos Externos g'Cuciles organismos extemos han realizado actividades relacionados con la produccio'n delfre'jol dentro de los ultimos cinco arias? #1: Nombre del organismo: Tipo de organismo: [ 1 ] Organismo del gobiemo [2]ONG [ 3 ] Empresas privadas (Consorcio) [99 ] Otros. Especifique: T igg dc £2020: [ l ] Semilla no / si [ 2 ] Crédito no / Si [ 3 ] Plaguicidas no / si [4 ] Comercializacion no / si [ 5 ] Investigacién no / si [ 6] Capacitacion no / si En que? [ 7 ] Asistencia técnica no / si En que? [ 8 ] Otro no / si Especifique: Duracion: Ailo que empezo: Aflo que termino: Observaciones: #2: Nombre del organismo: Tipo de organismo: [ 1 ] Organismo del gobiemo [2]ONG [ 3 ] Empresas privadas (Consorcio) [ 99 ] Otros. Especifique: T la de 2020: [ 1 ] Semilla no / si [ 2 ] Cre'dito no / si [ 3 ] Plaguicidas no / si [4 ] Comercializacion no / si [ 5 ] Investigacion no / si [ 6] Capacitacion no / si En que? [ 7 ] Asistencia técnica no / si En que? [ 8 ] Otro no / si Especifique: Duracion: Ado que empezoz Aflo que terminoz Observaciones 137 APPENDIX FIVE: DESCRIPTION OF THE UNSATISFIED BASIC NEEDS INDEX The Unsatisfied Basic Needs (UBN) index is used in this thesis as a proxy for household wealth. It is assumed that wealthier households will have met a greater number of basic needs, as defined in Figure A.5.1, than will have less wealthy households. The calculation of household-specific UBN indices was achieved by including relevant questions into the field survey (included in Appendix 3). Survey results fi'om 132 interviews found 46% of household classified as poor, and 17% of households classified as extremely poor. Figure A.5.l Description of the Unsatisfied Basic Needs Index Definition: The Unsatisfied Basic Needs (UBN) Index is an indicator of poverty used by the government of Ecuador to define the number of persons that live in poverty. A person is considered poor if they belong to a household that fails to meet specific standards of adequate building materials, water and sanitation services, employment, and education. Methodology: The index uses methodology developed by the Andean Community (Comunidad Andina) which establishes a household as “poor” if at least one of the following conditions is present and as “extremely poor” if they meet two or more of the following conditions: 1. Physically inadeguate characteristics: measured as having exterior walls constructed from disposable materials and/or with a dirt floor) . 2. Inadguate water or sanitation services: measured as having no connection to potable water and/or without a latrine or septic tank. 3. High economic dgpendence: measured as having more than 3 household members per occupied person and a head of household with two years of primary school education or less. 4. Lack of education: measured as having at least one child residing in the household between the ages of six to twelve that does not attend school. 5. Dwelling in a state of criticfial overcrowding: having more than three people per room, on average, without countifl the bathroom, kitchen, or principal living room. Source: SIISE (2001) 138 APPENDIX SIX: SUPPLEMENTAL REGRESSIONS AND REGRESSION DIAGNOSTICS FROM CHAPTER FIVE 139 Table A.6.1: OLS Regression Results for the Yield, Fungicide Demand, and Insecticide Demand Equations, Imbabura and Carchi, Ecuador, 2006 Yield Fungicide Demand Insecticide Demand Explanatory Variables by Category: (kg / ha) (kg AI / ha) (kg AI / ha) Treatment Eflect Variables: Adopted improved variety (1=yes) -35.1 (0.89) —16.9 (0.02)" -4.50 (0.01)*** High disease pressure (1=yes) -433 (0.15) -16.3 (0.03)M Adopted x Disease pressure (1=yes) 1102 (0.00)* 18.9 (0.09)* High pest pressure (1=yes) 122 (0.61) 1.861 (0.26) Production Inputs: Fungicide active ingredient (kg ha") 0.588 (0.98) Fungicide Al squared -0.048 (0.80) Insecticide AI (kg ha") 65.9 (0.14) Insecticide AI squared -0.102 (0.78) Foliar fertilizer A1 (kg ha") -149 (0.42) Foliar fertilizer AI squared 0.063 (0.77) Bean seed (kg ha") -0.982 (0.92) 0.553 (0.00)*** 0.220 (0.00)*" Bean seed squared 0.015 (0.72) Plot Characteristics: Plot size (ha) 21.8 (0.81) -1.42 (0.31) -O.255 (0.64) Altitude (m.a.s.l.) 0.334 (0.16) 0.015 (0.04)" 0.004 (0.15) Loam soil (1=yes) -12.1 (0.95) -2.85 (0.51) -0.451 (0.83) Irrigated plot (1=yes) -13.5 (0.97) -6.09 (0.64) -3.99 (0.39) Plot prev. cropped w/ beans (1=yes) -388 (0.06)* 2.03 (0.65) 2.25 (0.18) Sharecropped plot (1=yes) 240 (0.21) -8.94 (0.04)" -4.98 (0.04)" Rented plot (1=yes) —206 (0.55) -11.6 (0.14) 0.248 (0.94) Household Variables: Age of HH (years) 2.97 (0.68) 0.757 (0.00)"* -0.254 (0.01)*** Attended pest man. seminar (1=yes) 151 (0.43) 0.846 (0.87) -3.55 (0.07)‘I Symptom based pest man. (1=yes) 4.34 (0.49) 6.56 (0.07)‘ Poor household (1=yes) -1.83 (0.69) 1.09 (0.60) Price Variables: Market price for beans ($/kg) -22.4 (0.17) -6.57 (0.27) Cost of transport ($/qq) -3.51 (0.66) -4.08 (0.05)" Avg. price of fungicide ($/kg A1) -10.6 (0.00)"* Avg. price of insecticide ($/kg A1) -2.69 (0.30) Community-Level Variables: Chota valley (1=yes) 402 (0.04)” 8.97 (0.08)* 3.50 (0.05)" Prev. extension intervention (1=yes) 231 (0.19) 0.990 (0.82) -0.986 (0.61) Observations 85 73 73 R2 0.88 0.80 0.80 F(k,dj) 19.1 7.92 9.91 lr>Chi2 0.00 0.00 0.00 Notes: p-values parentheses; data is from the 2006 B/C CRSP and INIAP farm-level survey; firngicide and insecticide demand models estimated using White's heteroskedastic robust standard errors *** significant at 1%; ** significant at 5%; * significant at 10% 140 Table A.6.2: BSLS Regression Results for the Yield, Fun gicide Demand, and Insecticide Demand Equations, Imbabura and Carchi, Ecuador, 2006 (n=73) Yield Fungicide Demand Insecticide Demand Explanatory Variables by Category: (kg / ha) (kg Al / ha) (kg AI / ha) Treatment Eflect Variables: Adopted improved variety (1=yes) -117 (0.72) -1 1.6 (0.02)" -4.65 (0.01)*** High disease pressure (1=yes) -287 (0.58) -2.74 (0.60) Adopted x Disease pressure (1=yes) 1215 (0.01)*** 7.32 (0.31) High pest pressure (1=yes) -119 (0.79) 2.96 (0.04)" Production Inputs: Fungicide AI (kg ha") 5.39 (0.94) Fungicide A1 squared -0. 180 (0.74) Insecticide A1 (kg ha") 148 (0.40) Insecticide A1 squared -0.582 (0.66) Foliar fertilizer AI (kg ha") -272 (0.47) Foliar fertilizer AI squared 0.239 (0.63) Bean seed (kg ha") 3.49 (0.84) 0.545 (0.00)*** 0.219 (0.00)**"‘ Bean seed squared -0.012 (0.88) Plot Characteristics: Plot size (ha) 76.3 (0.50) -2.01 (0.36) -0.215 (0.81) Altitude (m.a.s.l.) 0.314 (0.38) 0.013 (0.02)" 0.003 (0.19) Loam soil (1=yes) 125 (0.60) -l .47 (0.74) -0.483 (0.79) Irrigated plot (1=yes) -437 (0.30) -12.6 (0.19) -4.219 (0.27) Plot prev. cropped w/ beans (1=yes) -284 (0.19) 3.71 (0.45) 1.778 (0.37) Sharecropped plot (1=yes) 231 (0.25) -9.73 (0.04)" -4.713 (0.02)" Rented plot (1=yes) -548 (0.30) -12.9 (0.16) 0.170 (0.96) Household Variables: Age of HH (years) 1.35 (0.87) -0.715 (0.00)*** -0.245 (0.00)*** Attended pest man. seminar (1=yes) 159 (0.39) -0.235 (0.96) -3.41 (0.08) Symptom based pest man. (1=yes) 5.98 (0.28) 7.14 (0.00)"* Poor household (1=yes) -2.99 (0.50) 1.06 (0.55) Price Variables: Market price for beans ($/kg) -26.2 (0.05)" -7.95 (0.15) Cost of transport ($/qq) -2.25 (0.73) -4.47 (0.09)* Avg. price of fungicide ($lkg A1) -1.33 (0.77) Avg. price of insecticide ($/kg AI) -0.751 (0.66) Community-Level Variables: Chota valley (1=yes) 401 (0.08)* 9.66 (0.08)* 3.56 (0.10)‘ Previous extension interv. (1=yes) 405 (0.04 ** -2.79 (0.54L -l .35 (0.45) 0.87 0.79 0.80 522 296 301 0.00 0.00 0.00 Notes: p-values parentheses; fungicide AI and insecticide AI variables treated as endogenous; data is from the 2006 B/C CRSP and INIAP farm-level survey **" significant at 1%; ** significant at 5%; * significant at 10% I41 Table A.6.3: OLS Estimates of the Linear Unit Variable Cost Function, Imbabura and Carchi, Ecuador, 2006 (n=73) Unit Variable Cost Explanatory Variables by Category: ($/kg) Treatment E fleet Variables: Adopted improved variety (1 =yes) 0.0121 (0.73) High disease pressure (1=yes) 0.0112 (0.78) Adopted x Disease pressure (1=yes) -0.0518 (0.37) High pest pressure (1=yes) -0.0520 (0.07)M Price Variables: Avg. fungicide price ($/kg ai.) 0.0247 (0.47) Avg. insecticide price ($/kg a.i.) 0.0490 (0.16) Bean seed price ($/kg) 0.1759 (0.00)*** Output Variable: Bean yield (kg/ha) -6.01x10'5 (0.00)*** Plot Characteristics: Plot size (ha) -0.0195 (0.11) Altitude (m.a.s.l.) -l.14x104 (0.01)*** Loam soil (1=yes) 0.0201 (0.42) Irrigated plot (1=yes) 0.0258 (0.65) Sharecropped plot (1=yes) -0.0173 (0.50) Rented plot (1=yes) 0.1030 (0.04)" Household Variables: Age of HH (years) -0.0011 (0.26) Attended pest man. seminar (1=yes) 0.0425 (0.1 1) Community-Level Variables: Chota valley (1=yes) -0.0415 (0.15) Prev. extension intervention (1=yes) -0.0904 (0.00)*** Constant 0.3374 (0.01)* "' * R2 0.59 (18, 54) 4.34 rob>F 0.00 p-values in parentheses * significant at 10%; ** significant at 5%; ***significant at 1% Table A.6.4: Variance Inflation Factors (VIFs) for the Quadratic Yield Model Explanatory Variables by Category: VIF Adopted improved variety (1=yes) 2.70 High disease pressure (1=yes) 3.52 Adopted x Disease pressure (1=yes) 3.39 High pest pressure (1=yes) 1.69 Production Inputs: Fungicide active ingredient (kg/ha) 86.57 Fungicide a.i. squared 172.16 Insecticide a.i. (kg/ha) 57.42 Insecticide a.i. squared 26.36 Foliar fertilizer a.i. (kg/ha) 32.84 Foliar fertilizer a.i. squared 86.00 Bean seed (kg/ha) 70.74 Bean seed squared 105.97 Plot Characteristics: Plot size (ha) 1.49 Altitude (m.a.s.l.) 1.47 Loam soil (1=yes) 1.49 Irrigated plot (1=yes) 1.58 Plot prev. cropped w/ beans (1=yes) 1.38 Sharecropped plot (1 =yes) 1.62 Rented plot (1=yes) 1.53 Household Variables: Age of HH (years) 1.59 Attended pest man. seminar (1=yes) 1.33 Community-Level Variables: Chota valley (1=yes) 1.58 Prev. extension intervention (1=yes) 1.25 143 Figure A.6.1: Scatter Plots of DFBETA Regression Diagnostic Statistics a) Scatter Plot of DFBETAS for Coeffcient on Adoption X Disease Pressure Interaction Term (Numbers indicate household ID) Pd .5 L .40 061 :0, 338-18403 312‘. 035arr a 1‘69 .29 042.50 33366.?989 .1. 110.119 :98100M‘6336 “9038805)?” {*Wm .... $188 94 Badge—esa Q ’21 .i991 0126 beeta adothprsdis 0 l -.5 I .9 . j T f 0 50 100 150 numero de la encuesta del hogar b) Scatter Plot of DFBETAS for Coeffcient on Disease Pressure (Numbers indicate household ID) .9 '9 .1 .11 41% .27711 .93 1g 21 ”655962 075 a o a ‘% Waffle-49 owwwm g .Zfiafi ’50 .5589 .l‘fioe .126 .43 .61 .121 "3 4 | I r 1 I l 0 50 100 150 numero de la encuesta del hogar M: Values on the Y-axis indicate the influence of each individual observation on the value of regression coefficients, measured as a proportion of one standard deviation (STATA 2006). Inclusion of the largest outlier (observation #9) decreases the coefficient on the interaction variable and increases the coefficient on the disease pressure variable. No remedial measure is implemented since this outlier has the effect of lowering expected incremental net benefits. 144 APPENDIX SEVEN: DATA ELICITATION WORKSHEETS FOR RESEARCH COSTS 145 e: 181:: u 8 an cease-nu - .. 83m c. 888 8: 88 mm 88.8 e. 9&8 8: 88 R 808 c. 88 8: 88 a 08...” c. 88 8: 88 cm 8...~ c. 88 8: 88 2 8de c. O88 8: 82 M: 80. E c. 288 8: 82 : 8o. 8 e. 888 8: 82 2 80.3 c. 888 8: c8. 2 28. E c. 88m 8: ma: : 28.8 c. 88¢ 8: 32 2 8de c. 898 8: .82 2 8o? c. 888 8: m8. _ _ 88.2 c. 888 8: 32 2 88.8 e. O83 8: 82 e 88.? e. 80:” 8: as: a 88.2 e. 885 8: 82 a 88.9 c. 805 8: E: c 88.8 c. 88:” 8: 82 m 08.8 c. 885 8: 32 v 88.3 e. 885 8: 33. m 08.8 e. 80:” 8: 32 m 83: e. o8; 8: 82 _ 46V 8 R: 6: 8V 2: 88288 cogs—«>0 Love: tong—6 88 .8583: :80... motor? 3 3823: =2:me + 888 5:250 80> cocoa. . R: we :DEDQDE coco—one .88. gag—3m .395 823...; 3 E53509 225.5 .aESxm can mameU 5322—0 in 82.3.83 8:0 5.583— ;.S. 2e»...— 146 AZV x G6 x 3v u 8: 8 88.8.5 .. t. A: x 8 x g n O: 8 88.380 - .. FNNdP 8. N8.“ m othN 8. 83 m 8: 88 8 8NK 8. 88 m 8N8 8. N86 8 8: 88 NN 88.8 8. 88 m a 8.8 8. 88 v 8: 88 .N 83 8. 88 m So.» 8. 88 v 8: S8 8 88 8. 8N.N m 85. 8. 8N.N m 8: 88 2 88; 8. 85 N 88 8. 85 m 8: 82 M: 88 8. N28 N 82 8. N28 m 8: 82 2 E8 8. m {.8 N 8.» 8 SN... m 8: 32 2 88.. 8. 88.. N n 58 8. 88. m 8: 82 2 m8...” 8. EN... N EN? 8. :86 v 8: 32 3 88.8 8. 88 N 88? 8. 88 v 8: 32 2 N8.N 8. N85 N .86 8. N88 m 8: 8.2 N_ NNtN 8. N8.N N 28 8. N8.N m 8: 82 _ _ «EN 8. 88 N 88 8. 88 m 8: 32 2 8o; 8. 88 P N86 8. 88 m 8: 82 a 8o.N 8. o8.N F 88 8. o8.N N 8: 82 w 83 8. o8.N F 83. 8. o8.N N 8: 82 h 85 8. 8N_N P 88 8. o8.N N 8: 82 o 83 8. 8P_N F 88 8. o8.N N 8: 82 m o8; 8. 8o.N r 88.” 8. o8.N N 8: £2 8 o8; 8. o8.N 8 88 8. o8.N N 8: 82 m 83 8. 88; F 8N.” 8. o8.N N 8: 82 N 83 8. 8: F 888 8. 8F.N N 8: S2 _ 1.6: 9: a): S: .0: A: 8 a: 6: Am: :3 88288 88280 :858 $3805-83 8822.8 UMWMMMWM SWMMM 88805 ~88 8818 8 8g 88 :0 wait; 88805 8 88286 8 88 :89 «o 8855 So 08 8858 .8 88068 Ad tau :258 :89 no . > u . m “moo Each. mo COS—Gaohn— 3:58 mo 595:7— umoo 730,—. 9.3 m0 3355 535:2 . .w>< 88085 .w>< 8:835— .525 8.82.5 3 83289 380 09.88% 8:5: ”nu 82.9.33 .80 8.383— 8: 2:2..— 147 APPENDIX 8 NPV AND IRR CALCULATION TABLES 148 149 dedddd dd dd_dddd d5: dddd dddd 833 Ed 32d dd dddm dd dfldddd dd dNEddd dud. dddd Nddd ddEd Ed 32d dd dddm E Emtdd dd 235 8d dddd Nddd ded Ed 32d dd Edd mm dd~.-dd dd ddmdddd d~d_ dddd Nddd added Ed Ed dd dddm Nd ESE; dd ESE; Ed dddd .ddd dEvd Ed 32d dd mddm _~ ddddtd dd ddddtd Ed dddd dddd «Red Ed ES dd 38 dd mmddzd dd fiddzd d3. dddd 33 Ed Ed 3:.d dd dddm 2 dmifld dd dmifld 2d dddd dEd dddvd Ed 32d dd 82 M: «3.2; dE._dd 332d 3d dddd dEd ddddd Ed 32d dd dde : dEdEd Eqddd dEd2d 8d dddd add and Ed 32d dd 82 2 dmmid d_d.ddd 223d 23 dddd :Vdd ddEd Ed 32d dd 82 2 23.3: dddddd ddmdad ddd dddd dddd Emdd Ed 32d dd 32 3 £2.85 2.de dddddmd mdd dddd dddd Eddd Ed 32d dd E2 2 dmddzd 2..de dE.E_d :d dddd dddd 33 Ed 32d dd 82 .2 added ddddmd ddmdzd Ed dddd :dd dEmd Ed 32d dd 82 : d:.dd_d Eqmmd ddddfld d: dddd dmdd dddmd Ed 32d dd 32 d_ Nddddd dmdfid 23.85 E dddd dmdd ddad Ed 32d dd ddd. d 323 ddmdmd dddddd d; dddd Edd dEd Ed 32d dd 32 d 23d ddddmd Eddd d; dddd dddd ddddd Ed 32d dd ddz d _ :3 33.3... _ _E.~_d- ddvdmd dddfid d_.d dddd dddd dEdd Ed Edd dd 52 d 5d: dEde- ddvdmd and d; dddd _dd.d dd_d.d Ed 32d dd Edd. d 22‘ 233—30 :— ddd.d~d- dddde dd d; dddd dddd ddddd Ed Edd dd 22 d _ $3 _ddd~dEd L ddddmd- ddddmd dd d; dddd dddd ddddd Ed E3 dd E2 d and: d_d.dmd- dz.de dd d_.d dddd dddd ddddd Ed 32d dd 82 m 8.: 3888 :— ddddmd- ddmvmd dd d; dddd dddd ddddd Ed 32d dd 32 _ 5: :2 m2 9: m2 m m M V SE B d w EN: 35% z .2 d v. d _ m o d m a u m < gcuéma- .8233.» 6232: cu.— m<—Z— 6:23.530 my: 25 >mz .5.— ota—Sum oE—odam "—.w.< «Sch .dvdzd d .dddzd dwE. ddEd .ddd ddEd Ndd EE..d dd Eddd Ed Edd.d.dd d Edddzd 2E. ddEd :3 Sad med EE..d dd Eddd E 352d d E.v.dd.d ddd ddEd dddd ddddd Ned EE..d dd Edd dd .dE.E.d d :Eddd ddd. ddEd dddd Eddd ddd EE..d dd dddd dd .dd..dd d .dd..dd :d ddEd dddd Ede 9d EE..d dd Nddd .N NEEdd d fidddd Ed ddEd dddd Sad 93 EE..d dd .ddd dd EE..dd d EE..dd ddd ddEd Rdd $de dad EE..d dd dddd d. Edda d dddddd d..d ddEd dddd EdEdd Ned EE..d dd ddd. d. .E..d.d ddd..dd dddddd :A ddEd dddd Edddd Ned EE..d dd ddd. d. Eddddd Edddd 235 SE ddEd dddd 8de ed EE..d dd ddd. E. :dddd d.d.ddd ddd..Ed dd.d ddEd :dd Edd Ned EE..d dd Edd. d. :dEd dddddd :Eddd ddd ddEd Edd Edddd dvd EE..d dd ddd. E ddEddd dddddd dEd.dEd and ddEd :3 32d dud EE..d dd Edd. d. 312d E. A—Z .3.— ctacoom o>=a>uod=c0 "N.w.< 035—. 150 d.ddddd d d..ddddd ddd. ddEd dddd ddddd ddd dd..d dd Eddd dd ddddddd d ddddddd d...E. ddEd dddd ddddd ddd dd..d dd Eddd dd ddddddd d ddddddd ddd ddEd dddd ddddd ddd dd..d dd Edd dd .dE.dddd d .dEddE ddd. ddEd dddd ddddd ddd dd..d dd dddd dd ddEdddd d ddEdddd d..d ddEd dddd ddddd ddd dd..d dd dddd .d ddddddd d ddddddd Ed ddEd dddd ddddd ddd dd..d dd .ddd dd d.E.dddd d d.E.dddd ddd ddEd dddd ddddd ddd dd..d dd dddd d. Edd.dddd d Edddddd d..d ddEd dddd ddddd ddd dd..d dd ddd. d. Eddddd ddd,.dd d.d.Edd .vd ddEd dddd ddddd ddd dd..d dd ddd. d. ddEdEdd Edddd .dd.dddd .d.E ddEd dddd ddddd ddd dd..d dd ddd. E. ddd.dddd d.d.ddd .d..dddd ddd ddEd dddd ddEd ddd dd..d dd Edd. d. ddd.dddd dddddd ddddddd ddd ddEd dddd ddEd ddd dd..d dd ddd. 3 E.d.dddd dddddd dddEdd ddd ddEd dddd .dEd ddd dd..d dd ddd. d. dddEdd E...ddd ddd.dddd ...d ddEd dddd ddEd ddd dd..d dd ddd. d. dEddEd Edd..dd dEdEdd dEd ddEd Edd Edddd ddd dd..d dd ddd. .. «dd..ddd dddddd d.d.dddd d..d ddEd dddd ESd ddd dd..d dd .dd. d. ddddddd ddd.:d ddddddd ddd ddEd .Edd EEd ddd dd..d dd ddd. d ddddE.d dddddd dd..ddd d..d ddEd dddd Edddd ddd dd..d dd ddd. d ddEddd ddd.Edd ddvddd d..d ddEd :dd ddddd ddd dd..d dd ddd. d. _ .dEE _dEd,dE.d _ddd,dd dddddd dddddd d..d ddEd Eddd ddddd ddd dd..d dd ddd. E Ed: Edd- dddddd ddEdd d..d ddEd dddd dd.dd ddd dd..d dd Edd. d Eddd 2.828 .: ddd.Edd- ddd.Edd dd d..d ddEd dddd ddd ddd dd..d dd ddd. d _ $3 Land—33 ddd.Edd- ddd.Edd dd d..d ddEd dddd ddd ddd dd..d dd ddd. d 5d: d.d,ddd- d.d.ddd dd d..d ddEd dddd ddd ddd dd..d dd ddd. d ddd. 2528 d. dddddd- dddddd dd d..d ddEd dddd ddd ddd dd..d dd ddd. . .5: 22 m2 E: E < m m Ed V dd? u: q d 53. 32$ 2 .). d v. d . m o d m d. o d. < Edddddd. 5.3:? 3.22.. d2 .32. .223325 5: d..d >._z 3. £358 .258. .ddd «.ddd 151 152 E.d.ddd dd E.d,dd dd.d ddEd dddd dd.d Ed dd..d dd .ddd dd E.d.ddd dd E.d.dd dd.d ddEd dddd dd.d Ed dd..d dd dddd Ed E.d.ddd dd E.d.dd dd.d ddEd dddd dd.d Ed dd..d dd dddd dd E.d.ddd dd E.d.dd dd.d ddEd dddd dd.d Ed dd..d dd .ddd dd E.d.ddd dd E.d.dd dd.d ddEd dddd dd.d Ed dd.d dd dddd dd E.d.ddd dd E.d.dd dd.d ddEd dddd dd.d Ed dd..d dd d.dd dd E.d.ddd dd E.d.dd dd.d ddEd dddd dd.d Ed dd..d dd d.dd .d E.d.ddd dd E.d.dd dd.d ddEd dddd dd.d Ed dd..d dd 28 dd E.d.ddd dd E.d.dd dd.d ddEd dddd dd.d Ed dd..d dd E.dd d. E.dddd dd E.d.dd dd.d ddEd dddd dd.d Ed dd.d dd d.dd d. E.d.ddd dd E.d.dd dd.d ddEd dddd dd.d Ed dd..d dd d.dd d. E.d.ddd dd E.d.dd dd.d ddEd dddd dd.d Ed dd..d dd d.dd E. E.d.ddd dd E.d.dd dd.d ddEd dddd dd.d Ed dd..d dd d.dd d. E.d.ddd dd E.d.dd dd.d ddEd dddd dd.d Ed dd..d dd :8 3 ddEdd dd ddEd dd.d ddEd dddd Ed Ed dd..d dd d.dd d. dEddd dd dEdd dd.d ddEd dddd dd.d Ed dd..d dd dddd d. dddddd dd ddddd dd.d ddEd .Ed dd.d Ed dd..d dd . dddd . . ddEd dd ddd.? dd.d ddEd Eddd dd.d Ed dd..d dd dddd d. ddEddd dd ddEdd dd.d ddEd d.dd d..d Ed dd..d dd Eddd d dd.dd dd dd.d dd.d ddEd Edd dd.d Ed dd..d dd dddd d .ddddd- Eddaddd ddd.. dd. ddEd .ddd .dd Ed dd..d dd Edd d 7 ed: _ ddd...ddd _ddd.ddd- ddd.ddd d R. ddEd dddd dd.d Ed dd..d dd dddd E 5d: E82588 a. ddd.:d- ddd.:d d dd.d ddEd dddd dd.d Ed dd..d dd dddd d d.E.d.d- d.E.d.d d dd.d ddEd dddd dd.d Ed dd.d dd .ddd v _ $2 _ d._..:d 4 ddddd- ddd.dd d dd.d ddEd dddd dd.d Ed dd..d dd dddd d and: d.dd—332.3 5 dEdddd- dEdddd d dd.d ddEd dddd dd.d Ed dd.d dd ddd. d ddd..dd- ddd..dd d dd.d ddEd dddd dd.d Ed dd.d dd ddd. . «S :2 m2 8. m2 a 5. Ed .d at Edd d Ed 35 z s. d V. d . m o d m d. u d. < d.dddddd. 5.8.3» 8.32.. «.dd.... .37.. “aegis 5: d..d >..z .8 380 5.38: ..< .3 2:2on 2:83 "d.dd asap 153 E.d.ddd dd E.d.dd dd.d ddEd dddd dd.d Ed dd..d dd dddd dd E.d,ddd dd E.d.dd dd.d ddEd dddd dd.d Ed dd.d dd dddd Ed E.d.ddd dd E.d.dd dd.d ddEd dddd dd.d Ed dd..d dd dddd dd E.d.ddd dd E.d.dd dd.d ddEd dddd dd.d Ed dd..d dd .ddd .Vd E.d.ddd dd E.d.dd dd.d ddEd dddd dd.d Ed dd..d dd dddd dd E.d.ddd dd E.d.dd dd.d ddEd dddd dd.d Ed dd..d dd d.dd dd E.d.ddd dd E.d.dd dd.d ddEd dddd dd.d Ed dd.d dd d.dd .d E.d.ddd dd E.dndd dd.d ddEd dddd dd.d Ed dd..d dd d.dd dd E.d.ddd dd E.d.dd dd.d ddEd dddd dd.d Ed dd. .d dd E.dd d. E.d.ddd dd E.d.dd dd.d ddEd dddd dd.d Ed dd..d dd d.dd d. E.d.ddd dd E.d.dd dd.d ddEd dddd dd.d Ed dd..d dd Edd d. E.d.ddd dd E.d.dd dd.d ddEd dddd dd.d Ed dd.d dd d.dd E. E.d.ddd dd E.d.dd dd.d ddEd dddd d..d Ed dd..d dd d.dd d. E.d.ddd dd E.d.dd dd.d ddEd dddd dd.d Ed dd . .d dd ..dd d. ddd..dd dd ddd..d dd.d ddEd dddd Ed Ed dd..d dd d.dd d. dE.ddd dd dE.dd dd.d ddEd dddd dd.d Ed dd..d dd dddd d. dddddd dd ddd.dd dd.d ddEd .Ed dd.d Ed dd.d dd dddd .. ddd.ddd dd ddd.? dd.d ddEd Eddd dd.d Ed dd..d dd dddd d. ddEddd dd ddEdd dd.d ddEd d.dd d..d Ed dd..d dd Eddd d dd..Ed dd dd..E dd.d ddEd Edd dd.d Ed dd..d dd dddd d Edd.d.d- dEd.d.d ddd.. dd.. ddEd .ddd .dd Ed dd..d dd Edd d T .ddd _ ddd.dddd LdEddd- dEdd.d d dd. ddEd dddd dd.d Ed dd..d dd dddd E and: Edddddadds d. ddddd- ddddd d dd.d ddEd dddd dd.d Ed dd..d dd dddd d Eddd- Eddd d dd.d ddEd dddd dd.d Ed dd..d dd .ddd d T .ddd _ dEE.Eddd _ddd.dd- ddd.dd d dd.d ddEd dddd dd.d Ed dd.d dd dddd d 5m: ddddddsdds 5 .dddd- .dd.Ed d dd.d ddEd dddd dd.d Ed dd..d dd ddd. d dEd.dd- dEd.dd d dd.d ddEd dddd dd.d Ed dd..d dd ddd. . 5: 32 m2 9: m2 m d M V SE 92 d Ls: 32$ 2 .2 d v. d . m o n. m a o d. < 33-x:— .do=e_..a> :0:ch 29::— mSZ— $532330 1%— 6:: >5: .8.— dddcu £233: @233). 29::— E 2.8:3m 3:025 dmdé. 039—. IIIIIIIIIIIIIIIIIIIIIIIIIIIIII ‘IJIIIHNIIIHIIIIIIIIIMIIHIIHIWIHM‘IIIHMIIHIHII