VIIIIIIIIIIIIIIIIIIIIIIIIII.II..... ti Rh’qd LIBRARY Michigan State A University ,' This is to certify that the dissertation entitled Factors Affecting the Demand for Fertilizer in Senegal's Peanut Basin presented by Valerie A. Kelly has been accepted towards fulfillment of the requirements for Ph.D. degreein Agricultural Economics Quanta Major professor Eric W. Crawford ‘ ‘1r C: Date March 8, 1988 . MS U i! an Affirmative Action/Equal Opportunity Institution 0-12771 ‘ MSU LIBRARIES RETURNING MATERIALS: Place in book drop to remove this checkout from your record. FINES will be charged if book is returned after the date stamped below. FACTORS AFFECTING THE DEMAND FOR FERTILIZER IN SENEGAL’S PEANUT BASIN By Valerie A. Kelly A DISSERTATION Submitted to Michigan State University in partial fulfillment of the requirements for the degree of DOCTOR OF PHILOSOPHY Department of Agricultural Economics 1988 ABSTRACT FACTORS AFFECTING THE DEMAND FOR FERTILIZER IN SENEGAL’S PEANUT BASIN By Valerie A. Kelly The design of an effective fertilizer policy for Senegalese rainfed agriculture has become a critical policy issue in recent years as farmers’ effective demand for the input has declined substantially since 1980. This dissertation examines the causes of low demand and identifies a number of options for making fertilizer policy more responsive to farmers’ needs and government objectives. The dissertation examines fertilizer policy issues from three perspectives, presenting (1) a review and evaluation of 1960-80 institutions for assuring fertilizer supply and stimulating demand, (2) a review of available evidence on fertilizer response and economic returns, and (3) an analysis of factors currently influencing fertilizer demand. A variety of data sources and analysis techniques were employed. Regression and factor analysis were used to examine secondary data on past fertilizer consumption. Regression analysis and, farm budgeting techniques were used to estimate fertilizer response from demonstration trial data, and to examine economic returns. Logit analysis and hierarchical decision models were used to examine survey data concerning farmers’ input investment behavior. The research found that the major constraints on fertilizer demand are: (1) low and uncertain fertilizer response, (2) lack of financial liquidity due to low agricultural productivity causing low farm targe‘ trade; I p . I GgrlCL resist incomes, (3) farmers’ perceptions of low fertilizer profitability relative to alternative investments, and (4) an unresponsive distribution system. It is recommended that the government begin an official policy of targeting fertilizer to zones of high fertilizer response and low risk by selectively introducing credit programs for both farmers and private traders. Su‘ch credit programs would stimulate demand and encourage greater private sector participation in fertilizer distribution. For zones of low fertilizer response, greater investment in basic agricultural research to find more productive technologies (drought resistant varieties, improved cultural practices, etc.) is recommended. This dissertation is dedicated to Badié, who got me into this mess, and to Da Monzon, without whose help I would never have gotten out. ii Th1: isdividu Hy assistan Jim Dem fine: a Sasques d‘isserta m, en: ACKNOWLEDGEMENTS This dissertation is the result of the efforts of a great many individuals at Michigan State University and in Senegal. My major professor, Eric Crawford, has provided invaluable assistance throughout my entire graduate program. Les Manderscheid and Jim Oehmke helped a great deal in both the design and the analysis phases of my research. Suggestions made by Jim Bingen, Jim Shaffer, Jacques Faye and Russ Freed, who read various working papers and dissertation drafts, were very useful. The United States Agency for International Development provided the funding for the field work conducted in Senegal. The in-depth interviews with farmers to collect information about decision making could not have been done without the assistance of Matar Gaye. The collection of general survey data was accomplished by Biran Diop, Moustapha Gaye, Emil Sene, and Amadou N’Diaye. Amadou Cissé deserves special thanks for making SODEVA data available to us. Guy Pocthier was extremely helpful in providing the Amélioration Fonciére data and offering observations on the analysis. Ibrahima Sene provided assistance in collecting information about the evolution of input distribution and extension policies in the Sine Saloum. The least rewarded but clearly most important contributors to this entire research program were the farmers in Gossas and Micro who lillirg process these 0 willingly participated in a relatively time-consuming interview process. Hopefully, the results of this research, when combined with those of the many other ISRA programs that make significant demands on farmers’ time, will contribute to the design of agricultural policies that better respond to farmers’ needs. iv LIST C LIST C LIST C Paw» TABLE OF CONTENTS LIST OF TABLES ' viii LIST OF FIGURES xi LIST OF ABBREVIATIONS AND ACRONYMS USED xiii I. INTRODUCTION 5—: A. PROBLEM SETTING l B. RESEARCH DESIGN AND ORGANIZATION OF THE DISSERTATION 4 II. AN INSTITUTIONAL ANALYSIS OF FERTILIZER DEMAND AND SUPPLY 9 A. DETERMINANTS OF FERTILIZER SUPPLY: 1949 - 1980 10 1. AGRICULTURAL RESEARCH DETERMINES FERTILIZER PRODUCTS SUPPLIED 10 2. THE AGRICULTURAL BUREAUCRACY TAKES CHARGE OF INPUT DISTRIBUTION ' 3O 3. FERTILIZER SUPPLIED BY A GOVERNMENT PROTECTED MONOPOLY 36 B. DETERMINANTS OF FERTILIZER DEMAND 37 1. GOVERNMENT CONTROLS PEANUT MARKETING TO GUARANTEE FARMERS FAIR PRICES 38 2. ”PROGRAMME AGRICOLE” CREDIT SPURS DEMAND FOR MODERN INPUTS 43 3. AGRICULTURAL EXTENSION INCREASES DEMAND BY IMPROVING KNOWLEDGE 47 C. INSTITUTIONAL PERFORMANCE 49 D. POST ”PROGRAMME AGRICOLE' INPUT DISTRIBUTION POLICIES 65 III. QUANTITATIVE ANALYSIS OF FERTILIZER CONSUMPTION: 1961-1980 73 A. CONCEPTUALIZATION OF FERTILIZER DEMAND DURING THE PA 75 B. ESTIMATION OF THE MODEL 80 C. FACTOR ANALYSIS 83 D. GENERAL CONCLUSIONS OF THE ANALYSES 91 1V, REVI SET A. ECC m-wNO—4 IV. REVIEH OF RESEARCH ON ECONOMIC RETURNS TO FERTILIZER USE IN SENEGAL 95 A. ECONOMIC ANALYSES OF FERTILIZER RESPONSE DATA 95 1. EARLY ISRA/IRHO CONTRIBUTION 95 2. FAO CONTRIBUTION 101 3. IFDC ANALYSIS OF IRHO/ISRA DATA 103 4. ECONOMIC JUSTIFICATION OF FERTILIZER SUBSIDIES 104 5. IFDC/SODEVA CONTRIBUTION SPURS GREATER ATTENTION TO ECONOMICS 105 6. THE POST-1980 SITUATION 113 B. LINEAR PROGRAMMING ANALYSES ‘ 116 C. BEHAVIORAL AND DESCRIPTIVE RESEARCH BASED ON FARM SURVEY DATA 117 D. SUMMARY 120 V. AN ECONOMIC ANALYSIS OF 'AMELIORATION FONCIERE” DATA 121 A. A DESCRIPTION OF THE DATA 121 B. ESTIMATION OF PRODUCTION FUNCTIONS 127 C. ECONOMIC ANALYSIS 138 1. BUDGET ANALYSES 138 2. INTER-ANNUAL VARIATION IN V/C RATIOS 144 3. DECISION ANALYSIS 149 D. EVALUATION OF THE AF DATA SETS 158 E. SUMMARY OF IMPORTANT POINTS 162 VI. SURVEY DESIGN, IMPLEMENTATION, AND ANALYSIS 164 A. DESIGN 164 1. SAMPLING PROCEDURES 164 2. QUESTIONNAIRE DESIGN AND SUB-SAMPLES 169 B. IMPLEMENTATION 172 C. ANALYSIS 175 D. ORGANIZATION OF THE DISCUSSION 175 VII. FACTORS HHICH INFLUENCED 1981-85 FERTILIZER PURCHASES AND USE 176 A. AGRICULTURAL PRODUCTIVITY AND ECONOMIC CLIMATE 177 1. AGRICULTURAL PRODUCTIVITY 177 2. ECONOMIC CLIMATE 180 B. FARMERS’ PERCEPTIONS OF FERTILIZER 182 vi C. TY hm; APPEHD l'ih Hf ‘ M ”Eng AFDCDH ‘ulu LIST 0 THE RELATIVE IMPORTANCE OF FERTILIZER PROBLEMS APPROPRIATE FERTILIZER APPLICATION TECHNIQUES ALTERNATIVE SOIL RENENAL TECHNIQUES RISK AND FERTILIZER INVESTMENT . FERTILIZER RESPONSE SUMMARY OF MOST IMPORTANT PERCEPTIONS C. TYPES OF ECONOMIC ANALYSES PERFORMED BY FARMERS 1. ECONOMIC THEORY OF INPUT DEMAND 2. TYPES OF ECONOMIC ANALYSES USED BY FARMERS D. FARMERS’ INVESTMENT PRIORITIES aim-bum.— O O O O O l. SEED 2. LIVESTOCK 3. 'BANABANA' E. FERTILIZER PURCHASES: 1982-85 VIII. DISTINGUISHING FERTILIZER PURCHASERS FROM NON-PURCHASERS AND MODELING THE PURCHASING DECISION A. LOGIT ANALYSIS: DISTINGUISHING PURCHASERS FROM NON- PURCHASERS B. HIERARCHICAL DECISION MODELS IX. SUMMARY OF SALIENT FINDINGS AND POLICY IMPLICATIONS A. SUMMARY OF SALIENT FINDINGS l. EVIDENCE OF FERTILIZER RESPONSE 2. ANALYSIS OF ECONOMIC RETURNS TO FERTILIZER 3. FACTORS INFLUENCING FERTILIZER DEMAND AND SUPPLY 8. POLICY IMPLICATIONS OF THE RESEARCH 1. CHOOSING REALISTIC POLICY OBJECTIVES 2. DESIGNING SHORT-RUN POLICY INITIATIVES 3. RESOLVING THE LONG-RUN POLICY DILEMMAS APPENDIX I: DATA FOR FERTILIZER DEMAND ANALYSIS APPENDIX II: CALCULATIONS FOR ESTIMATES OF TRACTION AND EQUIPMENT COSTS APPENDIX III: GENERAL SURVEY QUESTIONNAIRE APPENDIX VI: DESCRIPTION OF VARIABLES EXAMINED IN LOGIT A ALYSIS LIST OF REFERENCES vii 216 216 232 242 242 242 244 250 258 258 263 266 270 272 272 285 288 Table 1 Table 2 Table 3 Table 4 Table 5 Tai‘e 6 Tazle 7 Table 8 Table 9 Table 1: Table Table Table Table Table Table Table Table Table Table Table Table Table Table Table Table Table GMhWN N 10: 11: 12: 13: 14: 15: 16: 17: LIST OF TABLES Original ”Themes Légers” Fertilizer Recommendations Original ”Fumure Etalée" Fertilizer Recommendations Original IRHO Fertilizer Recommendations 1963 ”Themes Légers' Fertilizer Recommendations 1963 'Fumure Etalée" Fertilizer Recommendations ”Themes Légers" Fertilizer Recommendations -- Late 19605 ”Themes Intensifs” Recommendations: 1976 to Present Fertilizer Recommendations Mid-19705 to Present Percent of Total Agricultural Debts Reimbursed by Farmers 1966-80 Fertilizer Demand Model -- Preliminary Version Fertilizer Demand Model -- Final Version Varimax Loadings of Independent Variables Varimax Loadings of Dependent and Independent Variables Regression of Factor Scores on Quantity of Fertilizer Consumed 1971 Partial Budget Analysis Comparing "No Fertilizer" with "Themes Légers" and "Themes Lourds" Recommendations: Northern Peanut Basin 1971 Partial Budget Analysis Comparing "No Fertilizer" with "Themes Légers” and "Themes Lourds" Recommendations: Sine Saloum Summary of FAO Fertilizer Trial and Demonstration Results viii 12 14 17 18 19 21 27 29 63 81 82 85 89 9O 97 98 102 Table 18: Table 19: Table 20: Table 21: Tatle 22: Table 23: Table 24: Table 25: Table 26: Table 27: Table 28; 72518 29: Table 30: Table 31; 13le 32; Table 33: Table 34: Table 35. T 31119 35: Table Table Table Table Table Table Table Table Table Table Table Table Table Table Table Table Table Table Table 18: 19: 20: 21: 22: 23: 24: 25: 26: 27: 28: 29: 30: 31: 32: 33: 34: 35: 36: Cost/Benefit Analysis of Government Fertilizer Subsidies Frequency Distribution of Fertilizer Value/Cost Ratios Using Subsidized and Unsubsidized 1975 Prices Description of the Amélioration Fonciere Data Set ”Amélioration Fonciere' Peanut Treatments for the Sine Saloum 1964 - 1982 "Amélioration Fonciere' Cereal Treatments for the Sine Saloum 1965-1982, Nioro Peanuts Pooled Regression Model Nioro Peanut Hay Pooled Regression Model Nioro Sorghum Pooled Regression Model Boulel Peanuts Pooled Regression Model Boulel Peanut Hay Pooled Regression Model Boulel Millet Pooled Regression Model Summary of Cross-Sectional Analysis of "Amélioration Fanciere' Data Nioro 1987 Partial Budget Analysis Comparing "No Fertilizer” Hith "Themes Légers” and "Themes Lourds" Recommendations Boulel 1987 Partial Budget Analysis Comparing "No Fertilizer” Hith ”Themes Légers" and "Themes Lourds" Recommendations Frequency Distribution of Fertilizer Value/Cost Ratios Over Time - Nioro Frequency Distribution of Fertilizer Value/Cost Ratios Over Time - Boulel Expected Net Benefit Calculation of Five Investment Alternatives - Nioro Expected Net Benefit Calculations For Five Investment Alternatives - Boulel Characteristics of Respondents in 98-Farmer Sample ix 106 112 122 124 125 129 130 131 132 133 134 137 139 140 145 146 151 152 170 Table Table Table Table Table Table Table Table Table Table Table 37: 38: 39: 40: 41: 42: 43: 44: 45: 46: 47: Inter-Annual Changes in Key Productivity Measures for 41 Farms 1981/82 through 1984/85 Farmers’ Perceptions of Peanut and Millet Response to Fertilizer Compared to AF Estimates Prices Farmers Are Hilling To Pay For Fertilizer Farmers’ Criteria for Judging Fertilizer Cost Farmers’ Concepts of Acceptable Peanut/Fertilizer Ratios Quantity of Fertilizer Purchased 1981/82-1985/86 Official and Informal Market Prices of Fertilizer 1981 - 1985 Results of 98-Farmer Logit Model Group-by-Group Predictions of 98-Farmer Logit Model Compared to Actual Farmer Purchasing Behavior Results of 46-Farmer Logit Model Group-by-Group Predictions of 42-Farmer Logit Model Compared to Actual Farmer Purchasing Behavior 179 188 197 199 200 209 214 221 222 224 225 Figure 1: Figure 2: Figure 3: “Sure 8: illlre 9: ”sure 10; “We 11: ‘1 FlgUre 13: ltggre 14: p ' 9m 12: H3119 15: (I) Figure Figure Figure Figure Figure Figure Figure Figure Figure Figure Figure Figure Figure Figure Figure Figure Figure Figure OGNO‘U‘ 10: 11: 12: 13: 14: 15: 16: 17: 18: LIST OF FIGURES Map of Senegal and the Sine Saloum IRHO Mapping of Fertilizer Zones in 19605 Description of Major Government Institutions Serving the Agricultural Sector in the Post-Independence Period Fertilizer Distribution and Consumption, 1949/50 - 1986/87 Equipment Purchases in the Peanut Basin, 1960 - 80 Senegalese Peanut Production and Marketing, 1960-80 Evolution of Peanut Yield Per Hectare, 1935 - 1984 Senegalese Millet Production, 1960 - 1980 Farm Revenues from Official Peanut Sales, 1960 - 1980 Schematic Diagram of Factors Determining Fertilizer Demand and Supply during the "Programme Agricole" Variables Thought to Influence Fertilizer Consumption Map of Gossas and Nioro Villages Hhere Sample Farmers Are Located Hypothetical Questions Deserve Hypothetical Answers Peanut and Millet Planted and Harvested by SODEVA Farmers 1981/82 - 1984/85 Farmers’ Investment Priorities, 1985/86 Fertilizer Acquisition and Use, 1981/82 - 1985/86 Changing Patterns in Fertilizer Use, 1981/82 - 1985/86 Definition of Significant Variables in Logit Models xi 16 33 51 52 53 55 56 59 76 78 168 174 178 205 211 213 217 figure 19: HI Figure 20: H Figure 19: Hierarchical Decision Model of Fertilizer Purchase Decisions for 1983/84 233 Figure 20: Hierarchical Decision Model of Fertilizer Purchase Decisions for 1985/86 234 xii NE 52510 PADS BSD LE2 LERP CI (IPA) FC FCFA FTDES AF BCEAO BNDS BSD CER CERP CI CIRAD CNRA CPSP CRA CRAD ENB F0, 1, etc. FAO FC FCFA FIDES LIST OF ABBREVIATIONS AND ACRONYMS USED Amélioration Fonciere agricultural researach program Bureau d’Analyses Macro-Economiques, a branch of ISRA Banque Centrale des Etats de l’Afrique de l’Ouest Banque Nationale de Développement du Sénégal Banque Sénégalaise de Développement (predecessor of BNDS) Centres d’Expansion Rurale Centres d’Expansion Rurale Polyvalent Confidence Interval Centre de Cooperation Internationale en Recherche Agronomique pour le Développement Centre National de Recherches Agronomiques Caisse de Péréquation et de Stabilisation des Prix Centre de Recherches Agronomiques (predecessor of CNRA) Centres Régionaux d’Assistance au Développement Expected net benefits Used to represent different fertilizer themes; the greater the number the heavier the fertilizer dose Food and Agricultural Organization of the United Nations Fonds Commun Senegalese currency; one FCFA - 50 French francs Fonds d’Investissement pour le Développement Economique et Social (became Fonds d’Aide et de Cooperation, FAC) xiii M as; CCC 03 PA PA955 Alec FMDR GDP 605 GT HDM ICS IFDC IRAT IRHO ISRA LP MDR MFC MVP N-P-K NPA OCA OCDE/OECD ONCAD OS PA PAPEM SATEC Fonds Mutualiste de Développement Rural Gross Domestic Product Government of Senegal Groupe Typologique; a farm classification unit used in Senegal Hierarchical decision model Industries Chimiques du Sénégal International Fertilizer Development Center Institut de Recherches Agronomiques Tropicales Institut de Recherches sur les Huiles et Oléagineux Institut Sénégalais de Recherches Agricoles Linear Programming Ministry of Rural Development Marginal Factor Cost Marginal Value Product Nitrogen-Phosphorus-Potassium Nouvelle Politique Agricole Office de Commercialisation Agricole Organization for Economic Cooperation and Development Office National de la Cooperation et d’Assistance pour le Développement Organisme stockeur Programme Agricole (Senegal’s agricultural credit program 1960-80) Points d’Appui de Prévulgarisation et de l’Expérimentation Multilocale Sulfur Société d’Aide Technique et de Cooperation xiv SISHAR SCUEFTTEX SOCElA SCHACOS SCAAR $20. 1, etc UE USAID V/C VHP SEIB SIES SIP SISMAR SODEFITEX SODEVA SONACOS SONAR SPO, 1, etc. UE USAID V/C VMP Société Electrique et Industrielle du Baol Société Industrielle des Engrais du Sénégal Sociétés Indigenes de Prévoyance Sociéte Industrielle Sahélienne de Mécanique, de Matériels Agricoles et de Representations Société de Développement des Fibres Textiles Société de Développement et de la Vulgarisation Agricole Société Nationale de Commercialisation des Oléagineux du Sénégal Société Nationale pour l’Approvisionnement du Monde Rural Used to indicate different soil preparation themes; the greater the number the more intensive the theme Unites Expérimentales United States Agency for International Development Value/cost ratio, a measure of fertilizer profitability Value of the Marginal Product XV I. 121'. 1. PFC Richard I. INTRODUCTION A. PROBLEM SETTING The year 1990 will mark the 100th anniversary of’ Commandant Richard’s introduction of the peanut to Senegal and the launching of a still unfulfilled quest to modernize Senegalese agriculture. The modernization plans were built around a perceived ”technological imper- ative” -- agriculture could only become more productive through. the introduction and rapid expansion of modern technologies. The building blocks were improved seed, chemical fertilizers, and agricultural equipment. The cement which was to hold it all together in the long- run was farmer-run cooperatives. The scaffolding provided to facili- tate the modernization was a government administered program of agricultural sector interventions to encourage the cooperative movement and agricultural investment. One can point to numerous accomplishments during Senegal’s first century of peanut culture; however, it was exceedingly clear as the decade of the 19805 began that the hoped for modernization had not taken place and serious problems had to be redressed. Researchers’ failure to provide drought resistant technologies was eroding confi- dence in the technological imperative. The government scaffolding which was to have been a temporary, enabling structure had taken on a life of its own; it was siphoning off agricultural surpluses to sustain its own existence. Little progress was being made in building grass- roots farm organizations. The government is now grappling with the task of identifying the causes of these unanticipated results and seeking remedies. In 1984 it 1 :ert 0 r I r t (CH, u."‘ .07.“ 2 officially released the ”Nouvelle Politique Agricole" (NPA), a state- ment of new initiatives being undertaken to improve agricultural pro- ductivity. The transfer of many functions from government to the private sector and improved price policies are the cornerstones of the new program. The government has clearly stated that, despite current problems, it continues to support an active fertilizer policy, postulating that correct use of chemical fertilizers would increase crop yields by an average of 40 percent (Government of Senegal, 1984, p. 48). National fertilizer consumption goals elaborated in the NPA (Government of Senegal, 1984, p. 44) were: 40,000 metric tons for 1984/85 70,000 " " " 1985/86 90,000 ' " ” 1986/87 120,000 " " " 1987/88 This represents only a small share of the 285-300,000 tons of total estimated needs (assuming application of 150 kilos/hectare to 75 per- cent of all cultivated areas). Even these modest goals are proving elusive. To date, the private sector has shown a reluctance to accept the high risks associated with fertilizer marketing. At the same time, farmers have been unwilling to purchase quantities supplied, blaming unfavorable price and credit terms. To suggest remedial actions is no easy task. Fertilizer policy; is a complex subject for countries such as Senegal which have rainfed agricultural zones characterized by: (1) Crop varieties relatively unresponsive to fertilizer (Shalit and Binswanger, 1984); 3 (2) Changes in soil quality and rainfall patterns which bring into question continued use of fertilizer recommendations based on 15-20 year old agronomic trials (IFDC, 1980; Forest, 1982); (3) Unreliable rains with frequent, unpredictable droughts that expose fertilizer users to high levels of risk; (4) Low farm incomes which constrain investment in modern inputs; (5) Relatively high input distribution costs given the low and geographically dispersed demand (Crawford and Kelly, 1984). The government has made a political statement concerning the direction it plans to take in fertilizer policy and is slowly working out the details of implementation. There are three basic categories of fertilizer policy questions which must be answered before these details can be worked out: (1) Hhat formulas should be manufactured and what doses should be recommended to farmers? This question cannot be answered without careful study of manufacturing capacity and costs, agronomic considerations, extension and distribution considerations, and farmers’ willingness and ability to purchase. (2) What should fertilizer price policy be? Answering this ques- tion demands careful consideration of agricultural subsidy and tax policies, the relationship between fertilizer and product prices, the capacity of the agricultural product marketing system to pur- chase farm produce at profitable prices and in a timely fashion, and the issue of uniform prices across regions and products. (3) What type of fertilizer distribution system will best meet the needs of all participants in the system? The answer to this (1! 235m 1C, pt SETlOu $52295 tiles. 4 question requires information about the relative costs and benefits of private versus public sector distribution networks, decisions about who will bear the risks and provide needed credit, consideration of the need for incentives to encourage private sector involvement, and finally tough decisions about the most appropriate roles for both government and private sector organiza- tions. The above list is only partial, but it gives one a sense of the magnitude of the task. Hithin each basic question is nested a seeming- ly endless number of equally difficult sub-questions. Many of these questions are ones which have been answered in the past, but the answers must be constantly reexamined in light of the changing econom— ic, political, and ecological situation; others have not yet been seriously considered because the Senegalese have taken government involvement for granted for such a long time. 8. RESEARCH DESIGN AND ORGANIZATION OF THE DISSERTATION The present study provides information and analysis which can be used to examine the fertilizer policy issues outlined above. The intent is to help policy' makers judiciously' select objectives and successfully design fertilizer policies which will further these objec- tives. The research is focused on the following three questions: (1) What can we learn from past experience with fertilizer policy and input distribution programs that is relevant to improving the design of future programs? (2) What does analysis of agronomic data reveal about returns to fertilizer under various price, credit and ecological conditions? 5 (3) Hhat can we learn about farmers’ investment behavior that will help us to better understand effective demand? In light of the three general questions posed above, five specific objectives were delineated: (1) Describe and evaluate the performance of institutions and policies that shaped fertilizer demand and supply from 1949 to the present; 1 (2) Review and evaluate past research on economic returns to fer- tilizer in the Senegalese Peanut Basin; (3) Present an updated analysis of economic returns to fertilizer using partial budgets and investment decision models based on new estimates of fertilizer response; (4) Describe fertilizer acquisition and use under the input distribution policies introduced in the 19805, identifying major factors having influenced fertilizer demand during this period; (5) Develop logit and hierarchical decision models (HDM) that synthesize descriptive information on fertilizer acquisition and use thereby providing guidance for the design of policy instru- ments most likely to increase fertilizer demand; (6) Examine the implications of knowledge gained in meeting objectives 1-5 for design of Senegalese fertilizer policy. The study is largely micro-economic in terms of data collection as it concentrates on gathering information about farmers’ acquisition and use of fertilizer, their attitudes about fertilizer' and fertilizer policy, and their investment behavior. The policy analysis associated with the first and last objective, however, takes into consideration the inter; the intera the exlge problems, sector, et tural poli Sever first two fished pa; agronomic El‘S dlr‘ec policy al Programs 1 third objr 6 the interaction of macro- and micro-economic factors -- particularly the interactions between fertilizer price and distribution policies and the exigencies of the national budgetary crisis, balance of payment problems, poor performance of the agricultural product processing sector, etc. Also taken into account are the linkages between agricul- tural policy and the evolution of the Senegalese fertilizer industry. Several data collection and analysis methods have been used. The first two objectives are met by a review of government documents, pub- lished papers, and unpublished reports concerning fertilizer policy and agronomic research; discussions with government officials and research- ers directly involved in development and implementation of fertilizer policy also provided important insights. Ar time-consuming search through mostly unpublished documents on Senegalese fertilizer research programs was necessary to locate appropriate data sets for meeting the third objective. Regression analysis of this data provided the fertil- izer response estimates used to update the economic analysis. Data collected by the Société de Développement et de Vulgarisation Agricole (SODEVA) and information obtained from a series of farmer interviews are used to meet the fourth and fifth objectives. SODEVA, a rural development agency working throughout the Peanut Basin, made available several years of detailed input/output data for a sample of farmers located in the Sine Saloum. Time and resource considerations restricted the present study to Nioro and Gossas, two of the six departments in the Sine Saloum.l A map showing the locations of the 1From an administrative perspective, the Sine Saloum was one of the six government subdivisions called "regions”. From an agricultural per- spective, it is one of the major peanut producing areas in the Peanut 7 Sine Saloum, Gossas, and Nioro is presented in Figure 1. More detail on sample selection, survey design, and data analysis are presented in Chapter VI of the dissertation which precedes the discussion of survey results. The dissertation consists of nine chapters. Chapter II traces the evolution of agricultural policies affecting the demand and supply of fertilizer up to 1980, evaluates the effectiveness of these policies, and describes the series of ad hoc policy measures introduced after the 1980 collapse of the agricultural credit program. Chapter III de- scribes an attempt to model Senegalese fertilizer demand using knowl- edge obtained in Chapter II and time-series data for the 1961-80 period. Chapter IV reviews past analyses of the economic returns to fertilizer. Chapter V presents our own updated analysis of economic returns to fertilizer. Chapters VI through VIII deal with on-farm surveys conducted during the 1985/86 agricultural season. Chapter VI discusses survey methods. Chapter VII describes farmers’ knowledge and attitudes about fertilizer, economic analyses used by them for making input invest- ments, and how they acquired and used fertilizer during the 1980-85 period. Chapter VIII presents results of the logit and hierarchical decision modeling efforts. Chapter IX summarizes information presented in Chapters 11 through VIII and discusses the policy implications. Basin. In 1984 the region was divided into the Region of Fatick (which includes the Department of Gossas) and the Region of Kaolack (which includes the Department of Nioro). As our research covered both of these new regions, we continue to use the term Sine Saloum throughout the dissertation. Esopom acvm us» use Fmomcmm mo no: oeoooo\.. )V .u- can 6:3. 2.. an: 235 3.2.3 uo I” Jawrdm "H mcsmwn a. 32...... o Qv—ox O on nucaou¢fluh I I I he. JUOWGFIIV . Uwflflb .\\ 0’. 0‘. 000 Q 001‘ :8 9‘ .00 DO ‘0 PZLQFO \ ‘5 006‘ 000. 00‘. . . .33...” n. .. O 0' ad a a .03.. C 32:. new. Sac—om uc_m _moo:mm mo uwpasaom 11. AN INSTITUTIONAL ANALYSIS OF FERTILIZER DEMAND AND SUPPLY The roots of Senegal’s current fertilizer dilemma go back to poli- cies of state intervention introduced by the colonial government and pursued in earnest by the post-independence socialist government which sought to establish state control over all aspects of agricultural pro- duction, marketing, and processing. Our thesis is that the high level of state intervention in agriculture encouraged increased use of modern inputs but' led eventually to a complex, over-centralized, non- performing institutional structure which siphoned off agricultural surpluses for largely non-productive uses. The growth of this mono- lithic structure stunted the private sector, rendering it incapable of assuming its traditional input manufacture and distribution role when government finally did withdraw. Government’s heavy hand in agricul- ture has also alienated farmers and discouraged development of basic decision—making and management skills required for economically sound farm investments. This discussion is organized around the two themes of fertilizer demand and supply. He first trace the evolution of institutions and policies that shaped fertilizer supply between 1949 and 1980. Next we turn to policies and institutions which influenced farmer income, access to credit and knowledge of fertilizer during this same period-- i. e., those factors which influence demand. A look at institutional performance, particularly the tendency for input supply to exceed demand and fer farm revenues to stagnate, follows. The chapter ends with a description of ad hoc policy measures implemented to sustain the USE pro! Te Lio 10 use of modern inputs after the 1980 collapse of the agricultural credit program. A. DETERMINANTS OF FERTILIZER SUPPLY: 1949 - 1980 The major determinants of fertilizer supply to 1980 were: (1) Agricultural researchers who prescribed N-P-K formulas; 9(2) The agricultural bureaucracy charged with ordering and dis- tributing a quantity of fertilizer that would meet agricultural production objectives while satisfying farmer demand; (3) A Government of Senegal (GOS) - Société Industrieller des Engrais du Sénégal (SIES) agreement which stipulated minimum quantity and price conditions for GOS orders. He examine each of these determinants of supply in the next three sec- tions of this chapter. 1. AGRICULTURAL RESEARCH DETERMINES FERTILIZER PRODUCTS SUPPLIED Agricultural research conducted first by colonial research services and later by French and Senegalese institutes has been the greatest single determinant of the types of fertilizer supplied to Senegalese farmers. During the 40 years since fertilizer research began, research objectives and recommendations have remained relatively stable. Much of the current policy debate revolves around the question of whether long-standing fertilizer recommendations remain relevant in the 19805. For this reason, we present a detailed discussion of how recommendations. developed overtime and the .justifications. for ‘them. Economic analysis conducted by the Senegalese research establishment receives little attention here because Chapter IV treats the subject in-depth. 11 (A) EARLY CRA/CNRA AND IRHO RESEARCH Two different organizations conducted the earliest Senegalese fer- tilizer research, the Centre de Recherches Agronomiques at Bambey (CRA-Bambey),l which began fertilizer research in 1947, and Institut de Recherches pour les Huiles et Oléagineux (IRHO), which began trials in 1951. The programs of these two institutes developed independently until 1963 when the GOS requested a joint statement on fertilizer recommendations for the extension services. The most important part of the early CRA/CNRA research program was a series of multi-rate peanut trials conducted from 1950-57. The result was the ”themes légers" (low intensification theme) recommenda- tion shown in Table 1. Researchers proposed the "themes légers" to farmers about 1955. These recommendations were not designed to attain maximum crop produc- tion but to familiarize farmers with fertilizer technology without putting too great a drain on their finances: Il s’agit dans un premier temps jusqu’en 1963, tout en accroissant du facon significative et a faible coOt la pro- duction agricole, de familiariser le paysan sénéga ais ,a l'utilisation des engrais minéraux (ISRA, 1980, p. 1). 1Agricultural research began in Senegal at Bambey in 1921. In the 19505, the research station officially became the Centre de Recherches Agronomiques-Bambey (CRA-Bambey), responsible for coordinating agricul- tural research throughout French West Africa. At independence in 1960, the CRA-Bambey became the Centre National de Recherches Agronomiques du Sénégal (CNRA). The French Institut de Recherches Agronomiques Tropi- cales (IRAT) ran the CNRA under contract with the Senegalese government from 1960 to 1974 at which time the Institut Sénégalais de Recherches Agricoles (ISRA) was created. 2Translation: In the early years before 1963, although low cost fer- tilizer recomendations increased production significantly, the main objective was to familiarize the Senegalese peasant with the use of chemical fertilizers. 12 Table 1: Original "Themes Légers' Fertilizer Recommendations Recommended Dose “921111— 150 kg of 14-7-7.a Zone Formula P Louga 12-10-10 E Tivaouane 10-14-08 A Thiés 10-00-30 N Northern Sine Saloum 06-20-10 0 Southern Sine Saloum 06-20-10 150 150 150 150 150 MILLET: Recommendations were the same for all zones -- Source: ISRA (1980, p. l) and ISRA (1975, p. 2). 3During this early period millet trials were given low priority, never- theless, CRA-Bambey did conduct a limited number in the central and southern Peanut Basin. IFDC (1977) looked at some of this data and claimed that there was only an 80 percent chance that the responses noted were statistically significant. 13 ...on considere que les techniques culturales insuffisantes ne permettant pas a l’engrais d'amener les récoltes a leur potentiel; bien d’autres facteurs jouent le role de facteurs limitants pour la production agricole; les apports d’engrais sont donc volontairement limités de facon a ne pas trop gréver la trésorerie des paysans et a assurer cependant une marge de profit mfximum (debut de la courbe Mitscherlich). (ISRA, 1975, p. 1) From an economic perspective, it is interesting to note that the author of the second quotation considers a recommendation which maximizes profit as a compromise, the ultimate objective being to "bring crops to their full potential“ -- i. e., attain maximum yield. In 1957 researchers presented better farmers -- those who planted and weeded on time, using animal traction, seeders, and hoes -- with the "fumure étalée"2 recommendations that are presented in Table 2. The recommendations were designed for a four-year crop rotation of fallow or green manure followed by peanuts, cereal, and peanuts again. The objective was to encourage greater fertilizer use while taking advantage of Senegal’s rich phosphate resources. "Fumure étalée" employed local phosphates in a basal dose, thereby cutting down on phosphate imports in the form of N-P-K compounds. "Fumure étalée" was designed... non seulement a obtenir de hauts niveaux de rendements, mais aussi a maintenir et meme augmenter la fertilité du sol en 1Translation: "...given that poor cultural techniques did not permit fertilizer to bring the crops to their full potential -- many other factors acting as constraints on agricultural production -- doses were voluntarily limited so they would not unduly drain peasants’ resources but would, nevertheless, assure a margin of maximum profit (beginning of the Mitscherlich curve)." 2These recommendations have also been called "themes lourds" and ”themes semi-intensifs". 14 Table 2: Original "Fumure Etalée" Fertilizer Recommendations Year 1: 500 kg/ha of local tricalcium phosphate on a fallow or green manure Year 2: 50 kg of potassium (in the form of 85 kg of potassium chloride) on the first peanut crop following the fallow. lear_3: 60 kg of nitrogen (in the form of 300 kg of ammonium sulfate) on millet Year 4: Repeat year two recommendations for the second peanut crop Source: ISRA (1980, p. 1) and ISRA (1975, pp. 6-7). 15 assurant un bilanl mineral équilibré si non excédentaire. (ISRA, 1975, p. 4) The IRHO program complemented and, for the most part, confirmed CRA/CNRA results for peanut fertilization.2 IRHO mapped four fertil- izer recomendation zones for the Peanut Basin which are shown in Figure 2. Multi-rate trials conducted on research stations in Louga, Tivaouane, Bambey, and Darou provided the baseline data. ”Confirma- tion' trials and demonstrations conducted on farmers’ fields, primarily in the Sine Saloum, confirmed results of multi-rate trials and produced the recommendations in Table 3. With the exception of potassium in the North (IRHO - 0, CNRA -- 10), CRA/CNRA ”themes légers" and IRHO recommendations are similar. IRHO did not propose different doses for better farmers, insisting that all trials and demonstrations be conducted with a high level of crop management (use of animal traction, mechanical seeders, hoes, and fertilizer spreaders; mandatory five-year rotation with two fallows, and timely performance of all planting, weeding, and harvesting activities). (8) CONSOLIDATION 0F EARLY RESEARCH RESULTS Hhen CNRA/IRHO results were consolidated in 1963, the "themes légers" recommendations became those reported in Table 4. The "fumure étalée" recommendations were revised upward to those shown in Table 5 lTranslation: ”... not only to obtain high yields, but also to main- tain and even increase soil fertility by insuring that minerals added through fertilizer were equal to or in excess of mineral extractions." ZInformation about IRHO’s program is taken from Gillier and Prévot (1960). IRHO did no work on millet fertilization other than looking at the residual effect on millet crops following peanuts. 16 ZONE A :gQ. ZONE C 41 ‘ ‘ 50 7 0 7 50 58 339- 22 , , :; ZONE 8 //// ZONE 0 .:7 40-60 -20 xf/x’ 70-30-0 Source: Gillier and Prévot (1960, p. 138) Note: Numbers denote kilos of N-P-K, not percents. Figure 2: IRHO Mapping of Fertilizer Zones in 19605 Table 3: Original IRHO Fertilizer Recommendations Zane Egrmula uantit k ha North 15-10-00 100 Thies 11-00-30 100 Northern Sine Saloum 07-20-10 120 Southern Sine Saloum 06-22-09 150 Source: Gillier and Prévot (1960, p. 138). Note: These recommendations were translated to standard N-P-K format from information presented in Figure 2. Conversions are based on 22 percent nitrogen in ammonium sulfate, 40 percent phosphate in dicalcium phosphate and 34 percent in "phosphal" (calcined aluminium phosphate), and 60 percent potassium in potassium chloride. 18 Table 4: 1963 "Themes Légers" Fertilizer Recommendations m—Ic:z>m-u Zgng Form n it k ha Louga 12-10-10 120 Thies: 10-00-30 100 South Thies and Diourbel 6-20-10 120 Sine Saloum, Eastern Senegal and Casamance: 6-20-10 150 MILLET - Recommendations were the same for all zones: 150 kilos/ha of 14-7-7 Source: ISRA (1975, p. 3). 19 Table 5: 1963 "Fumure Etalee' Fertilizer Recommendations 500 kg/ha of local tricalcium phosphate on a fallow or green manure PLUS 100 kg ammonium sulfate on the green manure only 50 kg of potassium (in the form of 85 kg of potassium chlo- ride) on the first peanut crop following the fallow PLUS 50 kg uf ummuniuu sulfate 60 kg of nitrogen (in the form of 300 kg of ammonium sulfate) on millet Repeat year-two recommendations for the second peanut crop ISRA (1975, p. 6) and ISRA (1980, p. 3). €335 abou 1Tar Can 1 than 20 because researchers found nitrogen mineral balances to be negative with previous levels. (C) MODIFICATIONS TO ACCOMMODATE EXTENSION AND MANUFACTURING INTERESTS In the late 19605 researchers modified the "themes légers" to simplify distribution and extension. It was felt that a standard quantity of fertilizer per hectare would be easier to distribute and apply so recommendations were adjusted to the 150 kilos/hectare doses shown in Table 6. The ”fumure étalée” also underwent a few revisions. One was the replacement of ammonium sulfate by urea because the former caused acidification (ISRA, 1980, p. 3). Research on nitrogen leaching problems indicated that urea was most effective if applied in three equal doses at planting, tillering, and elongation. Artotal dose of 150 kilos/hectare, with one-third applied at each of these three inter- vals, was recommended for millet to satisfy the extension services’ demand for quantity uniformity and researchers’ desire to reduce leach- ing (ISRA, 1975, p. 7). Also to comply with the 150 kilos/hectare rule, the potassium chloride and ammonium sulfate application for peanuts became 150 kilos/hectare of 3-0-42. (D) GENERAL SATISFACTION WITH RESEARCH RESULTS By the end of the 19605, CNRA researchers were quite confident about fertilizer recommendations, particularly the ”themes légers". Les formules [themes légers] actuellement vulgarisées ne don- nent lieu a aucune contestation sérieuse,.... On déplorera seulement que moins de 10% des surfaces cultivées au Sénégal recoivent ces fumures annuelles... (Tourte et al., 1971, p. 641) 1Translation: ”The fermulas ["themes légers"] currently recommended can no longer be seriously questioned,... One only regrets that less than 10 percent of cultivated area in Senegal receives these 21 Table 6 "Themes Légers" Fertilizer Recommendations -- Late 19605 Zane ' Lemuel; mum Louga 10-10-8 150 kilos/ha Thies 06-10-20 150 kilos/ha South Thies and Diourbel 06-20-10 150 kilos/ha Sine Saloum, Eastern Senegal, Casamance 06-20-10 150 kilos/ha Source: ISRA (1975, p. 3). Resu‘. cooti 22 Results from 'prévulgarisation" demonstration trials which began in 1963 provided an important complement to on-station trials, increasing confidence in recommendations.1 Tourte et al. (p. 641) did foresee future changes in potassium and phosphate rates for soils in continuous cultivation because preliminary research showed that over time the soil’s ability to provide potassium declined while phosphate availability increased.2 In making these comments, the authors anticipated one of the highly debated issues of the 19705 and 19805 -- how to estimate correct potassium doses. (E) RESEARCH AND RECOMMENDATIONS IN THE 19705 According to ISRA (1980) the three major research concerns of the 19705 were: (1) Maintenance of the nation’s soil resources; (2) Food security; (3) Reduction of economic costs associated with fertilizer use. recommended doses..." l"Prévulgarisation" (literally, pre-extension) trials were conducted on researcher managed fields at various locations outside the major re- search stations using larger surface areas (400 square meters) than research trials. 2To support this forecast, the authors presented data on P and K ef- fects (kilos of added millet associated with each nutrient) for a field cultivated continuously during nine years. Years 1-5 received 15 kg/ha of K20 while years 6-9 received 30 kg/ha. Year 1 2 3 4 5 6 7 8 9 P effect 317 252 176 142 145 77 87 135 252 K effect 94 233 241 241 231 210 504 265 473 This same data is also presented in ISRA (1975, p. 12). H ,l 1'. '5 111. 23 Researchers stopped conducting multi-rate trials, focusing their attention on mineral balance and soil renewal studies. The inter-action between fertilizer and various agricultural practices (plowing, rotation, incorporation of organic matter) retained research- ers’ attention. Trials comparing ”themes légers" and "themes lourds" were common. Treatments were often some combination of the two prin- cipal factors -- fertilizer and soil preparation.1 Given the growing importance of government food security objectives, research on cereals grew in popularity. Researchers also showed a heightened desire to better understand fertilizer response under farmers’ conditions. This became possible in 1969 when ISRA created the "Unites Expérimentales" (UE), providing a structure which gave agronomists the opportunity to work with farmers in their own fields. The UE also facilitated the collection and analysis of data on socio-economic factors influencing farmers’ adoption behavior. Adjustments in fertilizer recommendations in the 19705 came in response to a variety of concerns expressed by researchers and non- researchers alike. Agronomists presented convincing evidence from mineral balance studies that continued use of "themes légers", which remained popular despite attempts to encourage the "fumure étalée", was 1An example is the protocol used by the ”Type de Fumure" program which had three treatments identified as follows: FOSPO: no fertilizer, no soil preparation FISP1: ”themes légers" fertilizer and light plowing F2SP2: "themes lourds” fertilizer and heavy plowing with incorporation of crops residues. This type of protocol made it impossible to do any type of response curve estimation or profit maximization analysis, but most researchers felt that previous multi-rate trials and analyses had more than ade- quately responded to such needs. 24 accelerating soil depletion. Another concern was that cereal yields possible with 14-7-7 were far too low to meet the government’s food security goals. The extension services continued to demand fewer and simpler recomendations and the Société Industrielle des Engrais du Sénégal (SIES) put pressure on agronomists to reduce the total number of fertilizer formulas required nationwide. By the mid-19705, fertil- izer reconmendations had been modified several times to ‘accommodate these varied interests. Shortly after the creation of SIES fabrication of N-P-K using local phosphates became possible; there was no longer an important justification for the separate phosphate applications associated with 'fumure étalée”. A one-time basal dose was still recommended for better farmers, however, particularly on newly cleared land that ex- hibited phosphate deficiencies. The basal phosphate recommendation was reduced from 500 to 400 kilos because extension agents claimed it was easier to teach farmers to apply two 50-kilo sacks per "corde"1. The 500-kilo dose required extension agents to explain application in terms of hectares (units farmers did not know well) or to deal with partial sacks (2% sacks per ”corde"). Changes in phosphate recommendations necessitated new N-P-K recom- mendations for peanuts (150 kilos of 7-21-29) and millet (150 kilos 10-21-21 plus 100 kilos of urea). The term "fumure étalée" was re- placed by the term "themes lourds”. The new millet formula contained substantially increased potassium levels. The justification was the introduction of the Souna III variety which had greater demand for this lA 'corde' is a measure equal to 1/4 hectare. n“ urie 25 nutrient. Within a few years the peanut formula changed from 7-21-29 to 8-18-27, a multi-purpose formula which further accommodated manufacturers and distributors because it served for peanuts as well as rainfed rice, and cotton. The 1970 changes for the "themes légers" concerned only peanuts and only the Sine Saloum, Eastern Senegal, and the Casamance.l These were the zones where soil depletion was accelerating because fallows were less frequent, crop residues were removed more completely, and farmers’ technical skills had improved. The 8-18-27 formula previously recomended only for "themes lourds" was now recommended for both "themes légers" and "themes lourds” peanuts in the three zones speci- fied above. One view of the rationale for this change maintains that: Il ne s’agit donc pas simplement de chercher a augmenter les rendements de facon a obtenir des profits immédiats, pour une dépense d’investissement minimum, mais plus profondement, d’approcher du potentiel de production des cultures et obte- nir regulierement des rendements élevés pendant une période indéfinie. (ISRA, 1980, p. 4) 1There is some disagreement in the literature on this point. ISRA 1975 insists that only peanut recommendations were increased, but ISRA (1980, p. 4) states that the 10-21-21 plus urea recommendation for millet was ”proposed by researchers" for "themes légers" farmers (im- plying, but not clearly stating that it was a recommendation actually used by the extension services). The following are the "themes légers“ recommendations which, according to ISRA (1980), research "proposed" in 1974: Eeenuts Millet North 10-10-08 + 14 S North 14-07-07 + 16 S Center North 06-20-10 + 8 S South 10-21-21 + 5 S South 08-18-27 + 5 S + 100 kg urea IFDC (1977) and IFDC (1980) both mention the 10-21-21 plus urea formula as the standard "themes légers" recommendation in the Sine Saloum. 2Translation: "It is not a question, therefore, of simply looking to increase yields so as to obtain quick returns to the smallest possible investment, but more profoundly, to approach the full productive 26 ISRA (1980) talks about ”themes intensifs“ introduced in 1972 for superior farmers capable of using even greater fertilizer application rates than the ”themes lourds", but it never clearly defines these rates. Documents on 'prévulgarisation" trials in the ”Amélioration Fonciere" program mention the introduction of ”themes intensifs" in 1976, but do not moment on any attempts to actually recommend this particular theme to farmers. Table 7 reports "themes intensifs" fer- tilizer doses. The intent of the "themes intensifs” was to completely replace all the nutrients consumed by the crops planted. Recent mineral balance studies in Senegal (P. L. Sarr, 1981, for example) and attempts to synthesize results from mineral balance stud- ies throughout the arid and semi-arid tropics (Pieri, 1985) have high- lighted the difficulties of accurately measuring and interpreting mineral balances. Some of this more recent work suggests that meeting soil maintenance objectives requires greater attention to nitrogen balances and organic matter than earlier research implied. In addition to looking at mineral balances, P. L. Sarr (1981) measured variations in mineral stocks for land that had been continuously cultivated during 17 years.1 His findings suggest that there is little difference in mineral stocks for land not fertilized and that fertilized with "themes potential of crops, obtaining high yields regularly during an indefinite period.” 1Mineral balances are calculated by measuring plant uptake of various minerals and subtracting that from nfinerals applied through chemical fertilizer and organic matter. Mineral stocks are obtained by measur- ing the level of minerals actually in the soil; this method is better able to deal with some of the more elusive factors influencing soil fertility (leaching, nitrogen fixation, mineralization, etc.). 27 Table 7: "Themes Intensifs” Recommendations: 1976 to Present Peanuts: 200 kilos of 8-18-27 250 kilos of "phosphogypse' 50 kilos of potassium chloride plus 400 kilos of tricalcium phosphate for basal phosphate application the first year the field is cultivated Millet: 150 kilos of 10-21-21 plus 150 kilos of urea Source: Unpublished ISRA notes on treatments for the "Amélioration Fonciere" demonstrations. 28 légers"; nitrogen and calcium decline substantially while available phosphorous and potassium increase slightly. Yield performance with the “themes légers”, however, was significantly greater; 28 percent increase for peanuts, triple yield for sorghum, and nine times greater yield for maize.1 Despite these new findings and the rather dramatic changes in rainfall patterns and fertilizer prices, no adjustment in fertilizer recommendations has occurred since the "Md-19705 when the ”themes légers“ and themes lourds” recommendations summarized in Table 8 were established. (F) SUMMARY By way of summary, it is important to note that a limited number of recurrent themes are responsible for most changes in fertilizer recommendations. Although agronomic considerations, particularly the results of mineral balance studies, take precedence in developing recommendations; one discerns a willingness to accommodate non- agronomic considerations. These include constraints faced by local fertilizer manufacturers, and the extension services’ desire to simpli- fy application procedures. On the other hand, a strong reluctance to base recomendations on economic criteria is evident. Agronomists prefer to give agronomic criteria priority, suggesting that adjustments in price policy or improvements in marketing strategies, not changes in recommended doses, are the best means of addressing the problem of low economic returns. lPieri (1985, pp. 11-15 and Annexe 2) presents a summary of P. L. Sarr (1981) from which this mineral stock information was taken. 29 Table 8: Fertilizer Recommendations Mid-19705 to Present "Theme" Zone Crop Formula Quantity 'Léger" North Peanuts 10-10-08 150 kilos Center North 06-20-10 " Southa 08-18-27 " All Cereals 14-07-07 " "Lourd" All Peanuts 08-18-27 " All Cereals 10-21-21 " 46-00-00 100 kilos Source: Unpublished ISRA notes on 1976 fertilizer recommendations. aSouth includes Tiénaba through Eastern Senegal. 30 2. THE AGRICULTURAL BUREAUCRACY TAKES CHARGE OF INPUT DISTRIBUTION The Senegalese penchant for state control of the agricultural input sector is very much a continuation of the colonial heritage. In the early 18005, before the technological imperative associated with the introduction of peanuts, Governor Roger saw government as the prime mover in Senegalese agricultural development. The following passage describes only a small part of Roger’s grand plan for the colony: ...a c6té des établissements du Gouvernement, véritables sta- tions d’essais et de vulgarisation agricole, les concessions agricoles mises en valeur par des particuliers, mais béné- ficiant d’un important appui gouvernemental: primes a la production, avances en outils, bestiaux, armes et munitions utiles pour la protection dans un pays qui n’a pas été domi- né, distribution gratuite de graines et de végétaux; fourni- ture de vivres de soudure dans l’attente de la récolte de mil sans charge de restitution, encadrement et orientation par des jardiniers d’Etat, travaux divers d’intéret commun: c16- tures de villages, tra aux de protection, puits, digues, etc... (Ly, 1958, p. 24) For those already familiar with Senegal’s post-independence credit and input distribution programs, this passage has a strong quality of "deja vu"; for those new to the topic, the similarities between Roger’s vision and post-independence programs will be apparent by the end of this chapter. Conceptually, the institutional structure for determining input supply and distribution has changed little over time. Since 1910, the 1Translation: "...along side the government establishments -- virtual agricultural research and extension stations -- farms being run by private individuals, but benefiting from important government support: economic incentives for productivity, advances in equipment, animals, arms and munitions useful for protection in a country not yet dominat- ed, free distribution of seeds and plants; gifts of food aid in the hungry season before the millet harvest, orientation and training provided by government agents, various activities to improve the com- munity: village enclosures, protection of wells, dikes, etc." 31 government has nurtured a network of farmer organizations which were to protect farmers from the vagaries of the market place and usurious money lenders. These organizations were designed to communicate farmers’ input needs to local authorities and to serve as conduits for allocation of credit and distribution of agricultural inputs. I The impetus for the creation of the first farmer organizations was an overwhelming concern on the part of colonial administrators for the Senegalese farmer’s inability to manage his own peanut seed. Ly pro- vides an excellent account of how these concerns were first expressed in 1893 (only three years after the introduction of peanut culture). After several initiatives to develop village storage projects and short-term credit programs, the ultimate solution was the 1910 decree establishing the Sociétés Indigenes de Prévoyance (SIPs). SIPs were a form of mutual aid society. Membership was contingent on payment of annual dues. Financing for activities came from dues, interest pay- ments, and a variety of government funds. SIPs went through a number of legislative reforms. The most important were the 1915 law making membership obligatory fer all heads of household ("chefs de carré"), and the gradual expansion of the SIP mandate from seed credit to a wide range of credit and mutual aid activities. Fertilizer marketing began in 1949. By 1957, SIP members had con- sumed 27,000 metric tons. Fertilizer was distributed on credit with the largely government financed Fonds d’Investissement pour le Dévelop- pement Economique et Social (FIDES) and Fonds Commun (FC) having sup- ported all transport costs and a significant share of purchase costs. Information about the actual mechanisms used for determining quantities 32 supplied and accomplishing delivery is sparse, but discussions of the SIP system in general imply that colonial administrators had a strong hand in all decisions and control over the various government funds supporting SIP activities.1 Although the same concept of farmer organization enabled by government services continued after independence, new institutions and procedures evolved. The most important new actors and their duties are summarized in Figure 3.2 The process of determining supply theoretically began at the farm. Farmers were to comunicate input needs to their cooperatives from September through November (before the current year’s crop was market- ed). It appears that even at this early stage the system was flawed. Schumacher (1975, p. 196) cites a 1963 study of 28 Sine Saloum coopera- tives where "more often than not, credit requests for a cooperative were estimated solely by its president.” Regardless of procedures, completed requests were passed on to CRAD agents. Having entered government channels, the requests were subject to review and revision by numerous technical and financial services. Among those mentioned by Schumacher (1975, pp. 193-94) are: Regional Office of Agricultural Services; Regional Development Committee; Agricultural Services Depart- ment, Ministry of Rural Economy, Dakar; Governing Board of the Fonds Mutualiste de Développement Rural (FMDR), and the Cabinet. Once the 1Information on SIP’s comes primarily from Ly (1958). Tignor (1987) also provides a thorough discussion of pre-independence farmer organi- zations in Senegal. 2See Poli (1970) and Cissé et al. (1967a) for more detailed descrip- tions of these institutions. 33 OCA Office de Commercialisation Agricole: charged with purchasing farmers’ crops and selling them to processors,ordering and arranging delivery of inputs, and importing and distributing rice CRAD Centres Regionaux d’Assistance au Développement: provided liaison between OCA and cooperatives; charged with keeping cooperative records and training cooperative leadership ONCAD Office National de la Cooperation et d’Assistance pour le Développement: created in 1966 to consolidate most OCA and CRAD functions CERP Centres d’Expansion Rurale Polyvalent: local administrative units charged with coordination of services provided by OCA and CRAD and provision of technical advice concerning agriculture, health, forestry, animal husbandry, and cooperatives BSD/BNDS Banque Sénégalaise de Développement (became Banque Nationale de Développement du Sénégal): headed consortium of banks providing credit for agricultural input distribution and peanut marketing FMDR Fonds Mutualiste de Développement Rural: source of guarantee for agricultural equipment debts and of credit for large producers, also financed input subsidies Figure 3: Description of Major Government Institutions Serving the Agricultural Sector in the Post-Independence Period cues 34 total value of each region’s envelope was approved (usually in March or April), special committees met in each region to decide how much would be allocated to each cooperative. Schumacher (1975, p. 193) claims that original cooperative re- quests were... ...adjusted somewhat arbitrarily by regional and central officials according to their own estimation of the farmers’ "real” needs and in view of various technical and financial guidelines established by government. Although this complex process of review and revision continued through March and April, OCA was obliged to place preliminary orders for ap- proximately 50 percent of anticipated needs in September. According to IFDC (1977, p. 40), OCA officials used fertilizer orders for the previ- ous campaign as a guide for estimating preliminary orders. OCA had to place final orders and make some advanced payment by the end of Decem- ber; 'The only farm-level intelligence that OCA had at this time was the statement of each cooperative’s expressed needs which CRAD agents had communicated to OCA. The consequences of over- or under- ordering had to be ironed out in regional meetings when final allocations were made, and within each cooperative. If deliveries to cooperatives were less than the original orders, the president or other cooperative officers made final allocation decisions. Cooperative leadership positions tended to be dominated by former 'traitants”1, religious leaders, and other influential individu- als whose decisions few dared to question. Invariably members whose l"Traitants" were traders who acted as intermediaries between the French commercial houses, which financed peanut marketing during the colonial period, and the farmers. 35 allotments were less than anticipated believed they were being discriminated against. In many cooperatives this became an ethnic issue (e. g., the Serer claiming that the Holof had disproportionate control). A series of bookkeeping entries at the Banque Nationale de Dével- oppement du Sénégal (BNDS) kept track of fertilizer program finances. Hhen OCA placed orders in January it received a twelve-month line of credit. Once OCA had delivered inputs to the local CRAD units, it turned its delivery receipts over to the BNDS who credited the OCA account and debited the CRAD accounts. The CRADs in turn delivered the fertilizer to the cooperatives, presenting delivery receipts to the BNDS who then transferred the amounts due to the individual cooperative accounts and removed the debt from CRAD ledgers. Although the BNDS performed most of the bookkeeping, the CRADs were responsible for recording their shipping and storage costs which were to be reimbursed directiy by OCA. According to Casswell (1984, p. 52) no systematic attempt was made to record individual debts until the early 19705 because it was thought that personalizing debt would hamper development of collective solidar- ity. Even after individual accounts were authorized, farmers as well as cooperative administrators had difficulty getting accurate informa- tion on account balances from the BNDS. A 1977 study cited by Casswell (pp. 52-53) found that 100 percent of individual accounts audited had errors -- 80 percent of the errors were to the detriment of farmers. The BNDS had little incentive to keep the records straight because collective funds from cooperative rebates and the FMDR. guaranteed 36 reimbursement. Poor accounting procedures encouraged fraud which, some claim, reduced farmers’ incomes by 5-10 percent between 1960 and 1980 (Casswell, p. 58). Government attempts tar improve agricultural sector' performance resulted in a number of relatively minor structural changes during the 19605-805. The most important for input supply was the consolidation of the CRADs and OCA when the Office Nationale de la Cooperation et d’Assistance pour le Dévelopement (ONCAD) was created in 1966. In the short-run, this relatively cosmetic change had little impact; in the long-run, however, ONCAD grew into a costly, unmanageable bureaucracy with a list of responsibilities so diverse that it was unable to per- form any of them well (see Frélastre, 1982, pp. 57-63; and Casswell, 1984, p. 42 for in-depth discussion of these many activities). 3. FERTILIZER SUPPLIED BY A GOVERNMENT PROTECTED MONOPOLY1 Senegal is one of the few African countries that has its own fertilizer industry. This has not been an unmitigated blessing for the agricultural sector. Prior to 1968 the fertilizer industry consisted of two phosphate extraction firms2 oriented toward the export market but providing small amounts of natural phosphates for "fumure étalée". In 1968 SIES,3 having negotiated a very favorable industrial agreement 1Information in this section comes primarily from USAID (1983, Annex F, pp. 8-9). 2Société Sénégalaise des Phosphates de Thies and Compagnie Sénégalaise des Phosphates de Taiba. 3SIES was indirectly owned by the French corporation Mines de Potasses d’Alsace which had a 90 percent interest in the Société Sénégalaise d’Engrais et de Produits Chimiques, the major shareholder in SIES. Other shareholders were the two local phosphate companies, the World Bank, and the BNDS. 37 with the GOS, began production of N-P-K compound fertilizer for the domestic market. The GOS/SIES agreement contained two clauses with the potential for stimulating an excess supply of fertilizer: (1) The GOS agreed to purchase a minimum of 30,000 tons per year come rain or shine; (2) A variable pricing arrangement was approved whereby the GOS would pay higher prices for any quantity less than 60,000 tons, thereby insuring SIES that fixed costs of production would be covered even when the plant was running significantly under capacity. He have been unable to find any documented evidence that this GOS/SIES agreement caused government fertilizer orders to be greater than anticipated farmer demand. Both the variable pricing clause and the minimum order requirement, however, had the capacity to stimulate over: consumption. Furthermore, the variable pricing clause and .a guaranteed six percent profit, kept the price of domestic fertilizer relatively high compared to world prices. Exports accounted for only 5-10 percent of SIES revenues because its products were not competitive in international markets, even though the GOS was assuring the coverage of fixed costs. In essence, the agricultural sector became dependent on a government protected monopoly for its fertilizer supplies. B. DETERMINANTS OF FERTILIZER DEMAND In liberal input marketing systems a farmer’s financial situation (income and access to credit) and his knowledge of economic returns associated with different inputs shapes his demand for particular prod- ucts. Although Senegal’s agricultural input sector was not a liberal 38 marketing system, planners believed that fertilizer consumption depended on these same factors. To insure a propitious environment for increased consumption of modern agricultural inputs, government became increasingly involved in peanut marketing, agricultural credit, and extension services. 1. GOVERNMENT CONTROLS PEANUT MARKETING TO GUARANTEE FARMERS FAIR PRICES Government control of post-independence peanut marketing was a reaction against the colonial past rather than a continuation of it. Colonial marketing was dominated by the French commercial houses1 because they were the only ones with access to the necessary commercial bank credit. Some historical accounts suggest that calculated efforts by French and Lebanese to restrict Senegalese participation in the colonial peanut trade were responsible for the post-independence pro- clivity toward nationalization of commerce (Amin, 1969).2 Most accounts, however, blame the usurious practices of Lebanese and 1Buhan et Teisseire, Deves et Chaumet, and Maurel et Prom were among the most active French commercial houses. 2Amin (1969) notes that the introduction of peanuts and the colonial pacification of the interior occurred concurrently. Prior to these two developments, the gum trade was the predominant economic activity in Senegal and the colonial government encouraged the participation of Senegalese traders because the interior was considered too dangerous for others. In 1900, 500 of the 800 registered traders in Senegal were Senegalese. The Senegalese gum trade experienced a crisis in the early 19005 and was virtually wiped out due to stiff competition from British Sudan. Rather than integrating the Senegalese gum traders into the newly established peanut trade, the colonial government and French commercial houses chose to reserve the peanut sector for themselves, thereby sowing the seeds for what outsiders considered an over-zealous post-independence program of "Senegalization' and ultimate nationaliza- tion because by independence there was no longer a Senegalese commer- cial class with resources to invest. 39 Senegalese 'traitants”, licensed traders who acted as intermediaries between commercial houses and farmers. To insure a reliable supply of peanuts at marketing time, "trai- tants' offered farmers credit for food or emergencies during the plant- ing season. The credit was secured by agricultural equipment or future harvests. Hhen future harvests served as collateral, the quantity of peanuts required to pay the debt always had a much larger market value than the debt being acquitted. In the case of equipment collateral, a trader would actually take possession of the item (usually a seeder which was not needed after planting). To recuperate his property, the farmer had to pay back his debt plus substantial interest in peanut seed. These credit practices drew criticism from colonial administrators who considered the interest charges usurious. In an effort to diminish farmers’ reliance on "traitants", the government authorized SIPs to market members’ peanuts in 1933; and in 1947 pressure from the French socialist party made it possible for farmers to form cooperatives for the sole purpose of peanut marketing. Commercial interests were ada- mantly opposed to these attempts at "liberalization", denouncing them as the first step toward collectivizing the peanut sector and predict- ing it would lead to ultimate destruction of rural commerce (Ly, pp. 37-8). SIPs and cooperatives never became a serious threat to the large commercial houses during the colonial period, but the pre-independence growth of farmer organizations, and efforts to make them more 40 democratic and to expand their mandates should not be dismissed.1 The network of cooperatives already established by the early 19605 served to expedite the post-independence drive to nationalize the peanut trade. Nationalization of peanut marketing was to (1) give farmers a fair price, thereby freeing them from their cycle of indebtedness; and (2) insure that peanut sector profits went either to farmers or government, thereby eliminating transfer of these revenues to ”unproductive" middlemen. The same institutions responsible for input supply (OCA, CRADs, BNDS, and cooperatives) were also assigned marketing functions. Through the 1966/67 campaign, OS ("organismes stockeurs" or licensed traders) performed first-handler functions along with cooperatives. OCA provided financing obtained through the BNDS. Both cooperatives and OS earned a 1.5 FCFA per kilo rebate on all purchases. The CRADs served as intermediaries between OCA and the cooperatives, facilitating cash flow, supervising weighers, and keeping cooperative accounts. The OS accounted for an important share of total peanut marketings when they were phased out (27 percent in 1966 and 36 percent in 1967) (Amin, p. 60). Farmers’ continued dealings with OS were probably linked to lLy (1958) presents an excellent discussion of farmer organizations and their relationship to the state during the colonial period, ending with his proposals for creating a truly democratic network of farmer cooper- atives capable of carrying out the socialist mandate of the post- colonial government. His criticisms of early farm organizations bear an uncanny resemblance to criticisms of the cooperative movement in the 19705 and 19805 (e.g., Government of Senegal, 1984; D. Sarr, 1985; Crawford et al., 1985) suggesting that farmer organizations have yet to meet the expectations of their earliest advocates. Tignor (1987) also provides a review of pre-independence cooperative movements in Senegal. 41 traditional credit transactions or a desire to circumvent obligatory credit repayment at their own cooperatives. After the 1966 consolidation of OCA and CRAD into ONCAD (to im- prove efficiency by placing often competing services within a single, unified decision-making structure), and the elimination of the OS, the rural end of the peanut marketing structure remained unchanged until the 1980 dissolution of ONCAD. The creation of the Société Nationale de Commercialisation des Oléagineux du Sénégal (SONACOS) in 1975 trans- formed the higher end of the marketing chain: SONACOS was a largely government owned peanut purchasing conglomerate which bought peanuts from ONCAD and hired processing services from private sector firms (Lesieur Afrique, Sodec, V. Q. Petersen, Société Electrique et Indus- trielle de la Casamance, etc.) that had formerly purchased their own peanuts. The Société Electrique et Industrielle du Baol (SEIB), a 51 percent government-owned processor was the only other purchaser of ONCAD peanuts beginning with the 1975/76 campaign. By 1980, however, the private sector processors, not happy with their "piece-work" stat- us, withdrew from processing activities. Each had signed agreements with the GOS whereby their facilities were rented to SONACOS, with options to purchase. SONACOS was now responsible for both marketing and processing functions. while expansion of state intervention in peanut marketing and processing activities was to have enhanced govern- ment ability to realize peanut sector profits, in the long-run it became a liability.1 Hhile increasing government revenues, the 1See Casswell (1984), Thénevin and Yung (1982), or Andersen, Gaye and Associes for more information on costs and benefits of GOS involvement in the peanut sector. 42 peanut marketing program was also to improve farmers’ ability to invest in agriculture. The 1.5 FCFA per kilo marketing margin earned by the cooperatives had a multitude of planned uses. Above all it served to guarantee cooperative seed and “Programme Agricole” (PA) debts.l Planners anticipated that margins would create a sizeable fund enabling cooperatives to become multi-functional engines of rural development, and permitting much of the government enabling structure to wither away. At the individual level, farmers would be guaranteed the official price (which they had not always received from 'traitants"). This price tended to be significantly less than the world market equiva- lent;2 but the profit was now going to the government who reinvested a share in agriculture through the FMDR. The FMDR acted as a conduit for financing agricultural subsidies -- particularly fertilizer -- and a variety of other rural development activities. In essence, government control of peanut marketing was to free farmers from debt not by increasing farm-gate prices but by reducing farmer dependence on un- scrupulous traders. In freeing farmers from nonproductive debt, the government hoped to channel resources into improved technology-- especially fertilizer and agricultural equipment. Consequently, the peanut marketing program went hand-in-hand with the PA credit program. 1See the next section of this chapter for a discussion of the PA. 2Caswell (1984, p. 48) presents data for the period 1965-80 comparing estimates of total peanut revenues received by farmers to those re- ceived by ONCAD. She concludes that producers’ receipts averaged about 50 percent of the final value of peanuts marketed by ONCAD. 43 Comments made above about ONCAD’s loose accounting system, accused of encouraging fraud in input supply activities, are even more relevant with respect to peanut marketing activities. This is the context in which most of the ONCAD fraud appears to have occurred. Schumacher and Casswell both provide ample discussion of these problems and the impact on farmer revenues.l In effect, the new system served to shift farm- ers’ perception of the 'bad guy": During the colonial period it was the "traitant'; after independence it became the government acting through ONCAD. In his analysis of state intervention in the Peanut Basin, Waterbury (1987, p. 190) points out that "... the relationship of cultivators to state agencies has been, in both directions, manipul- ative or adversarial, but seldom cooperative." 2. "PROGRAMME AGRICOLE” CREDIT SPURS DEMAND FOR MODERN INPUTS Agricultural credit programs in pre- and post-independent Senegal have been administered through so-called "farmer organizations" created in fact by administrative decree and run largely by government agents. Unfortunately, none of these programs has proven financially sound over time, all of them being plagued by farmer defaults ultimately covered by recourse to government funds. The Fonds Commun created for the SIPs in 1936 began running annual deficits of 100 million francs (Ly, 1958, p. 57) and by 1958, the out- standing SIP debt totaled 300 million FCFA (Schumacher, 1975, p. 187). The pre-independence cooperative movement sponsored by the socialist lCasswell (p. 58) notes, for example, that in 1976 16 percent of weigh- ers were fired by ONCAD for theft. She goes on to point out that those cases which involved weighers stealing from the cooperatives rather than from ONCAD were seldom brought to justice. 44 party also failed the test of financial viability. Ly (1958, p. 59) points out that by 1952, '11 fut établi que le mouvement coopératif ne vivait plus que grace a l’intervention des finances publiques;"1 and the movement was dissolved. Debts totaled some 292 million francs of which only about 15 million had been collected by 1956. By independence, the Senegalese believed they had learned from past failures. The key was to have truly democratic farmers’ organiza- tions formed by freely associating individuals, and some system of guarantee to cover debt defaults. The "Programme Agricole” (PA), administered through a new system of cooperatives and guaranteed by the peanut marketing rebate, which would not be paid to cooperatives until debts were recovered, was the proposed solution. The PA offered 5-year credit for equipment purchases and seasonal credit for fertilizer and I'nonproductive" subsistence loans. ONCAD administered seed credit directly without passing through the cooperatives. Seed defaults, however, were also backed by cooperative marketing rebates. The under- lying principles which made credit, cooperatives, and peanut marketing interdependent were: (1) The collective guarantee by the cooperative of all debts con- tracted by individual members; (2) The concept of debt-carrying capacity being established at the cooperative level rather than on an individual basis; (3) The link between credit repayment and peanut marketing, with lTranslation: "...it was determined that were it not for the injection of government funds, the cooperative movement would be dead.’I valu perc perc (Set were cate for equi vest stat debt COVE 45 debts collected by marketing agents and rebates serving as a guarantee. Cooperative borrowing limits were established at 25 percent of the value of the previous year’s peanut marketings and not more than 7 percent of total credit could be used for subsistence loans.1 The 25 percent limit was apparently seldom approached and never exceeded (Schumacher, 1975, p. 192).2 Early attempts to limit subsistence loans were unsuccessful. Such loans accounted for 1/2 to 2/3 of credit allo- cated until the 1966/67 campaign when strict enforcement began. Demand for credit and reimbursement rates fluctuated; but, judging by data on equipment sales and fertilizer orders, the PA appeared to increase in- vestment in modern inputs during the 19605 and 19705.3 At the same time the PA and ONCAD credit program were presenting statistics on diffusion of modern technologies, they were accumulating debt defaults. At first the defaults were relatively minor and easily covered by rebates. During the 19705, however, they became more common and government found itself "forgiving" farm debts after several par- ticularly bad seasons. By 1980, the government financial situation was so precarious that agricultural sector debts and subsidies could no longer be financed. This brought the PA to an abrupt halt in 1980 and 1Casswell (1984, p. 52) claims the credit limit was 25 percent of the average of the three previous years’ peanut marketings. 2Schumacher’s research does not go beyond the early 19705 and no other source of information on how well the 25 percent rule was followed in the late 19705 has been identified. 3Data supporting this claim are presented in Figures 4 and 5 with the discussion of institutional performance. led c0ul< erat ses. been pose exa: not C0111 min Sch 46 led to a doubling of fertilizer prices by 1983/84 as the government could no longer pay for the subsidy.1 The repeated failures of Senegalese agricultural credit and coop- erative programs have generated many published and unpublished analy- ses. Most suggest that farmer organizations through which credit has been administered are not authentic cooperatives. The solution pro- posed is to foster creation of ”authentic" cooperatives (see, for example, Ly, 1958; D. Sarr, 1985; Belloncle, 1980; and Gellar, 1987). Another perspective suggests that Senegalese rural society is not as conducive to socialist development as "political ideologues” and ”ad- ministrative framers” of the cooperative movement originally thought. Schumacher reviewed a number of studies expressing this point of view: ...several locality monographs and technical studies pointed to incompatibilities between the requirements for effective agricultural cooperatives and the traditional behavioral orientations of the Senegalese peasantry. Observers under- scored the gap separating formal prescriptions of democratic participation by the rank and file in cooperative affairs from the actual internal dynamics of most rural associations. Far more decisive than legislation and administrative regula- tions emanating from Dakar in determining the actual opera- tion of cooperatives were such factors as the existing vil- lage elite structure; customary social differentiations on the basis of age, sex, family, caste, and land-tenure rights; inter- and intra-village rivalries among political factions; and deep-rooted decision-making practices stressing the norm of unanimity as well as deference to elders. ...traditional values and solidarities in some ways clearly ran counter to forms of collective behavior which the cooper- atives were supposed to foster. Virtually all observers underscored, for example, the deep reluctance of ordinary members to contravene kinship ties or deferential norms to 1Casswell provides one of the better analyses of ONCAD’s financial situation in the late 19705 using records from the Caisse de Péréqua- tion et de Stabilisation des Prix (CPSP), the Banque Centrale des Etats de l’Afrique de l’Ouest (BCEAO), and the International Bank for Recons- truction and Development (IBRD). -_._Lr\A-‘nf' These for l attew and C the 1 to t1 bure ial stra 1954 demc lanl husl inc- Pro eSt 47 undertake action (legal or' otherwise) against. cooperative officers known to have uflsappropriated funds. Likewise, it was rare for legal sanctions or even social pressure to be applied against individuals defaulting on loan payments, despite the fact that the BNDS would debit the amounts owed from the rebates of all cooperative members. The limited relevance of traditional behavior orientations to the suc- cessful operation of cooperatives was also indicated by the fact that cooperative rebates were rarely used to finance collective investments initiated by the membership of these associations. (Schumacher, 1975, pp. 154-5) These observations, written in the 19605 and early 19705 remain valid for the village cooperatives (”sections villageoises") -- the newest attempt to create authentic farmer organizations (see 0. Sarr, 1985, and Crawford et al., 1985). Rather than protecting farmers from debt, the PA increased their vulnerability to it. Farmers were now in debt to the government as well as to local traders. 3. AGRICULTURAL EXTENSION INCREASES DEMAND BY IMPROVING KNOWLEDGE While ONCAD took care of inputs and peanut marketing, a separate bureaucracy evolved to handle extension activities. During the colon- ial period, limited extension efforts were mounted by colonial admini- strators, SIPs, and the Centres d’Expansion Rurale (CER) created in 1954. IRHO’s confirmation trials also served as extension demonstrations. At independence, a network of Centres d’Expansion Rurale Polyva- lante (CERP) provided agents in the fields of agriculture, animal husbandry, forestry, and health. Much of the credit for farmers’ increased familiarity with fertilizer must go to the FAO Fertilizer Programme conducted from 1961-66 in collaboration with the CERPs. FAO estimates that 46,500 farmers attended a total of 3,100 field days for fertilizer demonstrations (FAO, 1974, p. 35). The FAO program was first to ted taro In ment to SATEC's Basin 1 to com oed o SMEC' tiallj point it b. sopp evol fell 48 first to stress the importance of fertilizing cereal crops and motiva- ted farmers to increase the percent of fertilizer going to millet. In 1964 the government contracted with the French project manage- ment corporation ”Société d’Aide Technique et de Cooperation" (SATEC). SATEC’s primary objective was to increase productivity in the Peanut Basin by 25 percent over four years. This increased productivity was to compensate for the anticipated fall in prices when the French stop- ped paying preferential prices for Senegalese peanuts in 1967.1 SATEC’s arrival fueled a debate on whether extension should be essen- tially technical or conmunity development oriented. The technical point of view won.2 SATEC was so well endowed with financial and human resources, that it began taking on some of the responsibilities for determining input supply that originally belonged to the CRADs. Schumacher describes the evolution of SATEC’s responsibility for determining input needs as follows: ...in July 1965 the ministers of rural economy and of plan- ning jointly charged SATEC with the tasks of determining optimum farm input objectives for the entire project area and for individual farming units; providing individualized coun- seling to farmers in the preparation of their annual credit requests; and advising cooperative members at the general meetings where individual input needs and cooperative orders for equipment and supplies were determined....Equipped with special record books, the extension teams were instructed to make an inventory of each farm (areas cultivated, available labor force, fertilizer and seed consumption, previous 1As early as 1933, the French Government guaranteed its colonies a preferential peanut price by imposing a 15 to 20 percent tariff on peanut imports from non-French areas (notably, Nigeria). When France joined the European Economic Community, it was obliged to discontinue this preferential treatment. 2See Schumacher (1975, pp. 198-205) for a discussion of this debate. This SATEt input ed prov many age mode fron SODE Opir tech Hldg have ibre 49 investments in equipment and draft animals, etc.). In September, on the basis of this information, the village workers began 'door-to-door" consultations with farmers to encourage them to purchase the "optimal” quantity of farm inputs, though often this was done with insufficient regard to the financial situation of either the farmer or his cooperative. (Schumacher, 1975, p. 203) This policy of "door-to-door" counseling did not last long; by 1969 SATEC was instructing extension agents to rely on village meetings for input counseling because the former method had led agents to: 'propagandize in favor of increased purchases of farm inputs without considering the financial situation of each individ- ual or his cooperative as a whole. It also did not permit peasants to be satisfactorily informed of the economic and technical justification for their investments. (Schumacher, 1975, p. 217) In 1968, SATEC turned the extension program over to a newly creat- ed Senegalese rural development agency, SODEVA; but they continued providing technical assistance for a number of years. Recognizing that many of the new agricultural technologies were too expensive for aver- age farmers to adopt easily, SODEVA instituted a program stressing model farmers who were more prepared than others to realize profits from rapid adoption of technical packages. Another characteristic of SODEVA’s program was to avoid working through the cooperatives, devel- oping smaller producer groups oriented toward particular crops or technologies. Like ONCAD, SODEVA evolved into an immense bureaucracy, assuming a wide range of extension, research, and development activities; many have not been performed well. C. INSTITUTIONAL PERFORMANCE Before trying to evaluate the performance of institutions respons- ible for fertilizer demand and supply, it is important to consider the 50 performance criteria we will use. We have identified two categories of criteria: (1) those used to some extent by the government and (2) those which we think were relevant but ignored by the government. In the first category, the following three criteria were most popular: (1) Quantity of modern inputs distributed; (2) Growth in aggregate production of peanuts and millet; (3) Growth in government revenues from the peanut sector. We have identified two criteria which fall into the second category: (1) Growth in farmers’ personal income; (2) Ability of the system to equate fertilizer demand and supply. Our hypothesis is that the strong belief in the technological impera- tive led government to choose a narrow range of performance indicators. Farm income, the one variable which absolutely had to progress if growth in agricultural productivity was to be sustained, was given little if any attention by evaluators. During the 20 years that government sought to build a socialist agricultural sector and bureaucracies such as ONCAD and SODEVA grew by leaps and bounds, performance indicators used by the government did not always signal poor performance. Figures 4-8 present data on the evolu- tion of a number of productivity indicators up to 1980. Through the mid-19605 availability of credit and price subsidies were judged to be having a favorable effect -- use of fertilizer and purchases of modern equipment surged forward (see Figures 4 and 5). The impact on peanut production appeared positive as production increased from 892,000 metric tons in 1960/61 to 1,168,000 in 1965/66 and marketed peanuts went from 786,000 to 1,089,000 tons (see Figure 6). Much of the gain \ OAUO‘O - coon!“ . 51 .AcmmHV commewwm mu manage can Ammmfiv mmumwuomm< un< .Ammomfiv wmmmu Eon; twpwasou Nm\mmma-ao\ooma cowuaszmcou can coaosaeapmeo Lme_wptmc ”a weaned .mmwocmm< acmeaopm>wo pacaz new o_—mu mucmmwcgmc u_ ma—co mm\~mmfiafim\ommfi cow opampwm>a name comuan_cum_a .mcmscee co mm>_aocmaoou "mucaom on mmwcm>Fpou.mucmmmgqmc a? mzpco um\omoH-m~\esa~ com mpnapme>e name :o_uaszm=ou ”muoz mwmfiwp nxwm_. mwhmw— nuthP mwwmwp nuwmwp O 000"— 0001N 0000” 000.? OOONB COSQEJnCOU ooovo coznntuna D 00000 00000— coco". 132;”in to suol suiaw E F“ L\ {K 1 0000K. 52 OQICQGH .cwmam uacmwa up; cw mommzogzm “cup—item um 9:57.. .ANH mxmcc< .NmmHV acman—m>mc —eg:m mo xcumwcwz .Pmmmcmm mo ucmsccm>ou mep mhmw Ohmr mwmp Cmmr nu \ nunXUnuw OOOON OOOOM OOOO¢ OOOOfl OOOO® QOOON. ”muczom our) (sung pup ‘smold ‘saoH 'suapaag sapn pasouomd suun To JBQUJDN 1.51.“ F OQm~ L. 5 9L! omuoomfi .mcmumxcmz ecu comuuzuoca “scam; ommpmmwcmm no acumen .Amumfiv cucumsasum use .Aemmfiv —pmzmmmu .Amwmav ocumupwcu cw mama scum umpvasou owmr mhmp Ohm? mom? 53 UmumeOE mujcoml COZODUQkquCOOQ nuwmw— “mucaom o owe own We 0 n m fluemw “u D- S own w. W W. IJ oom m. l O on? w Wu: com, d 9 D U 031 m. 0.06? 00m, vas thr thousar vas a\ produc 514,0‘ exoar k110i 195C 0161‘. Ave be 54 was through increased planting (1,114 thousand hectares versus 977 thousand) but Figure 7 shows that increased productivity per hectare was also evident (1007 kilos per hectare versus 913). Aggregate millet production increased steadily from 392,000 metric tons in 1960/61 to 514,000 tons in 1964/65 (see Figure 8). The increase was due to expansion of area cultivated, as yields per hectare declined from 574 lcilos in 1960/61 to 508 in 1964/65. Estimates of average annual growth in gross industrial product due to agriculture was 4.4 percent for 1960-65 (Frélastre, 1982, p. 50, citing the Senegalese Fifth Develop- ment Plan). Indicators for the late 19605 and early 19705 were less favorable. Average annual growth in agricultural productivity was only 1.1 percent between 1965 and 1970 and 1.6 percent from 1970-1974 (Frélastre, 1982, D. 50). Peanut production for five of the ten years between 1966/67 and 1975/76 was lower than the 1960 production (Figure 6). The decline iii millet production was less dramatic; in only two years did produc- ‘tion fall below 1960 levels (Figure 8). The 1965-1975 period witnessed a sharp increase in the percent of fertilizer going to cereal crops Which could partially explain why millet production did relatively better than peanuts. Total quantities of inputs demanded declined, however, during this same period. Unfavorable changes in farm-gate prices and poor rains are the most commonly cited causes of this poor performance. Given the difficulty of determining to what extent poor performance was due to policy rather than exogenous price and weather effects, there was a tendency to place most of the blame on the latter, ignoring important signs that policy itself needed to be reconsidered. emofiummma .wcmuuw: can u—mw> uacmm¢.mo cowus—o>u ”u «gnaw; .Aomm: Ease: can .9me 2%: .3me 553;... 59C uquEou ”mucaom Dmmp flhmp Ohmp nmmp Ommp name onmp n¢mp Ovm— mamm— O Om, OCH 1‘.‘ 1'1 09¢ 000 55 OWN 00m O_DO:O>O uOc owlmnmp LO‘ OuOO CMOP BJDIOBH 19d sinuoad )0 so. OONF Owns comp \Ofiwo~ 56 nzwmwp mwhgwp owm~-oeaa .=o_su=uOLa am___z mmm_mmm=mm nuhgwp mxmeP ”m mcamwm .Ammmfiv mwumPQOmm< an< ”mucaom 00m? . my AUnUF Aunzw ll Mn nu nu oon m. nu nr 8 n:U¢. mr m Aunxw mm“ nu ma Aunxw nu n: w. nxun. W. m. nzumw 1+. nXUmw 000p ll peanut annual (Fréla fertil report Anon very 57 The 1974/75 harvest, however, seemed to signal a turning point; peanut production was on the rise and the World Bank’s estimate of annual agricultural growth for the 1974-77 period was 7.1 percent (Frélastre, 1982, p. 50). In the mid-19705 an IFDC mission gave both fertilizer policy and performance of the fertilizer sector a good report card. Senegal has the agricultural base for increasing fertilizer use. The costzprice relationship between fertilizer and crops is very favorable to the producer. Crop production programs are underway in all major regions of the country and organizations are structured to deliver credit and production inputs and collect farmer produce. The study team estimates that fertilizer use will increase at a rate of 8 to 12% per year for the next 5 to 10 years, provided the above institutional factors favoring fertilizer use are maintained (IFDC, 1977, p. 13). Annong the institutional factors referred to by the IFDC report was a very important double subsidy: 1) The government paid SIES higher than world market prices 2) Farmers purchased fertilizer at less than world market prices1 No formal study of price policy was conducted, but IFDC presented several justifications fOr' both subsidies. The industrial subsidy increased domestic employment, saved foreign exchange, and provided a hedge against fluctuating world market supplies and prices. This latter point is subject to debate given that all raw materials but phosphate still had to be imported. The farm subsidy increased crop production, particularly cereals, which reduced dependence on imported food. The report notes that between 1962 and 1976 fertilizer applied 1IFDC (1977) reported that the 1976 subsidy to the fertilizer manufac- turing industry totaled 1.84 billion FCFA (about $8 Inillion); the subsidy paid to farmers in the same year was 1.34 billion FCFA (about $6 million). to mi For 5 pound due 1 rate The dena can; Peri PCT. sel sut OEt tu' BS 58 to millet increased from 9.3 to 30.4 percent of national consumption. For Senegal as a whole fertilizer use increased at an annually com- pounded rate of 20 percent between 1964 and 1967, declined from 1968-70 due to drought and producer price effects, then climbed at a compound rate of 40 percent annually between 1970 and 1976 (IFDC, 1977, p. 33). The 1975/76 campaign had achieved record harvests and record fertilizer demand. Orders for equipment reached a record high the following campaign. During the 1970/71 through 1973/74 campaigns the Caisse de Péréquation et de Stabilisation des Prix (CPSP) realized 24 billion FCFA in revenues from the agricultural sector (mostly on profits from selling peanut and cotton products bought by ONCAD and SODEFITEX) while subsidizing consumers to by 17.5 billion FCFA (Abt, 1985, p. 13, citing OECD sources).:l From a government revenue perspective, the agricul- tural sector appeared to be fulfilling its mission. It is difficult to assess the impact of increased use of modern inputs and growth in aggregate production on farm income because no appropriate price index is available. Figure 9 shows farm revenue from peanut production in nominal 1960-80 prices and compares it to an estimate of real revenue for the 1965-78 period.2 Figure 9 shows that, except for 1975, real farm income remained below the 1965 level for the entire period. Not only does real income fail to rise, but both real and current revenues 1The CPSP was officially created in 1973, to assume the function of price stabilization which were previously carried out by a number of smaller, product-specific, price stabilization boards. SODEFITEX, Société de Développement des Fibres Textiles, is a rural development agency similar to SODEVA; it specializes in cotton production. 2The price index is based on urban consumption for African families living in Dakar. Although not ideal for deflating farmer revenues, it is the only price index available in Senegal. The base year is 1970. 59 ommuuommfl .mmpmm uzcmma meowweo scam mmacm>mm Exam “m mcamwu .mxooacmm> pmowumwumum .z.= Edge ma xmu:_ woven "Acmmflv P_ozmmou ecu Amuafiv cogumszgum cm umaoac mmucaom oo .. .05 OZ grooooo T0006? gr0000“ ext frc na: do be f0 er ex 19 re re bi As it 60 exhibit increasingly greater inter-annual fluctuations as one moves from the 19605 to the 1970s and 19805.1 Although criticisms of the ONCAD/SODEVA/Cooperative Movement’s management ability and cost structure ‘were topics of conversation during the 1965-75 period, and farmers’ defaults on PA debts were becoming more frequent, the relatively good harvests of 1975 and 1976 fostered a general belief that all was going well. In 1978/79, howev- er, the government entered a prolonged period of severe economic crisis exposing many of the previously hidden weaknesses in the system. By 1980 the government. was subsidizing the peanut sector: rather ‘than realizing revenues from it. Criticism of agricultural policy and the institutional structure responsible for its implementation started coming from all quarters-- bilateral and multilateral donors as well as the Senegalese National Assembly and the MDR itself. Criticisms were leveled primarily against the three main actors -- ONCAD, the Cooperative Movement, and SODEVA. The most common charges leveled against government services (corrup- tion, mismanagement, over-centralization, sky-rocketing costs) and against the Cooperative Movement (non-democratic administration and an absence of solidarity among members) have been mentioned above and are well documented in numerous reports and studies (Schumacher, 1975; Casswell, 1984; Government of Senegal, 1984; Frélastre, 1982; Waterbury 1 A close look at peanut yield trends shown in Figure 7 suggests that yield variability is at least partially responsible for the income variability. While the average yield in the post 1960 period is better than that for the pre-independence period, yields per hectare demon- strate increasing inter-annual variability in the seventies and eig ties. and gener dena isti to ecor als and abc the agr 811 H] 61 and Gersovitz, 1987). While the shortcomings noted in these more general critiques of institutional performance had their impact on the demand and supply of fertilizer, it is important to look more closely at exactly how demand and supply were affected by particular character- istics of the system. The preceding discussion has shown that researchers gave priority to increasing productivity and maintaining soil fertility, relegating economic analysis and studies of farm income to the back burner. We also noted that the structure for ordering fertilizer forced farmers and ONCAD to estimate quantities needed at a time when information about prices and farm incomes was unavailable. Thirdly, we saw that in the case of fertilizer, industrial interests carried more weight than agricultural interests, with SIES winning concessions which did not encourage it to prodUce at the lowest possible cost. On the demand side, we found that extension services did a reason- able job of informing farmers about the benefits of fertilizer but a poor job of advising them how to determine economically appropriate quantities to employ. Policies to insure adequate farm incomes for reinvestment in agriculture failed due to corruption and mismanagement as well as the complexity of the credit/subsidy/tax interactions which made it difficult to assess what was really happening. Finally, the cooperative-based structure for agricultural credit exhibited all the weaknesses of its colonial predecessors, raising the question of the appropriateness of the cooperative movement itself. In assessing the extent to which the system was able to equate demand and supply, we must examine two different expressions of demand: (11 or: re; cas les sub Tat 19} det que .l'f III. M 62 (1) the ”anticipated” demand represented by that quantity of fertilizer ordered by and delivered to farmers, and (2) the "effective" demand represented by that quantity of fertilizer actually paid for. In both cases we find that supply tended to exceed demand. Statistics on the level of unpaid PA debt illustrate that "effective" demand was substantially less than quantities actually delivered to farmers (see Table 9). After the 1970/71 campaign all PA debts were forgiven. In 1972/73 and 1973/74, the government assumed a large share of unpaid debts. In 1977/78, government had a sliding scale for debt forgiveness based on estimated losses due to drought.1 In addition to these fre- quent debt cancellations, estimates of cooperative debt outstanding to ONCAD in 1980 were in the 30 billion FCFA range (Sheldon and Le, 1982). There is also evidence that quantities which ONCAD made available to farmers exceeded the "anticipated" demand. A 1965 inspection of OCA and the CRADs revealed that in July 1964 undistributed farm implements, fertilizer, and fungicides attained a valued of 323 million CFA francs (Schumacher, p. 168). Following the 1974/75 campaign, ONCAD’s cost for over-stocking inputs was estimated at 1,200 million CFA (Casswell, 1984, p. 53). Even as late as 1985, one could still see evidence of fertilizer ”graveyards" in rural areas where excess ONCAD stocks had been set to rest. There is no single scapegoat for the system’s penchant for over supply; hindsight shows that the cooperative and government institu- tions designed to communicate farmer demand to manufacturers were 1It is difficult to get precise information on the government’s assump- tion of agricultural debts. Tuck (1983 and 1987) and Casswell (1984) provide two of the better accounts. 63 Table 9: Percent of Total Agricultural Debts Reimbursed by Farmers 1966-80 FERTILIZER AND EQUIPMENT SEED 1966 90 72 6 1967 91 83 1968 85 62 1969 89 76 1970 49 34.5 1971 96 95.8 1972 51 44.3 1973 73.5 57.6 1974 80 92.7 1975 89.9 88.2 1976 82.5 72.3 1977 NOT AVAILABLE 27.8 1978 54.7 62.1 1979 8.3 32.5 1980a *--- 2.2 ----- * Source: Casswell (1984, p. 47, from BCEAO sources). aData is unavailable for the separate categories; 2.2 is the percent repayment for seed, equipment, and fertilizer. clear tions that into: poll! the PT: ab' en ma EC 64 clearly deficient in both conception and performance, as were institu- tions designed to insure farmers the fair prices and access to credit that would permit agricultural investment and, theoretically, increased income. One particular view blames the failure on fertilizer price policy: The GOS has been unable to develop and implement a consistent policy with respect to relative prices of agricultural inputs and output. It was the government's intention, through the use of guaranteed producer prices and the subsidization of inputs, to encourage farmers to adopt modern cultivation techniques (e. 9. increased use of fertilizers, agricultural tools and heavier equipment). Subsidized inputs were design- ed to minimize farmer risks particularly associated with the purchase of expensive fertilizers, which are ineffective in drought years. In practice, however, fertilizer use does not appear to have been especially sensitive to price and the financing of subsidies out of earnings from the sales of export crops has tended to reduce producer prices. (USAID, July 1983, p. 5) In our opinion, however, incorrect prices are a symptom rather than a cause of poor performance: It is extremely difficult to get prices right if the requisite economic and financial data is not avail- able. One seldom mentioned weakness which seems to have permeated the entire system from the highest echelons right down to the farmer was a marked tendency to give very low priority to all types of financial and economic analysis. A few examples illustrate the point: (1) Researchers gave low priority to economic criteria in recom- mending fertilizer products; (2) Extension services encouraged farmers to increase fertilizer orders with little analysis of debt carrying capacity and no thought to the relative returns of alternative investments; (3) Input manufacturers forced ONCAD and farmers to indicate afa abov ONE! obs; tecl to 900 ver con NPA Sho Wi] fin 65 input needs at a time when they lacked adequate information on farm income and prices; (4) The cooperatives, ONCAD, and the BNDS kept such poor accounts that rigorous financial or economic analysis was impossible and fraud was encouraged; (5) Tax, subsidy, and price policies were so complex and inter- twined that it was difficult to trace profit and losses of dif- ferent sectors and of participants in each sector. This lack of attention to financial and economic analysis fostered a false sense of security (exemplified by the IFDC analysis discussed above, which condoned fertilizer subsidies but had no analysis to back it up). All the furor over the collapse of the PA and dissolution of ONCAD has tended to focus attention on institutional inadequacies, obscuring the equally important question of how the pursuit of this technological imperative has affected farm incomes and how it is likely to affect them in the future. The NPA statement on fertilizer is a good example -- fertilizer consumption goals are set at levels which were seldom achieved under much more favorable climatic and price conditions (see p. 2 above). There is absolutely no suggestion in the NPA that the economic returns to fertilizer (or any other modern input) should be examined, despite ample evidence that most farmers are un- willing to pay the real cost and the government is not in a position to finance the subsidy. D. POST ”PROGRAMME AGRICOLE" INPUT DISTRIBUTION POLICIES In 1980, in response to its own assessment of the situation but also under pressure from major donors (France, the United States, the 66 International Monetary Fund, and the World Bank), the GOS began a series of fairly drastic agricultural sector reforms. One of the four primary objectives of the reform was to better insure the availability of agricultural inputs by using more adequate financing and better management (Government of Senegal, 1984, pp. 26-7). Two underlying principles guiding reforms were the desire to: . 1) Curtail direct government intervention in the agricultural sector while encouraging private sector actors (both comercial and cooperative). 2) Eliminate government subsidies and taxes to the greatest extent possible. While the need to curtail government intervention in the agricultural sector is recognized by many Senegalese, the fairly rapid pace at which government has withdrawn from certain activities is more a result of bilateral and multi-lateral donor pressure than Senegalese initiative. ONCAD was dissolved in 1980. SEIB and SONACOS, the oil processing and marketing firms which had previously purchased peanuts from ONCAD, were forced to assume the added responsibility of supervising first-handler collection activities.1 In 1985 a selected number of 1The decision to pass more responsibility on to the oil processors was not welcomed with open arms. If one uses the balance sheet of the Caisse de Péréquation et de Stabilisation des Prix (CPSP) as an indica- tor of the relative health of different sectors, one finds that the oil processing account balances had been negative in 9 of the 12 years from 1973 to 1984, with a net balance for the entire period of -25,819 million FCFA, while the peanut production balance had only been nega- tive since 1980 (5 ,years) with an accumulated 1973-84 balance of +23,379 million FCFA (Martin, 1986, Table 8). At the same time the government was forcing more responsibilities on the unprofitable pro- cessing sector, it declared that the CPSP would no longer cover loses; beginning in 1986 the oil companies would have to turn profits or swallow their own losses. This resulted in several cost saving meas- 'org from cont 1962 the SOC' inp ext but eat of ha 19 be to YC 67 ‘organisateur stockeurs' (05) received licenses to buy peanuts directly from farmers, thereby competing with the cooperatives. The rules concerning the OS were comparable to those in effect during the early 1960’s (see p. 40 above) -- financing was provided by the BNDS through the oil companies and OS margins were strictly controlled. Following the dissolution of ONCAD, the government created the Société Nationale d’Approvisionnement du Monde Rurale (SONAR) to assume input distribution functions. SONAR’s role was defined to a large extent by changes in the PA and mechanisms for financing input distri- bution. A different set of rules evolved for each of the three major categories of inputs. In 1980 all equipment credit was discontinued, with no intention of beginning a new program for at least five years. SONAR, therefore, had no equipment responsibilities whatsoever. Seed credit continued in 1980/81 with the government halving the interest rate from 25 to 12.5 percent. Despite this conciliatory overture, only 14 percent of the total debt was recovered. In 1981/82 seed credit was shifted from voluntary reimbursement proportional to quantity of seeds received to_ the ”retenue" system which imposed an across the board tax of 10 FCFA per kilo of peanuts marketed by farmers selling through official (the only legal) channels. The seed "retenue' system continued through 1984/85, with minor modifications. The system never paid for itself.1 ures, among them a sharp reduction in the number of peanut collection points thereby shifting a part of transportation costs to farmers. 1The MDR estimated that a minimum of 650,000 tons of peanuts had to be marketed annually for the seed ”retenue" to break even. This goal was never attained during the entire 1981-85 period. In dis ket off sup rai $012 by cot 198 ize sma 68 In 1985/86 the government dissolved SONAR and instructed SONACOS to distribute 75,706 tons of seed free of charge to farmers who had mar- keted 1985 peanuts in official channels. In 1986 there would be no official distribution. PA credit for fertilizer continued on a restrained scale in 1980/81 and 1981/82. Attempting to maximize returns to limited 1981/82 supplies, the government permitted distribution only in zones of better rainfall. In 1982 all PA fertilizer credit ended. Since that time, SODEVA, SODEFITEX, and the confectionery peanut program (administered by SONACOS) have offered credit through a limited number of maize, cotton, or confectionery peanut contracts. Farmers wanting to obtain fertilizer without signing contracts could make cash purchases from SONAR in 1982/83 and from SODEVA in 1983/84. The results of these two forays into cash-and-carry fertil- izer distribution were abysmal -- quantities placed were extremely small, the number of sales points was limited, information about price and sales locations was not well publicized, and price was almost doubled between 1982/83 and 1983/84.1 1This scenario of a sharp drop in demand after sizeable price hikes and the elimination of credit interestingly has an historical precedent which can be discerned from the following set of data from Cissé et al. (1967a): C n m .* Fert. Price Credit 1957 5569 tons 5 F/kg yes 1958 5661 tons 5 F/kg no 1959 3990 tons 10 F/kg no 1960 3795 tons 10 F/kg yes 1961 13262 tons 10 F/kg yes * Includes both peanut and millet fertilizer Although demand responds in a lagged fashion, the sequence of policy witl 5 F! 1982 er, ed; no I rece pers beca subs full 1985 quad diss Sene Indu bring nativ mark; Chang 1Ics 1! a IVOry and 1 al 3 Vario 69 This general decline in fertilizer use provided the government .-with an excuse to extend the ”retenue” system to fertilizer in 1984. A 5 FCFA/kilo obligatory fertilizer "retenue" was withheld from farmers’ 1983/84 and 1984/85 peanut revenues. Unlike the seed "retenue”, howev- er, the fertilizer 'retenue' was proportional to fertilizer distribut- ed;_those paying more would receive more. The fertilizer retenue was no more successful than the seed retenue. Quantities of fertilizer received by individual farmers were insignificant from an agronomic perspective (averaging 15 kilos/hectare in the Sine Saloum in 1984/85) because the government lacked resources to sustain the fertilizer subsidy. This meant that the small ”retenue" withheld had to cover full costs -- approximately 105 FCFA/kilo in 1984 and 90 FCFA/kilo in 1985. In the short span of four years farm-gate fertilizer prices had quadrupled. The fertilizer "retenue” died a quiet death in 1985 when SONAR was dissolved, with no viable alternative replacing it. In the same year, Senegal’s fertilizer manufacturing industry took on a new look when Industries Chimiques du Sénégal (ICS) came on line, absorbing SIES and bringing an official end to the GOS/SIES agreement. ICS is a multi- national endeavor with considerable attention being given to export markets.1 Despite the export orientation, ICS responded to a request changes and consequences were very similar to those of the 19805. 1ICS was created with assistance from the World Bank. It is essential- ly a private firm with ownership shared by the Senegalese (23 percent), Ivory Coast (9 percent), Nigerian (9 percent), Cameroonian (9 percent) and Indian (8 percent) governments (Kingsbury, 1985, p. 18). Addition- al shareholders are the India Farmers’ Fertilizer Cooperative and various fertilizer manufacturers who were previously active in SIES. cr ef BS se me It DU ab' ing ext com dot ity 70 by USAID and the GOS to produce a less expensive binary fertilizer for the Senegalese market. In an attempt to increase interest in produc- tion and marketing of this binary fertilizer, USAID offered to finance a subsidy on all cash sales made by the private sector (SONACOS, Cooperatives, and comercial outlets). The private sector, however, was generally unwilling to assume the risks of costly storage, trans- portation, and distribution given the uncertainty about future fertil- izer price and credit policies. Furthermore, the absence of a farmer credit program made it virtually impossible to accurately estimate effective demand; and the few independent traders who expressed inter- est claimed a major constraint was lack of commercial credit. Due to the absence of effective participation by the private sector, the 1986 "solution” again fell to SONACOS (a 50 percent govern- ment-owned corporation) which purchased 7500 tons of fertilizer from ICS, and the Union Nationale des Cooperatives Agricoles (UNCA) which purchased 610 tons. Both SONACOS and UNCA made the fertilizer avail- able to farmers at selected peanut collection points during the market- ing period. Cash sales by both organizations were practically non- existent; SONACOS reported 189 tons sold and UNCA 10 tons.1 Discouraged by very poor sales of the binary formula, an ad hoc committee established to reexamine fertilizer distribution policy decided that a major cause of poor sales was farmers’ lack of familiar- ity with the new product. The committee recommended a return to the 1Data on fertilizer orders and sales in this paragraph are from Groupe de Reflexion (12 November 1986, p. 6). old Ré.‘ sec Giv rec brc the cha Pre C00; tlm com C00; 71 old standbys of 6-20-10 and 14-7-7 for the 1987/88 campaign (Groupe de Reflexion, December 1986). . The government program for withdrawal from the input distribution sector calls for giving increased responsibility to cooperatives. Given the poor performance of previous cooperative movements, the GOS recognized a need to revamp cooperatives so that they could assume a broader role in general community development. Government efforts took the form of a drive to restructure the agricultural cooperatives into ”integrated" cooperatives which were to have the following characteristics: A l’opposé de la cooperative conventionnelle, la cooperative intégrée s’avere tres dynamique dans la mesure ob l’interven- tion de l’Etat se limite seulement a asseoir un cadre juri- dique et a mener des actions de contr6le, d’éducation et de formation. La responsabilisation des membres étant effec- tive, leur participation devient déterminante. Ainsi, la coopérative intégrée ne saurait se limiter a la commerciali- sation de produits de rente,.... (Government of Senegal, 1984, p. 31)1 Unfortunately, old habits are hard to break. (Early indications are that the basic building block of the "integrated" cooperative-- the village cooperative ("section villageoise") -- is no more a group of freely associating rural producers than previous cooperatives were. Pressure on the ”Service de l’Aide a la Cooperation” to get the reform moving resulted in the boundaries of most village cooperatives being 1Translation: In contrast to the conventional cooperative, the in- tegrated cooperative turns out to be very dynamic; State intervention is limited to the establishment of the legal framework within which the cooperative functions and to the conduct of certain educational and training programs. The process of increasing the effective participa- tion of members having been accomplished, cooperative decisions are now controlled exclusively by the membership. Hence, the integrated cooperative is no longer limited to the narrow task of marketing cash crops. admini how vi policp The 0 but t The 1 econo and t the cons izer 72 administratively defined. Many farmers have little, if any, idea of how village sections differ from former cooperatives.l In sum, agricultural policy in general, and input distribution policy in particular, has been in a state of relative chaos since 1980. The GOS has a stated policy, particularly with respect to fertilizer, but to date it has been unable to implement many facets of the policy. The rules for input distribution, shaped to a large extent by GOS economic constraints, have change radically from year to year. Farmers and distributors have received little advanced warning of changes. At the farm level, the end result has been sharply reduced fertilizer consumption and the development of new strategies for acquiring fertil- izer and compensating for diminished access. 1See 0. Sarr (1985), Crawford et al. (1985), Gaye (1986), and Gellar (1987) for a more detailed discussion of problems encountered with the launching of the new village cooperatives. III whi con par ab‘. abl mar the for to of na As an 131‘ pr st de 11C 15 III. QUANTITATIVE ANALYSIS OF FERTILIZER CONSUMPTION: 1961-1980 .Using what we have learned about the institutional environment in which demand and supply were determined, we now examine fertilizer consumption data to see if we can acquire any further insights. We are particularly interested in the extent to which price and income vari- ables explain consumption during the PA given that these are the vari- ables most frequently employed in models of input demand for liberal marketing systems. Tomek and Robinson (1977) point out that it is theoretically possible to model fertilizer demand from production function data1 but that use of time-series data for variables thought to influence demand has been more popular and successful. The absence of adequate time-series data in many developing countries, however, has led economists to estimate fertilizer demand using a variety of alter- native techniques. David (1976) estimated fertilizer demand for the Asian rice economy using a combination of time-series consumption data and cross-sectional production data. Ogunfowora (1972) used linear programming results from farm survey data to develop a regional model of fertilizer demand for the North Central State of Nigeria. Although many documents discussing Senegalese fertilizer policy present data on fertilizer consumption parameters, few attempt any statistical analysis, and none try to develop an econometric model of demand. This is understandable. First, different data sets are fre- quently in conflict and poorly documented. For example, it is often not clear if data represents quantity of fertilizer ordered by ONCAD 1See Chapter VII for a discussion of this method. 73 CO da' in C02 the 74 (i. e., supply) or quantity actually consumed by farmers.1 Second, the relationship between price and demand is difficult to discern; price was exogenously determined by government and not always announced early enough to influence demand. Third, data on some explanatory factors (e. 9., amount of credit available, dates when price changes were announced, carryover stocks held by ONCAD) are not readily available. Despite these problems, it seemed that a systematic examination of fertilizer data for the PA period might still prove useful: (1) It would illustrate the extent to which PA demand can be explained by variables normally thought to influence demand in more liberal fertilizer marketing systems; 2) To the extent that the analysis sheds light on the relative importance of different variables in determining past demand, it can be a useful complement to the decision-making information obtained from farmer interviews (see Chapter VII and VIII); 3) Problems encountered in collecting and analyzing past data can provide invaluable guidance for the types of data collection and analysis that would be most useful in the future. The analysis begins with a standard econometric approach-- conceptualization of a theoretical model and selection of appropriate data series to estimate the model. A variety of problems prevented us from obtaining a model with a good fit and statistically significant coefficients on all the variables suggested by theory and knowledge of the Senegalese fertilizer sector. Given the problems encountered, the 1See Figure 4 for an illustration of the differences between distribu- tion and consumption. ecor data larg sma' of sic so; on pl; def te th GT. 017 IE 75 econometric analysis is complemented by a factor analysis of the same data. The major contribution of the factor analysis is that it takes a large set of interdependent explanatory variables and factors out a smaller set of underlying orthogonal factors. These factors are then regressed on the dependent variable and the relative importance of each factor in explaining the variability in the dependent variable is ascertained. A. CONCEPTUALIZATION OF FERTILIZER DEMAND DURING THE PA Figure 10 presents a schematic summary of the preceding discussion of factors influencing fertilizer demand and supply. The left-hand side identifies nine separate influences on the quantity of fertilizer supplied (QFS). It is important to keep in mind the timing of ONCAD orders. A preliminary order for 50 percent of estimated needs was placed in September, before the current year’s harvest, marketing, and debt repayment performance could be taken into consideration. ONCAD tended to use the prior year’s consumption as the primary gauge for these preliminary orders. The final order and payment had to be made in December. At this time preliminary information on cooperatives’ anticipated needs was available; but the marketing season was not yet over, and the final accounting of outstanding cooperative debts was not yet available. Figure 10 shows that the quantity of fertilizer supplied (QFS) is equal to the quantity of fertilizer consumed (QFC) plus ONCAD stocks (S). Since fertilizer consumption was not determined by the give and take of market forces, we make a distinction between quantity of fer- tilizer eonsumeu and quantity of fertilizer demanded (QFD). Note that VVVVV VVVVV nausea cmepwucmu m:W:PELmumo mcopumu eo aucune: u_umem;um gun 43;: aa— pgg was; «3.! g VVVVV «33 g. nun—4:82 8 >2; :mpo0wcm< mesmcmocq= mgu mcwcao xPaazm use VVVVV 83v g nun—fl an. 30 5:3 3v 8. «Us bum-Id a; as AAAA. «was .31: 838 38 p.12:- aw: AAAAA AAAAA .................. «.8. :35. an...“ .32 Snug o ........ 8:580 33. 8 8:15!“ 33 IIIIIIIIIIIIIIIII h—gfi. .................... a-.£ 8—t 3:; nuns—plus .0124; a oooooooooooooooooooo 8-.! But a; 38 I ........ 5!“! 3338 e "OH mcamwu AOOAA ............. mg“ c-§ Guzman» 3 03;: a ugsaua— 935.533 0 ......... :88 32.8.2: . .................... hay; in; can‘t—2 p! o if . QFS con ana 198 var and lis var detr mod: TBVt slo; DOS‘ inCr Ship PFRt :1 betw used CDmb 77 if stocks are greater than zero, QFS > QFD; but zero stocks do not mean QFS - QFD, since farmers could have had unfulfilled demand. We are interested in examining how well factors thought to influ- ence fertilizer demand (the right side of Figure 10) explain fertilizer consumption data (QFC) during the PA. The QFC data used for this analysis is peanut fertilizer consumed in the Sine Saloum from 1961- 1980. The types of data available to represent the three conceptual variables thought to influence fertilizer demand (prices, expectations, and financial resources) are shown in Figure 11. Appendix 1 provides a list of data and sources used. After an examination of graphs of variables over time, scatter plots, and the correlation matrix; we determined that the choice of variables best reflecting the theoretical model was: QFC - 80 + 81 Mt-1 + 82 Mt-2 + 83 PFRt-1 + B4 PFRt_2 + Error Initial attempts to estimate QFC as a linear function of Mt-1 revealed that the relationship is parabolic, exhibiting a negative slope between 100,000 and 300,000 metric tons of peanut sales and a positive slope for larger sales. The movement of the two variables is in the anticipated direction (increases in Mt-1 are accompanied by increases in QFC) in only 12 of the 18 years. The illogical relation- ship for certain years plus problems of multicollinearity between PFRt-2 and Mt-1 prevented us from obtaining a reasonable model with Mt-1 present; it was finally dropped. Given the high correlation between Mt-2 and QFC, and the approximately linear relationship, we used Mt-2 as a proxy for the farmer income variable rather than a combination of Mt-1 and Mt_2. 78 PRICES FP - Fertilizer Price: current period, t-l, and t-2 PV . Peanut Value (producer price): current period, t-1, and t-2 PFR - Peanut/Fertilizer price Ratio (PV/FP): current period, t-1, and t-2 VC - Value/Cost ratio (calculated assuming that additional yield due to fertilizer is 400 kilos/ha on average)a: current period, t-l, and t-2 FARMERS’ EXPECTATIONS R - Rainfall: t-l and t-2 PY . Peanut Yield/ha.: t-1 and t-2 FARM INCOME M - Quantity of Peanuts Marketed: t-l and t-2 VM - Value of Peanuts Marketed (M * PV): t-l and t-2 PP - Estimated Peanut Production: t-l and t-2 aThe 400 kilo estimate based on farmers’ perceptions in Nioro see Chapter VII). Figure 11: Variables Thought to Influence Fertilizer Consumption 79 Correlations between QFC and price ratios were generally greater than those between QFC and prices of fertilizer or peanuts separately. Furthermore, peanut and fertilizer prices were highly correlated sug- gesting that some index of the two would be appropriate. Use of a ratio also compensates somewhat for the lack of an adequate deflator variable. As the correlations between PFR and VC were high (.99), we elected to use the PFR series which did not require any questionable assumptions about increased yield due to fertilizer. The low and negative correlation between current ratios and QFC suggests that current prices were usually unknown at the time fertilizer orders were placed, therefore only the lagged ratios were retained. Due to problems of multicollinearity between PFRt-1 and PFRt-2 we used an average of the two (PFRA), believing it was important to keep both variables in the model. Even though neither was statistically significant when both were simultaneously included in the model, each one was significant when employed individually. No variable was retained for expectations because the data (past rainfall and yields) were highly correlated with the other independent variables.1 The elimination of expectations as a separate variable means that the coefficients in the model reflect the influence of prices and income as well as expectations. 1For example, the correlation coefficient for Rt-2 and PFRt-1 was .49; for PYt-2 and Mt-2 it was .79. 80 B. ESTIMATION OF THE MODEL A least squares regression using data for 18 annual observations (1963-1980, 1961 and 1962 being lost due to lags) gave the results shown in Table 10. Coefficients on Mt-2 and PFRA were statistically significant at better than .05 and the F-test for the model was significant at .003; however, the adjusted R2 was relatively low (.47) and the residual for case 5 representing 1967 was an outlier. A reexamination of the data revealed that during 1961-67 QFC climbed regularly although PFRs (current and lagged) remained constant or declined. Given that this period repre- sented the early years of the PA, when most farmers were just learning about fertilizer, we introduced a dummy variable -- LEARN -- to distin- guish the learning period from the rest of the PA. Estimation of the revised model gave the results in Table 11. The adjusted R2 is much higher (.62) with the LEARN variable, but the price variable (PFRA) is now only significant at the .08 level. The problem of the outlier has not been eliminated causing the residuals to exhibit a non-normal distribution. The Durbin-Watson statistic is again in the indeter- minate range, but an auto-correlation plot of residuals did not reveal any serious problem so no correction was made. Although one must use caution in interpreting the regression results, certain conclusions appear warranted. The standardized coef- ficients suggest that changes in Mt-2 have the greatest effect on QFC, followed by the LEARN effect and then the PFRA effect. The explanatory power of lagged marketing and price variables1 suggests that QFC was 1Remember that PFRA is the average of PFRt-1 and PFRt-2. 81 Table 10: Fertilizer Demand Model -- Preliminary Version DEP VAR: QFCa N: 18 MULTIPLE R: .730 SQUARED MULTIPLE R: .533 ADJUSTED SQUARED MULTIPLE R: .471 STANDARD ERROR OF ESTIMATE: 6.203 sro TOLER- IPROB VARIABLE core 55 coer ANCEP T 2 TAIL CONSTANT -24.061 10.757 0.000 1.00 -2.237 0.041 M - c 0.050 0.014 0.628 .99 3.556 0.003 PFRA 11.799 5.025 0.415 .99 2.348 0.033 *ANALYSIS OF VARIANCE* SOURCE S-OF-S DF MEAN-SQUARE F-RATIO PROB REGRESSION 659.790 2 329.895 8.574 0.003 RESIDUAL 577.154 15 38.477 WARNING: CASE 5 IS AN OUTLIER (STUDENTIZED RESIDUAL = 3.762) DURBIN-WATSON D STATISTIC 1.107 FIRST ORDER AUTOCORRELATION .430 Source: Estmated using data in Appendix I. aQFC data was recorded in thousands of tons. bTolerance is a measure of multicollinearity; the closer the number is to 1, the less multicollinearity present in the data (see SYSTAT, Inc., 1986, p. MGLH-4). cMt.2 data was recorded in thousands of tons. 82 Table 11: Fertilizer Demand Model -- Final Version DEP VAR: QFCa N: 18 MULTIPLE R: .829 SQUARED MULTIPLE R: .687 ADJUSTED SQUARED MULTIPLE R: .62 STANDARD ERROR OF ESTIMATE: 5.257 STD TOLER- PROB VARIABLE COEF SE COEF ANCEb T 2 TAIL CONSTANT -16.787 9.529 0.000 1.00 -1.762 0.010 M - C 0.040 0.012 0.504 .91 3.211 0.006 PERA 8.510 4.439 0.299 .92 1.917 0.076 LEARN 7.868 2.999 0.425 .85 2.624 0.020 *ANALYSIS OF VARIANCE* SOURCE S-OF-S DF MEAN-SQUARE F-RATIO PROB REGRESSION 850.043 3 283.348 10.253 0.001 RESIDUAL 386.901 14 27.636 WARNING: CASE 5 IS AN OUTLIER (STUDENTIZED RESIDUAL = 3.104) DURBIN-WATSON D STATISTIC 1.205 FIRST ORDER AUTOCORRELATION .297 Source: Estimated using data in Appendix 1. aQFC data was recorded in thousands of tons. bTolerance is a measure of multicollinearity; the closer the number is to 1, the less multicollinearity present in the data (see SYSTAT, Inc., 1986, p. MGLH-4). th-2 data was recorded in thousands of tons. 83 more a reflection of ONCAD’s supply' decisions, which incorporated information about current farmer demand in a lagged fashion. Farmers’ actual decisions about fertilizer consumption were certainly based more on current income (Mt-1) than past income (Mt-2). At the time ONCAD placed preliminary orders, however, information about the current year’s final production and marketings was not available forcing them to use QFCt-1 as a guide. If we assume that QFCt-1 was influenced by farmers’ income in t-2, we have a plausible explanation for the impor- tance of Mt-2 as an explanatory variable. Use of PFRA rather than the individually lagged PFR variables prevents us from examining the relative influence of t-l and t-2 lags. The fact that t-l prices were certainly known by the time ONCAD placed orders suggests that they as well as t-2 price effects (reflected in t-2 fertilizer consumption data) should have influenced QFC. c. FACTOR ANALvSIsl An alternative way of examining this data is to use exploratory factor analysis rather than trying to select the most appropriate variables. The purpose of factor analysis is to collapse a large set of variables linearly to obtain a smaller number of factors which can be used in place of the variables. The method of extraction used in this analysis is principal components which transforms the original variables into a set of composite variables (principal components) that are orthogonal to each other. The same 18 years of data used in the econometric model were used in the factor analysis. The rainfall 1This analysis is based on a model of factor analysis presented in Faruqee 1979. 84 variables originally considered for the econometric model were elimi- nated from the factor analysis because they resulted in nonsensical groupings On certain factors; this is possibly because spurious correl- ations between rain and price ratio variables were greater than those between rain and QFC.l All other explanatory variables listed in Figure 11 above were used in the factor analysis. The factor matrix resulting from a verimax rotation is shown in Table 12. Each entry in the matrix is a “factor loading" representing the relative effect of the underlying factor on the corresponding variable. The square of each factor loading is the percent of varia- tion in the relevant variable accounted for by the corresponding fac- tor. Since factors are uncorrelated, the sum of the squared loadings for a given variable represents the percent of its variability explain- ed by the four factors (comparable to an R2 in regression analysis). We used an eigenvalue of one as the cut-off point for identifying factors.2 Four factors were identified in this manner. The boxes in each factor column mark the variables which are assigned to that factor (a variable is assigned to the factor for which it has the highest factor loading). Factor 1 explains 31 percent of variation in the data set. It is highly correlated with the current and lagged values of the fertilizer and peanut price variables (FP and PV). 1The following correlation coefficients illustrate the problem: Rt-z/PFR - .52; QFC/Rt-1 - .43; Rt-2/PFRt-1 - .49; QFC/Rt-2 - .17 2Factors with an eigenvalue smaller than one usually explain such a small part of the variability in the data that little is gained by examining them in detail. 85 Table 12: Varimax Loadings of Independent Variables F A C O R 1 2 3 4 F1? 0.979 0.083 0.102 0.032 Pvt-,2 0.972 0.099 0.157 0.082 v FPt..1 0.971 0.076 -o.103 -0.112 Pvt.1 0 . 938 -0 . 085 0 . 309 -0 . 002 A ret_2 0.928 -0.096 -0.254 -0.175 W 0.880 -O.143 0.373 -0.205 R . . 1?sz 0.012 0.953 -0.212 0.111 I 9'sz -0.178 0.918 0.067 0.168 Flt-2 0.178 0.909 -O.199 0.184 A PFRt_1 0 . 078 -0 . 296 0 . 851 0 . 175 B vet.1 0.092 -0.302 0.847 0.172 L -2 0 . 162 0. 340 0 . 780 0 . 398 0.034 -0.432 0.676 -0.505 E PFR 0.044 -0.422 0.674 -0.515 s PPt_1 -0.086 0.239 0.108 0.935 PYt_1 -0.223 -0.079 0.211 0.908 PERCENT OF ‘IUI‘AL 1 2 3 4 VARIANCE EXPIAINED BY EACH m: 30.814 19.714 22.815 19.798 Source: EstinatedusingdatainAppendixI. 86 Factor 2 explains 20 percent of the variability and is most close- ly correlated with the productivity variables for t-2. The loadings on all price ratio variables are much lower than the production t-2 vari- ables, but they are all above .30. The current and t-l price ratio variables exhibit negative signs. As more recent prices cannot have influenced past production, the only conclusion one can draw is that the GOS had a tendency to adjust prices in response to low productivity 50 that ratios moved upward to entice farmers toward increased produc- tivity. Price relationships in t-2 are positively correlation with the productivity measures of that period. Factor 3 explains another 23 percent of the variation. All of the price ratio variables have their highest loadings here. PV and Pvt-1, the peanut price variables, have the next highest loadings. Factor 4 explains 20 percent of the variation. The three meas- ures of agricultural productivity for period t-1 -- peanut yield/ha. (PY), estimated total peanut production (PP), and peanuts marketed (M) -- exhibit their highest loadings on this factor. Current price ratios exhibit loadings in the -.50 range, again suggesting that prices may have been adjusted in response to production. The price ratio vari- ables for t-2 have relatively high loadings (.40) on Factor 4 while t-1 ratios do not (.17). A major task in factor analysis is properly identifying and label- ing the factors produced by the analysis. Factor 1 is clearly a price movement variable reflecting the changes in both peanut and fertilizer prices. Correlations between Factor 1 and variables other than fertilizer and peanut price are all 87 less than .30, implying that there is little relationship between the price movement factor and other variables. Factor 3 is a price ratio variable which stresses changes in relationships between fertilizer and peanut prices or value and cost of fertilizer. As the ratios increase, farmers can anticipate greater returns on fertilizer investments (assuming comparable climatic condi- tions); hence, this factor can be considered a proxy for fertilizer profitability. Current and prior year producer price (PV and PVt-1) are the only variables other than the price-ratio variables with a loading greater than .30; this suggests that changes in peanut prices have had a greater impact on price ratios than changes in fertilizer prices. Factor 2 and 4 represent agricultural productivity in t-2 and t-1 because the variables measuring agricultural productivity all have their highest loadings on these two factors. A number of price-ratio variables also have relatively high loadings on these productivity factors. Current ratios are the most highly correlated with both factors. In both cases the correlation is negative. Ratios for t-2 have the second highest correlations and they are positive. Standard economic theory would expect favorable (high) price ratios in the current period to encourage increased use of fertilizer and increased production -- 'i. e., a positive correlation between productivity and price ratios. In the Senegalese context, current prices were not always known when fertilizer investments were made, so it would not be unusual for lagged price ratios to also affect current production. The t-2 price ratios exhibit the anticipated relationship, but the current 88 ratios do not. A negative sign on a smaller correlation could be dismissed, but the correlations for current ratios are in the .40 to .50 range. The only explanation we can provide is that the government responded to poor productivity by making prices more favorable to farmers a year or two later. In other words, the direction of the relationship is from productivity to price change rather than prices having encouraged increased productivity. Having thus identified the factors, the next step is to examine the relationship between these new composite variables and the depen- dent variable. We do this in two steps: (1) By adding the dependent variable to the independent ones and performing another factor analysis; (2) By regressing the factor scores obtained from the analysis of independent variables on the dependent variable and assessing the explanatory power of the factor model. Table 13 illustrates that the same basic factors are identified when the dependent variable is included in the analysis, but the order in which they are identified changes. The dependent variable has the highest loading on the t-2 productivity factor (now Factor 3) and the second highest loading on the t-l productivity factor (now Factor 4), suggesting that these factors are strongly correlated with PA fertil- izer consumption. ‘The QFC loadings (n1 price trends and price ratios are all quite low suggesting little correlation between price movements and fertilizer demand during the PA. In regressing the factor scores on QFC we obtained the results in Table 14. Table 13: 89 VarimxloadjngsofnepetflentandIrdeperdentVariabla F A C 0 R S 1 2 3 4 PP 0.981 0.089 0.076 0.031 FPt.1 0 . 972 '0 . 111 0 . 048 .0 . 118 V Wt.1 0 . 937 0 . 314 '0 . 066 0 . 007 FPt..2 0 . 924 '0 . 242 7'0. 125 '0 . 182 A W 0.879 0.385 '0.131 '0.192 R PFRt..1 0 . 076 0 . 873 '0 . 211 0 . 205 VCt.1 0.089 0.870 '0.217 0.202 I .2 0.152 0.740 0.385 0.405 .2 0.174 0.731 0.386 0.411 A VC 0.032 0.718 '0.399 '0.473 PFR 0.042 0.716 '0.388 '0.484 B ‘F'——* L Pit-2 '0 . 163 ‘0. 009 0. 917 0 . 147 Mt-Z 0 . 188 '0 . 267 0 . 902 0 . 155 E m "0.107 0.204 0.664 0.454 S PPt.1 '0 . 079 0 . 058 0 . 262 0 . 932 Ht.1 ‘70.221 0.188 '0.041 0.915 Mt. 0.031 0.134 0.296 0.897 1 W 01" 'IUI'AL 1 2 3 4 VARIM EXPLAINED BY EACH FACIOR: 29 . 278 22 . 234 20 . 486 19 . 649 Source: EstimtedusingdatainAppendixI. 90 Table 14 RegressimofFactorScoraQOntityofFertilizercm'smned IEP VAR: QFC N: 18 um R: .770 SQIAREDMTIPIER: .592 WWWR: .467 m Em 0F ESTDM'E: 6.228 STD ‘IOIER- PROB VARIABLE (DEF SE (DEF ANCE T 2 m1. (INSTANT 15.056 1.468 0.000 1.00 10.256 0.000 PRICE WVENENTS -0.893 1.511 -O.105 1.00 -0.591 0.564 mecrlvm ‘152 4.841 1.511 0.568 1.00 3.205 0.007 PRICE RELATIGJS 1.935 1.492 0.227 1.00 1.281 0.223 PEDEJCTIVITY T-l 3.888 1.492 0.456 1.00 2.574 0.023 *ANALYSIS OF VARIANCE* SCIJRCE SUM-OF-SQERES DF MEAN-SQJARE F-RATIO PROB WIQJ 732.638 4 183.160 4.721 0.014 RESIIIDSL 504 . 306 13 38 . 793 W: (ESE 5 IS AN (IH'LIER (S'IUWI'IZED RESIIIIAL= 3.061) IIJRBIN-MTSON D STNI‘ISI‘IC 1.677 m ORDER WON .106 Source: Estimated usir'mg data inAppendix I. 91 The model has some problems. The residual for 1967 is an outlier with a studentized residual greater than 3. The Durbin/Hatson statis- tic (l.677) is also inconclusive, suggesting that some autocorrelation may be present. Despite the problems, the significant coefficients on both productivity factors confirm the importance of prior productivity in explaining current fertilizer consumption. The absence of signifi- cant coefficients on the two price factors leaves the importance of price as a determinant of fertilizer demand in doubt. D. GENERAL CONCLUSIONS OF THE ANALYSES What tentative conclusions about determinants of PA fertilizer demand can be drawn from the two analyses presented here? (1) Both econometric and factor analyses suggest that t-2 produc- tivity is the strongest single influence on demand. (2) Although the econometric models using t-l productivity (in the form of peanuts marketed) were not entirely satisfactory, the factor analysis suggests that t-l productivity was the second most important influence on QFC. (3) We have not established a clear picture of significant and direct price or profitability influences on PA fertilizer demand. Attempts to use peanut and fertilizer prices directly and to separate the relative influences of PFRt_1 and PFRt-2 were complicated by multi- collinearity in the econometric model. The price movement and price ratio factors did not explain any significant amount of the variability when we regressed all four factors on QFC. The factor analysis does suggest, however, that price ratios are correlated with productivity measures. Lagged ratios tend to be positively correlated with current 92 or subsequent production suggesting that increases in price ratios (i. e., making prices more favorable to farmers) are associated with subse- quent increases in productivity. (hi the other hand, current ratios have a negative correlation with past production, suggesting that government made prices more favorable to farmers in response to low productivity. The fact that price variables in general do not explain fertilizer consumption and do not have a clearly discernable relation- ship to productivity makes it difficult to sort out what really was happening during the PA. What are the implications of these results for fertilizer policy? The strong explanatory power of t-Z variables during the PA offers evidence that available fertilizer consumption data are more a reflec- tion of ONCAD supply estimates based on ”old information" rather than current farmer demand. In other words, ONCAD did not respond quickly to changes in price and productivity which are normally responsible for changes in farmers’ demand. One is inclined to blame the heavy bureau- cratic structure of ONCAD for this sluggishness and anticipate that the more liberalized marketing system envisioned by the NPA will respond faster. The review of fertilizer ordering procedures presented above, however, makes it clear that the crux of the problems was not ONCAD bureaucracy but the SIES requirement that ONCAD place initial orders before the end of the previous season. Moving from the centralized ONCAD system to a network of private distributors may not improve the system’s ability to respond to changes in demand if the current fertil- izer manufacturer, Industries Chimiques du Sénégal (ICS), is not able to act with a shorter lead time. 93 The muddled picture on how prices influence fertilizer demand suggests a need for more information on the role prices and profitabil- ity measures play in farmers’ fertilizer investment decisions. Schumacher’s observations about extension agents’ neglect of financial analysis in their counseling sessions offers a partial explanation for the lack of a significant price effect. Another explanation is that farm incomes were generally so low that lack of purchasing power con- strained fertilizer investments irrespective of price relationships. The results of farmer interviews presented in Chapter VII offer some concrete evidence in support of this last explanation as well as some additional insights about the role of prices in farm investment decisions. In examining this data we have identified a number of problems inherent in the data itself. Documents reporting fertilizer consump- tion data have not clearly defined what they mean by consumption, leaving the reader to guess whether the numbers represent end use by farmers, deliveries to cooperatives, or intermediate purchases by distribution agencies. Price series also are not always in agreement and the lack of information on when prices were announced versus when they became effective poses serious problems for models using prices as explanatory variables. The lack of data on credit availability pre- vents us from examining a potentially important determinant of fertil- izer demand, particularly during the last few years of the PA when limits were imposed. Another problem is poor documentation of tempor- ary exogenous influences on fertilizer demand. An example would be the fact that fertilizer price became a political campaign issue in 1983 94 when the opposition candidates promised they would reduce fertilizer prices if elected. The uncertainty about prices led many farmers to take a “wait and see" attitude, forcing fertilizer purchases below what they might have been otherwise. If such information becomes readily available, models of fertilizer demand can be adjusted accordingly and seemingly perverse economic behavior better explained. For better or worse, econometric methods employing time-series data remain the primary tools used by economists for estimating input demand. Fifteen to twenty years of data is usually considered the minimum for a meaningful model. Given the 605’ interest in moving towand a liberalized input marketing system, it is not too early to start improving Senegal’s agricultural statistics. IV. REVIEW OF RESEARCH ON ECONOMIC RETURNS T0 FERTILIZER USE IN SENEGAL Precise, generally accepted estimates of economic returns to fertilizer are difficult to obtain, yet the design of a sound fertil- izer policy requires some assessment of the potential returns at both the farm and the national level. In this chapter' we review ‘the evidence available from economic analyses of fertilizer response data, linear programing models, and survey research. A. ECONOMIC ANALYSES OF FERTILIZER RESPONSE DATA 1. EARLY ISRA/IRHO CONTRIBUTION Available documents on fertilizer research during the 19505 and 19605 pay only limited attention to economic analysis of fertilizer, per se. An article by Tourte et al. which appeared in Agrongmig Tropi- galg in 1971 used data from ”prévulgarisation” trials conducted between 1965 and 1969 to present a detailed partial budget analysis comparing the economic returns of the "themes légers" and "themes lourds" techni- cal packages.l Partial budgeting is a decision-making tool whereby the gross margins associated with alternative courses of action are com- pared. The decision rule used in partial budget analysis is to select the alternative which provides the largest gross margin.2 1The data used for this particular analysis were not based on actual recommendations mentioned in Tables 5 and 6 (Chapter II) but on varia- tions of them which were used in "prévulgarisation" trials. For example, the "themes lourds" dose of basic phosphate at the beginning of the rotation was 700 kilos of tricalcium phosphate in the Tourte et al. £1371) data set whereas recommendations to farmers never exceeded 500 i 05. 2Gross margins are total revenues generated by a particular alternative minus the variable costs associated with that alternative. While a helpful tool, one must use caution when comparing gross margins of 95 96 The Tourte et al. budget analysis looks at the entire technical package (animal traction, equipment, and fertilizer), making the point that an economic analysis of individual components of the package (e. g., fertilizer only) is "difficult" and even "dangerous" (Tourte et al., p. 6). The comparison is based on economic returns for the full four-year rotation. The partial budgets developed for the Northern Peanut Basin (Louga, Tip, Tiénaba, Forbote) and the Sine Saloum (Boulel, Nioro, Keur Samba, Toubacouta, Keur Yoro Dou) are reproduced in Tables 15 and 16. The authors’ interpretation of these budgets illustrates the tendency of ISRA researchers to give soil maintenance considerations greater weight than economic factors when justifying fertilizer recommendations. This preference is particularly well illustrated by the following list of reasons justifying the "themes lourds" in the North, even though they were not more profitable than the ”themes légers”: -- a failure to do so would lead to catastrophic soil degradation and present a serious danger of soil acidification -- the "themes légers” satisfy the soil needs for N and P but not K (i. e., the K mineral balance is negative) -- at the level of production obtained with "themes légers" one approaches yields which require increased levels of potassium -- the "themes lourds” provide more nutrients than needed and therefore set off the process of soil enhancement under economic conditions which bring a profit, although it is a lower profit than that obtained with the "themes légers" (translated and para- phrased from Tourte et al., 1971, p. 650). Part of the dilemma faced by Tourte et al. (1971) is their inability to alternatives that have significantly' different demands on limiting resources whose costs are not considered in the analysis (e.g., cash or labor). 97 Table 15 1971 Partial Budget Analysis Carpering "No Fertilizer“ with "Theme Légers" and “Than Lourds" Reconnendatims Northern Pecut Basin (in rem (LOJGA, m, TIENABA, FORBOTE) rose1° 716111b ADDED ADDED 111:1 BENEFIT wc ROTATIOI new VALUE YIELD VALLE 911000011011 COST 11011111011 RATIO “11.011 ---------------------- mm 950 17,570 1090 20,160 mm 630 7,310 700 11,900 ZNDPEANUT 960 17,760 1170 21,660 oosrsmeuems 01111001 60051017 62,660 53,700 11,060 10,725 335 1.03 oosrsleeuems 1.1111150166066101 11,060 5,600 5,660 2.05 ros1>2c rzsezd FALLOJ ---------------------- 2,600 is! PEANUT 1060 19,260 1160 21,090 MILLET 690 6,330 1020 17,360 12,630 21109610101 960 17,760 1070 19,790 COSTS/BENEFITS 01111001 suesnov 65,330 56,220 12,690 16,930 - 2,060 .66 COSTS/BENEFITS 111111 501: SIBSIDY 12,690 7,520 5,370 1.70 SQRCE: Adapted from Tourte et al. (1971, p. 669). aFOSPI REPRESENTS no FERTILIZER AND LIGHT PLGIING. bF‘ISI’I REPRESENTS "TNEIES LEGERS" (LIGHT FERTILIZER AND LIGHT PLGIING). CFOSPZ REPRESENTS no FERTILIZER AND PLGIING WDER 0F FALLUJ. c'FZSI’Z REPRESENTS ”TREES LQRDS" (PLWING UIDER 0F FALLGI AID HEAVY FERTILIZER) 98 Table 16 1971 Partial Budget Analysis Comparing "No Fertilizer“ with “Thénea 1.69616" and “Themes Lourds" Recounmdations Sine Salon (in FCFA) (NIMO, MEL, KELR SANBA, KELR YGO 0011, TGJDACGITA) 10511' 115111b 60060 60060 1161 66116111 VIC 601611011 YIELD VA1116 1161.0 V6106 111000611011 COST 1101611011 RATIO 1611.011 ---------------------- 1511661611 1560 26,660 1790 33,115 506616111 1160 19,360 1710 29,070 216116611111 1390 25,715 1710 31,635 00515/661161115 1111110111 51165101 73,955 93,620 19,665 10,725 9,160 1.65 00515/661161115 111111 5011 51165101 19,665 5,600 16,665 3.70 105112c 125112d 1611.011 ---------------------- 6,320 15116611111 1660 31,060 2060 36,660 506611111 1360 22,760 2330 39,610 16,960 21101661611 1500 27,750 1960 36,630 COSTS/BENEFITS 11111100151165101 61,610 116,720 33,110 19,300 13,610 2.60 00515/661161115 111111501151165101 33,110 9,960 23,130 3.30 SGRCE: Adapted from Tourte et al. (1971, p. 652). .FDSPT REPRESENTS 110 FERTILIZER AND LIGHT PLGIING. bFlSPl REPRESENTS “THEFES LEGERS" (LIGHT FERTILIZER AND LIGHT PLGIING). cF0892 REPRESENTS ID FERTILIZER AND PLWING lllDER 0F FALLGI. dFZSPZ REPRESENTS "THEIES LGRDS“ (PLWING UIDER 0F FALLGI AID HEAVY FERTILIZER) 99 quantify the future benefits accruing from soil enhancement associated with the ”themes lourds". If this were possible, one might be able to show that gross margins of the ”themes lourds" were more attractive.1 Although the budgets presented by Tourte et al. (1971) analyze returns to the entire rotation, the authors do comment on the profita- bility of individual components of ‘the rotation. ”Themes lourds" cereal fertilizer was profitable in all zones while "themes lourds" peanut recommendations increased yields but not always profits. The absence of increased profits does not lead researchers to reexamine recommendations but rather to suggest that the government reexamine price structures and marketing policies, including marketing of crop residues for animal feed.2 Although the authors are content to base fertilizer recommenda- tions on a simple comparison of gross margins, a number of other fac- tors should be considered when interpreting the results of partial budgets. Of particular importance in the comparisons shown in Tables 15 and 16 is the greater labor requirement for the heavy plowing associated with "themes lourds" fertilizer. Plowing under the fallow 1P. L. Sarr (1981) studied physical and chemical changes in soil resulting from 17 years of cultivation using no fertilizer, "themes légers", and ”themes lourds". To date, no attempt has been made to assess the economic implications of these changes. Given the lack of quantitative analysis on long-run costs of not using fertilizer, it would appear to be a worthwhile endeavor. 2The specific comment made was: "Il convient de noter que l’action de la fumure forte (donc sa rentabilité) est en premier lieu assurée par la céréale (et plus généralement par les cultures autres que l’ara- chide, qui cependant profite de cette fertilisation forte), ce qui posera, au moment de la vulgarisation, le probleme de sa commercialisa- tion et de son prix, donc celui de son débouché, l’alimentation animale ne devant pas étre exclue;" (Tourte et al., 1971, p. 644). 100 not only requires approximately 80 more hours of labor, but the recommended time for doing this is during the harvest when the labor constraint for many farmers is most severe. A second difficulty with this partial budget is the choice of options being compared. The authors compare gross margins/hectare for different levels of fertil- izer and plowing technology. Another option not considered is to increase acreage planted rather 'than intensifying current acreage. Using the Sine Saloum example, a farmer who invested the 9,980 FCFA required to move from FlSPl to FZSPZ recommendations in an additional .8 hectares of production at the FlSPl level would realize a gross margin of 26,733 rather than 23,130 FCFA. This reflects the fact that the rate of return on investment is greater for the "themes légers”. While this might not seem like much of a temptation given the addition- al labor and land requirements, it is an option which warrants more careful analysis because many Senegalese farmers have consistently demonstrated a preference for expanding low-technology cultivation across larger cropping areas. He will reexamine this issue in Chapter V when we look at more up-to-date estimates of fertilizer response. Bockelee-Morvan and Vaillant (1967) estimate the economic returns to fertilizer using 1959-65 farm survey data. The data were collected from 10 fertilized and 10 unfertilized fields selected randomly each year in four villages participating in the IRHO l‘Village Témoin” pro- gram. The authors conclude that farmers in the study area (Ndoffane, Southern Sine Saloum) were realizing a net benefit of 5 to 9 FCFA per 1 FCFA invested in fertilizer. Noting a high correlation between fertil- izer derived net benefits per active worker and investment in 101 agricultural equipment and livestock, the authors further concluded that economic returns to fertilizer were financing. other productive investments. 2. FAO CONTRIBUTION Although CNRA and IRHO were the main actors in fertilizer research during the 19505 and 19605, they were not the only ones. FAO super- vised 410 fertilizer trials and 2950 extension demonstrations in Senegal from 1961-66 as part of their worldwide Freedom from Hunger Fertilizer Programme.1 A major consideration in FAO work, which was usually absent in CNRA and IRHO studies, was the identification of the most profitable treatments and a comparison of the dollar returns per dollar investment in each nutrient. FAO also emphasized cereal rather than peanut pro- duction. Table I7 summarizes the principal FAO conclusions. The FAO results had no apparent influence on the CNRA research program and recommendations. In comparing FAO trial procedures with those of Senegalese research institutions, one can easily understand why the latter would not consider FAO results as sufficient justifica- tion for changing recommendations: I) Millet demonstration results differed from trial results. 2) Trials from year to year did not use the same fields so it was not possible to consider residual fertilizer effects nor cumula- tive soil depletion effects due to continuous cultivation. Although research protocols and recommendations differ from those ‘ 1Only 394 of the 410 trials and 2268 of the 2950 demonstrations Ilroduced usable results (FAO, 1974). 102 Table 17 Summary of FAO Fertilizer Trial and Demonstration Results Highest Kilos/Ha. Number of Net Returns N-P-Ka Observations Millet Trials $33.84/hectare 22.5-22.5-22.5 83 Returns/S of N $2.70 P $2.70 K $6.00 Millet Demos $11.95/hectare 22.5-22.5-0 901 Peanut Demos $22.42/hectare 16-25-18 460 Source: FAO (1974). Some earlier FAD documents present results of analyses performed on fewer observations and their conclusions dif- fer from those shown above. We assume that the more recent docu- ments supersede the previous ones. aThese fertilizer levels are reported as kilos of nutrient per hectare and are not comparable to the N-P-K formulas shown in other tables. In the latter case, each number represents the percent of a given nutrient in the recommended dose. For example, ISRA’s 150 kilo/hectare recom- mendation of 14-7-7 for millet would be 21-10.5—10.5 in FAQ notation. 103 of ISRA and IRHO, FAO's work provided further evidence that Senegalese farmers using fertilizer at subsidized prices could obtain yield in- creases that insured reasonable economic returns on investments (value-cost ratios greater than 2).1 3. IFDC ANALYSIS OF IRHO/ISRA DATA - Although ISRA and IRHO were reluctant to use multi-rate trial re- sults for rigorous economic analyses, an IFDC mission in the early 19705 used these data sets to estimate nitrogen, phosphorous, and po- tassium production functions. IFDC (1977) presents the estimated mathematical functions and the value/cost ratio given a series of cost/price ratios ranging from 1 to 4.2 The IFDC noted that the opti- mum peanut recommendations using either IRAT or IRHO data were similar and that varying the cost/price ratio generally had little effect on the optimum (profit maximizing) level of fertilizer (IFDC, 1977, p. 23). In general, the IFDC found farm-gate cost/price relationships very favorable3. 1The documents consulted did not specify, but it appears that only subsidized prices were considered in the economic analyses. 2The value/cost ratio is the total value of added product divided by the cost of the fertilizer treatment; the cost/price ratio is the cost of a kilo of nutrient (not a kilo of fertilizer) divided by the per kilo value of output -- peanuts, in the Senegalese case. As the cost/price ratio rises, the value/cost ratio declines and the invest- ment becomes less profitable to farmers. 3Cost/price ratios for then current reconlnendations calculated with subsidized prices ranged from .91 for 8-18-27 to 1.72 for 10-10-8. Unsubsidized prices had cost/price ratios ranging from 2.19 to 4.14 (IFDC, 1977, Table 27, p. 24). 104 4. ECONOMIC JUSTIFICATION OF FERTILIZER SUBSIDIES In a 1969 paper published in Agrgnomjg Trgpicalg, Bray addresses the macroeconomic issue of the costs and benefits of government fertilizer subsidies.1 Fertilizer subsidies began in 1951 after a 1950 attempt to sell fertilizer on credit at real prices was labeled a failure because farmers did not reimburse their debts. The subsidy was intended as a temporary measure to insure that farmers just learning fertilizer technology would realize a profit. As techniques improved the subsidy would be diminished. In the late 19605, the government began reducing subsidies.2 Bray, then Director of Agriculture, opposed these price changes, believing that they discouraged agricultural investments which brought the government more in tax revenues than the cost of the subsidy. Much of the incentive for the Bray analysis was the desire to convince government that the recent reduction in subsidy was unwarrant- ed. He believed the increase in farm-gate fertilizer prices had de- pressed demand and was lobbying for a return to previous prices. His analysis compares aggregate data on peanut yields for 2 exceptionally good years (1956 and 1965) and two unusually bad years (1958 and 1967). He assumes the difference in production between the 19505 and 19605 is lBray published several very similar papers on this topic. This discussion is based on his 1969 presentation. The data used in his 1970 paper are different but the conclusions are the same. 2This background information on fertilizer subsidies is from Cissé et al. (1967a). 105 due to technological advance. Applying a rule of thumb1 that 35 per- cent of total yield increase associated with new technologies is due to fertilizer, Bray estimates that fertilizer response in a good year is 229 kilos/ha, in a bad year 139 kilos/ha, and on average 200 kilos/ha. Using then current prices, subsidy policies, and estimates of how increases in peanut production influenced gross domestic product and subsequently tax revenues, Bray shows that the government realizes a net profit on the fertilizer subsidy of 692 FCFA/ha. He also shows that the farmer averages a net benefit of 1200 FCFA on a 2400 FCFA investment. Table 18 illustrates how the author arrived at these figures. Bray argued that although the farmer realized a profit on average with the 16 FCFA/kilo fertilizer price, it was necessary to reduce the farm-gate price to 11 FCFA so that even in bad years a farmer would be insured of a value/cost ratio no lower than 1.5. Although the economic analysis is not sophisticated and the data (by the author’s admission) are ”fragile", the argument presented here appears to have convinced policy makers that fertilizer subsidies were in the economic interests of the government. Prices were reduced and remained highly subsidized until the 19805. 5. IFDC/SODEVA CONTRIBUTION SPURS GREATER ATTENTION TO ECONOMICS By the nfid-l9705, there were some individuals and organizations who felt that economic considerations deserved more attention than past fertilizer research had afforded them. The World Bank financed a project from 1975-78 in which SODEVA and the IFDC combined forces to lBray claims this is a rule of thumb established by "research", but offers no more specific explanation. In other versions of his analysis he uses 40 percent. 106 Table 18 Cost/Benefit Analysis of Government Fertilizer Subsidies (FCFA/hectare) Farm Level Returns; Cost of Fertilizer: 16 FCFA * 150 kilos - 2,400 FCFA Added Product: 18 FCFA * 200 kilos . 3,600 FCFA Benefit to farmers 3,600 - 2,400 FCFA - 1,200 FCFA MW Estimate of indirect effect on GDP of increased production: 3,600 FCFA * 2.2 = 7,920 FCFA Revenue from taxes on increased GDP: 7,920 * .22 = 1,742 FCFA Cost of fertilizer subsidy per hectare: 7 FCFA * 150 kilos = 1,050 FCFA Benefit to government: 1,742 - 1,050 FCFA - 692 FCFA Source: Bray (1969, p. 1106). 107 conduct multi-rate trials on farmers’ fields. IFDC analyzed the fer- tilizer response data and developed zone-specific fertilizer recommend- ations based exclusively on economic criteria. The research program was carried out despite serious protestations from ISRA about (1) the adequacy of the research protocol1 and (2) the inappropriate use of limited funds on a topic which ISRA had already thoroughly researched.2 The fertilizer responses obtained in the IFDC/SODEVA trials, when subjected to regression analysis and profit maximizing decision rules did produce a set of fertilizer recommendations which differed from ISRA’s in several respects. The most noteworthy differences were: Peanuts: N Eliminated in all zones but the Center (Thies/Diourbel) P Increased in Sine Saloum; slight decrease in other zones K Eliminated in the North and Center Millet: N Increased in the North and Center; decreased in Sine Saloum P Increased in the North, Center, and N. Sine Saloum (all departments but Nioro) * 1TWO of the major criticisms were that the protocol called for joint aOaTysis of the data from different regions and for too few observa- t101115 to obtain a coefficient of variation within reasonable limits (ISRA. 1975, pp. 16-28). ZISRA implied that those responsible for the IFDC/SODEVA research p'I‘c’lnasal were guilty of: "...reconduction indéfinie des memes ques- 2th et des mémes expérimentations, par la reexamination des problemes r 301u5;" (ISRA1975, p. 59). 108 K Eliminated in the North and Center, increased for S. Sine Saloum (Nioro)l IFDC brought to the surface again the issue of K fertilization which had been alluded to in prior research by IRHO (no peanut response to K in the North) and FAO (no response to K in millet demonstration trials). ISRA reacted quite negatively to the IFDC results, believing that reconlnendations based on only two years of analyzable results could in no way be superior to then current recomendations based on some thirty years of research. Seules les recommandations actuelles de l’ISRA, obtenues apres de nombreuse études menées aussi bien en station que chez le paysan permettent de répondre aux objectifs du Gou- vernement: assurer l’autosuffisance alimentaire, tout en luttant contre la aégradation des sols et la desertification. (ISRA, 1978, p. 4) ISRA’s arguments against decreasing potassium levels were based on results showing that potassium had many beneficial effects and must not be applied at lower rates than 30 kilos per hectare:3 1) It increased the speed of millet growth, heading, elongation, and flowering 1These differences are based on information contained in IFDC (1980). ISRA (1979), which was in response to a preliminary report of IFDC results, does not mention exactly the same changes. For example, ISRA (1979) claims millet K was reduced 80 percent and peanut K 40 percent for the Thies/Diourbel zone. IFDC (1980) eliminates K entirely in this zone. It is assumed that IFDC (1980) contains the final IFDC recom- mendations, therefore, they are the ones presented here, even though no ISRA response to the 1980 document has been found. 2Translation: ”Only the current ISRA recommendations, based on numerous studies conducted on-station as well as on-farm, respond to Government objectives of insuring food self-sufficiency while waging the war against soil degradation and desertification." 3These arguments are presented in ISRA (1979) and are based on research conducted by Pieri and reported in Pieri (1979). 109 2) It increased the number of fertile heads per hectare and the ratio of kilos of grain per head by 12 percent 3) It increased available dry matter for millet 4) It increased yield stability for millet 5) It increased the amount of peanut hay produced IFDC presented their counter argument as follows: The general argument for K fertilization in Senegal is that K in cropped soils is being depleted and, therefore, main- tenance supplements of K should be made. Another factor cited in favor of K fertilization is that research results indicate that K tends to give earlier head formation, more heads, and more grain per head. Thus, K may be applied more as an insurance factor. Our views are that there is no need to apply K until it is demonstrated that farmers receive economical responses to applied K. It is agreed that with continuous cropping at an intense level and high yields, K fertilization will be needed on a larger scale. Thus, a program is needed to continuously monitor the occurrence of K responses in farmer fields. (IFDC, 1980, pp. 30-31) ISRA also strongly opposed the reduction in N on the grounds that chemicalm nitrogen was needed to compensate for Senegalese farmers’ failure to incorporate sufficient amounts of organic matter in their soils. The IFDC/SODEVA research had no immediate effect on fertilizer recommendations made by ISRA and implemented by the rural development agencies such as SODEVA. The Ministry of Rural Development (MDR) was unwilling to accept the IFDC/SODEVA results without ISRA concurrence. In their 4-page response (ISRA, 1979) to IFDC's preliminary report of results, ISRA conceded that the IFDC method was of scientific interest but stressed that the results could not be relied on for agricultural policy decisions because the coefficients of variation were usually greater than 30 percent and the economic analysis did not consider the 110 soil as capital which would be irredeemably depleted if IFDC recommend- ations were adopted. Dans le raisonnement économique, le sol doit étre pris en compte en tant que egpjtgl fenejer, et toute agriculture qui se veut mmderne ou rationnelle doit prendre en compte l’en- t ’ mélioration de ce imoin r. La non restitution des éléments minéraux exportés par les cultures sur des sols pauvres et fragiles est a déconseiller dans une agriculture moderne se voulant intensive. Il n’est donc pas normal de ne prendre en compte que l’effet économique inlné- diat, ce qui risque d’aboutir a la longue a un désastre. (ISRA, 1978, p. l) The problem with this line of reasoning is that it cannot be con- sidered in isolation from social and legal institutions which determine who has access to what land and for how long. Poli (1970) describes the current land ownership laws and what they were meant to accomplish. Cissé et al. (1967b) discusses problems of introducing ”themes lourds" -- particularly a respect for the four-year rotation -- that are asso- ciated with social conventions concerning land tenure. Access to farm land is a complex topic which cannot be addressed adequately in this dissertation. It should be noted, however, that the G05, not the farmer, owns Senegalese farmland. This raises legitimate questions about who should bear the costs of soil maintenance and improvement. Recomendations based on ISRA’s analyses of fertilizer response data tend to assume that both the government and the farmer will be willing to invest limited resources in fertilizer investments that do not 1Translation: ”Economic reasoning requires that the soil be considered as capital and any agriculture aspiring to be modern and rational must consider the main 6 n e an im rove en this it . The failure to replace minerals which have been used by crops planted on poor and fragile soils is not recommended for an agriculture wanting to moder- nize and intensify. It is, therefore, not normal to consider only the short-run economic effect when this will result in long-run disaster. 111 guarantee a long-run payoff and do not always bring in short-run profits. The brouhaha raised by IFDC’s economically based fertilizer recom- mendations appears to have heightened ISRA’s awareness of the need for more rigorous economic analysis in fertilizer research. A lengthy 1975 document (ISRA 1975) criticizing the proposed IFDC/SODEVA program made an effort to consolidate previous research on economic returns to recommended doses and presented ISRA’s only attempt to consider returns over time. The authors used 15 to 19 years of data (depending on the zone) from IRHO confirmation trials and then current prices to obtained the frequency distributions of value/cost ratios shown in Table 19. The table shows that the prognosis for the Sine Saloum was quite favorable using unsubsidized prices. Hhen subsidized prices were considered, the We ratio was consistently greater than 2 for the entire period and greater than 8 for more than 50 percent of the time. The results for the North and Central regions were less spectacular. When more favorable subsidized prices were considered, the North real- ized a value/cost ratio equal to or greater than 2 only 63 percent of the time and the Center just 80 percent of the time. Researchers maintained, however, that the long run consequences of not using fer- tilizer in these zones would be more costly than their recommendations. An article published by three researchers (Pieri, Ganry, and Siband) in the 1978 issue of Agrgnomie Trgpicele proposed a method for developing fertilizer recommendations which would give equal considera- tion to three decision rules often considered separately: * that the value/cost ratio be > 2 112 Table 19 Frequency Distribution of Fertilizer Value/Cost Ratios Using Subsidized and Unsubsidized 1975 Prices Percent of Years That V/C Has Attained Zone V/C Subsidized Prices Unsubsidized Price Sine Saloum < 1 -- 5% (19 years) 1-2 -- 21% 2-3 5% 32% 3-4 11% 42% 4-6 11% -- 6-8 21% -- 8-10 47% -- > 10 5% ~- North 0 6% 6% (16 Years) < 1 6% 31% 1-2 25% 25% 2-3 -- 31% 3-4 6% 7% 4-6 19% -- 6-8 13% -- 8-10 19% -- > 10 6% -- Center < 1 -- 33% (15 years) 1-2 20% 40% 2-3 13% 27% 3-4 13% -- 4-6 27% -- 6-8 20% -- 8-10 7% -- > 10 -- -- Source: ISRA (1975, p. 36). 113 * that the marginal revenue be 2 marginal cost * that the mineral balance be positive By way of illustration, the technique was applied to data for millet response to nitrogen and potassium. The authors concluded that a feasible solution existed for nitrogen; but no dose of’ potassium existed which satisfied all three constraints unless all dry matter from the previous crop had been plowed back into the soil. There is no evidence that this method was ever used to develop Senegalese fertilizer recommendations, but the article does illustrate that some researchers were searching for analytical techniques which would help reconcile soil maintenance objectives and economic considerations. 6. THE POST-1980 SITUATION In 1980 some ISRA scientists, basing their recomendations on results from a variety of unspecified ISRA research programs, were ready to move in a direction similar to that suggested by the IFDC. The major recommendation for change was the elimination of the "start- er” dose of nitrogen on peanuts. This change left the following P, K, and S (sulfur) levels (in kilos/hectare): P K 5 North 15 12 12 Center 30 30 12 Center South 27 4O 15 Proposed millet reconlnendations were stated in terms of production objectives. A farmer wanting to produce one ton of millet should use 60 kilos of N, 31 kilos of P, and 10 kilos of 5 per hectare. Farmers 114 aiming for two tons of production were to use the above nutrient levels plus 30 kilos of K (ISRA 1980)., The important change in millet was the elimination of K for the lower productivity farmer. The decision to recommend the total elimination of peanut N is stronger than the posi- tion taken by IFDC which retained some N in the Thies/Diourbel zone. The reduction of K for millet yields of one ton or less is difficult to compare to IFDC recommendations which varied by zone, but it is clear that there is some movement in the IFDC direction. This was not true for peanut K recommendations which remained unchanged.' The importance of production constraints faced by fertilizer manufacturers and the desire of extension services to limit the number of available formulas were recognized by this group of researchers when they recommended the above changes. They illustrated how the new recommendations could be met with only 3 different products: * 0-23-20 +12 5 * Singlesuper P (0-18-0) * Urea (45 or 46-0-0) The authors did not use economic criteria to justify these changes; but they did present a brief analysis showing that their rec- ommendations would be profitable (i.e., value/cost ratio greater than 2) to most farmers at current, subsidized prices. The only exception was peanut production in the North where the value/cost ratio was only 1.6.1 1The authors also commented on the economic returns to existing fer- tilizer recommendations. Their analysis identified profit maximizing doses of fertilizer using production functions which had been estimated from fertilizer response data collected in the "Approche Minérale" program prior to 1968. The authors pointed out that existing 115 No action was taken on this specific reconmendation. In 1985, however, USAID agreed to subsidize distribution of imported urea and a binary fertilizer (0-15-20), manufactured with less costly bulk blend- ing procedures newly implemented by the Industries Chimiques du Sénégal (ICS). ISRA reluctantly agreed to the change. Some researchers felt that the binary formula required extensive testing before being distri- buted to farmers. The concurrence was based on economic realities rather than agronomic considerations; researchers believed farmers would be unwilling to pay for any more expensive fertilizer given the lack of credit and recent series of poor harvests. The "Doubting Thomases” were vindicated after the 1986/87 campaign when the Groupe de Réflexion sur l’Engrais (an ad hoc committee of government, private sector, and donor agency representatives) convinced the govern-ment to return to the old standbys (6-20-10 and 14-7-7) because farmer response to the new formula had been so abysmal (Groupe de Réflexion, December 1986). Recent discussions with a number of soil scientists and agrono- mists revealed that ISRA is continuing research on binary fertilizers. These researchers believe that binary fertilizer is a feasible means of reducing fertilizer costs without compromising productivity and feel that the poor 1987 marketing results were due to factors other than the fertilizer’s performance. In sum, whether the decision rule in economic analysis was profit maximization or value/cost greater than 2, all analyses reviewed con- cluded that economic returns to fertilizer (at subsidized prices) were recommendations were well below profit maximizing levels and therefore insured remunerative returns to investment. 116 sufficient to stimulate on-farm use. Fertilizer consumption, however, has never met expectations, even in the mid-19705 when it reached an all-time high before plummeting in the wake of recurrent drought and policy changes that eliminated credit and subsidies. th is consump- tion so low if economic analyses have been consistently favorable? It is in reviewing linear programing analyses and social science research that we find some clues to why farmers did not purchase as much fertil- izer as anticipated. 8. LINEAR PROGRAMMING ANALYSES Linear programming (LP) has been the most popular method used to identify optimal input combinations for Senegalese farmers, despite difficulties in formulating coefficients for production relationships. LP analyses have helped researchers recognize constraints that keep farmers from adopting the full range of recommended practices. Hopkins (1975) used a profit maximizing objective function to compare the optimal results obtained with her LP model to farming methods recom- mended by SATEC and found that given existing constraints (particularly labor) farmers who opted for non-SATEC methods were making economically rational choices. Barnett (1979) used goal programming to compare optimal resource allocation given multiple objectives with that obtain- ed using a single profit maximizing objective. He found the results of the multiple objective function only slightly more realistic than those of the profit function. Both credit and labor were identified as con- straints by his model. The amount of fertilizer called for in optimal solutions was greater than that currently employed by farmers. Cereal was found less responsive to fertilizer price increases than peanuts 117 (the author notes, however, that favorable model assumptions about cereal marketing gave cereals an artificial cash crop status). Several researchers have used the Model '4-S" LP programs to identify optimal cropping patterns. Each cropping pattern, however, is restricted to the recomended amount of fertilizer and therefore the model offers no insight about the economic tradeoffs between fertilizer and other inputs (see Richard, Fall, and Attonaty, 1976). While LP models provide some guidance concerning incentives and disincentives associated with modern inputs, they suffer from the poor quality of quantitative data on input/output coefficients (see Barnett, 1979, pp. 80-81, for a discussion). Furthermore, LP models tend to tell us what farmers should do given the mitigating assumptions of the model, but they do not actually tell us what farmers will do nor why they do it. This latter type of information is of great value to those trying to market inputs or develop government agricultural policies. C. BEHAVIORAL AND DESCRIPTIVE RESEARCH BASED ON FARM SURVEY DATA Brochier (1968) reported results of research conducted to explain the gap between yields actually obtained by farmers using "themes lourds” and those anticipated by research results. Among the causes cited by Brochier was the fact that farmers did not use recommended levels of fertilizer. Respondents maintained that it was too expensive for general use but of particular value on very poor soils. These views were expressed at a time when fertilizer was highly subsidized; in 1966, for example, the peanut/fertilizer price ratio was 1.6 (cur- rently, it is about .9). In other words, despite researchers’ ex-post profitability analyses, farmers had different perceptions and made 118 _ input decisions accordingly. Brochier concluded that these perceptions were often related to a poor understanding of fertilizer technology. Venema (1978) reports on a study conducted in the early 19705 on the relationship between social structure and rural development in the 'Unités Expérimentales“. The author noted that early investment in agricultural equipment probably resulted in growth in farm incomes due to extensification. At the time of his study, however, land was becom- ing scarce making investment in equipment profitable only if sufficient attention were paid to techniques for intensification (destumping, timely sowing, sowing in straight lines, use of fertilizer, thinning, and frequent manual weeding). Venema observed that such techniques were not being applied, causing yield increases to be less than anticipated. Nguyen Van Chi-Bonnardel (1978) touches on a number of issues related to input acquisition in her study about the exchange of goods in Senegalese society. Using data from case studies of household revenues and expenditures as well as other less formally collected data, the author concludes that the introduction of modern inputs placed the Senegalese farmer in a vicious spiral of increasing indebt- edness. Of particular interest are the data which illustrate the ex- tremely high inter-annual variability of net revenues, a situation which does not foster a consistent pattern of input use and debt repay- ment. Another observation concerns the relative willingness to seek credit for various needs: most individuals interviewed felt compelled to seek credit to cover costs of social obligations such as weddings, baptisms, and funerals but not necessarily for' production related 119 investments (fertilizer, agricultural implements). Research by Tuck (1983) found attitudes and behavior toward credit similar to those discussed by Nguyen Van Chi-Bonnardel. A recently published review of Unités Expérimentales (UE) research conducted between 1969 and 1976 (Benoit-Cattin, 1986) highlights the difficulty of using farm survey data to draw conclusions about economic returns to fertilizer. ISRA agronomists anticipated that UE results would provide statistically significant confirmation of experimental results. This expectation remains unfulfilled. The major impediments cited by the authors were: 1) Too few observations coupled with extremely high coefficients of variation both within and across years; 2) Interdependence between fertilizer and other agricultural techniques making fertilizer response partially dependent on the organization of work, the allocation of land, and access to equipment; I I 3) Social factors causing less than optimal use of modern tech- nologies by some farmers. The authors argue that available evidence, though not statistically significant, suggests that farmers using the "themes lourds" obtained higher yields than others and experienced less yield variability.l 1These conclusions are based on a comparison between 30 peanut and 13 millet fields farmed with "themes lourds" techniques and 171 peanut and 35 millet fields farmed with "ordinary" techniques (a hybrid of "themes légers" and traditional practices). 120 0. SUMMARY The economic analyses of agronomic data reviewed above suggest that fertilizer is potentially profitable (at least at subsidized prices) while the social science research confirms that farmers have not yet been able to realize the full benefits of the input. A major weakness in the agro-economic research reviewed, however, is the fact that all of the fertilizer response data used was vintage mid-19705 or older and the economic analysis has not been updated using post-1980 prices. Given changes in rainfall patterns, agronomists’ claims that soil quality is rapidly declining, and fourfold increases in fertilizer prices, more regular updating of fertilizer response data and economic analyses appears warranted. Another problem is a failure to adequately address problems of risk and uncertainty in past analyses and a tenden- cy to assess fertilizer profitability in absolute terms (value/cost ratio equal to 2) rather than in relative terms (the profitability of fertilizer versus that of alternative investments). We attempt to rem- edy these lacunae in the next chapter where our own analysis of 1964-82 "Amélioration Fonciere“ fertilizer response data is presented. V. AN ECONOMIC ANALYSIS OF 'AMELIORATION FONCIERE” DATA This chapter presents the results of regression analyses used to examine fertilizer data from the IRAT/ISRA Amélioration Fonciere (AF) programs in Boulel and Nioro.1 The objectives of the analyses are to: (1) Estimate time-series production functions showing average yield response over time, and annual cross-sectional functions which illustrate inter-annual variability in fertilizer response; (2) Use the regression analysis results to evaluate economic returns to current fertilizer recommendations; (3) Evaluate the appropriateness of the AF data set for the analyses performed. The chapter begins with a description of the data which is followed by a discussion of regression results. The next section presents the results of several economic analyses that make use of the data and regression models. We conclude with our observations on the advantages and disadvantages of the AF data set. A. A DESCRIPTION OF THE DATA Table 20 provides the principal characteristics of the four data sets employed. The objective of the AF program was to compare crop response to three levels of fertilizer (F0, F1, F2) in combination with three types of soil preparation (SPO, SP1, SP2) for a total of nine separate treatments. The program was conducted at several off-station 1The data were made available by Guy Pocthier who is currently the CIRAD representative in Senegal. He worked on the AF program during the entire 1964-1982 period. 121 122 . Table 20 Description of the Amélioration Fonciere Data Set Data Zone Set Crop Variety Product Dates Nioro: (I) Sorghum 51-69 Grain 1965-74; 1976-82 CE-lll Grain 1975 (2) Peanuts (II)a 28-206 Peanuts 1965-1982 Hay 1965-74; 1976-82 Boulel: (3) Millet Souna Gam Grain 1973-1980 8001 Grain 1981-82 (4) Peanuts (II)a 28-206 Peanuts 1964-1982 Hay 1965-1982 aThe Roman numeral identifies the place of the peanut crop in the four-year rotation: fallow, peanuts (I), cereal, peanuts (II). 123 locations called 'PAPEM'l. Researchers prefer to call AF activities demonstrations rather than trials because no repetitions were conducted and the size of each treatment plot was 400 square meters -- consider- ably larger than that used in trials. There were two major changes in treatments (1972 and 1976) and a number of minor ones (during the first 5-7 years). Tables 21 and 22 describe the major characteristics of the treatments; details of minor changes have been accounted for in coding but are not mentioned in these Tables. Fertilizer application rates shown in these figures were broken down into their N, P, and K components and coded as kilos of nutrients applied per hectare. In any given year data are available for only three different fertilizer doses; due to changes in protocol, however, the full data set contains 4-5 application rates per nutrient. Dummy variables were used for the four soil preparation levels and for the combined inter-annual effects of such uncontrollable variables as rainfall, insect infestations, and government policy changes. The biggest problems encountered in using these data sets to estimate production functions were: (1) Poor documentation, (2) Deciding how to allocate the basal phosphate dose across the four-year rotation, (3) Multicollinearity. lPAPEMs (Points d’Appui de Prévulgarisation et de l’Expérimentation Multilocale) were the off-station locations where "prévulgarisation" trials were carried out. PAPEMs were located in a number of different agro-climatic zones on land offered to researchers by local canmufities. 124 Table 21 "Amélioration Fonciere' Peanut Treatments for the Sine Saloum 1964 - 1982 DATES LABEL TREATMENT (per hectare) 1964-71 FO . No fertilizer F1 150 kilos of 6-20-10 F2 'Fumure étalée"a SPO No soil preparation SP1 Light soil preparation using horse traction and a hoe SP2 Medium soil preparation using oxen traction to plow under the residue of the fallow at the beginning of the rotation 1972-75 "F ann l ' re 1 'fumu ta 6 ' for ll h r r rem i e n ha . F2 150 kilos of 7-21-29 in 1972 and 1973 replaced by 8-18-27 in 1974 1976-82 F1 and SPO eliminated; F3 and SP3 jgtroduced F3 150 kilos of 8-18-27 plus 50 kilos of potassium chloridg (60% K) and 250 kilos of 'phosphogypse' SP3 Heavy soil preparation using oxen traction to perform several deep (20 cm.) plowings during the course of the 4-year rotation aExact amounts of basal phosphate, amonium sulfate, and potassium chloride changed annually; this is accounted for in the coding. bDue to coding problems the 'phosphogypse' was not considered in the current analysis. It would be useful to examine the effect in the future as data from the ”Régénération des Sols" program did not show any significiant effect. See Pocthier, Pieri, and Mara (1977) for more discussion of this analysis. 125 Table 22 "Amélioration Fonciere' Cereal Treatments for the Sine Saloum 1965-1982 DATES LABEL TREATMENT (per hectare) 1965-7O FO No fertilizer F1 150 kilos of 14-7-7 F2 'Fumure étalée"a SPO No soil preparation SP1 Light soil preparation using horse traction and a hoe SP2 Medium soil preparation using oxen traction to plow under the residue of the fallow at the beginning of the rotation 1971—75 ”Egmgre annuelle' replaeeg 'fgmgre etalee" for F2 1 the m n han d. F2 150 kilos of lO-21-21 plus 100 kilos of urea in Boulel and 150 kilos in Nioro 1976-82 F1 and SPO eliminated; F3 ang SP3 introduced F3 150 kilos of 10-21-21 plus 150 kilos of urea SP3 Heavy soil preparation using oxen traction to perform several deep (20 cm.) plowings during the course of the 4-year rotation aExact amounts of basal phosphate, ammonium sulfate, and potassium chloride changed from year to year; this is accounted for in the coding. 126 Coding treatments was extremely difficult because there is no systema— tic documentation on year-to-year changes in fertilizer application rates. This necessitated lengthy discussions with researchers involved in the AF program to obtain accurate information on application rates. Discussions with agronomists revealed that there is no clear concensus on the best way to allocate the quadrennial phosphate dose to each crop in the rotation for ”fumure étalée“ treatments. We decided to use the amount of P contained in the F2 applications recommended when ”fumure étalée" was replaced by 'fumure annuelle". This decision was based on agronomists’ assurances that the change from “fumure étalée" to "fumure annuelle" did not entail any significant change in application rates.1 The next problem was the high multicollinearity among the N, P, and K variables which precluded estimation of significant coefficients for the separate nutrients.2 It was also responsible for large changes in the sign or value of N, P, and K coefficients when slight model changes were introduced. To get around this problem, we used the sum of N+P+K as the independent variable and reported the combined effect.3 1In fUture analyses of AF data, it would be worthwhile to examine alternative procedures for coding these phosphate treatments and test the sensitivity of results to changes in coding procedures. 2In all data sets correlations between N, P, and K were in the .80 to .98 range while correlations between the dependent variable (yield) and these explanatory variables were in the .25 to .45 range. It should be pointed out that the AF program was not designed to produce data for production function estimation and separation of N, P, and K effects. It is in trying to do more with these data than originally planned that the correlations among different levels of N, P, and K posed a problem. 3For example the variable "NPKSUM" equals N+P+K. Summing variables in this manner is frequently done when one believes that the coefficients for all the summed variables are identical. This is net the case here; the summing was strictly a means of handling multicollinearity problems. 127 B. ESTIMATION OF PRODUCTION FUNCTIONS The data were analyzed from two different perspectives: (1) as a pooled time-series and cross-section analysis and (2) as a simple cross-section analysis for each year. The model used for the pooled data set was: Yield - Bo_+ BlNPKSUM + BzNPKSUM2 + B3SPO + B4SP1 + B5SP2 + 85Y2 + 87Y3 + .... + BnYN + Error Where Yield - yield in kilos/hectare; Bo - the without fertilizer yield in the earliest year for which data were available (Y1) using very intensive soil preparation (SP3); NPKSUM - sum total of NPK nutrients applied per hectare; NPKSUMZ - NPKSUM squared. Oummary variables in the model were: SPO - 1 if no soil preparation, zero otherwise; SP1 I if light soil preparation, zero otherwise; SP2 = 1 if heavy soil preparation, zero otherwise; SP3 - 1 if very intensive soil preparation, zero otherwise; Y1 = 1 for all observations in year one, zero otherwise; Y2 - 1 for all observations in year two, etc. A quadratic model was selected for ease of estimation and because it was the model most frequently used in earlier analyses of fertilizer response in Senegal. If the quadratic term was not significant at .05 or better, NPKSUM2 was dropped and a linear‘ model was estimated. Initially, SP dummies were analyzed separately as shown above. A series of F-tests, however, indicated that the coefficients for SPO and 128 SP1 were not significantly different. This was also true for SP2 and SP3. In the final models a composite SP variable was used: SPOI (equal to I if either SPO or SP1 were equal to 1). This means that the constant (80) in the final models reflects the without fertilizer yield in the base year using either SP2 or SP3.1 The pooled models explain a large part of yield variability for peanuts, peanut hay, and cereal grains; exhibiting adjusted Rz’s rang- ing from .73 to .84. F-tests for the models and t—tests on fertilizer coefficients are all significant at better than .001. Taken as a group, the year dummies are also highly significant (better than .001) in all models. The soil preparation dummies are also significant at .05 or better in all models except that for Boulel peanut hay. The Durbin-Watson statistics fell in the indeterminate zone for most models, but an auto correlation plot of residuals did not reveal any problems warranting adjustment of the models. Tables 23-28 present the results of the pooled analyses, showing the estimated coefficients and standard statistics. The~ estimated fertilizer effect for each data set with its 95 percent confidence interval, and the estimated yields associated with different fertilizer and soil preparation levels are also presented. The confidence inter- vals on the fertilizer response coefficients are quite narrow in most cases. This, unfortunately, is not true for the overall yield esti- mates which exhibit 95 percent confidence intervals with widths of 1There was one exceptional case, Boulel peanuts, where no preparation and very intensive preparation appeared to have a similar effect while the two middle levels of preparation were not significantly different. In this case SP1 and SP2 were combined, with SPO and SP3 being reflect— ed in the constant. 129 Table 23: Nioro Peanuts Pooled Regression Model DEPENDENT VARRIABLE: PM YIELD ’ N: 162 ADJUSTED R SQUARED: .840 SEE: 222.657 PROB 2 TAILED VARIABLE COEFFICIENT STD ERROR TOL* T-TEST CONSTANT 2326.362 82.272 1.0 000 NPKSUMZ -0.043 1.169 000 NPKSUM 9.601 0.010 000 T01 -75.704 37.110 043 Y66 -556.333 104.962 000 Y67 -lO46.556 104.962 000 Y68 -667.655 104.989 000 Y69 -ll9l.099 104.989 000 Y7O -IlOl.550 104.964 000 Y7l -39l.327 104.964 Y72 -384.193 104.964 Y73 -490.526 104.964 Y74 -33l.474 104.980 Y75 -978.761 104.983 Y76 -483.235 107.738 1727.569 107.738 Y78 -467.235 107.738 Y79 -949.791 107.738 1509.791 107.738 Y81 40.320 107.738 Y82 -763.457 107.738 SIGNIFICANCE LEVEL OF F-TEST FOR MODEL: .000 SIGNIFICANCE LEVEL OF F-TEST FOR Y66 - Y82: .000 ..< N \l 1 ..< 00 O 1 OOOOOOOOOOOOOOOOOOOOO O O O mmmmwmmmwmmmmmmmmmuw .OOO ESTIMATED FERTILIZER RESPONSE AND ESTIMATED AVERAGE YIELDS (KILOS/HECTARE) TREATMENT FERT. RESP. 95% 0.1. AVER. YIELD NO FERTILIZER/SPOI NA NA 1528 150 KILOS 6-20-10/SP01 393 325-461 1921 150 KILOS 8-18-27/SP23 491 432-551 2096 *Tolerance is a measure of multicollinearity; values close to zero are associated with high levels and values close to one with low levels of multicollinearity. Table 24: 130 Nioro Peanut Hay Pooled Regression Model DEPENDENT VARRIABLE: HAY N: 153 ADJUSTED R SQUARED: .816 SEE: 299.751 PROB 2 TAILED VARIABLE COEFFICIENT STD ERROR TOL* T-TEST CONSTANT 2394.434 111.425 1.0 0.000 NPKSUMZ -0.066 0.013 .1 0.000 NPKSUM 17.310 1.614 .1 0.000 T01 -352.235 51.407 .9 0.000 Y66 571.222 141.304 .5 0.000 Y67 -733.889 141.304 .5 0.000 Y68 -310.353 141.341 .5 0.030 Y69 -1058.464 141.341 .5 0.000 Y70 -728.328 141.308 .5 0.000 Y71 144.005 141.308 .5 0.310 Y72 -487.054 141.307 .5 0.001 Y73 77.835 141.307 .5 0.583 Y74 -289.419 141.329 .5 0.043 Y76 -902.146 145.109 .5 0.000 Y77 -642.591 145.109 .5 0.000 Y78 -185.035 145.109 .5 0.204 Y79 -265.591 145.109 .5 0.069 Y80 -670.368 145.109 .5 0.000 Y81 27.298 145.109 .5 0.851 Y82 -786.480 145.109 .5 0.000 SIGNIFICANCE LEVEL OF F-TEST FOR MODEL: .000 SIGNIFICANCE LEVEL OF F-TEST FOR Y66 - Y82: .000 ESTIMATED FERTILIZER RESPONSE AND ESTIMATED AVERAGE YIELDS (KILOS/HECTARE) TREATMENT FERT. RESP. 95% C.I. AVER. YIELD N0 FERTILIZER/SPO] NA NA 1675 150 KILOS 6-20-10/SP01 742 617-867 2406 150 KILOS 8-18-27/SP23 959 749-1169 2961 *Tolerance is a measure of multicollinearity; values close to zero are associated with high levels and values close to one with low levels of multicollinearity. 131 Table 25: Nioro Sorghum Pooled Regression Model DEPENDENT VARRIABLE: SORGHUH YIELD N: .162 ADJUSTED R SQUARED: .768 SEE: 435.682 PROB 2 TAILED VARIABLE COEFFICIENT STD ERROR TOL* T-TEST CONSTANT 1474.853 159.727 1.0 .000 NPKSUMZ -0.049 1.554 .000 NPKSUM 17.258 0.008 .000 T01 -345.972 72.614 .000 Y66 23.357 205.522 Y67 607.024 205.522 Y68 165.516 205.395 Y69 -26.039 205.395 Y70 206.575 205.478 Y71 391.838 205.539 Y72 -143.829 205.539 .910 .004 .422 .899 .316 .059 .485 mmmmmmmmmmmmmmu‘mmmt—H OOOOOOOOOOOOOOOOOOOOO Y73 -63.385 205.539 .758 Y74 267.838 205.539 .195 Y75 -843.162 205.539 .000 Y76 -708.027 213.861 .001 Y77 -426.694 213.861 .048 Y78 289.417 213.861 .178 Y79 -966.361 213.861 .000 Y80 -364.916 213.861 .090 Y81 -1328.583 213.861 .000 Y82 -1457.361 213.861 .000 SIGNIFICANCE LEVEL OF F-TEST FOR MODEL: .000 SIGNIFICANCE LEVEL OF F-TEST FOR Y66 - Y82: .000 ESTIMATED FERTILIZER RESPONSE AND ESTIMATED AVERAGE YIELDS (KILOS/HECTARE) TREATMENT FERT. RESP. 95% 0.1. AVER. YIELD NO FERTILIZER/SP01 NA NA 915' 150 KILOS 14-7-7/SP01 638 480-797 1524 150 KILOS 10-21-21 AND 150 KILOS OF UREA/SP23 1478 675-2281 2710 *Tolerance is a measure of multicollinearity; values close to zero are associated with high levels and values close to one with low levels of multicollinearity. 132 Table 26: Boulel Peanuts Pooled Regression Model DEPENDENT VARRIABLE: PN YIELD N: 171 ADJUSTED R SQUARED: .759 SEE: 226.716 PROB 2 TAILED VARIABLE COEFFICIENT STD ERROR TOL* T-TEST CONSTANT 853.996 81.551 1.0 000 NPKSUM 3.317 0.387 000 T12 93.018 36.778 1. 012 Y65 770.778 106.875 Y66 483.889 106.875 Y67 ~161.889 106.875 Y68 547.272 106.881 Y69 28.716 106.881 Y70 328.128 106.876 Y71 237.350 106.876 Y72 -261.239 106.876 Y73 -559.683 106.876 Y74 545.728 106.881 Y75 323.535 106.896 Y76 -394.153 107.290 Y77 109.069 107.290 Y78 233.180 107.290 Y79 -543.931 107.290 Y80 -552.487 107.290 Y81 233.624 107.290 Y82 183.291 107.290 .090 SIGNIFICANCE LEVEL OF F-TEST FOR MODEL: .000 SIGNIFICANCE LEVEL OF F-TEST FOR Y65 - Y82: .000 mmmmmmmmmmmmmmmmmmom OOOOOOOOOOOOOOOOOOOOO O O—I Ch ESTIMATED FERTILIZER RESPONSE AND ESTIMATED AVERAGE YIELDS (KILOS/HECTARE) TREATMENT FERT. RESP. 95% C.I. AVER. YIELD NO FERTILIZER/SPIZ NA NA 1023 150 KILOS 6-20-10/SP12 197 137-221 1202 150 KILOS 8-18-27/SP12 264 202-325 1287 *Tolerance is a measure of multicollinearity; values close to zero are associated with high levels and values close to one with low levels of multicollinearity. 133 Table 27: Boulel Peanut Hay Pooled Regression Model DEPENDENT VARRIABLE: HAY N: 162 ADJUSTED R SQUARED: .786 SEE: 339.199 PROB 2 TAILED VARIABLE COEFFICIENT STD ERROR TOL* T-TEST CONSTANT 695.967 122.430 1.0 0.000 NPKSUM 5.777 0.590 .9 0.000 T12 32.741 56.533 1.0 0.563 Y66 2181.000 159.900 .5 0.000 Y67 663.444 159.900 .5 0.000 Y68 1691.112 159.910 .5 0.000 Y69 -54.221 159.910 .5 0.735 Y70 223.334 159.901 .5 0.165 Y71 471.778 159.901 .5 0.004 Y72 786.889 159.901 .5 0.000 Y73 73.222 159.901 .5 0.648 Y74 969.666 159.910 .5 0.000 Y75 -72.554 159.933 .5 0.651 Y76 -256.140 160.543 .5 0.113 Y77 483.638 160.543 .5 0.003 Y78 -280.695 160.543 .5 0.083 Y79 459.083 160.543 .5 0.005 Y80 88.305 160.543 .5 0.583 Y81 461.749 160.543 .5 0.005 Y82 354.083 160.543 .5 0.029 SIGNIFICANCE LEVEL OF F-TEST FOR MODEL: .000 SIGNIFICANCE LEVEL OF F-TEST FOR Y66 - Y82: .000 ESTIMATED FERTILIZER RESPONSE AND ESTIMATED AVERAGE YIELDS (KILOS/HECTARE) TREATMENT FERT. RESP. 95% C.I. AVER. YIELD N0 FERTILIZER/SPIZ NA NA 1187 150 KILOS 6-20-10/SP12 312 280-344 1499 150 KILOS 8-18-27/SP12 460 412-506 1646 *Tolerance is a measure of multicollinearity; values close to zero are associated with high levels and values close to one with low levels of multicollinearity. 134 Table 28: Boulel Millet Pooled Regression Model DEPENDENT VARRIABLE: MILLET YIELD N: 90 ADJUSTED R SQUARED: .731 SEE: 278.497 PROB 2 TAILED VARIABLE COEFFICIENT STD ERROR TOL* T—TEST CONSTANT 301.541 105.128 1.0 0.005 NPKSUM 6.118 0.482 .9 0.000 T01 -151.300 62.274 .9 0.017 Y74 597.333 131.285 .6 0.000 Y75 375.222 131.285 .6 0.005 Y76 -211.131 133.981 .5 0.119 Y77 509.758 133.981 .5 0.000 Y78 184.980 133.981 .5 0.171 Y79 362.647 133.981 .5 0.008 Y80 292.980 133.981 .5 0.032 Y81 492.425 133.981 .5 0.000 Y82 551.980 133.981 .5 0.000 SIGNIFICANCE LEVEL OF F-TEST FOR MODEL: .000 SIGNIFICANCE LEVEL OF F-TEST FOR Y74 - Y82: .000 ESTIMATED FERTILIZER RESPONSE AND ESTIMATED AVERAGE YIELDS (KILOS/HECTARE) TREATMENT FERT. RESP. 95% C.I. AVER. YIELD N0 FERTILIZER/SPIZ NA NA 472 150 KILOS 14-7-7/SP01 257 216-297 723 150 KILOS 10-21-21 AND 100 KILOS 0F UREA/SP23 759 639-878 1376 *Tolerance is a measure of multicollinearity; values close to zero are associated with high levels and values close to one with low levels of multicollinearity. 135 approximately 800 kilos/hectare for Boulel peanuts, 1200 for Nioro peanuts and Boulel millet, and 2200 for Nioro sorghum. While the production functions all confirm that there is a statistically signifi- cant fertilizer effect (the NPKSUM coefficients do not equal zero), the confidence intervals do not provide convincing evidence that the yield response a farmer can anticipate with F1 or F2 will be significantly different than that obtained with F0. The method used for estimating these time-series production func- tions described above provides a single estimate for the average fer- tilizer effect over the entire period. This abstracts from the fact that fertilizer response is unlikely to remain constant from year-to- year. An alternative way of examining year-to-year changes in crop response is to fit production functions to cross-sectional data for each year. This data consisted of one observation per treatment for a total of ‘nine observations per year. For each year, however, there were only 3 different levels of fertilizer -- the absolute minimum number of observations for estimating a quadratic function. Two models were tested: (1) Yield - Bo + BlNPKSUM +BZSP01 + ERROR (2) Yield - 80 + BlNPKSUMZ + BZNPKSUM + B3SP01 + ERROR] The quadratic was selected over the linear model if the t-test on both fertilizer coefficients were significant at .10 or better. In a number of cases this rule selected models where the signs on the fer- tilizer variables were reversed (NPKSUM negative and NPKSUM2 positive). The models were retained as the F-tests, adjusted Rz’s, and plots of 1The SP variable for Boulel peanuts was SP12 as in the pooled models. the da' 136 the data suggested that these models presented accurate pictures of the data. Apparently crops responded poorly to low doses in these years and did not exhibit diminishing returns to higher doses. Table 29 presents a frequency distribution of model characteris- tics obtained from the cross-sectional analysis. Given the limited number of observations per year, linear models were more common than quadratic ones -- particularly in Boulel. We were able to show a significant fertilizer effect about 80-90 percent of the time for the two cereal crops and Nioro peanuts. Only sixty-three percent of the time were we able to confirm a statistically significant fertilizer effect for Boulel peanuts. The adjusted Rz’s were relatively high for cereals and Nioro peanuts, but quite low for Boulel peanuts. Although fertilizer coefficients were significant at .10 or better, they had relatively large confidence intervals. This was particularly true for quadratic models. In an attempt to obtain smaller confidence intervals we also examined the following model: Yield . BllYl + B21Y1*NPKSUM + B31Y1*NPKSUM2 + B41Y1*SP01 + .... + BlnYN + BZnYN*NPKSUM + B3nYN*NPKSUM2 + B4nYN*SP01 + Error In theory, estimation of this model with the entire multi-year data set should reduce standard errors for the individual coefficients and for the model as a whole while providing the desired information on inter- annual variability in without-fertilizer yields, fertilizer response, and soil preparation effect. Attempts to use the model for Nioro and Boulel peanuts were thwarted by error messages indicating singular matrices. Quadratic models estimated for Nioro sorghum and Boulel 137 Table 29 Summary of Cross-Sectional Analysis of "Amélioration Fonciere" Data -- Numbers Represent the Percent of Total. Annual Observations Falling Into Each Categor -- BOULEL NIORO PEANUTS MILLET PEANUTS SORGHUM YP F 00 N0 SIGNIFICANT MODEL 37 10 22 17 LINEAR MODEL 42 70 50 33 QUADRATIC MODEL 21a 20a 28 50 ncv o STRI u ON or ADJUST 3L5 R NT F N 0 S <.75 58 22 21 13 (.90 17 44 43 27 2.90 25 33 36 50 R T 0F SIGNIFICANT OD S TH S GNI ICANT COEFFI NT 0 R A T ON 67 22 29 53 Source: Estimated from AF data. aThree of the peanut models (1976, 1978, and 1982) and one of the millet models (1979) have the signs of the NPK coefficients reversed from what is normally sxpected in fertilizer response functions (NPKSUM is negative and NPKSUM is positive). 138 millet, and a linear model for Boulel peanuts gave approximately the same results as the year-by-year models. The standard errors for the fertilizer coefficients, however, tended to be larger in these models than in the simple cross-sectional models. C. ECONOMIC ANALYSIS We use the AF data and the pooled regression models to perform three types of economic analyses. The first is a partial budget which compares returns obtained without fertilizer to both light and heavy fertilizer doses for an entire 4-year rotation in Nioro and Boulel. This is followed by an examination of the variation in v/c ratios over time. Finally, we use our knowledge of the variability in yields and v/c ratios to develop a number of decision models which examine altern- ative investments. 1. BUDGET ANALYSES Results of partial budget analyses are presented in Tables 30 and 31. The yield responses in these tables are estimated using the pooled production function results, current "themes légers'I and "themes lourds” recommendations, and 1987 prices.1 The SP1 versus SP2 differ- ential in equipment and animal traction costs for the entire rotation is shown in the "added costs“ column of the SP2 line marked "fallow". Equipment prices used were obtained from SISMAR (Senegal’s only agri- cultural equipment manufacturer) in November 1987. Information from Raymond, Monnier, and Cadot (1974) was used to estimate animal feed lPeanut (II) yields are usually lower than Peanut (I) yields. As we had no data on Peanut (I) crops in Boulel and Nioro, we estimated peanut (1) yields as 110 percent of Peanut (11) yields. This is approximately the difference shown in the budgets presented in Tourte et a . 1971 . 139 Table 30 Nioro 1987 Partial Btdget Analysis Cmrim "Mo Fertilizer“ with “Themes Légers" and “Themes Lourds“ Reed-mendation (FCFA/hectare) ro-sm r1-s111 “1.1100. ADDED NET. BEN. v10 1101111011 11510 mus 1151.0 mus moucT. COSTS 1101111011 11111110 11111011 1111111 (1) 11015 1660 151200 2113 190170 36970 11111 1643 46075 2647 66175 20100 11660 47190 4.97' 901101131 915 64050 1524 106660 42630 12315 30315 3.46 1111111 (11) 11015 1526 137520 1921 172690 35370 111111 1675 41675 2406 60150 16275 11660 41765 4.523 COSTS/BENEFITS 1111110111 SUBSIDY 440720 596065 155345 36075 119270 4.31 oos1s1uuerx1s 111111 SUBSIDY 155345 26675 126470 5.36 ro-spz rz-spz 11111011 0 13772 -13772 0.00 911111 (1) 1101s 1764 156760 2306 207540 46760 11111 2230 55750 3257 61425 25675 15150 59305 4.91‘3 9011611011 1232 66240 2710 169700 103460 23130 60330 4.47 PMUT (11) ms 1604 144360 2096 166640 44260 , 11111 2027 50675 2961 74025 23350 15150 52460 4.46a COSTS/BENEFITS . 1111110111 suas101 495765 741330 245545 67202 176343 3.65 mmwuunm 111111 5065101 245545 57602 167943 4.26 Source: Adapted from Tourte et al. (1971) using yields estimated from AF data. aThe value/cost ratio is calculated using the conbined value of added peanuts and hay divided by the fertilizer costs. 140 Table 31 Boulel 1987 Partial Bulget Analysis Conquering “NO Fertilizer“ 11ith "Thales Légers" and “Theses Lourds“ Racouendationa (FCFA/hectare) ro-sm F1-SP1 VAL. ADD. ADDED 1151 BEN. VIC nOTATIa1 YIELD VALtE YIELD VALLE PnooucT COSTS ROTATION RATIO FALLGJ --- PNUT (I) NUTS 1125 101250 1322 1189i) 1TBO RAY 1306 32650 1649 41225 8575 11880 14425 2.218 M! LLET 472 33040 723 50610 17570 12315 5255 1.43 P11111 (ll) HITS 1023 92070 1202 108180 16110 11A11 1167 29675 1499 37475 7600 11660 12030 2.011 COSTS/BENEFITS 1111111111 SUBSIDY 288685 356470 67785 36075 31710 1.88 COSTS/BENEFITS 111111 8188101 67785 28875 38910 2.35 FO-SP2 F2-SP2 FALLGJ ---- ---- ---- ---- 0 13772 -13772 0.00 P1101“) WTS 1125 101250 1416 127440 26190 MAY 1306 32650 1811 45275 12625 15150 23665 2.568 MILLET 617 43190 1376 96320 53130 20220 32910 2.63 P1101 (11) 11115 1023 92070 1287 115830 23760 5 MAY 1187 29675 1646 41150 11475 15150 20085 2.33 mSTS/BENEFITS 111111011 918is 298835 426015 127180 64292 62888 1.98 OOSTS/BENEFITs 111111 SlBSlOY 127180 55492 71688 2.29 Source: Adapted from 1ourte et al. (1971) using yields estimated from AP data. aThe value/cost ratio is calculated using the cabined value of added peanuts and hay divided by the fertilizer costs. 141 costs. Appendix 11 contains detailed information on the estimation of various parameters used in the budgets. To facilitate comparison with .earlier economic analysis we followed the budget format used by Tourte et al. (see Tables 15 and 16 above). The major difference between our analysis and earlier work is that we have considered economic returns to increased peanut hay production associated with fertilizer use.1 An examination of the v/c ratios calculated at real prices for each crop in the Nioro rotation (Table 30) reveals that the ratio is generally greater than four. There is no difference between v/c ratios for light and heavy fertilizer doses on peanuts. Sorghum, however, exhibits a v/c ratio of 4.47 for the heavier fertilizer application versus 3.46 for the lighter dose. The v/c ratios associated with the entire "themes légers” rotation are greater than those for the "themes lourds” using both real (4.31 versus 3.65) and subsidized (5.38 versus 4.26) prices. Using real prices, the "themes lourds" net benefit is 59,073 FCFA more than the less intensive theme; using subsidized prices increases the additional net benefit by less than 2,500 FCFA. This revenue represents returns on the cash investment as well as additional labor required for the "themes lourds”. The extra labor would be about 80 hours required during cereal harvesting time to plow under the fallow, and the time necessary to harvest, thresh, and transport the added peanut and sorghum production. 1Some data on peanut hay prices were collected by the ISRA/BANE cereals marketing program between 1985 and 1987. Although price varies by zone and season, we have used a price of 25 FCFA/kilo which is representa- tive of the average price that farmers can reasonably expect to receive. 142 The economic picture for Boulel (Table 31) is less favorable than that for Nioro, with most individual-crop v/c ratios just reaching the minimum acceptable level of 2 when real prices are used. The v/c ratios for the entire rotation, however, are below 2 for both themes. In the case of light fertilizer, the poor millet performance (v/c ratio - 1.43) is responsible. In the case of heavy fertilizer, the low overall ratio is due to the additional equipment costs. The current fertilizer subsidy increases the v/c ratio for both themes to slightly greater than 2. In reviewing these results with ISRA agronomists, we found general agreement that the estimated yields for all treatments in Nioro were higher than those generally obtained by farmers. There was a general acceptance of the Boulel results, which many consider to be a reflec- tion of recent declines in agricultural productivity due to lower rainfall and declining soil fertility. With respect to both Nioro and Boulel, agronomists expressed concern that the relatively' high re- Sponses associated with ”themes légers" should not be extrapolated to farming situations where the fallow is not respected and little effort has been made to maintain soil fertility in recent years. It is gener- ally thought that 14-7-7 and 6-20-10 will elicit little response under such conditions; but, to the best of our knowledge, this has not been confirmed by any trial or demonstration data.l lPreliminary analysis of 1986 researcher-managed trials, conducted on farmers’ fields in 2 zones of the Sine Saloum characterized by differ- ent soil types, shows a statistically significant fertilizer response ,with only a 50 percent application of the recommended dose for millet (the recommended dose was 150 kilos 0-15-20 plus 100 kilos of urea). The difference between yields for fields receiving 50 percent and 100 percent of the recommended dose was not significant in one zone but was 143 Some may be inclined to blame the poor peanut performance in Boulel on continued use of the lZO-day variety despite the shorter Boulel rainy season which has become common in recent years. Given the very long data series (1964-82), this can only be a partial explana- tion. There was no significant difference in the fertilizer coeffi- cients estimated for the data from 1964-1973 versus those for 1974- 1982. Furthermore, a 105/110-day peanut variety was tested from 1975- 1982 at Boulel. Preliminary analyses of these data suggest that the shorter cycle variety performed less well than the lZO-day variety, both with and without fertilizer. Earlier budget analyses by Tourte et al. did not factor in the economic returns associated with peanut hay. Although the importance of the hay crop was recognized in the early 19703, it is only in the last 5-10 years that farmers have actively begun marketing a large portion of their hay production and price data have become available. To facilitate a comparison between the Tourte et al. analysis and our own, it is necessary to remove the economic returns due to peanut hay , and consider only the returns to seed. In doing this we find that v/c ratios (real prices) for Nioro peanuts remain greater than 2 (2.92 to 3.28); on a crop-by-crop basis they are greater than those associated with the Tourte et al. analysis.l Value/cost ratios for the entire rotation, with either subsidized or unsubsidized prices, continue to be in the other. Source: Personal communication, A. Thiam, ISRA-Kaolack. 1Table 5 does not present calculations of v/c ratios for each year of the rotation; but available information permitted us to estimate ratios of 1.19 and 1.65 for ”themes légers” peanuts (I) and (II), and 1.64 and 2.12 for "themes lourds”. 144 as good or better than averages reported for the Sine Saloum by Tourte et al., even without considering income from peanut hay. The case for Boulel is much bleaker. The peanut v/c ratios range from 1.36 to 1.73 at 1987 real prices if hay is not considered. The overall v/c ratios for the entire rotation are 1.43 for the "themes légers“ and 1.6 for the ”themes lourds”. At 1987 subsidized prices, the ”themes légers" v/c ratio attains the 2.35 level but the ”themes lourds” ratio remains less than two. Current Boulel results resemble the Tourte et al. results for the North (Table 15) rather than those for the Sine Saloum (Table 16). Some caution should be used in making these comparisons as the Tourte et al. budget for the Sine Saloum averaged results from Boulel, Nioro, Keur Samba, Toubacouta, and Keur Yoro Oou. The fact that our current Nioro analysis produces better results than Tourte et al.’s average Sine Saloum analysis does not necessarily mean that the situa- tion in Nioro has improved, nor that the situation for Boulel has deteriorated. In doing our analysis on a site-by-site basis, we are able to illustrate the heterogeneity in Senegalese agriculture -- even within the same geographic zone. It is our opinion that averaging such diverse results together masks important differences which must be taken into account in fertilizer policy analysis. 2. INTER-ANNUAL VARIATION IN V/C RATIOS The budget analyses presented above were based on estimates of average yields and fertilizer response during 1964-82 and do not ad- dress the problem of inter-annual variability. Tables 32 and 33 pre- sent frequency distributions of annual v/c ratios associated with a 145 Table 32 Frequency Distribution of Fertilizer Value/Cost Ratios Over Time Nioro 1967 SUBSIDIZED 1987 RE L NOMINAL PRICESb PRICESc PRICES CROP ma 3 cuu z CUM % cuu PEANUTS <1 6 6 6 6 22 22 1965 1-2 22 26 26 34 17 39 TO 2-3 17 45 11 45 6 45 1982 3-4 6 51 6 51 11 56 >4 so 101 so. 101 44 100 AVERAGE: 5.11 AVERAGE: 3.69 AVERAGE: 3.1 PEANUT <1 0 0 o o SEED 1-2 6 6 16 16 AND HAY 2-3 16 24 6 24 1965-74 3-4 6 3o 16 42 AND >4 71 101 59 101 1976-62 AVERAGE: 6.4 AVERAGE:5.13 SORGHUM <1 0 o o o o o 1965 1-2 0 o 5 5 5 5 TO 2-3 5 5 o 5 5 10 1962 3-4 0 5 11 16 21 31 7o F/KG >4 95 100 64 100 66 99 AVERAGE: 11.35 AVERAGE: 7.45 AVERAGE: 6 SORGHUM <1 5 5 5 5 1965 1-2 0 5 21 26 TO 2-3 32 37 21 47 1962 3-4 16 53 16 63 4o F/KG >4 37 100 37 100 AVERAGE: 4.26 AvERAGE:3 43 Source: Estimated using AF data. aData included in each category are equal to or greater than the first number and less than the second number unless noted otherwise. bNominal prices are those faced by farmers during 1965-82 for "themes légers" fertilizer doses. cPeanut price . 9O FCFA; cereal fertilizer . 9,915 FCFA/ha; peanut fertilizer . 9,480 FCFA/ha. dPeanut price - 90 FCFA; cereal fertilizer = 12,315 FCFA/ha; peanut fertilizer - 11,880 FCFA/ha. 146 Table 33 Frequency Distribution of Fertilizer Value/Cost Ratios Over Time Boul el 1987 SUBSIDIZED 1987 RE L NOMINAL PRICESb PRICES CROP V/Ca % CUM x CUM x. CUM PEANUTS <1 53 53 53 53 53 53 1964 1-2 0 53 11 64 16 69 T0 2-3 16 69 11 75 16 85 1982 3-4 11 80 ll 86 O 85 >4 21 101 16 102 16 101 AVERAGE: 2.1 AVERAGE: 1.8 AVERAGE: 1.4 PEANUT <1 44 44 50 50 SEED 1-2 11 55 11 61 AND MAY 2-3 6 61 6 67 1965 3-4 6 67 11 78 T0 >4 33 100 22 100 1982 AVERAGE: 2.4 AVERAGE: 1.8 MILLET <1 10 10 10 10 10 10 1973 1-2 0 10 0 10 10 20 T0 2-3 0 10 20 30 30 50 1982 3-4 10 20 30 60 20 70 70 F/KG >4 80 101 40 100 30 100 AVERAGE: 6.7 AVERAGE: 4.3 AVERAGE: 3.5 MILLET <1 10 10 20 20 1973 1-2 30 40 50 70 T0 2-3 30 7O 0 70 1982 3—4 0 7O 20 9O 40 F/KG >4 30 100 10 100 AVERAGE: 2.5 AVERAGE: 2 Source: Estimated from AF data. aData included in each category are equal to or greater than the first number and less than the second number unless noted otherwise. bNominal prices are those faced by farmers during 1964-82 for ”themes légers" fertilizer doses. cPeanut price - 90 FCFA; cereal fertilizer - 9,915 FCFA/ha; peanut fertilizer - 9,480 FCFA/ha. dPeanut price - 9o FCFA; cereal fertilizer = 12,315 FCFA/ha; peanut fertilizer - 11,880 FCFA/ha. 147 move from FO-SPI to FI-SPI cultivation practices for Nioro and Boulel. Fertilizer response is based on actual data rather than regression coefficients for the cross-sectional analysis.1 Value/cost ratios presented reflect (1) nominal prices paid and received by farmers during the 1964-82 period (2) 1987 subsidized prices and (3) 1987 real prices. Peanut v/c ratios are estimated with and without hay to illus- trate the role peanut hay can play in reducing the risk of low returns to fertilizer investments. The nominal price v/c ratios do not include returns to hay because historical price data are not available. The 1987 ratios for cereal crops are calculated using both the official 70 CFA/kilo price and a 40 CFA/kilo price. The latter price is a rough estimate of the minimum prices being paid in zones where the government has not been able to support the official price. A quick look at the tables shows that Nioro farmers who invested in fertilizer between 1964 and 1982 faced average v/c ratios of 5.11 for peanuts and 11.35 for sorghum. Sorghum in Nioro performed well consistently. At official prices the risk of the added value being less than fertilizer investment costs is zero and the risk of a v/c ratio less than 2 is only 5 percent. This is surprising given that this sorghum variety is a 135-day crop. Many farmers in the Sine Saloum sharply reduced sorghum areas during the late 1970s and 1980s, believing that Shorter cycle cereals were more reliable under then 1He decided to use actual data rather than the cross-sectional fertil- izer response estimates because the small number of annual observations made it difficult to obtain production functions that provided reliable fertilizer response estimates with reasonable confidence intervals. 148 prevailing drought conditions.1 When one considers the lower 40 CFA per kilo producer price, v/c ratios less than 2 occur only 5 percent of the time if fertilizer is subsidized. The removal of the subsidy in- creases the probability of ratios less than two to 26 percent. At nominal and subsidized prices there is only a 6 percent chance of V/c ratio being less than one for peanuts. Moving to real prices increases the probability of experiencing a ratio less than one to 22 percent; the probability of a ratio less than two changes to 39 per- cent. Factoring in returns to hay significantly reduces risk: zero probability of a ratio less than one and only 18 percent probability of a ratio less than two. The average peanut plus hay v/c ratio is 5.13 at real 1987 prices -- 60 percent greater than the without-hay ratio. The average V/C ratio for Boulel peanuts was much lower than in Nioro (2.1 versus 5.11) at nominal prices. Risk of peanut v/c ratios being less than I is 53 percent under all price assumptions, clearly a serious impediment to increasing fertilizer consumption. Factoring in returns to peanut hay reduces the probability of ratios less than one by only 3 percent at real prices. The data do not support the hypothe- sis that hay responds well to fertilizer even in years when peanut response is low. Increased hay production due to fertilizer had a value (at 1987 prices) greater than 10,000 FCFA only 27 percent of the time. In seven of the eighteen years fertilized peanut yields were less than unfertilized yields; but in only one of these seven years was lIn splitting the AF sorghum data set into two parts (1965-73 and 1974- 82) we did find a significant difference in the fertilizer coefficients with the more recent years exhibiting a smaller response. Even with the smaller response associated with more recent years, the v/c ratios remained greater than 2. 149 there a compensatory (i.e., relatively high) hay response to fertiliz- er. In five of the eighteen years fertilized hay yields were less than unfertilized yields -- four of these five years coincided with very low (or negative) peanut response to fertilizer. Boulel millet performs better than peanuts, but less well than Nioro sorghum. The average v/c ratio at historical prices was 6.7. A We less than 1 occurred only 10 percent of the time under all price assumptions. Using I987 official and real prices, we can anticipate a We ratio less than two 20 percent of the time. A drop in producer price from 70 to 40 CFA/kilo, however, has an extremely depressive effect on v/c ratios, raising the probability of a v/c less than 2 to 40 percent at subsidized prices and 70 percent at real prices. Table 19 presented a similar analysis of the distribution of v/c ratios for peanuts using 15-19 years of data from IRHO confirmation trials begun in the 1950s. It should be noted that fertilizer doses were similar but not identical and that IRHO employed a 2-year fallow in lieu of the 1-year AF fallow. Our analysis shows a much greater probability of V/c ratios less than two for Nioro peanuts (34 percent versus zero percent at subsidized prices and 39 percent versus 26 percent at real prices). Boulel peanut results are less favorable than those previously reported for the North: 64 percent probability of a We less than 2 versus 37 percent at subsidized prices; 69 percent versus 62 percent at real prices. 3. DECISION ANALYSIS Using the actual data from AF yields, we have developed a set of decision models which use the probability of obtaining different 150 fertilizer responses to estimate the expected net benefits (ENB) of alternative investments. Models of this type are frequently used to help farmers examine the consequences of risk and uncertainty associ— ated with agricultural investments. The analysis can also assist in the design of policies dealing with the use of risky inputs such as fertilizer. - We examine the tradeoffs between selected combinations of five specific alternatives: (1) Plant cereal without fertilizer; (2) Plant peanuts without fertilizer; (3) Plant cereal with "themes légers" fertilizer; (4) Plant peanuts with ”themes légers" fertilizer; (5) Invest the cost of peanut fertilizer in additional peanut seed rather than in fertilizer. Tables 34 and 35 illustrate how ENB were calculated for each alternative in Nioro and Boulel, respectively. A number of assumptions made in developing the models must be clarified before proceeding with a discussion of the analysis. (I) The probabilities for good, average, and bad years are based on fertilizer response, not on total yield. In some models, expected yield is greater in an "average" year than in a "good" year; this is because the fertilizer response in the ”average" year was lower than in the "good" year, but the overall yield was greater. The rules used for designating fertilizer_rg§pgn§g as good, average, or bad were: (a) Peanuts: >300 kilos/hectare - good >175 kilos/hectare - average 151 Table 34 Expected Net Benefit Calculation of Five Investment Alternatives Nioro State of Prob- Expected Net Expected CHOICES: Nature ability Yield Benefit Net Ben Plant Good .56 1001 69,720 Sorghum Average .33 1113 77,560 69,789 V/O Fert. Bad .11 674 46,830 Invest Good .56 2442 158,275 In Sorghum Average .33 1756 110,255 131,564 Fertilizer Bad .11 1031 59,535 Plant Good .61 1712 140,580 Peanuts Average .06 1837 151,830 128,336 V/O Fert. Bad .33 1277 101,430 Invest Good .61 2305 182,070 In Peanut Average .06 2075 161,370 151,752 Fertilizer Bad .33 1326 93,960 Invest Good .61 3201 -258,798 In Peanut Average .06 3435 279,858 235,916 Seed Bad .33 2388 186,628 Source: Estimated using AF data. 152 Table 35 Expected Net Benefit Calculations For Five Investment Alternatives Boulel State of Prob- Expected Net Expected CHOICES: Nature ability Yield Benefit Net Ben Plant Good .50 707 49,140 Millet Average .40 581 40,320 42,343 V/O Fert. Bad .10 240 16,450 Invest Good .50 1572 97.375 In Millet Average .40 997 57.125 72,161 Fertilizer Bad .10 270 6,235 Plant Good .32 954 72,360 Peanuts Average .15 1178 92,520 70,280 H/O Fert. Bad .53 847 62,730 Invest Good .32 1595 118,170 In Peanut Average .15 1394 100,080 78,775 Fertilizer Bad .53 826 48,960 Invest Good .32 1784 131,268 In Peanut Average .15 2203 168,978 127,385 Seed Bad .53 1584 113,268 Source: Estimated using AF data. 153 $175 kilos/hectare - bad (b) Millet: >500 kilos/hectare - good >225 kilos/hectare 5225 kilos/hectare - bad (c) Sorghum: >750 kilos/hectare >500 kilos/hectare 5500 kilos/hectare - bad average good average The expected yield for each category is a simple average of estimated yield for all observations assigned to that category. (3) All fertilizer alternatives are based on the ”themes légers" recommendations of 150 kilos of 6-20-10 for a hectare of peanuts and 150 kilos of 14-7-7 for cereals. (4) Net benefits are net of both seed and fertilizer costs. Per hectare costs are 350 FCFA fOr cereal seeds and 13,500 FCFA for pea- nuts. Fertilizer prices are 1987 real prices (11,880 FCFA for a peanut treatment and 12,315 FCFA for cereals). We consider only real prices because the subsidy is scheduled to be phased out by the end of the 1988/89 campaign and because we believe analysis with real prices is more appropriate for policy analysis. (5) Producer prices are the official 1987 prices: 90 FCFA/kilo for peanuts and 70 FCFA/kilo for cereals. (6) Additional seed cost for the "investment in peanut seed" option is based on the assumption that a farmer uses the 11,880 FCFA cost of fertilizer to purchase peanut seed at 200 FCFA/kilo of shelled seed. The 11,880 FCFA purchases 59.4 kilos which plants .87 hectares. 154 (7) The net benefit for the "invest in peanut seed" option is net of seed, traction, and equipment costs for the additional ,87 hggtgrgs as well as seed costs for the initial hectare of unfertilized peanuts. The opportunity cost for the additional land and labor is not directly factored into the analysis. Using the information presented in Tables 34 and 35, we examine five specific investment decisions which Sine Saloum farmers regularly confront. In general, we favor a decision rule which opts for the greatest ENB; however, we do discuss how factors such as attitudes toward risk, labor availability, and access to land might influence a farmer’s investment choice. (A) DECISION MAKING IN NIORO DECISION 1: Plant sorghum with or without fertilizer. The choice here is clear; a farmer using fertilizer can anticipate a net benefit 61,775 FCFA greater than his counterpart planting without fertilizer. Even in bad years, the fertilizer user averages greater net benefits than the non-user (59,535 FCFA versus 46,830). DECISION 2: Plant peanuts with or without fertilizer. The best choice is less clear here. ENB with fertilizer are approximately 20,000 FCFA greater than without fertilizer; but one year in three the fertilizer user will earn 7,470 FCFA less than the non- user. Given the relatively small additional net benefit, a risk-averse farmer may prefer to forgo peanut fertilizer. DECISION 3: Fertilize peanuts or fertilize sorghum. There are two ways of looking at this choice. The first is a choice between one hectare of fertilized peanuts with an ENB of 151,752 155 FCFA and one of fertilized sorghum with an ENB of 131,1564 FCFA. A farmer seeking to maximize income per hectare would plant fertilized peanuts as the ENB is about 20,000 FCFA greater. The second way of looking at the same issue -- fertilized peanuts versus fertilized sorghum -- is the case where a farmer has one hec- tare of each crop and only enough resources to fertilize one or the other crop. If he fertilizes peanuts, total revenue for 2 hectares is 221,541 FCFA. Fertilizing sorghum, however, nets 259,900 -- 38,359 FCFA more ENB for the two hectares. He believe the second way of assessing cereal versus peanut fertilizer better reflects the type of decision-making processes used by farmers. Virtually all farmers plant both cereals and peanuts, and few have resources to fertilizer all cultivated area. Furthermore, recent data on fertilizer acquisition and use presented in the next chapter suggests a growing trend to use limited quantities of fertilizer on cereals. DECISION 4: Plant unfertilized peanuts or unfertilized sorghum. A hectare of unfertilized peanuts still provides almost two times as much revenue as 6 hectare of unfertilized sorghum (128,336 versus 69,789 FCFA). This is a striking illustration of how unrealistic Senegal’s current food self-sufficiency goals are. Given existing production possibilities and price relationships, Sine Saloum farmers are not likely to increase cereal production at the expense of peanuts. The most probable scenario would be to produce family cereal needs on less land by applying fertilizer and use the freed-up land to produce more peanuts. 156 DECISION 5: Invest in peanut fertilizer or use the cost of fer- tilizer to purchase more peanut seed. A farmer opting for extensive cultivation by purchasing peanut seed and expanding peanut area by .87 hectares can anticipate 84,164 FCFA more net benefit than one investing in peanut fertilizer. This additional revenue, however, represents returns to land and additional labor. The extensive option brings only 126,158 FCFA versus 151,752 FCFA/hectare for the fertilizer option -- a 20 percent drop in returns to land. Using a rough approximation of 600 hours of labor/hectare,1 the extensive option brings in 210 FCFA/hour while the fertilizer option realizes 253 FCFA/hour -- 'hi both cases returns per hour are greater than the current minimum wage of 158.6 FCFA/hour which a farmer might be able to earn if he sought urban employment. The choice of fertilizer or peanut seed clearly depends on a farmer’s access to land and labor. Many Sine Saloum farmers have exhibited a strong preference for extensive peanut production -- reduc- ing fallows and cultivating increasingly marginal lands. One would expect fertilizer use to increase as land pressure becomes greater. A mitigating factor, however, is the limited employment alternatives for rural youth. As population increases and non—agricultural pursuits remain limited, household-heads are under growing pressure to provide land and peanut seed to dependents. Traditional forms of labor con- tracting, remuneration of family labor, land allocation, and agricul- tural income distribution do not encourage intensification of 1The estimate of labor hours is based on information provided by Raymond, Monnier, and Cadot. 157 production through fertilizer use, even though it is the more profit- able option.1 (B) DECISION MAKING IN BOULEL The preferred options are the same in Boulel as in Nioro when examined strictly from a perspective of maximizing ENB. In a number of cases, however, the risk factor is much greater and the difference in benefits much smaller than in Nioro, making it unlikely that risk- averse farmers would prefer the revenue maximizing choice. DECISION 1: Plant millet with or without fertilizer. The fertilized cereal option promises 29,818 FCFA more revenue than the non-fertilizer option with a 10 percent chance of the fertil- izer user earning 10,000 FCFA less than the non—user in a given year. Most farmers would be inclined to take the fertilizer option. DECISION 2: Plant peanuts with or without fertilizer. Even though the ENB for fertilized peanuts is 8,495 FCFA greater than for unfertilized peanuts, the 53 percent probability of the fer- tilizer user earning 13,770 FCFA less than the non-user makes the fertilizer investment relatively unattractive, particularly fOr risk- averse farmers. DECISION 3: Fertilize peanuts or fertilize millet. (a) One hectare of either fertilized peanuts or fertilized millet. The ENB of the peanuts is only 6,614 FCFA greater than that for a hectare of millet. There is a 53 percent chance of peanut revenues being less than 48,960 FCFA versus a 90 percent chance that millet revenues will be better than 57,125 FCFA. Farmers trying to reduce the 1This topic is discussed again in Chapters VII and VIII. 158 frequency of low revenue years might be inclined to opt for millet despite the slightly lower ENB. (Clearly, if the millet price falls below the official 7O FCFA/kilo price, as it has in recent years, peanuts become much more attractive.) (b) Grow one hectare of millet and one of peanuts; fertilize only one crop. The Boulel farmer, like his Nioroicounterpart, will do better (21,323 FCFA more net revenue) if he fertilizes his cereal crop and grows peanuts without fertilizer. DECISION 4: Plant unfertilized peanuts or unfertilized millet. As in Nioro, without fertilizer a hectare of peanuts is clearly more profitable than 6 hectare of millet (27,937 FCFA extra benefit). DECISION 5: Invest in peanut fertilizer or use the cost of fer- tilizer to purchase more peanut seed. The extensive peanut option returns 68,120 FCFA/hectare versus 78,775 FCFA for the fertilizer option -- a 16 percent lower return to land. Returns to labor for both options are considerably lower than in Nioro: 131 FCFA/hour using fertilizer and 114 FCFA/hour without. In both cases the per hour returns are considerably lower than the minimum wage in urban areas. As in Nioro, a farmer's Choice between peanut seed and fertilizer would depend on access to land, labor, and -- given the low per hour returns to labor -- opportunities for non-agricultural revenues. 0. EVALUATION OF THE AF DATA SETS Our general conclusion is that the AF data sets can provide a wealth of much needed information about fertilizer response. There are 159 those who will not be content with this analysis and the conclusions drawn about economic returns to fertilizer because AF data does not come from trials or demonstrations conducted on farmers’ fields. It is our feeling that the AF data provides a picture of fertilizer response which is intermediate between strictly controlled on-station trials and highly variable on-farm demonstrations or survey data. Data from on- farm trials are much more expensive to collect and tend to exhibit so much variability that statistically verifiable conclusions are extreme- ly difficult to draw.1 On the other hand, the on-station results reflect production possibilities which most farmers can never hope to attain.2 The AF results are obtained under conditions more favorable than those experienced by many farmers but they are conditions to which most farmers can realistically aspire. As the above analysis Shows, the data contain more unexplained variability than one would like; however, it was still possible to obtain some statistically meaningful results. The length of time covered as well as the geographic and crop diversity represented in the data are valuable resources which warrant more attention then they have been given in the past.3 1Pocthier, Pieri, and Mara (1977) present a good discussion of this problem in their analysis of "Régénération deS Sols' data from farmer managed trials. 2This can even be true of on-farm trials. The IFDC/SODEVA study was conducted on farmers’ fields but over 50 percent of the millet fields received heavy doses of insecticide (IFDC, 1980). This is a luxury which few farmers can afford. 3Pocthier (1983) presents a general analysis of 1976-82 AF data in his report on the 1982 campaign. No analysis combining data from the entire program has been found. 160 No data set, however, is without problems -- particularly when one attempts to perform analyses for which the data were not originally .. designed. Limiting the protocol to three levels of fertilizer serious- ly compromised our ability to estimate cross-sectional production functions with reasonable confidence intervals on the fertilizer coef- ficients. This raises the question of whether our method of analysis should be changed or the number of treatments increased. Research by the Kaolack farming systems team using one-half the recommended ”themes légers“ dose has provided interesting results (see footnote on page 143). This suggests that a protocol of F0, 1/2 F1, F1, 1/2 F2, F2 might be a reasonable alternative that responds to both agronomic and economic concerns. AS farmers tend to partially adopt soil preparation themes, adding two intermediate preparation levels may also be appro- priate. Hhile this would increase the number of treatments, the bene- fits could far outweigh the costs if we obtained data permitting more accurate estimation of production functions while also being more representative of current farming practices. Our decision to use actual data rather than cross-sectional re- gression results in the economic analysis of inter-annual variability provides us with some idea of fertilizer response across time; however, there is no statistical justification for generalizing these results to a larger population. Furthermore, certain of the average v/c ratios estimated in this fashion were quite different from those Obtained with 161 the pooled regression analysis.1 This poses the dilemma of deciding which of the analyses provides a more reliable picture. The problem of multicollinearity which precluded the estimation of separate N, P, and K effects has already been mentioned. The large number of treatments one would have to add to resolve this problem leads us to conclude that examination of separate effects Should be done in conjunction with multi-rate trials on smaller plots, not as part of demonstration trials conducted on 400 square meter fields. Use of a 400 square meter plot is considered by most agronomists to be an important step in approximating farmers’ yields. On smaller plots, cultivation activities such as weeding and thinning are not performed in the same manner as a farmer would normally perform them. Estimation of joint-NPKSUM effects will not help agronomists identify optimal N-P- K combinations, but it does provide sufficient information about the generally recommended formulas to perform useful economic analyses. A final question which needs to be raised is the wisdom of running the AF program without repetitions. Opinions are varied. Some believe that the size of the plot compensates for the lack of repetitions, but others would still prefer to see at least one repetition. Most thought that decreasing the Size of the plot to less than 400 square meters would compromise the objective of simulating actual farm conditions. 1The most striking example is millet in Boulel. The average value/cost ratio obtained from the pooled analysis for F1 versus F0 was only 1.43 (Table 31) while that obtained by calculating the v/c directly from the data was 3.5 (Table 33). The difference was also large in the case of Nioro sorghum (3.46 versus 6); as both Nioro ratios are significantly greater than 2, the problem of interpretation is less critical than in the Boulel case. Peanut value/cost ratios obtained from the two meth- ods were relatively close. 162 From the perspective of our own analysis, we would like to find a way of decreasing the large confidence intervals on the yield estimates. If doubling the number of observations could significantly reduce these intervals, it would be a useful modification to consider. To increase repetitions and maintain the same surface areas, however, would double the costs of the program; the benefits would have to be substantial to justify this. E. SUMMARY OF IMPORTANT POINTS Before moving on to a discussion of farmers’ decision making with respect to acquisition and use of fertilizer, it is helpful to identify what we consider the most important points made in the review of eco- nomic analyses and presentation of AF results. (1) There have always been zones and crops which consistently exhibited v/C ratios less than two at real prices, even though past analyses have Shown fertilizer V/c ratios to be generally favorable. (2) Continued use of fertilizer when v/c ratios were less than two has been justified on the grounds that a disastrous decline in soil quality would ensue if fertilizer were not used; . (3) Lack of precise knowledge about the cumulative yield effect of long-term cultivation without fertilizer precludes economic analysis that incorporates the costs and benefits of maintaining soil capital; (4) Current analysis of AF data suggests that v/c ratios are now less than 2 in zones that were previously thought to exhibit profitable responses; 163 (5) Factoring in the value Of peanut hay production does not substantially improve V/c ratios or reduce risk in zones of poor re- sponse; (6) Important inter-zonal variability exists in fertilizer re- sponse; this variability has been masked in past analyses which aver- aged data from different sites; (7) Inter-annual variability in fertilizer response is very high and often masked by use of V/c ratios averaged over time; (8) Value/cost ratios are good baseline indicators of whether fertilizer should be recommended to farmers, but they are poor indica- tors of potential fertilizer demand, because they do not provide any information about profitability of fertilizer relative to other investments. These findings have important implications for the Idesign Of fertilizer policy. Many of the policy recommendations which could be made on the basis of information presented in Chapters IV and V alone, are made much stronger by examining survey data on farmers’ attitudes about fertilizer, their past experiences using the input, and how they make input investment decisions. For this reason, we proceed directly to a discussion of our farm-level studies, reserving the discussion of policy options for the concluding chapter of the dissertation. VI. SURVEY DESIGN, IMPLEMENTATION, AND ANALYSIS A. DESIGN The next two chapters of this dissertation draw on data from a series of formal and informal surveys designed and implemented in a collaborative manner by the two principal researchers of the ISRA/BAME program entitled "Etude de l’Obtention et de l’Utilisation des Intrants dans le Sine-Saloum'. As such, the survey had to respond to a spectrum of concerns broader than those dictated by the specific objectives of this dissertation. The program was to generate (I) a number of timely, policy oriented ISRA/BAME working papers analyzing input distribution issues associated with the NPA; (2) a doctoral dissertation for the principal researcher on farmer decision making and input investment behavior, with emphasis on fertilizer; and (3) a research paper on the topic of peanut seed policy under the NPA for the second researcher (the paper would be used to determine the researcher’s eligibility for a permanent researcher position at ISRA). Before the design process was completed, interest by USAID and their willingness to fund the fieldwork increased reporting require- ments and broadened somewhat the scope of research. USAID was specifi- cally interested in information on the performance of input distribu- tion systems and farmers’ response to recent policy changes; they also demanded faster analysis and reporting of this information than origin- ally planned. I. SAMPLING PROCEDURES Funding and time constraints rendered the objective of establish- ing our own sampling frame impractical. The best alternative was to 164 165 coordinate our work with that of SODEVA who had been regularly collect- ing input/output data on a sample of Sine Saloum farmers Since 1981. SODEVA agreed to give us full access to their l981/82-1984/85 question- naires. This collaboration offered several advantages: 1) He would be spared the time and expense involved in selecting our own sample; 2) He would have access to longitudinal input/output data and general descriptive information for farmers in our sample; 3) We could foster collaboration between research and extension services, a collaboration which is frequently demanded but seldom in evidence; These advantages outweighed the disadvantages which were primarily the difficulty of working with unfamiliar data of an unknown quality which many of our ISRA colleagues considered to be dubious at best. The sampling method used by SODEVA is described in Gazagnes and D’Hiver (1978). It is based on a system of farm1 classification which was developed fer the Senegalese Peanut Basin by the authors. Using factor analysis, the authors boiled down a myriad of farm characteris- tics to two basic indicators which were used to assign all farms to one of five “groupes typologiques” (GT). The typology is based on total hectares cultivated and the number of hectares cultivated per active worker. The five groups are: 1N6 use the word farm to represent the Senegalese terminology "exploi- tation" which is the basic production unit in the Peanut Basin. See Benoit-Cattin and Faye (1982) for a discussion of the "exploitation" and why it is a more appropriate unit of analysis than the "concession" or the "ménage". 166 Group Cultivated Hectares Active Horkers/Hectare I < 6 < 1.5 II . < 6 2 1.5 III 2 6 and < 17 < 2 IV 2 6 and < 17 z 2 V 2 17 no distinction The original SODEVA sample drawn in 1980 was based on a census of each Department and reflected the relative importance of each GT in each department. Over time purposive changes were made in the sample to rectify identified shortcomings or to accommodate changes in SODEVA extension programs. Non-programmed changes were made -- usually by interviewers -- to compensate for unCOOperative interviewees, deaths, or departures. SODEVA personnel assured us that the underlying struc- ture of the sample was not distorted by these changes. The criteria used for selecting a subset of the Sine Saloum sample were: I) Desire to study at least two zones characterized by different (a) ecological conditions (b) peanut varieties (an important issue in the seed research) and (c) ease of access to fertilizer; 2) Desire to have 4 yearS’ longitudinal data on as many farms as possible; 3) The feasibility of covering all farmers with a limited number of interviewers and mobylettes. The samples for the Departments of Gossas and Nioro met these criteria. Ecologically, Gossas has lower rainfall, a shorter growing season, and sandier soils than Nioro. These conditions limit Gossas farmers to millet, oil peanuts, and some cowpea production while their Nioro counterparts produce millet, sorghum, maize, oil and 167 confectionery peanuts, and cotton. Oxen traction is relatively more important in Nioro than Gossas. Livestock revenues account for a much larger portion of total farm income in Gossas than in Nioro. Given the relatively better rainfall in Nioro during the recent past, government policy has favored this zone over Gossas, providing the former with a larger share of officially distributed seed and fertilizer. Nioro has also benefited from low cost contraband fertilizer smuggled from neigh- boring Gambia. Figure 12 Shows the locations Of the Nioro and Gossas villages where the respondents in the 98-farmer sample lived.l Due to funding problems, the 1984/85 SODEVA sample had been com- pressed, hence we used the larger sample from I983/84 as a base. This sample included 60 Nioro farmers (20 with data over 4 years) and 45 Gossas farmers (24 with 4 years’ data). Only 98 of the original 105 were still available in 1985. It made no sense to replace these farm- ers as none of the longitudinal data would have been available. The Gossas sample only lost one farmer. The Six lost in Nioro pose some problems for generalization of results as four of the six were from small, marginal farms (GT 11). The losses were due to death, departure from the village, or abandonment of farming. Much of our analysis and description paints a gloomy picture of recent agricultural productiv- ity. Since the farmers dropped from the original sample were generally less productive, this gloomy picture cannot be attributed to departures from the original sample. 1Figure I, presented in Chapter I, was a map of Senegal which showed the Sine Saloum and the Departments of Gossas and Nioro. 168 capoOOA use mcoscau o—asom mews: 2...! 333.5. £63 2.8.. 6.23.528. . 3.3 .53. use... aum< gummc Lang neaaan-qaw-p nip-a Jugs: Pica-s a: .33: 2...... 6 2e33- 265 . «lg—Lang moau__w> oco_z can mummou to ac: ”NH ocamwm ago—n no wen-aL-doa acumen ea “ea-accuse 169 Table 36 provides descriptive statistics on the 98 farmers retain- ed for the general survey. The statistics are based on the 2 to 4 years’ of SODEVA data available for each farmer. A chi-square test was performed to see if there was a significant difference between Nioro and Gossas fermers with respect to these five characteristics. The only significant finding (.005) was that Gossas farmers tend to have more cultivated area per active worker than those in Nioro. Trying to design agricultural surveys based on sampling frames which justify generalization of results to a larger population is a notoriously difficult task in countries such as Senegal. We believe that the SODEVA effort was based on carefully thought out procedures which were subsequently modified to meet the realities of implementing field research on limited budgets. We are confident that results based on the survey of 98 farmers are a fair representation of agricultural performance and farmer attitudes in the zones of Gossas and Nioro. Generalization beyond these two zones is tempting, but not justified. 2. QUESTIONNAIRE DESIGN AND SUB-SAMPLES Given the multiple objectives of the research program, several levels of survey were planned for various subsets of the sample. An interviewer-conducted general survey (see Appendix III) of the full 98- farmer sample collected the following information: a) Farmers’ perceptions of their major constraints b) Farmers’ opinions about recent policy changes c) Input acquisition and use for the l985/86 campaign d) 1980-85 changes in farm assets and liabilities (i. e., recourse to credit, disinvestment, dissavings) 170 Table 36 Characteristics of Respondents in 98-Farmer Sample SAMPLE CHARACTERISTIC AVERAGE MEDIAN MINIMUM MAXIMUM Age of principal farmer 49 49 21 79 Total number of persons/farm 10.2 8.5 2.33 37.33 Number of active workers/farma 4.79 4.3 .87 19.6 Cultivated hectares/farm 9.48 7.67 .77 39.86 Hectares cultivated/active worker 2.02 1.9 .38 5.11 Source: SODEVA data. aISRA norms used; male 15-59 years - 1; female 15-59 = .5; male 8-14 - .5; female 8-14 . .25; all others = 0. 171 e) Access to non-agricultural revenues f) Farmers’ knowledge and attitudes about fertilizer use 9) Farmers’ anticipated behavior given a number of hypothetical input investment scenarios Another interviewer—conducted survey about seed issues was administered to a subset of farmers who had been interviewed by SODEVA in both 1983/84 and I984/85 (two years where seed policy was rapidly changing). Finally, researcher-administered informal interviews with 46 farmers provided detailed information about factors influencing fertilizer and agricultural equipment investment decisions. This subset consisted primarily of farmers for whom 4 years of data were available; excep- tions were made, however, to increase sample size and include more fertilizer users than were contained in the 4-year group. In selecting subsets for' more' detailed study' of 'farmer input acquisition, we departed from the rules of random sampling to make use of available longitudinal data. Although technically we cannot claim the samples used are representative of the general population, a com- parison of descriptive statistics for the 98-farmer sample, the strict 4-year sample of 41 farms, and the modified 4-year sample used for the informal interviews reveals no serious over- or under- representation of particular groups in the subsets analyzed for this dissertation. He employed a variety of questionnaire techniques in both the formal and informal surveys. Straight-forward factual questions pro- vided information to complement and verify SODEVA data. Both open- ended and multiple-choice questions elicited farmers’ opinions about recent changes in agricultural policy. Repertory grids, described in 172 Franzel (1983), helped to organize information about farmers’ percep— tions of the relative advantages and disadvantages of various soil renewal techniques. Questions designed to understand farmers’ deci- sion-making heuristics were based on previous work with hierarchical decision models reported by Franzel (1983) and Gladwin (1976). Farm- ers’ risk attitudes were studied using a technique employed in India by Binswanger (see Binswanger 1980). B. IMPLEMENTATION Funding problems delayed implementation of the general survey until August, providing an unusually long period for questionnaire and informal interview design and testing. During this time, SODEVA data were coded and analyzed. Preliminary analysis provided a good picture of the types of farms we would be studying and signaled a number of interesting cases which were later examined in-depth. The analysis also revealed some weak spots in SODEVA’S data that we compensated for in our own surveys. The general survey was conducted in August and September by four interviewers who were required to have a minimum level of education at the DEF level (9 years of fermal schooling) and prior experience in rural surveys. The seed survey was conducted concurrently. Only one visit per farmer had been planned. In most cases, however, a follow-up Visit was required to resolve internal inconsistencies in the question- naires or blatant discrepancies between farmers’ responses to our questionnaires and previously recorded SODEVA data. The second Visit resolved many but not all of these problems, bringing us to realize that the SODEVA longitudinal data could be a curse as well as a 173 blessing. The question of what to do with unresolved differences was vexing; should we rely on SODEVA data and assume that farmer recall was faulty or our interviewers less adept than their SODEVA counter- parts? Ultimately, the decision was made on a case by case basis using information obtained by the principal researchers during the informal interviews and in discussions with SODEVA personnel familiar with the relative reliability of data collected by different interviewers. In some cases we decided not to analyze particular sets of data because we were uncertain of the reliability. Design of informal interviews was the responsibility of the prin- cipal researcher. Although all interviews addressed an established set of themes, each one was tailor-made to fit the particular respondent’s situation as revealed in SODEVA data and responses to the general survey. The probing nature of the questions and underlying objective of understanding what type of economic reasoning farmers use when making input investment decisions placed heavy demands on the language skills of the second researcher who served as interpreter as well as economist colleague. Although only one interview was planned per farmer, it quickly became apparent that two interviews were necessary if both fertilizer and equipment issues were to be covered. The two-interview procedure also had the added benefit of permitting internal checks on consistency of farmers’ responses. Many of our questions were hypothetical ques- tions about what the farmer would do in particular circumstances. AS Figure 13 suggests, some individuals believe that hypothetical ques- tions should get hypothetical answers -- the two-interview procedure IADH)Y'(D\P1’ 174 Hypothetical Questions Deserve Hypothetical Answers Figure 13: I75 helped us to know the farmers better and to assess the reliability of answers received. C. ANALYSIS The analysis presented in this dissertation is a refinement and extension of analyses already presented in USAID and ISRA reports (Kelly and Gaye, 1985; Kelly, 1986a and 1986b). Some Of the findings vary from those previously reported because more time has been devoted to cleaning and verifying the data. The major additions are: (I) the use of logit analysis to provide a better picture of what distinguishes fertilizer users from non-users and (2) a more thorough development of hierarchical decision models. 0. ORGANIZATION OF THE DISCUSSION The presentation of survey results is divided into two chapters. Chapter VII is essentially a descriptive Chapter. It presents a variety of information about factors which influence farmers’ input purchasing behavior and concludes with a description of fertilizer purchases actually made from 1981 through 1985. Chapter VIII is a more analytical chapter, using information presented in Chapter VII to develop logit and hierarchical decision models that help us separate the more important determinants of fertilizer demand from the less important. VII. FACTORS VHICH INFLUENCED l981-85 FERTILIZER PURCHASES AND USE This chapter is divided into five sections, the first four discuss the principal factors that influenced fertilizer purchases in the early 19805, and the fifth is a description of actual purchases made during the 1981-85 period. The four factors influencing fertilizer purchases are: (1) General agricultural productivity and economic climate in the zone under study; (2) Farmers’ perceptions about fertilizer and aJternative soil renewal techniques; (3) Types of economic analyses performed by farmers; (4) Farmers’ investment priorities. The chapter draws on information from (I) the SODEVA data base for 41 farmers who were interviewed consistently every year between 1981/82 and 1984/85, (2) the full survey of 98 farmers, and (3) in-depth inter- views with 46 farmers. Only the 98-farmer sample meets statistical criteria permitting generalization to a larger population. The 41- farmer SODEVA data set appears to have the same general farm character- istics as the larger sample but it was not randomly selected. The smaller data set from in-depth interviews contains more variables, is coded more consistently from farmer to farmer, and is generally more reliable data than our 98-farmer survey data; the smaller sample, however, is not completely random. A. AGRICULTURAL PRODUCTIVITY AND ECONOMIC CLIMATE To interpret the information on fertilizer purchases which we present in this Chapter, it is necessary to have some understanding of 176 177 the general agricultural and economic context in which these purchases took place. SODEVA data make it possible to show trends in agricul- tural productivity between 1981 and 1985 for 41 farmers in Gossas and Nioro. Data on food shortfalls, dissaving, disinvestment, and credit Obtained from our 98-farmer survey complement the SODEVA data providing a picture of the general economic climate during the early I980s. 1. AGRICULTURAL PRODUCTIVITY Figure 14 graphs inter-annual changes in peanut and millet produc- tion reflected in the SODEVA data. Table 37 illustrates inter-annual changes in key input/output relationships. The data reveal a serious decline in productivity. The most striking features of the decline are: (1) Extremely low yields in 1983/84-84/85 (a) Millet production/person was less than the FAO consump- tion norm of 200 kilos/person in 3 of the 4 years (b) ‘The value of peanut production/active worker was less than 30,000 per year in 1983/84 and 1984/851 (2) Very low fertilizer use, particularly in Gossas2 (3) A sharp decline in 1984 Gossas peanut plantings. 1The combined value of peanut and millet production per active worker was about $150 over the four-year period; this was significantly less than the estimated $370 per_gapita 1985 income for Senegal. 2As will be seen later, our surveys indicate that farmers did use more fertilizer than reported by SODEVA. It appears that SODEVA inter- viewers tended to record only fertilizer from the official distribu- tion, missing most of the informal purchases which were made at various times throughout the season. 178 paanpOJd 1110116105 )0 suol paanpOId smuoad 10 6001 22:3. 56:90. 33 333:: ecu .635: $2.3: ecu gamma 9:32 182mg tome nmmp N009 ramp 0 1 11 1 1| 0 mend. ON 33 ........................... $3.593“ v0 ICOP ON . noodn .. can cu av .3 3 .. 0.. \I/U“O.OV 9. on 33 / .3 on on uoacoi 0050.00... NO on f i a. 003 .3 .3 no. on. f .3 ON. O: 33 33 o: of 33 .. cm. on. 3. 33 on. OOu.. ..oou smmr nmmp Nmmp romp O - 1 1 o on .3 ans“ 33 on as 3 ........ 3300 neodn 00 e 50 usuau.0...l acor no 33 on g /l 6.. u on . 63:0... .1306: No“ on 306 00.. .3 .. no 33 00. on ONF % .. .. .3 ON, a: . Knew. ..o: 00.. .3 .. on. on, 3 33 on. oou 33 .. oou AmEcou m C oco_Z paiuold u1n1161os )0 661013611 paiuold smuoad 1o 9101:1311 99909941 1811111 10 $1101 paonpOId sgnuoad 10 81101 "cg 0L:m_m .6666 <>ucom ”ouaaom tmmp nmmp Nmmp Pmmp o o 11 O hand. on 33 ........... 3 ON 0.. 33 n; on nun.vn....... .3 ov H 36...: .o 2.6» a On 33 .. .3 on n... no.3 m on A. .. on a 0 I, no; 3: .300. W. N. on. 33 .\\//_ an. .391.- 1 uoaco_l\:coauux /\\1\1\\1 d 0.... 33 _ \ an. to: w 2.... 184. D: 00.33 3306. 00.33 300. 091.33 33cc“ ¢mmw nmmr Nmmp F009 0 - 1 1 - o on 3 39“ «nude 0' 3. ............ 3 CV man . .. P 06 33 tn 3 .3 on m a . S on 3. auacoan. .o....nc0h .3 00 mm. 00' LT ”OP ...4 'OF.'°P Av 00' ad our .7 03:20. «10.001 ..x. 3ON— 151 01. .. / . 330.; H . 0 [DDR3 .1431 00.. 3 an. 3m, 3306. W 003.3 .00. 0913 ..oou Amrccou an mommoo 179 Tab1e 37 Inter-Annual Changes in Key Productivity Measures for 41 Farms 1981/82 through 1984/85 G 0 S S A S N I 0 R 0 (23 Farms) (18 Farms) Productivity Measure 1981 1982 1983 1984 1981 1982 1983 1984 P Kilos/hectare 1002 930 311 373 1315 1436 278 596 E Kilos/kilo of seed 15.5 14 4.5 NotAv 21 20 4 NotAv A Kilos/active N worker 960 1262 422 345 1145 1536 342 418 U Kilos of fertil- izer/hectare 0 0 0 0 64 4 1.5 11 T M Kilos/hectare 507 267 175 208 656 462 346 421 I Kilos/active L worker 571 309 174 244 513 253 189 274a L Kilos/person 275 147 84 120 243 125 94 134a E Ki]os of fertil- izer/hectare 8.5 0 0 .45 65 4.5 2.5 3.5 T Source: Ana1ysis of SODEVA data. aThe 1981-84 farm population was fairly stab1e; however, the Nioro group had a 12 percent increase in active workers and a 16 percent increase in number of persons per farm between 1983 and 1984. This increase was accompanied by a decrease in cu1tivated area. 180 2. ECONOMIC CLIMATE Low agricultural productivity was translated into repeated food shortfalls which had serious repercussions on farm debt and equity. Ninety-four percent of the 98 farmers surveyed in 1985/86 claimed that their cereal production was insufficient for their personal consumption needs at least one of the previous five years; more than 50 percent had two or three deficit years. Only 9 percent were able to compensate fully with government food aid. Among the others, 13 percent sold peanut seed 24 percent sold one or more agricultural implements1 36 percent sold one or more traction animals2 71 percent sold small ruminants 31 percent sold beef cattle The five different types of sales were consistently made by a greater percent of the Gossas than the Nioro sample, but only the case of small ruminants was statistically significant.3 Food shortages increased indebtedness for the 9 percent who ob- tained credit from local traders and the 32 percent who borrowed from relatives. An additional 9 percent sought credit but were unable to obtain it. Only one farmer in Gossas managed to feed his family during the 5-year period without disinvestment or recourse to credit while 16 in Nioro claimed to have done so. lFifteen carts and 8 cases each of hoe and seeder sales were reported. 2Twenty-three sold horses, 11 oxen, and 9 donkeys. 3Chi-square test at .05 level of significance. 181 Some farmers also sold personal property to finance seed pur- chases; 9 percent sold traction animals, 8 percent sold cattle, and 5 percent sold small ruminants. Those selling traction animals often marketed a good quality or preferred animal (oxen, horse) in order to buy a less expensive animal (donkey or younger horse/oxen) and used the remaining cash for seed. Credit was also used to increase access to seeds, mostly through official programs.l Although sales of assets and credit were used primarily to remedy food and seed shortages, additional sales took place to meet pressing social obligations. Two percent of farmers sold machinery, 12 percent sold traction animals, and 23 percent marketed other livestock. Oisinvestment and indebtedness of this magnitude jeopardize farm- ers’ capacity for future investment in agricultural inputs, particular- ly non-essential fertilizers. One must ask whether the disinvestment is part of an inevitable evolution in Senegalese agriculture where the small, non-viable farmers will be forced out, or if it is so pervasive that it is significantly reducing the productivity of all classes of farmers. He found no statistically significant, difference between those disinvesting and not disinvesting with respect to general charac- teristics such as number of hectares cultivated, total farm population, active farm population, number of hectares per active worker, and age of the principal farmer. This result implies that the problem is generalized. 1Only 2 percent of farmers obtained seed credit from traders and 6 percent from relatives. 182 8. FARMERS’ PERCEPTIONS OF FERTILIZER This section discusses farmers’ perceptions of: (1) the relative importance of fertilizer problems, (2) appropriate fertilizer applica- tion techniques, (3) the effectiveness of fertilizer compared to other soil renewal techniques, (4) the relative risk of fertilizer versus other techniques, and (5) peanut and millet response to fertilizer. 1. THE RELATIVE IMPORTANCE OF FERTILIZER PROBLEMS Not a single farmer considered a lack of fertilizer to be his main production constraint during the 1981-85 period. Poor rains and lack of peanut seeds were the most frequently cited constraints. On the other hand, 28 percent did rank fertilizer problems among their three major bottlenecks. Farmers who note the absence of fertilizer as a constraint are more likely to have purchased the input during the last five years than those not mentioning a fertilizer problem. Surprisingly, however, farmers with access to organic fertilizers and fallow land are as likely to mention fertilizer constraints as those without such access. 2. APPROPRIATE FERTILIZER APPLICATION TECHNIQUES Many fertilizer policy and research documents suggest that the low demand for fertilizer is linked to Senegalese ‘farmers’ failure to follow recommended practices. It is argued that farmers apply inap- propriate amounts, at incorrect times, on the worst soils. The medio- cre results obtained lead farmers to discount the value of fertilizer and choose other investments. Since most farmers do not follow ISRA’s recommendations (spreading fertilizer before soil preparation and inc to af la $0 on we th 183 incorporating it into the soil with a light plowing), it is important to understand how and why fertilizer is actually applied. Over 50 percent of farmers prefer to spread peanut fertilizer after the first weeding, and millet fertilizer after thinning -- much later than the recommended dates. Actual application dates reported in SODEVA data for Nioro were even later than these expressed preferences; only 20 percent of peanut fields received fertilizer before the first weeding and only 10 percent. of' millet fields were treated before thinning. The timing of fertilizer application is influenced to some extent by labor and equipment constraints, but personal beliefs about agrono- mic relationships and risk avoidance strategies are also important factors which influence application practices. For example, there is a commonly held belief ‘that fertilizer should not. be Tapplied before weeding and thinning because early application "wastes” fertilizer on weeds and millet plants which are later removed. With respect to risk avoidance strategies, we found that most farmers are unwilling to apply fertilizer before plant emergence due to the risk of losing the entire investment from bad germination. Another agricultural practice that may also be responsible for less than optimal returns to fertilizer investments is farmers’ failure to use particular formulas as intended. Farmers know that fertilizer formulas differ for peanuts and millet but consider the two more or less interchangeable and thereby may get less than maximum possible returns (by using too much nitrogen on peanuts or not enough on millet). .Al pori til‘ fre< 3. 184 ALTERNATIVE SOIL RENENAL TECHNIQUES Senegalese farmers have three primary options for soil renewal: (1) Chemical fertilizers (2) 'Fumier': Spreading manure from animals kept near the family compound. This technique requires a great deal of labor for transporting and spreading the manure on selected fields. (3) 'Parcage": Application of manure by cattle "parked" on a field for an extended period of time. This method usually pro- vides more manure per unit of area and is less labor-intensive than 'fumier”. Farmers’ comments concerning these three techniques provide im- portant insights about their criteria for evaluating alternative fer- tilization methods and the factors that inhibit greater use. The most frequently cited observations were: (1) 'Parcage" increases cereal yields more than 'fumier" and fertilizer. (2) 'Parcage” has a four to five year carry-over effect. (3) Chemical fertilizer increases peanut yields more than other techniques. Organic fertilizers tend to increase the yield of hay but not that of peanuts; they also increase pest problems. (4) Horse manure increases ”striga' weeds in Gossas, hence farm- ers in that zone prefer fertilizer to 'fumier” on millet. Nioro farmers are less concerned with “striga' and rank "fumier" higher than fertilizer. (5) Fertilizer reduces "striga' problems. 185 (6) Chemical and organic fertilizers are substitutes, but farmers disagree about the rate of substitutability.1 (7) Labor shortages, lack of transport, and insufficient fodder to keep animals in the compound year round restrict the use of 'fumier'.2 (8) Insufficient pasture near the fields needing manure and fear that animals grazing on distant fields are easy prey for thieves restrict the use of 'parcage'. (9) "Parcage" causes the most crop damage if the rains are not good and is therefore considered the most dangerous technique to use when viewed from the perspective of a single crop year. The overall risk, however, is attenuated by the significant carry-over effect. 4. RISK AND FERTILIZER INVESTMENT The issue of risk and its impact on fertilizer investment is complex. Despite the greater danger of losing one’s crop if rains fail in the year of application, farmers continue to use ”parcage". The high risk of crop loss if rains are poor is compensated for by the carry-over effect which contributes to a good crop in subsequent years. With chemical fertilizer, the initial crop loss may be less but most farmers perceive little carry-over effect. Furthermore, where risk is 1For example, about half of the farmers believe that 50 kilos of chem- ical fertilizer could correctly fertilizer more area than the manure produced annually by one horse, and about half believe the opposite. 2Easy access to chemical fertilizers during the credit program could also be considered a constraint on the use of ”fumier" as 35 percent of farmers admitted that they did not make full use of available manure when they could afford chemical fertilizers. 186 involved farmers follow two different strategies. They are willing to employ fully inputs that have few alternative uses (organic fertil- izers, family labor or animal traction) in the hape that Allah will bring the necessary rains. When a cash investment such as fertilizer is involved, however, farmers more cautiously consider the consequences of poor rains since cash can be used to purchase food if harvests are bad. Nith investments like fertilizer where the risk of losing a cash investment is high, a farmer’s decisions are shaped by his personal preferences for risky undertakings. There is an extensive literature on eliciting information about an individual’s risk preference function for use in analysis of investment decisions (see, for example, Anderson, Dillon, and Hardaker, 1977). Models incorporating risk preference information have also been suggested for estimating input demand (see Antel, 1987). Given the importance of understanding how farmers perceive risk and react to it, we used a technique described by Binswanger (1980) to elicit farmers’ risk preferences. The results were generally disappointing. Binswanger’s method has the potential to classify farmers into six different risk preference categories. The best we could do with our results was classify 31 percent of farmers as consistently risk preferring, 18 percent as consistently risk averse, and 51 percent as inconsistent.l 1A short paper describing the problems encountered and evaluating the pros and cons of trying to elicit risk preference functions for Senegalese farmers is planned. We do not pursue the matter any further in this dissertation as it adds nothing to our current understanding of farmers’ fertilizer demand. 187 A second important component of decision making in risky situa- tions is the farmer’s expectation of probable returns to the investment under consideration. The decision analysis presented in Chapter V used observed yields and probabilities from the AF data. In general, it is assumed that farmers are guided in their fertilizer investments by some concept of the added product they will obtain using fertilizer under various states of nature. Ne turn now to a discussion of survey re— sults describing farmers’ perceptions of fertilizer response. 5. FERTILIZER RESPONSE Farmers’ perceptions of fertilizer yield response were elicited because (1) we wanted to compare farmers’ perceptions with results from researcher-controlled trials and demonstrations, and (2) the value/cost ratio used by researchers to assess the economic appeal of fertilizer assumes that farmers have some concept of yield response. Table 38 compares the median1 value of fertilizer yield response reported by farmers with results from our analysis of AF data presented in Chapter V. We expected AF data to show a greater value added than farmers’ perceptions because the AF program followed a strict 4-year rotation with fallow and generally assured that critical cropping activities were performed in a timely manner. The opposite was true -- farmers’ perceptions exceeded research results. Table 38 reveals that farmers’ expectations in Boulel are about 50 kilos/hectare greater for both peanuts and millet. In Nioro, farmers perceived 200 kilos/hectare lMedian values were used because we wanted to reduce the influence of extreme perceptions, which we thought were due more to farmers’ poor grasp of measures used (kilos and hectares) than to actual yield response. 188 Table 38 Farmers’ Perceptions of Peanut and Millet Response to Fertilizer Compared to AF Estimates ADDED ADDED KILOS PER KILOS KILO 0F NUTRIENTa P E A N U T S: NORTHERN SINE SALOUM GOSSAS FARMERS’ PERCEPTIONS 300 5.5 BOULEL lZO-OAY PEANUTS 264 4.9 SOUTHERN SINE SALOUM NIORO FARMERS’ PERCEPTIONS 600 11.1 NIORO lZO-DAY PEANUTS 393 7.2 M I L L E T: NORTHERN SINE SALOUM GOSSAS FARMERS’ PERCEPTIONS 300 7.1 BOULEL MILLET 257 6.1 SOUTHERN SINE SALOUM NIORO FARMERS’ PERCEPTIONS 400 9.5 NO COMPARABLE MILLET DATA AVAILABLE Source: Farmers’ perceptions are from l985/6 survey data for 98- farmer sample; Boulel and Nioro estimates are based on analysis of AF data presented in Chapter V. aFarmers’ perceptions of response were based on a variety of fertilizer application rates ranging from one to four 50-kilo sacks/hectare. The ”added kilos/kilo of nutrient” is a more accurate means of comparing farmers’ perceptions with AF data than the "added kilos" column. Both columns, however, lead to the same conclusion -- farmers’ perceptions of fertilizer response are somewhat greater than AF results in the Northern Sine Saloum and much better in the Southern Sine Saloum. . 189 greater peanut response than that estimated from AF data. Unfortunate- ly, we have no AF millet data to compare with the perceptions of Nioro farmers. Peanut and millet value/cost ratios calculated with farmers’ perceptions and using 1986/87 prices (90 FCFA/kilo for peanuts, 70 FCFA/kilo for millet, and 64 FCFA/kilo for subsidized fertilizer) range from 2.6 to 5.6. In all cases the perceived value/cost ratio exceeded 2 -- the level of economic returns which researchers have consistently maintained is sufficient to incite farmers to invest in fertilizer. The argument that farmers do not use fertilizer because they do not appreciate its yield potential is not supported by these results. Farmers’ perceptions of fertilizer response suggest that net benefits per hectare will be greater if fertilizer is used on peanuts (the peanut response per hectare is equal to or greater than the millet response and peanut prices are higher). Informal discussions with respondents revealed, however, that they prefer to use fertilizer on millet. Three factors seem to explain this preference: (I) It is easier to apply small quantities of fertilizer around selected millet plants than around peanut plants; (2) Given the small difference between peanut and millet prices, the need to assure cereal crops is dominant; and (3) The percentage increase in cereal yields is greater than that for peanuts. This last point requires some elaboration. It is our impression that many farmers consider relative rather than absolute fertilizer response ‘ in deciding which crops receive limited supplies of fertilizer. Gossas 190 farmers perceive a 75 percent and Nioro farmers a 67 percent increase in millet yield compared to only 31 and 46 percent increases in peanut yields. In the minds of many this means that cereals respond better to fertilizer than peanuts, therefore, fertilization of cereals is given priority. Hhatever the underlying reasoning, the decision to fertilize cereals is supported by the analysis of AF data in Chapter V which found that a farmer with insufficient resources to fertilizer all crops would be economically better off if he fertilized cereals rather than peanuts. A word of caution is in order concerning inferences drawn from farmers’ perceptions of fertilizer response. Yield increases reported in Table 38 are median values of highly variable responses. Further- more, the data are based on answers from a small part of our sample; only 50 percent of farmers were able to quantify millet response while 67 percent were able to do so for peanuts. Many farmers interviewed discussed the difficulty of estimating response for an ”average" year. Many were unable to answer due to imprecise recall and a reluctance to analyze agricultural outcomes in probabilistic terms (everything de- pends on Allah). Those who were unable to quantify yield response generally believed that it was positive. 6. SUMMARY OF MOST IMPORTANT PERCEPTIONS (1) Approximately 50 percent of farmers are unable to quantify fertilizer response; farmers who are able to do so believe response is greater than that obtained in AF demonstrations. (2) Value/cost ratios calculated with farmers’ perceptions of 191 yield response are greater than 2, yet these farmers have purchased very little fertilizer in recent years. (3) Farmers’ tendency to give other purchases priority (food, seed, equipment) constrains fertilizer purchases more than a belief that fertilizer response is poor. (4) Farmers prefer to use limited supplies of fertilizer on cereals rather than on peanuts. (5) Risk avoidance strategies influence farmers’ choice of soil renewal technology and fertilizer application procedures: (a) Although exclusive use of organic fertilizers is not considered adequate, when rains become very uncertain, farmers avoid the risk of losing the cash investment associated with chemical fertil- izer by relying entirely on organic fertilizers. (b) Farmers tend to purchase and apply fertilizer later than recomended to avoid losing their fertilizer investments due to poor plant emergence. . l C. TYPES OF ECONOMIC ANALYSES PERFORMED BY FARMERS Projections of fertilizer demand made by Senegalese policy anal- ysts -- even if not derived strictly from profit maximizing criteria-- are based on a belief that demand is influenced by the level of economic returns. The accuracy of demand forecasts developed from measures of economic returns such as value/cost ratios is dependent on the extent to which farmers’ economic evaluations resemble those of analysts. In this section we present our findings on the types of economic reasoning farmers do and do not use. Before presenting our 192 findings, however, we review some of the literature on the economic theory of input demand and its relevance to peasant agriculture. 1. ECONOMIC THEORY OF INPUT DEMAND Theory maintains that a profit maximizing farmer who has access to perfect knowledge and is operating in a competitive market will pur- chase fertilizer until the marginal factor cost of the last unit used equals the value of the marginal product. If a mathematical formula- tion of the fertilizer production function is available, the optimal dose is identified by solving the following constrained maximization problem which takes into account product and input prices: Max TT - PyY - PxX - K subject to Y - f(X) where TT - profit or gross margin Py - unit price of output Y - output Px - unit price of variable input X - variable input K - fixed costs One can estimate a farmer’s demand for fertilizer by solving this optimization problem with a variety of input/output price combinations; aggregate demand can be estimated by summing the individual demand and adjusting for cost-price effects. An extensive literature, often referred to as the "poor but effic— ient'I school of thought, has evolved since 1964 when Schultz argued convincingly that peasant farmers (often accused of behavior failing to conform to standard economic theory) did indeed allocate resources by equating MFCs with VMPs. Many economists developed mathematical models 193 of peasant behavior using cost-route survey data to provide statistical evidence of this phenomenon (Nelsch, 1965; Luning, 1967; Norman, 1972). This 'poor but efficient“ school of thought has not gone unchal- lenged. Luning (1967) claimed that survey data were an inappropriate means of verifying that MVP-MFC because standard errors of regression coefficients are large, making it unlikely that a t-test would show MVP to be significantly different than MFC. Shapiro (1983) showed that type II errors in many tests of allocative efficiency were extremely high (90 percent chance of falsely failing to reject the null hypothe- sis). In 1968 Lipton pointed out that Schultz’s model failed to incor- porate risk; if it were considered, one would find that a peasant’s objective function was better described as risk minimization rather than profit maximization. Holgin (1975) and McPherson (1983) contended that if farmers’ allocation of resources was influenced by individual risk perceptions and preferences, MFC could only equal VMP for the rare case of a completely risk neutral farmer. Others claimed that peasant behavior should be modeled at the household rather than the individual level, thereby taking into account multiple objectives of the household as a unit (Binswanger et al., 1980; Evenson, 1981). Farming systems research has also broadened the horizon of factors to reckon with when studying input decisions by paying greater heed to ecological and social constraints (Norman, 1980; Shaner et al., 1982). Economists unhappy' with the rigid, unrealistic assumptions of standard economic theory have produced an extensive literature on ”satisficing". Herbert Simon, the best known spokesman for this group, points out that decision makers have neither perfect knowledge nor the 194 mathematical skills required to maximize profits for every investment decision. Consequently, they develop rules of thumb or heuristics which simplify the decision process. It is our hypothesis that Senegalese farmers have also developed certain rules of thumb for evaluating alternative investments and that understanding these rules of thumb will provide more relevant policy insights at this time than attempts to develop econometric models of fertilizer demand. 2. TYPES OF ECONOMIC ANALYSES USED BY FARMERS He asked farmers a series of questions that forced them to examine economic returns to fertilizer. The reasons for forcing farmers to assess economic returns were multiple: (1) He wanted to evaluate farmers’ ability to conceptualize the issues and perform the necessary mathematical calcula- tions; (2) He wanted farmers’ to support their claims that I'fertil- izer is too expensive" with some type of economic logic; (3) He wanted to assess the extent to which farmers’ anal- yses of returns to fertilizer were likely to lead to the same conclusions that researchers obtain using profit maximizing algorithms or value/cost analyses. The four principal questions used to assess farmers’ ability to perform economic analyses and the responses received are described in the next few paragraphs. (1) He asked farmers who could quantify fertilizer response to 'calculate the value added per hectare (i.e., multiply number of extra kilos of peanuts times the price of peanuts) and then determine if any 195 profit would be realized if the fertilizer treatment cost 18,000 FCFA.l The intent of the question was to evaluate mathematical skills and get an idea of what farmers considered “profitable". Our conclusion is that farmers generally lack the mathematical skills to perform this type of analysis. After posing the question to several farmers who took an inordinate amount of time to calculate incorrect responses and became somewhat flustered in the process, thereby putting the rest of the interview in jeopardy, we decided to drop the question. One farmer looked at us in shock when asked the question, saying: "My friends, I would have to know arithmetic to answer that question!“ -- the implica- tion being that one should not expect a peasant farmer to be able to do basic arithmetic.2 (2) Having abandoned tests of mathematical skills, we continued to seek some indicator of what farmers would consider an acceptable level of returns to fertilizer. He asked respondents to estimate the income they would have to be able to expect from a peanut field before investing 18,000 FCFA in fertilizer. This estimate was converted to kilos, and the without-fertilizer peanut yield previously cited by the respondent was subtracted, the difference being used to calculate the farmer’s minimum acceptable value/cost ratio. Responses were not much more satisfactory than those to the previous question. Nine farmers gave a value which was lower than their perceptions of 118,000 FCFA was an approximation of the real, unsubsidized price of fertilizer in the mid-19805. 2Ne do not want to imply that none of the farmers in our sample could perform simple mathematical calculations. He did, however, have only two of 98 farmers who had any formal education other than religious training in Arabic schools. 196 without-fertilizer peanut .yields, making it impossible to estimate meaningful ratios. The large number of invalid responses erodes our, confidence in the apparently valid responses: are they really the result of thoughtful analysis or merely wild guesses which happened to be larger than the previously stated values for unfertilized peanut fields? Ratios calculated ranged from 1.1 to 11.7 with a median value of 4. (3) Our series of questions on economic analyses moved from the complex to the simple. Farmers were next asked the maximum price that they would be willing to pay for a sack of fertilizer. For those farmers who had been able to quantify crop response to fertilizer, this provided us with another method for estimating an acceptable value/cost ratio. Responses provide further evidence that farmers are unaccus- tomed to evaluating fertilizer in very strict value/cost terms. In those cases where it was possible to calculate value/cost ratios and compare them to ratios obtained in the previous question, many of answers differed substantially. This discrepancy may be partially due to the nature of the maximum price question. Several farmers offered very low prices, obviously taking their current financial situation into consideration. Other farmers, concerned that their responses might be used to justify higher prices, either refused to answer or gave lower prices than they might really be willing to pay.1 Table 39 summarizes the responses. The 2000-2500 FCFA range was the median as well as the modal response. Farmers in Nioro had a 1The typical statement made by farmers refusing to answer was: "It’s up to the authorities to set fertilizer prices.” 197 Table 39 Prices Farmers Are Willing To Pay For Fertilizer (Percent of Farmers) FCFA/50 Kilos Peanuts Millet <- 1000 6 6 1001 - 1500 14 6 1501 - 2000 14 18 2001 — 2500 34 29 2501 - 3000 11 18 3001 - 3500 ll 15 3501 - 4000 6 6 4001 - 4500 3 3 Source: Survey data 46-farmer sample. 198 tendency to be willing to go higher than their Gossas counterparts. Of the seven Nioro farmers suggesting that they were willing to go higher than 3000 FCFA, 5 bought fertilizer during the last five years, which lends some credibility to their answers. Our impression is that farm- ers’ maximum price is based on a combination of perceived economic returns and ability to pay without credit. Those who are regular users -- while not always able to quantify yield response -- have a sense of potential gain which shapes their willingness to pay. Many users also have access to non-crop revenues which enhances their ability to pay higher prices. (4) An open-ended question asking farmers to state their criteria for judging the relative expense of fertilizer was the last technique used to elicit farmers’ ideas about economic analysis. These criteria are listed in Table 40. Three of the criteria (1, 3, and 4) all relate in some way to the problem of financial resources. Farmers do not believe that fertilizer is too expensive because average returns do not justify the investment, but because their agricultural earnings do not even cover other more important agricultural and non-agricultural needs. The reference to peanut seed in criteria 3 and 4 is a reflection of farmers’ dependence on peanuts for cash. Even though most of the fertilizer since 1980 has been applied to cereals, the peanut/fertil- izer price ratio provides many farmers with a yardstick for evaluating fertilizer investments. Table 41 presents a frequency distribution of acceptable price/cost ratios (price of peanuts per kilo divided by the price of fertilizer per kilo) suggested by farmers. During the credit 199 Table 40 Farmers’ Criteria for Judging Fertilizer Cost Percent of Criteria Respondents 1) It is expensive when you do not have the means to buy it at current prices 40 2) If we could count on the rains we would not say it was expensive l7 3) Price of fertilizer should be judged by the price of peanuts (5 responses) or the price of millet (1 response) 14 4) If you do not have peanut seeds fertilizer is expensive at any price 10 5) At current prices fertilizer absorbs what it brought 10 6) Expense depends on the intensity of your needs; those who have poor soils and no organic fertilizer will be willing to pay more 7 7) One judges the price by past experience and today’s price is 4 times what we were accustomed to during the credit program 2 Source: Survey data from 46-farmer sample 200 Table 41 Farmers’ Concepts of Acceptable Peanut/Fertilizer Ratios Ratio Percent of Farmers <1 4 1-1.99 25 2-2.5 54 >3 18 Source: Survey data for 46-farmers. 201 program the ratio only exceeded 2 four times, yet 72 percent of the farmers queried think it should be greater than 2. There is no evid- ence that fertilizer consumption was constrained in years that the peanut/fertilizer ratio was less than 2 and credit was available; whether this will also be true in a no credit system remains to be seen. , Risk is the second most prevalent criterion for judging fertilizer expense. Seventeen percent of farmers stated that fertilizer would not be expensive if it rained, thereby implying that even at current prices‘ it would be profitable if the rains were good. While most farmers lack the concepts of probability necessary to analyze risky investments systematically, they do classify investments into categories of more or less risky. Most farmers consider livestock and "banabana" (petty commerce) investments less risky than fertilizer.l Although most farmers do not use marginal analysis to judge fer- tilizer, 17 percent did provide answers which suggested that they do entertain certain concepts of marginal analysis. These farmers stated that expense should be judged in terms of the added value that an investment brings (criterion 5, Table 40); those who are likely to get more added value are likely to pay higher prices (criterion 6, Table 40). For this group of farmers, the constraint on economic analysis is not at the conceptual level but in estimating fertilizer response and accurately calculating marginal returns. None of the questions described above produced responses lending themselves to statistical analysis and clear-cut statements about how 1Alternative investments are discussed more fully in section "D" below. 202 farmers evaluate economic returns and what impact that analysis is likely to have on demand. 0n the other hand, farmers did provide us with a wealth of anecdotal information which reveals the great diver- sity of attitudes toward fertilizer investment decisions. Responses varied considerably partly because farmers have different perceptions of yield response or different perceptions of what return is “profit- able", and partly because some have an instinctive feeling about fer- tilizer costs irrespective of potential returns on investment. A comparison of four comments illustrates the range of responses as well as the farmers’ ability to conceptualize the issue. One farmer believed that 2 sacks of fertilizer (100 kilos) produce 300 to 400 kilos of peanuts on average. He claimed that if the fer- tilizer costs 5,000 FCFA per sack, the investment eats up all the profits. Assuming the 1985/86 peanut price of 90 FCFA/kilo, the net benefit for 300 kilos of added product is 17,000 FCFA. In his example, it is obvious that all profits are not eaten up, yet he firmly believes that 5,000 FCFA is too much to pay for fertilizer. A second farmer said that any price of fertilizer which left him with at least 2,000 FCFA net benefit per hectare in an average year would be acceptable. This represents a return of 11 percent on his cash investment; but, if other costs such as additional labor are factored in, the rate of return would be considerably lower. This farmer is content with much lower average returns than most other farmers interviewed. A third farmer said that the gross returns on any investment should be two times the amount of the investment -- once to cover the 203 cost of the investment and once for profit. This rule of thumb re- sembles the agronomist’s minimum value/cost ratio of 2. The farmer went on to point out, however, that the cost of the fertilizer invest- ment in this case was not only the cost of fertilizer but all addition- al costs of labor, transportation, etc. Value/cost analyses performed by ISRA do not consider these latter costs. ' The final example is a farmer who was unwilling to commit himself to a price beyond his current rather limited means -- his maximum price was 1000 FCFA per sack and he would not consider whether a higher price could be profitable. The evidence presented in the last several pages leads us to conclude that: (1) Few farmers have both the ability to quantify fertilizer response and to accurately perform simple calculations of net benefits and value/cost ratios; (2) Most farmers justify their claim that fertilizer is too expensive from an ability-to-pay perspective rather than from an analysis of economic returns; (3) Given points (1) and (2) above, it is unlikely that farmers demand for fertilizer’ will be as robust as that anticipated by researchers’ projections based on value/cost and profit maximizing analyses. 0. FARMERS’ INVESTMENT PRIORITIES The interview methods used above provided information primarily on what farmers do not do when evaluating fertilizer investments. Having established that most farmers do not perform economic analyses likely 204 to inspire the same level of demand suggested by agro-economic anal- yses, the importance of understanding exactly how farmers do make investment decisions becomes paramount. He developed a second series of questions designed to elicit such information. The data was col- lected using a set of questions about hypothetical investment choices as well as questions about actual investment behavior between 1981 and 1985. He found that most farmers have distinct priorities for disposing of their limited incomes. Following poor cereal harvests, food natur- ally takes precedence over all other purchases and investments. In years when food needs are assured, investment priorities are shaped by rudimentary analyses of investment profitability and considerations such as keeping children on the farm, and maintaining sufficient liquidity to cope with emergencies or unexpected social obligations. Farmers consistently mentioned three types of investments which com- peted with fertilizer: peanut seeds, livestock, and 'banabana' (petty comerce). The discussion which follows is organized around these three alternatives. 1. SEED The 98 farmers interviewed in the general survey were asked how they would have invested 15,000 FCFA had they had that amount of cash available in May 1985 (i.e., just before the rains). Fifty-two percent said they would have invested in seed and 40 percent chose food. Of those mentioning food first, 59 percent listed seed as their second priority. In general, both second and third choices continued to indicate low priority for fertilizer (see Figure 15.) These answers 205 Wofmmm) mspmee Pattern I First Choice Seeds (51) ”alanine: (22) (12) (7) (10) 1111131 Choices ~1de —-Food (7) pm (3) investment (10) —I-‘erti.1izer (4) (2) —Livestock vestock pm'diase (4) investment (1) km (1) madam L-Save (1) animal (2) —Save Mmey (2) F-Banabana (2) We rapair (1) Laura Food (1) Mspaise Patten: II First Choice Food (39) Second cwices -5eeds -'rractim Annual W fidvostock Inves. (23) (5) (3) (3) ’mini Choices W (5) -Save they (2) -Sawe May (2) -More Animals (1) -L1ve£todc (5) r-Banahana (2) Lfiestas (1) —save they (1) ~Repair Mach (3) m (1) W (1) —Fertilizer (2) -Save Romy (2) v-Trac. Animal (1) m (1) Samoa: Survey data true 98-famr ample. Figure 15: Fanners' Investment Priorities 1985/86 206 suggest that seed would be given priority over equipment and traction animals as well as fertilizer. A glance at investment data for 1985/86 shows, however, that many farmers did purchase equipment or traction animals, but not fertilizer, before acquiring what they considered their "minimum” seed needs. Most of these purchases were replacements; but a few were additions to existing holdings to accomodate an in- crease in the number of family members wanting to farm. The high percent of farmers giving seed priority over fertilizer makes it clear that future fertilizer demand will be determined largely by farmers’ concepts of minimum peanut seed requirements and ability to secure these seeds.1 During in-depth interviews 69 percent of farmers stated that if they had sufficient food but no peanut seed or fertil- izer at the beginning of the agricultural season, and total cash on hand was 20,000 FCFA, they would spend the full amount on seed. Seven percent would buy some seed and save the rest for day-to-day needs (kola nuts, unexpected emergencies, etc.) while 24 percent would buy seeds and some fertilizer (seed purchases ranged from 50 to 95 percent of the available cash). Seventy-seven percent of farmers specified a minimum quantity of seeds they would have to obtain before thinking about fertilizer. Replies ranged from 100 to 2000 kilos of shelled seed. The median reply was in the 300- to 500-kilo range which represents roughly 4 to 7 hectares of peanuts, exclusively for the household head. Amounts given 1Answers to questions about priorities are clearly biased toward peanut seed due to the government’s decision to discontinue official seed distribution the year of our study. Nevertheless, we do think it is an important priority which competes with fertilizer. 207 were generally several times greater than quantities actually planted in 1985. Discussions with farmers suggest that labor contracting procedures and obligations to provide peanut seeds to family members may be perpetuating extensive agricultural techniques and constraining fertilizer purchases. More than 70 percent of farmers would not buy fertilizer before procuring a large quantity of seed even though they believe that a fertilizer investment would increase peanut yields more than an equal amount invested in seed. This is because the head of household must be able to offer land and peanut seeds as part of the contract when hiring seasonal laborers. Similarly, if he wants to keep his sons at home, he must offer them land and seed capable of producing agricultural revenues greater than their expectations of earnings in Kaolack or Dakar. 2. LIVESTOCK Responses to a series of questions concerning farmers’ willingness to transform livestock resources into fertilizer investments once more pressing food and seed requirements were met suggest that 61 percent would be willing to sell some animals to pay for fertilizer.l It should be kept in mind, however, that the probability of the first two conditions being met will be relatively low for the next few years.2 1This is somewhat lower than the 81 percent who said they would be willing to convert cash holdings to fertilizer once minimum seed and food needs were assured. 2Most farmers came nowhere close to procuring their minimum seed re- quirements in 1985/86 and most are likely to experience the same pheno- menon in 1986/87 (see Gaye, 1986; and Niang and Sarr, 1986). Survey data for the 1981-85 period show that 94 percent of the 98-farmer sample experienced food deficits at least once; more than 50 percent have had 2 to 3 deficit years. Fifty percent of farmers in the 46- farmer sample feared that they would experience cereal shortfalls again 208 The return on livestock investments requires further examination. It is difficult to generalize from the studies which have been done because they cover very specific zones and activities. The most recent and reliable data on cattle fattening activities in the north-central peanut basin show that in only 75 days farmers realized net profits of more than 50 percent on investments (Faye and Landais, 1984, p. 16).1 Farmers’ perceptions of their profits are even greater for they do not consider the opportunity cost of crop residues used as feed. In addi- tion to specific fattening activities, farmers think that investments in small ruminants or cattle, kept in a herd or allowed to graze in nearby fields, are dependable investments. Animals increase in value through growth and reproduction and the danger of disease or death is less than the risk of loss associated with fertilizer. Furthermore, animals can be sold at any time, thereby providing liquidity. These perceptions of the relative profitability of animal and fertilizer investments explain why 38 percent of farmers are unwilling to sell animals in order to purchase fertilizer. Even those willing to transform animal investments into fertilizer will not do so until the end of the dry season after food and seed is assured and fodder becomes scarce making animals more expensive to maintain. 3. "BANABANA' ”Banabana” is an alternative to fertilizer which is difficult to define and more difficult to evaluate. In general, the term is before the 1987 harvest despite the relatively good rains and high hopes for the 1986 season. 1The costs of animals and feed were both included in the value of the initial investment. 209 synonymous with petty comerce. For our purposes, we distinguish between two different types of 'banabana": (1) when an individual earns a commission by finding a client for someone who has something to sell and (2) when an individual invests personal resources to buy goods in one market and sell them in another. Hhile both types of "banabana" are being used increasingly to supplement farm revenues, it is the 'banabana' which requires an investment that competes with fertilizer. To the best of our knowledge, there is no information on economic returns to 'banabana'. All we can say is that several farmers in our sample claim that a significant part of their income is derived from 'banabana' activities and -- at least during the dry season -- they prefer to pursue these activities rather than immobilize cash by making early fertilizer purchases. E. FERTILIZER PURCHASES: 1981-85 Having discussed factors that influence fertilizer purchases and use, we now turn to a description of purchases actually made during the 1981-85 period. He asked respondents in the 98-farmer sample to recall all fertilizer purchases made outside official distribution programs during the 1981-85 period. Twenty-nine remembered at least one pur- chase. Table 42 is a frequency distribution of the number of sacks purchased over the entire period. Table 42: Fertilizer Purchases 1981/82-1985/86 Number of Number of SO-kilo Sacks Farmers 1-5 15 6-10 4 11-20 7 >20 3 210 Fifty-nine percent of farmers purchasing bought from other farmers. Prices paid ranged from 10 to 63 FCFA/kilo with the 20-30 FCFA range predominating. Sources of money used were: 48 percent agricultural revenue 24 percent non-agricultural revenues 24 percent sales of animals Thirty-eight percent of those having purchased were from Gossas and 35 percent of the 285 sacks reported were bought in that region. A total of 285 sacks of fertilizer purchased by 98 farmers over a five-year period is an insignificant amount. 0n the other hand, 30 percent of the sample did make some effort to obtain fertilizer when it was not freely available and agricultural revenues were relatively low. In-depth interviews with members of the 46-farmer sample provided detailed information on cash as well as contract and ”retenue" fertil- izer acquisitions and use for the same period. Location and timing of transactions were also discussed. In any given year between 1981 and 1985, 70 percent of the 46 farmers made neither cash nor credit purchases. Figure 16 summarizes information on total quantities of fertilizer acquired through con- tracts, cash purchases, and the "retenue”, showing the percent from each source used on different crops. 'Retenue' fertilizer accounted for 22 percent of all acquisitions while contracts represented 26 percent and cash purchases 52 percent. Total quantities used were very small, averaging two sacks per farm per year. Distribution among farmers was very uneven. The share of fertilizer acquired each year by 211 mwxmmonn~m\~mo~ .mn: vac covu—mwacu< Lw~3~wucmg no" ogsavu .ucmmmp as» :3 namepuonm on: as» Low wean—gen Lo~._3ucme m.coma guau be :o3ucoaoca mxogm oc3guuoz Hugo: m650 oo\nlow "00.500 nonnOO l monocutan. r300 9.22 l mononucnl :nOU 212 the largest purchaser ranged from 48 to 89 percent in Gossas and 25 to 56 percent in Nioro. Figure 17 illustrates changing trends in fertilizer use, particu- larly farmers’ growing preference for fertilization of cereal crops. The increasing share of Nioro fertilizer going to corn is due to expan- sion of areas cultivated whereas increased application on millet is a reflection of attitudes already described on pages 189-90 above. Table 43 shows that fertilizer prices paid by sample farmers have gradually risen but remain below current official prices. Farmers tended to purchase fertilizer close to home. Forty-five percent of all fertilizer transactions took place in weekly markets, 22 percent in a purchaser’s own or a nearby village, and 12 percent in major urban centers (Gossas, Nioro, Kaolack). Forty-eight percent of purchases were negotiated with other farm- ers who were trying to sell "retenue", contract, or smuggled Gambian fertilizer while forty percent were bought from traders. Official sales (by SODEVA or a cooperative) account for only 12 percent of the transactions. Only five farmers claimed that all fertilizer purchases were paid for with peanut revenues. Seven mentioned revenues from animal sales but in most cases the animal was not sold to finance the fertilizer purchase. Farmers were obliged to liquidate animals to buy food; some money was left over and they used it for small fertilizer purchases. 213 owxmmmnuwmxammm .um: Lo~3pwugam cm mcgmuuaa mcvmcacu "ma ogamvu .mcgEcoe we to» menu xm>cam “ouczom .nocu uo_m3uaam o» Le~333ucoh auscucou x—nqo anus xF—ouwumcomgu «costs» we cogmcwmcou Lo~3—_ugmm =oacouug. ecu mmugucaa cmuu xpco “maoz mflflflCr mfi<flCfl p OOP.“ Qflu IOFIG I...“ «I‘d n..« in RICA Ian nlId . - . . .... ......... .. .3 m \ . a m / 1...... Lr a“ \\ s. .r Q“ m .111. .......... .3 nun. \\\\ ‘21, u .3 nun" .. . o I N .. / .3 on .0 I .\ // .. .3 on W“ /.1\\ \ l l .3 0‘ ( /.... f 0‘ l..\ 00000006\sss aloooo ses.-.0 L1 °‘ N R. 4' an m .. .3 on m) . ../ .3 on m) .3 an. A ...... ... / .3 Oh I was. m 5.3.: it / M g.) a--- t O. m “Sgt .III co... +3 0. m 9323.... II. .32. m m 6 hectares Seed is a priority No cash purch. during PA No non-crop revenues IV. FS < 6 hectares No seed priority Made cash purch. during PA No non-crop revenues V. FS 2 6 No seed priority No cash purch. during PA No non-crop revenues VI. FS 2 6 Seed is a priority Made cash purch. during PA No non-crop revenues VII. FS 2 6 Seed is a priority No cash purch. during PA Has non-crop revenues VIII. FS < 6 No seed priority Made cash purch. during PA Has non-crop revenues .00 .001 .00 .034 .125 .096 .50 .459 .67 .719 .75 .773 .80 .813 1.00 .972 continued on next page 226 Table 47, continued Actual Probability Predicted Group Purchasing Probability IX. FS 2 6 1.00 .988 No seed priority Made cash purch. during PA No non-crop revenues X. FS 2 6 1.00 .993 Seed is a priority Made cash purch. during PA Has non-crop revenues XI. FS 2 6 1.00 1.00 No seed priority Made cash purch. during PA Has non-crop revenues Source: Estimated from survey data. 227 with approximately 100 percent probability of purchasing.1 Derivatives show that FS has the greatest influence on increasing the probability of purchase, followed closely by OTHREV, and CASHPA. An increase in SONLY reduces the probability of purchasing. Hhat do these results imply for those interested in encouraging fertilizer use? He respond to this question separately for each inde- pendent variable. M In the 98-farmer model, the derivative suggests that for small increases in FS the probability of purchasing increases .29. The effect in the 42-farmer model is much more dramatic; the derivative indicates a 1.1 increase in the probability of purchasing. The tenden- cy of small farmers to be non-purchasers could be. a reflection of limited economic means. An alternative view would be that the PA favored large farmers, therefore small farmers had little exposure to fertilizer during the credit program and are less inclined to experi- ment when credit is unavailable. In either case, it signals an impor- tant dilemma for Senegalese agriculture, since it is farmers with limited access to land who need to intensify the most if they are to provide their families with a decent standard of living. w d Fer zer Res 0 DONTKN : DONTKNO was only marginally significant (.10) in the 98-farmer model and not at all significant in the smaller model; it also was not significant in our 1Given that we have defined a ”purchaser" as any farmer who purchased once between 1981 and 1985, this probability must be interpreted as the probability of purchasing at least once in a five year period, not as the probability of purchasing in any given year. 228 preliminary two-by-two table analysis.1 The derivative for the large- sample model suggests that an inability to quantify fertilizer response decreases the probability of purchasing .13. This relatively small effect and lack of significance in the smaller model suggests that programs to increase farmers’ knowledge of fertilizer response will have little, if any, effect on the probability of a farmer purchasing fertilizer. The remainder of our analysis suggests that variables reflecting farmers’ economic circumstances are much more important than knowledge of fertilizer response. Purchasing Behavior During thelPA (CASHEAl: During the PA it was not uncommon for farmers to obtain fertilizer on credit and resell it at a loss to obtain quick cash. In-depth interviews revealed that a number of farmers regularly bought fertilizer resold by other farmers. The CASHPA variable is coded as one for all farmers who made such purchases during the PA. The derivative suggests that purchasing for cash during the PA increases the probability of current purchases by .87. From a policy perspective it is difficult to transform this infor- mation into a policy initiative -- we have no way of changing what a farmer did in the past. It would be helpful to know what other charac- teristics are correlated with this variable. It could simply be that those who purchased for cash during the PA were wealthier farmers or farmers who had non-crop revenues that gave them greater liquidity. 1In the small sample, 82 percent of farmers who have a 90 percent or better probability of purchasing were able to quantify response; this suggests that there is a correlation between better knowledge and purchasing behavior even though it is not statistically significant. 229 Unfortunately, this type of information for the PA period is not available. Access t9 Beljahle Nan-grep Begeages (QIHBEI); The non-crop revenue variable is highly significant; it increases the probability of purchasing by .93. The distinction made between non-crop revenues and non-agricultural activities (which was not significant in either model) is important. The latter only indicates that a farmer participates in some non-agricultural activity (cattle fattening, blacksmithing, petty conlnerce, ”maraboutage", etc.). The revenue variable is based on information obtained during in-depth interviews about the relative importance of crop versus non-crop revenues, types of purchases made with non-crop revenues, and the .inter-annual reliability of ‘these revenues. The coding is somewhat subjective as no precise quantitative data was collected on these revenues, but a value of one for the vari- able implies regular non-crop revenues which comprised a substantial share of total income during 1981-85. The variable is significant perhaps because our data cover 5 years of very low agricultural reve- nues which did not permit fertilizer investments; given better agricul- tural revenues, OTHREV might have lost some of its significance. Unfortunately, we did not have good data on total crop or total farm income during this period. If this data were available, it would be useful to test the relative influence of various sources of revenues and total revenue. A narrow interpretation of this variable as strictly non-crop revenues suggests that policies to foster diverse economic activities in rural Senegal will increase the probability of fertilizer 230 purchasing. Nithout such opportunities, farmers are driven into a downward spiral by recurrent droughts and poor harvests; agricultural investment and productivity decline continuously. A broader interpretation of the OTHREV variable would be to con- sider it a proxy for income in general. In this case, any policy measure that could increase general farm income would increase the probability of fertilizer investments. The hierarchical decision models discussed in the next section justify, to a certain extent, this broader interpretation. h-t 'e. -‘d Ha P in 0 e - i-- ON Y : The SONLY variable reflects a preference for extensive over intensive agriculture -- a certain minimum quantity of peanut seed must be ob- tained before any investment in fertilizer is considered. It also embodies the belief that one has to grow peanuts to earn a living in agriculture. Farmers do not yet consider cereal production a cash crop alternative. The decision analysis of AF data presented in Chapter V provided evidence that the preference for peanut over cereal crops can be justified on economic grounds, and the preference for peanuts over fertilizer can be justified on economic grounds for farmers who have easy access to land and labor. AJthough the variable is significant and reduces the probability of fertilizer purchases, the results on Table 47 (Groups VI, VII, and X) show that farmers who give priority to seeds can still exhibit a high probability of purchasing fertilizer if economic variables (CASHPA and OTHREV) are favorable. The major contributions of the logit analysis over the simple chi- square analysis of two-by-two tables are: (1) it shows the relative 231 importance of different farmer characteristics when present at the same time, and (2) it permits us to classify farmers into groups for which we can predict the probability of purchasing fertilizer. The fact that economic variables have more influence than attitude and knowledge variables suggests that policy initiatives to increase fertilizer consumption will have a greater impact if they aim at improving farm income rather than attempting to change knowledge and attitudes. The ability to divide farmers into groups according to the probability of purchasing fertilizer provides useful information for the design of distribution programs as vendors can target their efforts toward farm- ers with a higher probability of purchasing. An example of such targeting might be to have fertilizer available at markets where livestock are sold just before the beginning of the rainy season. Since income appears to be an important determinant of fertilizer demand, rapid publication of statistics on harvests and marketings by region could provide traders with valuable information on which zones would be most likely to purchase fertilizer for the next agricultural campaign. Farmers in zones closer to urban areas might also have greater demand if their proximity to urban areas means they have greater access to non-farm revenues. Before going on to our discussion of HDM, it is important to remind readers that the small sample size of the 42-farmer model does not meet minimum sample size requirements for use of the maximum like- lihood algorithm employed in the logit analysis. In an attempt to double check our results, we also used the Grizzle, Starmer, and Koch weighted least squares algorithm to estimate the same models presented 232 above. This algorithm also requires a large sample size; but the fact that we obtained almost identical results using both methods does increase our confidence somewhat in the small-sample models. Further- more, the importance given to economic variables by these models is corroborated by the HDMs which follow. 8. HIERARCHICAL DECISION MODELS Our decision to use hierarchical decision models (HDM) is based on our belief that farmers use a 'satisficing' rather than a profit maxi— mizing approach to resolution of investment problems. HDMs represent the 'satisficing" decision process as a hierarchical series of less complex decisions which lead ultimately to the resolution of the pri- mary decision under consideration. Attitudes and knowledge which determine preferred behavior as well as constraints on farmers’ ability to implement preferred courses of action are easily discerned from the models. HDMs have been used by Gladwin (1976) and Franzel (1983) in studies about farmers’ adoption behavior. Gladwin proposes HDMs for predicting adoption behavior; Franzel uses them only as a tool for identifying leverage points where farming systems research programs might find technical solutions to improve adoption rates. We use the information presented here to identify policy interventions capable of increasing fertilizer consumption. Ne have developed two HDM; one for the 1983/84 campaign following a very good 1982/83 harvest and one for the 1985/86 season following the extremely poor 1984/85 harvest. The decision process used by 28 of 46 farmers interviewed was identical for the entire 1981-1985 period. This is reflected in the upper half of Figures 19 and 20 which 233 ARE Till m TD 7!!"le Immn m (‘0) m (5) >» DU'T M I I 18 m Ila)! GENERALLY n W THAT VG! CONNIE! (1) Too rifiy fElTILIZEI A um rm CAIIOT "mi (1) n 0011 i0 tee fertile; fertilizer givee 0e weede (3) I heve never med; eetiefied with arsenic fertilizere TE: (10) m (30) I I M'T IN “KI ORIENT "IE AD CREDIT ”JCT, N rm TNIK FERTILIEI m0 EVE! It A ETTEI INVEST!!!" to run Til” EEO, EINFEIT, C ”mm m (9) YES (21) I I com I." N VIII TMIK TlflT FEITIUZEI IE m A ence lMETIENT THAT m ”CHASE IT mu MIT YEAR? m (17) TEE (0) I I N rm TMIK TllAT FElTlLIZEI CAI IE A m [WT IIT TIAT Ti!!! Al! once “HINTS I." we rem nues wince some u tive: mmm m (17) m m me one: mum commas I0 it‘s/n res‘Im' to (z) (12) Food (12) Sociel Miution (3) Seed (2) Eminent (2) Trection Anieel I hhfib AFTER “I“ "1C1" EXPEDITIRES, 010 um MAW m CASH LEFT rm GAVIN: a meme" YES (8) we (9) I I I DID Yul HAVE Alf “GENT Ila): rem III-FAD! ACTIVITIES? DID 1w TNIIK TNAT ”CHASING PEITILIZEI BETTIE TNE IAIIY m It TIE KIT mu?! use or THIS mum ««««< YES (i) D (8) I it: (I) m (5) I I I I an IIIAT DID VIII “SIDE! A ETTEI 18E? I I I I (3) Hold Ceeh (Z) Liveetock I I I I >»»»»»»»»»»» HAVING SEEN TIE IEGIIIIIII OF THE IAIIIT SEASOI use: vou BILLING T0 WT AIIV or fun SAVINGS a IINESTIEIITS IIITO fElTILIZER? YES (4) n (9) I I (i) Millet not 00 good 00 it ileum be; it looks like it neet fertilizer M'T In (1) Millet sergeme looks very good; were there i0 hope it ie worth it to imeet are ' (1)1100 one ceeh 0ft0r eetiefyino peerut 000:! meet 00 I Mt fertilizer (l) l pleated corn; everyone hiewe com deee not eueceed Inleee well fertilized I In 'Tetel m of tieee thet eech Mitre wee Initiated lo M in perotheeem Figure 19: Hierarchical Decision Model of Fertilizer Purchase Decisions for 45 Farmers in 1983/84 234 HE TN m T0 FITILIIER IMGT‘IT? N (41) TE! (5) --> NI'T III I I II m I“ mu N LG! THAT TO) MINI (1) Too rieky fEITlLIlEA A um VIII MT AFM? (1) Hy 0011 ie tee fertile; fertilizer givee ee weede (3) I heve never med; eetiefied with ereenic fertilizere TE! (10) N (31) I I ”'T an am “ET PRICE AN ”IT F1107, N TN TIIK FERTILIIER NAD m E A NTTI INVEST!" roe VIII THAI N, ENIPEMT, C LIVESTNK? N (9) TEE (22) I I ”'T I." N '01 TIIK TMT FERTILIZER I8 0n A m [mm TIIAT W news: IT mu MRT YEAR? D (i8) YES (4) I I N TN TMIK THAT FERTILIZER CAI IE A N IlVEETKMT IIT THAT TIER! ARE OTIIR IMSTIEHTS IN N PEASOIAL m lliICH Bill” I GIVE! RIGITT‘I YES (18) DID Yul HIE m PHIGITY MEDITIIES IH TIN/86? TES‘IIDI' lo (0) (16) Food (0) Sociel mliptim (0) Seed (3) Emipnt (3) Trectien Mieel I 0-0-0-0- AFTER MAIN FRICTTT MNIMES, DID TN It! AIT CASH LEFT m RAVI“ N "KM“? TED (4) no (it) I I I DID Tm HAVE MT WIT I“ no Hal-FAD! ACTIVITIES? DID VIII THIIK THAT ”Mlle FERTILIZEI DEme M IAIHY m we: TIE “it man USE OF THIS I!" <-------- YES (i) D (13) I YES (i) m (4) I I I I In UHATDIDTwwISIDEIADETTEIIBE? i I I I I (i) Dew-10 (2) Hold Ceeh (i) Liveeteck I I I I ------------------------ > HAVIHD SEEN TIE DEDIHHIN OF THE DAIHY SEAS!!! ltlE rm HILLIHG To MEAT MT U m IAVIHGS (I IIVESTIEHTS INTO FERTILIZER? TE! (4) m (13) I I (1)Hillet mt 00 000d 00 it “Id be; it leeke like it I!“ fertilizer M'T 0.11 (i) fertilizer for eeritrect corn 0. lete: obliged to eell willet to W000 fertilizer (i) Hed eae eeeh in head men fereer trying to eell 'reteme' fertilizer (i) m ectiwitiee eermd and: revuuee thet I said efferd fertilizer I ‘Totel W of tieee thet eech Minn-e wee lentil-ted i0 doe-0i in W. Figure 20: Hierarchical Decision Model of Fertilizer Purchase Decisions for 46 Farmers in 1985/86 235 show that 24 farmers never give fertilizer investment serious consider- ation while 4 farmers purchase consistently every year. Farmers not purchasing fall into four distinct categories: (1) Two farmers who have tried fertilizer and found it unsatisfactory. They do not consider it a viable investment alternative although they are farmers who would normally have the financial means to purchase it. (2) Three farmers who have never tried fertilizer and are not interested in trying it. All of these farmers are Serer and all use organic fertilizers extensively. (3) Ten farmers who used small quantities of fertilizer under the PA and liked it but are marginal farmers who have difficulty meeting minimum subsistence needs. They would like to use fertilizer but consider it a luxury they cannot afford. (4) Nine farmers who used small to average amounts of fer- tilizer during the PA but consider investment in extensive agriculture (more seeds, equipment, and animal traction) or livestock more profitable given current price and credit policies. Seventy-eight percent of these farmers have regu- lar access to "parcage" which may partially explain their cavalier attitude toward fertilizer. The characteristic tying these four groups of farmers together is a mind set that fertilizer is not for them; little, if any, thought is given to fertilizer investment on a year-to-year basis. 236 There is only one category of consistent purchasers. These are four farmers who do not bother to ask themselves whether or not fertil- izer should be purchased but begin by asking how much they can afford and what resources they can use to make necessary purchases. All of these farmers are in Nioro. Two of them have regular access to non- farm revenues. One is clearly a wealthier farmer than most of those interviewed (though the source of his wealth is not clear). The last is a farmer who needs fertilizer because he produces corn every year. He finances fertilizer investments by selling his pair of oxen at the beginning of the season, using these revenues to purchase a less expen- sive traction team plus food, seed, and fertilizer. The bottom halves of Figures 19 and 20 depict the sequence of decisions made by the 18 farmers who consider fertilizer an appropriate investment in some years but not in others. In most cases, the deci- sion not to buy is determined by available resources and perceptions of minimum requirements for purchases that have priority over fertilizer. Farmers allocate resources according to the following priorities: (1) food, (2) minimum complement of equipment and traction to assure work for all family members, and (3) minimum acceptable level of peanut seed. Social obligations, however, muddy the picture in many cases. Certain social obligations arrive unexpectedly and cannot be avoided. Cash for meeting planned social obligations (marriages, debt repayment, baptisms) is usually allocated immediately after harvest and does not pose the same problem as unexpected obligations. Marriages and debt 237 repayment to relatives and friends frequently take precedence over all other concerns, including food.1 To simplify the model we have placed ”priority purchases” together in one decision step. Hhile one can see which categories of expendi- tures were most comonly used, the model fails to present a clear picture of when the farmer ran out of funds and which categories of purchases were not satisfied. The infrequency of agricultural invest- ments compared to food and social obligations suggests that even after the relatively good harvest of 1982/83, this particular group of farm- ers invested most of its resources in labor (through food purchases) and very little in capital or other productive inputs. If a farmer has cash left after insuring minimum food, seeds, social obligations, traction animals and equipment repair, he then assesses alternative uses such as mere seed, new equipment, traction animals, cash savings, animal savings, livestock fattening, “banabana”, or fertilizer.2 Some opt for fertilizer at this time; others prefer to 1As in most societies, the amount of money required for different occasions varies by social status of the family involved. Systematic collection of data on expenditures for social obligations was not one of our objectives, but some farmers volunteered the information. One farmer spent 30,000 FCFA for the marriage of his daughter while another married two daughters in a single year, spending about 250,000 FCFA for each wedding. A young farmer who was essentially acting as the head of household for his aged father claims to have kept agricultural invest- ments to a nfinimum fer several years so he could save up the 400,000 FCFA which he needed for his marriage. These examples suggest that the sums of money required for certain rites of passage are far greater than those required for most agricultural investments. 2The main distinction between farmers in group 4 of the non-purchasers and farmers placed in the bottom half of Figures 19 and 20 is that the former will continue to invest in equipment, animal traction and seed after the minimum needs have been assured while the latter will consid- er fertilizer investment. 238 wait until they see the countenance of the newly arriving rainy season. Some of those who opt for savings or alternative investments before the arrival of the rains may later transform these investments/savings into fertilizer. There are also a few who increase cash holdings through off-farm activities or liquidation of assets and use this money for fertilizer if the season appears promising. Because fertilizer is a low priority investment, it is easily pre- empted by innumerable emergency cash needs that occur between peanut marketing time and about one month into the rainy season. In some cases even the actual purchase of fertilizer does not signal the end of the decision process because an unexpected emergency can lead the farmer to reconvert his purchase to cash, usually at substantial loss. While the resource constraint remains paramount, there are other considerations which influence the final decision. Cropping intentions is one; a decision to plant corn is frequently accompanied by a deci- sion to purchase fertilizer as farmers recognize that corn requires very fertile soil. Many farmers like to see how their crop debuts before making a fertilizer investment. To some farmers a good plant emergence is a signal that fertilizer investment is worthwhile (i.e., there is hope of a good harvest); to others, a crop that does not look as good as it should after the germination is a sign of infertile soil spurring farmers to invest in fertilizer. Another factor is access to contract fertilizer; those receiving some fertilizer through contracts may decide to use remaining resources for other investments. The only striking difference between the 1983/84 and 1985/86 models is the higher frequency of food expenditures in the latter. The 239 number of actual fertilizer purchases as well as other agricultural investments remained relatively stable and low for this group of 18 farmers despite the good 1982/83 harvest.l An intervening variable which could explain the relative constancy of fertilizer purchases despite lower incomes is availability. In 1983/84 there was no offi- cial distribution; farmers could purchase from a limited number of SODEVA distribution points for 45 FCFA/kilo. SODEVA sales were not well publicized and stocks were removed before the beginning of the rainy season. In 1985/86 fertilizer from the "retenue" distribution was available at the village level and in local markets; prices varied but were relatively low (20-30 FCFA/kilo). Availability was occasion- ally mentioned as a constraint by farmers, but in most cases the prin- cipal constraint was lack of funds. Availability, however, appears to be an important determinant of purchases by a small group of farmers whom we have labeled impulse purchasers. Five of the 18 farmers depicted in the bottom halves of Figures 19 and 20 did not consciously plan their fertilizer purchases. They happened to have some money in their pockets and no other pressing purchases to make when they encountered someone trying to sell fertil- izer. The other 13 farmers in this group of occasional purchasers all decided they wanted fertilizer. Having made that decision, they man- aged to find a source even in years of relatively scarce supply. Use of HDM has helped us identify seven categories of farmers: (1) Confirmed non-purchasers with past experience; 11982/83 was cited as the best year by more farmers than any other year and 1984/85 as the worst year. 240 (2) Confirmed non-purchasers without past experience; (3) Farmers who consider themselves too poor to use fertilizer; (4) Farmers who prefer extensive agriculture; (5) Confirmed purchasers who buy every year but, due to resource constraints, seldom purchase as much as desired; (6) Farmers who are impulse purchasers; (7) Occasional purchasers frequently constrained by lack of funds. This type of model underscores the fact that there are many different farmers with different perceptions and Idifferent needs. Policies designed to increase fertilizer consumption need to take this into consideration. Having identified these seven groups permits us to select target groups and design policies likely to respond to their needs. For example, it is probably best to let some non-users remain non-users. The confirmed non-users in groups one and two are few in number and not likely to change attitudes quickly. One can also argue that the farmers in the resource poor group should not be given priori- ty; both SODEVA data and our survey show that they are not good farmers in a technical sense, suggesting that much of the potential yield response would be lost. The more relevant policy question concerning this group of farmers is how to develop non-agricultural alternatives so that non-productive farmers can move away from agriculture. The case is not so clear-cut for farmers in group 4. As discussed above, we do not really have adequate knowledge of the relative profit- ability of different agricultural investments. Without this knowledge 241 it is difficult to make judicious policy choices. Six of the nine farmers in this category farm in Gossas, a zone similar to Boulel where AF peanut data showed a poor response to fertilizer. The AF analysis suggests that farmers’ analysis of’ investment alternatives may be correct. Fertilizer consumption by farmers in groups 5-7 would be enhanced by credit, price, and marketing policies which eased the resource con- straints, by extension programs which helped them do more accurate evaluations of fertilizer investments, and by improved distribution programs which lowered costs of finding fertilizer and increased the probability of purchases by those who buy more by impluse than by plan. Most of the farmers in groups 5-7 are from Nioro, the zone where AF data suggest that fertilizer use remains profitable at current prices. The concentration of these farmers in Nioro suggest that one could economize on policy interventions by targeting them to zones of better fertilizer response where farmers are more likely to invest. IX. SUMMARY OF SALIENT FINDINGS AND POLICY IMPLICATIONS This chapter provides a summary of major research findings and discusses the policy implications of these findings. The review of major findings is divided into three sections: (I) evidence of fertil- izer response, (2) analysis of economic returns to fertilizer, and (3) analysis of factors influencing fertilizer demand and supply. The discussion of policy implications is divided into three sections: (1) choosing realistic policy objectives, (2) designing short-run policy initiatives, and (3) resolving the long-run policy dilemmas. A. SUMMARY OF SALIENT FINDINGS 1. EVIDENCE OF FERTILIZER RESPONSE Evidence of fertilizer response obtained on farmers’ fields is limited due to the difficulty of conducting on-farm trials and demon- strations that produce statistically significant results. The conse- quence has been a tendency to rely on on-station trials or highly supervised on-farm trials and demonstrations. Most currently available data on fertilizer response come from on-station multi-rate trials conducted during the 19505 and 1960s and researcher-managed demonstra- tion plots which provided response data up through 1982. The NPA bases its fertilizer policy on a belief that fertilizer can increase yields 40 percent (GOS, 1984, cited in Chapter I above). Our review of past fertilizer research (Chapter IV) showed that this was only true of selected zones and crops. ISRA demonstration trials have consistently shown peanut response to fertilizer to be less than 40 percent. In the late 19605 it averaged 10-22 percent in the north 242 243 of the Peanut Basin and 15-32 percent in the Sine Saloum (Tourte et al., 1971). Cereal response was better, averaging 15-74 percent in the Sine Saloum and 63-109 percent in the North.1 Chapter V presented an analysis that used 1964-82 data for Boulel and Nioro to estimate fertilizer production functions. This analysis improves on past analyses as it includes data for drought years in the late 19705 and early 19805, and it examines inter-annual variability in fertilizer response. ‘The data come from “Amélioration Fonciere (AF) researcher-managed demonstrations at PAPEMs and are generally consider- ed to exhibit an intermediate level of fertilizer response between highly controlled on-station trials and highly variable on-farm trials or survey data. Our analysis shows an average Boulel response of 17-26 percent for peanuts and 53-191 percent for millet. Results for Nioro are somewhat better: peanut yields increase 26-37 percent and sorghum yields 67-196 percent. These results suggest that average peanut response in the northern Sine Saloum (Boulel) is about 10 percent less than in the southern Sine Saloum (Nioro), and cereal response is 5-15 percent less. Current and past analyses confirm that average peanut response is less than the 40 percent referred to in NPA documents, but average cereal response is better. There are two dangers in looking at only average percentage yield increases: (1) it provides no information about economic returns and (2) averages tend to mask inter-annual 1The lower end of the response range is for "themes légers" fertilizer doses and the higher end is for "themes lourds” doses. 244 fluctuations which have an important impact on farmers decisions to use fertilizer. The next section of this summary deals specifically with economic returns to fertilizer. In this section, it is sufficient to point out that the lower the without-fertilizer yield, the more likely one is to obtain a high percent increase. For example, a without-fertilizer millet yield of 150 kilos/hectare which doubles to 300 kilos/hectare exhibits a 100 percent increase but the increase in revenues at current prices is only 10,500 FCFA -- less than the current price of the fer- tilizer treatment. Percent increases in yields can provide misleading policy guidance. In examining inter-annual variability we were unable to confirm a statistically significant response for 37 percent of the Boulel peanut crops between 1964 and 1982. The relatively low average response in Boulel, coupled with the high probability of obtaining no response in a given year goes a long way toward explaining low fertilizer consumption in this zone. For Boulel millet, Nioro peanuts, and Nioro sorghum we were able to confirm a significant response 90 percent of the time. Again, caution must be used; a significant fertilizer response does not necessarily mean it is economically profitable. 2. ANALYSIS OF ECONOMIC RETURNS TO FERTILIZER Value/cost ratios estimated with fertilizer response data from IRHO confirmation and ISRA’s PAPEM demonstrations have been the most popular type of analysis. Although past analyses showed fertilizer value/cost ratios to be generally favorable, there have always been zones and crops which consistently exhibited value/cost ratios less 245 than 2 at real prices. Ratios calculated with subsidized prices were greater than 2 in all zones. Comon arguments justifying fertilizer subsidies were: (1) farmers needed added incentive to use fertilizer in dryer zones where v/c ratios using real prices were frequently less than 2, (2) indirect benefits of the subsidy (due to growth in GDP and tax revenues) exceeded costs, and (3) without fertilizer soils would suffer irreversible degradation. Research conducted by FAO and IFDC during the 1960s and 1970s in the Peanut Basin and Casamance confirmed the economic profitability of fertilizer. Formulas and dosages recommended by FAO and IFDC, however, differed from those used by ISRA. In Chapter V we used the Boulel and Nioro AF data and our esti- mated production functions to perform three types of economic analysis. These analyses improve on past analyses by (I) incorporating more recent fertilizer response data, (2) considering economic returns to peanut hay, (3) using nominal prices over the 1964-82 period to depict the situation actually faced by farmers, (4) using current prices to provide an up-to-date picture of economic returns, and (5) examining the impact of variability in fertilizer response on economic returns and investment decisions. In examining v/c retios calculated with nominal prices we found that: (l) Nioro farmers experienced average peanut v/c ratios of 5.11 during the 1964-82 period and average sorghum ratios of 11.35; the probability of the peanut ratio falling below 2 was 28 percent and that for sorghum was zero. 246 (2) Boulel farmers experienced average peanut v/c ratios of 2.1 and average millet ratios of 6.7; the probability of a peanut ratio less than 2 was 53 percent and that for millet 10 percent. Using current real prices for “themes légers“ fertilizer doses and official prices for peanuts and cereal we found the following picture of economic returns when value/cost ratios were calculated directly from AF data:1 (1) The combination of low response and high inter-annual vari- ability makes investment in peanut fertilizer relatively unattrac- tive in Boulel (average v/c of 1.4, increasing to 1.8 if peanut hay is considered; the 69 percent chance of experiencing a ratio less than 2 if only peanuts are considered drops to 61 percent if returns to hay are factored in). (2) Results on Boulel millet suggest that the average value/cost ratio is 3.5 with a 20 percent chance of it falling below 2. (3) Returns to Nioro peanut fertilizer are good (v/c 3.1 without hay and 5.13 with hay; the 39 percent probability of a ratio less than 2 declines to 18 percent when hay is considered). (4) Average ratios for Nioro sorghum are 6 with only a 5 percent chance of it falling below 2. Decision models which took into account the probability of good, bad, and average fertilizer response exhibited in AF data for Nioro 1Value/cost ratios based on fertilizer response estimated from pooled time-series and cross-sectional data tended to be lower for millet and sorghum than those calculated directly from the data. See Chapter V for a discussion of these differences. 247 provided the following insights about the relative profitability of different agricultural investments.1 (1) Expected net benefits (ENB) for fertilized sorghum are 62,000 FCFA. more than unfertilized sorghum; and the fertilizer user always has a greater net benefit than the non-user, even in bad years. (2) ENB for a hectare of fertilized peanuts are only 23,000 FCFA more than unfertilized peanuts with a 33 percent chance of the user earning less than the non-user. (3) Hith or without fertilizer, ENB for a hectare of peanuts is greater than ENB for a hectare of sorghum. Nithout fertilizer peanut ENB is 128,000 versus 70,000 FCFA for sorghum. Hith fer- tilizer the gap is reduced but still important -- 152,000 for peanuts and 132,000 FCFA for sorghum. (4) A farmer who grows both a hectare of peanuts and a hectare of sorghum but has enough money to fertilize only one crop will be 38,000 FCFA better off by fertilizing the sorghum (if the 70 FCFA/kilo cereal price is effective). (5) A farmer who uses the cost of a recommended fertilizer dose to buy more peanut seed instead of fertilizer earns 84,000 FCFA more than the fertilizer purchaser. Returns per hour (210 versus 253 FCFA) and per hectare (126,000 versus 152,000 FCFA) are lower for the peanut seed purchaser, but returns per hour remain signif- icantly greater than the current minimum wage (158 FCFA/hour). We 1These models are all based on "themes légers” fertilizer doses and returns to peanut hay are ac; included. 248 conclude that extensive peanut production is a reasonable and profitable investment for a farmer with labor and land that would be otherwise unemployed. The more profitable options were the same in Boulel as in Nioro, but the amount of additional net benefit associated with the fertilizer options was much less and the risk of a fertilizer user earning less than a non-user in a bad year was much greater. For example: (1) Fertilized cereal in Boulel promises 30,000 FCFA more net benefit than unfertilized, but fertilized sorghum in Nioro brings an additional 62,000 net benefit. (2) The ENB for fertilized peanuts in Boulel is only 8,000 FCFA more than that for unfertilized peanuts and the probability of a fertilizer user earning 14,000 FCFA less than a non-user in a bad year is 53 percent. (3) Both fertilized and unfertilized peanuts are more profitable than fertilized and unfertilized millet. For unfertilized crops the added net benefit for peanuts is 28,000 FCFA/hectare. For fertilized crops, however, it is only 7,000 FCFA greater and risk of low revenue years with peanuts is greater than with millet. (4) A farmer fertilizing one hectare of millet and growing one hectare of peanuts without fertilizer will earn 21,000 FCFA more for his two hectares than another farmer who fertilizes only the peanut crop. In Nioro the additional benefit was 38,000 FCFA. (5) A farmer choosing the extensive option of peanut seed rather than fertilizer can anticipate 49,000 FCFA more total revenue. Returns per hour for both extensive (seed) and intensive 249 (fertilizer) options, however, are below the minimum wage (114 FCFA/hour for seed and 131 FCFA/hour for fertilizer). This result is perhaps more noteworthy for what it reveals about generally low economic returns to agriculture in the zone than what it says about intensive versus extensive cultivation. Further evidence of the relatively low economic returns to agri- culture in Boulel comes from a comparison of ENB for all decision options examined in Chapter V (both with- and without-fertilizer op- tions). The comparison reveals that ENB/hectare are 64-82 percent greater in Nioro than in Boulel. Although we have been able to provide some concrete evidence on the economic returns to fertilizer use in two zones of the Sine Saloum, it is important to point out some factors which have prevented rigorous analysis in the past and other factors which make it difficult to obtain statistically reliable results at present: (1) Past analyses have averaged fertilizer response and value/ cost ratios across sites and through time. While averages provide useful information, they tend to mask important sources of vari- ability in fertilizer response and economic returns that should be considered in design of fertilizer policy. (2) Mineral balance studies replaced multi-rate trials in the early 1960s. Since then there has been a significant deteriora- tion in environmental conditions that influence fertilizer re- sponse, but there has been no systematic program to monitor the deterioration in response. Hithout updates of multi-rate trial 250 data it is impossible to identify profit-maximizing fertilizer doses relevant to current soil and climatic conditions. (3) High variability in PAPEM and on-farm data makes it difficult to get statistically significant estimates of fertilizer response approximating farmers’ yields. (4) Lack of data on anticipated yield decline due to soil deter- ioration in the absence of fertilizer makes it impossible to do economic analysis incorporating the long-run benefits of maintain- ing soil quality. (5) Lack of data on economic returns to investments which compete with fertilizer has resulted in an over-reliance on v/c ratios which fail to provide information on the profitability of fertil- izer relative to other investments. 3. FACTORS INFLUENCING FERTILIZER DEMAND AND SUPPLY Our discussion of fertilizer demand and supply was divided into three parts: (1) a historical overview and evaluation of institutions which determined demand and supply, during the "Programme Agricole" (PA), (2) a statistical analysis of PA fertilizer consumption data, and (3) a discussion of factors having influenced demand and supply since the 1980 dissolution of the PA. (A) HISTORICAL OVERVIEH During the PA fertilizer demand was shaped by a credit program administered through farmer cooperatives, high fertilizer subsidies, a government marketing program that was to guarantee farmers a fair price for their peanuts, and an extension program designed to increase farm- ers’ investments in modern inputs. Supply was shaped by agricultural 251 researchers who specified types of fertilizer to be distributed, a government bureaucracy charged with ordering and distributing fertil- izer to farmer cooperatives, and a GOS/SIES industrial agreement which specified minimum quantities and prices for government fertilizer orders. During much of its 20-year lifetime, the PA appeared to be ful- filling its mission. Performance indicators used by the GOS such as increases in aggregate peanut and millet production, increases in government revenues from the agricultural sector, and increases in sales of modern inputs, were usually moving in the right direction. Ultimately, however, the state-run structure, designed to train farmer cooperatives and then I'wither away", took on a life of its own. He concluded that the high level of state intervention encouraged in- creased use of modern inputs but led eventually to a complex, over- centralized, non-performing institutional structure which siphoned off agricultural surpluses for largely non-productive uses. The growth of this monolithic structure stunted the private sector, rendering it incapable of assuming its traditional input manufacture and distribu- tion role when government finally did withdraw. Government’s heavy hand in agriculture has also alienated farmers and discouraged develop- ment of basic decision-making and management skills required for eco- nomically sound farm investments and cooperative management. The literature offering PA post-mortems has placed most of the blame on (1) the overly centralized bureaucratic structure, (2) too much government and too little democracy and solidarity in the coopera- tive movement, and (3) corruption and general mismanagement in both 252 government and cooperative institutions. While we agree with these critiques, our historical review suggests another important weakness-- a failure to collect and analyze information that could have provided a true picture of the economic and financial health of the agricultural sector. A tendency to ignore economic and financial analysis was pervasive: (1) Researchers gave low priority to economic criteria in recom- mending fertilizer products; (2) Extension services encouraged farmers to increase fertilizer orders with little analysis of debt carrying capacity and no thought to the relative returns of alternative investments; (3) Input manufacturers forced ONCAD and farmers to indicate input needs at a time when they lacked adequate information on farm income and prices; (4) The cooperatives, ONCAD, and the BNDS kept such poor accounts that rigorous financial or economic analysis was impossible and fraud was encouraged; (5) Tax, subsidy, and price policies were so complex and inter- twined that it was difficult to trace profit and losses of dif- ferent sectors and of participants in each sector. Analysts looking at agricultural productivity and input purchases thought all was well. Had they examined instead growth in farm income or the system’s ability to equate input supply with effective demand, a more pessimistic but realistic picture might have emerged sooner.l 1He use the term effective demand to mean inputs that farmers actually paid for; this excludes the inputs delivered but not paid for due to debt defaults. 253 Senegal’s failure to pay attention to economic indicators did not end with the PA; the habit of ignoring economic analysis persists today. The NPA fertilizer objectives are a good example -- fertilizer consumption goals are set at levels which were seldom achieved under much more favorable climatic and price conditions. There is absolutely no suggestion in the NPA that the economic returns to fertilizer (or any other modern input) should be examined, despite ample evidence that most farmers are unwilling to pay the real cost and the government is not in a position to finance the subsidy. (B) STATISTICAL ANALYSIS OF PA FERTILIZER CONSUMPTION DATA Estimating fertilizer demand and designing price or credit poli- cies requires some knowledge of how farmers respond to changes in price and income. The major constraint to demand analysis in Senegal at present is that the current distribution system has no historical precedent; therefore, there is no way to develop a model which can be used to predict demand. Statistical analysis of fertilizer consumption data for the agri- cultural credit program (1960-1980) suggests that income was the most important explanatory variable and changes in price relationships (e.g., value/cost or cost/price ratios) had little influence on changes in fertilizer consumption. Great caution must be used in making infer- ences about current demand from what happened during the PA, but this does not necessarily mean that all PA information is irrelevant. Combining insights about what happened during the PA with current information from farm surveys can provide useful policy guidance. 254 For example, both logit and hierarchical decision analysis of survey data suggest that income is an important determinant of fertil- izer purchases. The logit analysis found that access to non-crop revenues was the most significant variable in trying to distinguish between fertilizer purchasers and non-purchasers. The hierarchical decision analysis illustrated that fertilizer is not a priority invest- ment for most farmers; therefore, when no credit exists, purchases tend to be made only following years where income is sufficiently high to cover other needs for cash that are considered more important than fertilizer. Among the most commonly mentioned priorities are consump- tion purchases (food, clothing), savings (either cash or as livestock), other agricultural investments, and social obligations. (C) FACTORS CONSTRAINING POST-1980 FERTILIZER PURCHASES Our farm-level surveys suggest that low fertilizer* demand in recent years has been due to (1) low farm incomes, (2) low and uncer- tain fertilizer response, (3) farmers’ belief that fertilizer is not an essential input, (4) farmers’ preference for alternative investments considered more profitable and less risky, (5) the lack of credit (viewed by many farmers as a form of insurance), and (6) an unrespon- sive distribution system. During the 1980-85 period the binding constraint for most farmers was low farm income. Extremely poor cereal harvests forced farmers to use available cash for food. Many were obliged to sell agricultural equipment or animals to obtain money for food. The dissolution of the agricultural credit and government peanut distribution programs put further pressure on limited cash resources. 255 For farmers who managed to cover food and minimum agricultural input needs, the binding constraint tended to be perceptions that other investments were less risky and more profitable than fertilizer. The decision to invest money elsewhere was reinforced by a belief that fertilizer is a bit like salt in a sauce -- a nice touch, but not an essential ingredient. The alternative investments which competed with fertilizer were (1) more peanut seed, (2) livestock, and (3) ”bana- bana'. A preference for keeping some cash liquidity to cover unexpect- ed emergencies and social obligations also kept farmers from purchasing fertilizers even though they had cash available. Part of the preference for alternative investments is probably related to the difficulty farmers have in assessing economic returns to fertilizer. Fifty to sixty percent of farmers are unable to quantify the fertilizer response of peanuts and millet crops for an average year. Those who were able to quantify response believed it was better, on average, than AF results, but many are still reluctant to purchase fertilizer. Part of the constraint here is farmers’ inability to perform simple economic analyses of average returns to fertilizer use. High inter-annual variability in yields and farmers’ conviction that only Allah can predict rains contributes to the problem. Many farmers prefer to rely on inadequate supplies of organic fertilizers, which entail as much risk of crop loss as chemical fertilizers if rains are poor, but do not also carry the risk of losing a cash investment. Farmers who did purchase fertilizer during 1981-85 operate farms larger than 6 hectares, have greater access to non-crop revenues and better knowledge of fertilizer response than non-purchasers. Some of 256 the purchasers and non-purchasers share similar ideas about obtaining minimum supplies of peanut seed before considering fertilizer pur- chases, but the economic situation of the purchasers permits them to meet these seed needs more often than the non-purchasers. Farmers with no non-crop revenues who were accustomed to purchasing all fertilizer through the PA credit program are unwilling to purchase now without some type of credit. Logit analysis suggested that changes in economic variables (particularly access to non-crop revenues) have a much greater impact on increasing the probability that a farmer will purchase fertilizer than changes in knowledge and attitudes (knowledge of fertilizer response and attitudes about minimum peanut seeds). Of farmers with a greater than 90 percent probability of purchasing fertilizer at least once in a 5-year period, - 91 percent have farms larger than 6 hectares, - 45 percent give priority to minimum peanut seed needs before considering fertilizer, - 100 percent paid cash for fertilizer sometime during PA, - 64 percent currently have access to a reliable source of non- crop revenues. Of farmers with a less than 10 percent chance of purchasing fertilizer, 53 percent have farms less than 6 hectares, 100 percent give priority to peanut seeds over fertilizer 18 percent paid cash for fertilizer sometime during PA 0 percent (none) currently have access to a reliable source of non-crop revenues . 257 Our hierarchical decision models (HDM) showed that 52 percent of farmers never seriously considered fertilizer purchases during the 1981-85 period while 9 percent bought regularly, and 39 percent evalu- ated the pros and cons of buying every year and made occasional pur- chases. Those not considering purchases were primarily small (less than six hectares) farmers who considered fertilizer a luxury they could not afford or farmers who believed that extensive agricultural production better met their needs given current fertilizer prices and response. The HDM suggests that the decision to purchase fertilizer in some years but not others is influenced by cash availability, percep- tions of expenditures which should be given priority over fertilizer, how good the rains and crops look at the beginning of the season, and fertilizer availability. Since 1980, the fertilizer distribution system has been unrespon- sive to farmers needs. The "retenue" system provided farmers with insignificant quantities of fertilizer at prices about four times higher than they were accustomed to paying during the PA. Supplies were not delivered on time and great confusion existed over how each farmers’ entitlement was calculated. The program did little but in- crease the already high level of mistrust between farmers and govern- ment services. Cash-and-carry distribution systems run by SONAR and SODEVA had very low sales because farmers were not informed of sales points, confusion existed over prices, and supplies were removed too early. The private sector distribution system which the GOS and for- eign donors have been trying to develop since 1985 has had a very slow start because no one (the GOS, BNDS, private banks, foreign donors, or 258 private businessmen) wants to accept the risk of supplying fertilizer to a geographically dispersed clientele which has shown very little interest in purchasing fertilizer given current price and credit terms. Although our research suggests that the major constraint to in- creasing fertilizer consumption at present is on the demand side, it is clear that inprovements in fertilizer supply must accompany efforts to stimulate demand. Farmer interviews did suggest two ways of making supply systems more responsive to farmers’ needs: (1) fertilizer must be available for the entire period from the harvest through the first weeding of the following season to accommodate the variety of purchas- ing habits exhibited by farmers in our sample and (2) fertilizer should be available in weekly markets which are regularly frequented by farm- ers and served by public transportation rather than at more distant and ' less accessible Rural Community' Cooperatives and peanut collection points. 8. POLICY IMPLICATIONS OF THE RESEARCH 1. CHOOSING REALISTIC POLICY OBJECTIVES Our research suggests that at least five of the current NPA objec- tives are either unrealistic or so poorly defined that successful implementation is not possible. He examine each of the five in turn. (A) FERTILIZER CONSUMPTION OBJECTIVES The fertilizer consumption objectives stated in the NPA are clear- ly unrealistic given farmer demand under present price and credit policies. A more important finding of our research, however, is that a global fertilizer policy for the entire country is misguided. Current economic and production relationships render fertilizer much more 259 profitable and less risky for some crops and zones than for others. It is our opinion that the GOS must begin an official policy of targeting fertilizer to zones of greater potential profitability and low risk. Given Senegal’s socialist orientation and past policies of providing agricultural inputs nationwide at uniform prices, any official program to begin targeting fertilizer would have to be carefully designed, with particular attention being given to the implications for soil degrada- tion in zones not targeted and the regional economic and social consequences.1 Targeting could be accomplished through selective introduction of fertilizer credit programs in those zones which have greater potential. In the long-run, a policy of targeting fertilizer could lead to the gradual phasing out of agricultural production in certain zones. Nhile a difficult option to face, it is certainly one which deserves serious consideration given the very low returns per hectare and per labor-hour obtained in the Boulel decision analysis. (8) BUILDING A PRIVATE SECTOR INPUT DISTRIBUTION SYSTEM The objective of building a nationwide private sector fertilizer distribution system is unrealistic because current demand is insuffi- cient to justify costs of setting up such a diffuse network. Again, selection of target zones where demand should be greater would be a more realistic objective. It is only by working in areas of high demand that the distributors will be able to provide services that appear to encourage increased fertilizer purchases (e. 9., making 1Although the GOS has not followed an official policy of targeting fertilizer, severe economic constraints since 1980 have resulted in a de facto policy of favoring zones with higher rainfall. 260 fertilizer available in weekly markets and during the entire period from harvest through the first rains). (C) REFORMING THE COOPERATIVE MOVEMENT The Senegalese cooperative movement has existed in various forms for almost eighty years, yet the same problems of poor management, corruption, government interference, lack of member solidarity, and poor debt repayment have been present in every reincarnation including the present one. It is time the Senegalese looked for alternative approaches to farmer organization. The current policy of administering credit through the village cooperatives ("sections villageoises”) will have the same long-run consequences of debt default characteristic of past cooperatives. History suggests that agricultural debt accounts should be individual- ized or credit offered to small groups of freely associating individu- als so that those who repay benefit from their repayment. At the same time, provision should be made for extending payment periods in the case of crop failures so that past policies of debt “forgiveness” are not repeated. The current policy of considering the cooperative movement as a "private sector” participant in the input distribution program is dangerous. The long and continued history of government involvement in cooperatives gives them an edge over the traditional private sector in trying to get a foothold on the input sector. There is a risk here of the cooperative movement establishing a monopoly over input distribu- tion which will be as inefficient as past government monopolies. 261 Farmer illiteracy and innumeracy are a constraint to improved agricultural management at the farm-level as well as at the coopera- tive. So long as the majority of cooperative leaders and members remain illiterate and unable to maintain or verify simple records of individual and cooperative transactions, cooperatives will remain extensions of various government services rather than the independent, dynamic engines of development described in the NPA document. If the government persists in its support of a cooperative movement, important investments must be made in training farmers to manage these organiza- tions. Past efforts have been extremely inadequate. (0) FOOD SECURITY OBJECTIVES The decision analysis presented in Chapter V clearly illustrates that increasing cereal production beyond personal consumption needs is not in the farmer’s economic interest given current production and price relationships. In both Boulel and Nioro, it is more profitable to produce a hectare of peanuts. It is difficult to envision a feas- ible policy change which could make cereals more profitable. Producer prices have been raised significantly in recent years. Unless rice prices were increased simultaneously, raising coarse grain prices any higher would make imported rice more attractive to consumers thereby reducing demand for locally produced cereals. It is the low cereal productivity per hectare rather than the price which appears to be the more appropriate leverage point. This means that variety research and programs to improve cultural practices (e.g., water retention, incor- poration of organic matter, etc.) are more likely to improve 262 productivity than changes in price and marketing policy.l An alterna- tive, however, would be to re-think the food self-sufficiency goals given the economic relationships between cereals and peanuts. (E) AGRICULTURAL INTENSIFICATION OBJECTIVE Another GOS agricultural goal is to increase productivity per hectare through more widespread adoption of intensive technologies such as fertilizer. Many Sine Saloum farmers, however, continue to prefer extensive peanut production. Trying to understand why this is true even in zones where analyses have consistently shown fertilizer v/c ratios in the 3 to 4 range is not easy. Our analysis of the peanut seed versus fertilizer option provides a plausible explanation for the tendency of some farmers to pursue extensive rather than intensive peanut cultivation. So long as land and labor are available, the extensive option increases total farm income more than the fertilizer option. Although land is becoming scarce and quality appears to be declining, surplus labor and the obligation to provided family members with a source of personal revenue pushes farmers to expand rather than intensify. I The solution here must be multi-sectoral. Agricultural research must find technologies which are not marginally more profitable than extensive production but so much more profitable that farmers cannot resist. There is also a role for sociological research to examine the extent to which labor contracting practices for both family and lEicher (1982) stressed the need for African countries to intensify varietal research given the rather poor results obtained from efforts to transfer or adapt "Green Revolution" varieties from other parts of the wor d. 263 non-family labor will have to change to accomodate more intensive production techniques. At the same time, the non-agricultural sector must start providing more employment opportunities for the rapidly growing rural population or the pressure for each dependent to have his/her own peanut field will continue to foster extensive practices. 2. DESIGNING SHORTfRUN POLICY INITIATIVES In our opinion the most important steps in fertilizer policy for the immediate future are: (1) clearly identify the zones where eco- nomic returns to fertilizer are high and risk is low, (2) launch a trial program in a limited number of these zones to foster growth in fertilizer demand and encourage the development of a private sector distribution system, and (3) monitor the performance of the trial program, taking necessary corrective measures to improve performance, and evaluating the feasibility of extending the program elsewhere. The identification of zones should be done by assembling available information on fertilizer response which is currently scattered among various sites and institutions. The AF program produced longitudinal fertilizer response data for seven sites in addition to Nioro and Boulel. Data are available on maize, cotton, and rice, as well as peanuts, millet, and sorghum. A first step would be to analyze these data. The second step would be to consolidate data from other similar multi-year, multi-location programs such as ”Type de Fumure" and "Régé- nération des Sols", evaluating the extent to which analysis of these 264 data substantiates AF findings.1 The value in the "Type de Fumure" data is that they represent four-repetition trials conducted at the PAPEMs using the same fertilizer formulas as the AF program. Com- parison of these data with AF could help us assess the value of in- creasing repetitions in future programs and provide an additional source of data for confirming AF yield responses. A final source of current information on fertilizer response which should be rapidly analyzed and consolidated is that collected by farm- ing systems teams in Djibelor and Kaolack. Much of this data is from on-farm trials and farm surveys. Given ISRA’s discontinuation of the AF program in favor of moving to on-farm work, it would be worthwhile to compare the types of analysis and statistical validity of conclu- sions that can be drawn from farming systems work with those that we have been able to extract from AF data. Such a comparison would pro- vide useful information for the design of future fertilizer programs. Logit and HDM suggest that changes in economic variables will be much more effective in increasing demand than changes in attitude and knowledge variables. Increasing producer prices is not a feasible option at present given world market cereal and peanut prices.2 Reduc- ing fertilizer prices through subsidies is also not feasible. The IMF, World Bank, and other donors are adamant about the GOS not financing such subsidies from the treasury. Hhile the GOS might find other 1A recent visit to Senegal indicated that there may be serious problems to overcome in locating these data bases and finding researchers knowl- edgeable enough about the work to assist in an analysis. This could also become true of the AF data sets if an effort is not made to ana- lyze them quickly. 2See Martin (1988) for a discussion of price issues in Senegal. 265 sources to fund temporary subsidies, the amount of the subsidy would probably be relatively small and have little impact other than to anger farmers when it was removed. Our analyses suggest than changes in income and cash availability have a ‘larger' effect on demand that changes in fertilizer price. The only short-run policy initiative which could improve cash availability seems to be credit. Agricultural credit, however, is fraught with problems in Senegal. In establishing criteria fer identifying trial zones, information on past credit repayment performance should go hand-in-hand with informa- tion on economic returns to fertilizer. Several ISRA/BANE programs have been collecting information on informal credit arrangements used by farmers and village cooperatives since the dissolution of the PA. Records of the PA might also provide some guidance. The success of a trial program and its replicability will hinge on (1) whether farmers who purchase fertilizer realize an economic return on their investment, (2) whether the farmers perceive the economic benefits to be greater than what they would have realized with other investments and, (3) whether credit at all levels of the system is reimbursed. At the same time a program of targeting fertilizer to zones of high response is launched, a parallel effort must be made to increase farmers’ incomes in zones where fertilizer response is low. Our re- search suggests that many farmers in these zones lose an important share of cereal crops to disease and insects. Agronomists point out that better quality peanut seed can increase peanut productivity more than fertilizer in these zones. Given farmers’ current preference for 266 livestock over fertilizer investments, programs to improve livestock rather than crop production might be appropriate. 3. RESOLVING THE LONG-RUN POLICY DILEMMAS The fertilizer policy dilemmas that defy short-run solutions tend to be ones that require more research and better knowledge of produc- tion and economic relationships. (A) BETTER UNDERSTANDING OF THE CUMULATIVE EFFECTS OF FARMING NITHOUT FERTILIZER IN ZONES OF LOH RESPONSE Our Boulel analysis already indicates that there is little eco- nomic incentive for using fertilizer on certain crops and in certain zones. Agronomists have argued, however, that fertilizer should be used to replace minerals consumed by a crop, even if it is not profit- able in the short-run, because the long-run costs of not using fertil- izer will be greater. The best way to resolve this argument would be to design studies that permit us to compare economic returns over time for fields receiving and not receiving fertilizer. It is possible that some data may already exist from PAPEM demonstrations or UE work which would permit such analysis. Both of these programs, however, have been terminated, which means that some systematic method of collecting such data should be instituted imediately, particularly in those zones which have not received fertilizer for a number of years and are not likely to use much in the future. This type of research will be less critical for zones of high profitability and low risk if the GOS does manage to renew fertilizer consumption in these areas. If not, the same type of research should be conducted in all zones. 267 (B) FINDING ALTERNATIVES FOR ZONES OF LON FERTILIZER RESPONSE Targeting limited supplies of fertilizer to areas of.high profit~ ability does not help farmers in zones of low profitability. The GOS has three options in these zones: (1) search for new crops and vari- eties which are more profitable, (2) look for ways of increasing income through non-crop and non-agricultural activities, and (3) relocate the population to more productive zones. The agricultural research estab- lishment must shoulder the responsibility for discovering solutions which respond to the first option; pursuit of the remaining two op- tions is largely beyond the competence of agricultural researchers, although there are areas of overlap.1 (C) MONITORING FERTILIZER USE AND RESPONSE IN TARGETED ZONES In retrospect, one of the major shortcomings of Senegalese fer- tilizer research programs was that fertilizer response studies and systematic updating of economic analyses were virtually abandoned after the 1960s. We have tried to compensate for this omission by using the AF data sets to update economic analyses and examine risk. The discon- tinuation of the AF program in 1982 has eliminated the last source of data available for monitoring fertilizer response over time and space. He believe it is essential to reinstate some type of research program which responds to the needs of both economists and agronomists to evaluate inter-annual and inter-zonal variability in fertilizer response. 'h) assure that such a program is sustainable from year to year, great care should be taken in designing it to be low cost. This 1For example, more work needs to be done on the potential for increas- ing livestock production in zones which are no longer very profitable crop producers. 268 will require close collaboration between agronomists and economists to design a program that collects only the most essential information and limits collection to zones where fertilizer monitoring is deemed criti- cal. Nhether these data should be collected through trials, demonstra- tions, or on-farm surveys must be discussed. Information obtained from the data analyses recomended above should provide guidance on this matter. Hhether the monitoring program should be conducted by ISRA or other organizations such as the regional development agencies must also be evaluated. The extent to which the effort could be piggy-backed on to other on-going programs requires investigation. A program to monitor fertilizer response should be complemented by research to collect and analyze data on investments that appear to compete with fertilizer, so policy makers can be presented with a better picture of the constraints to increasing fertilizer consumption. Careful thought should be given to the types of analyses which will be performed before the data collection effort is launched. The economic and risk analyses used in this paper are relatively easy to perform. More sophisticated methods of modeling risk and its effect on input demand are available. Linear programming is a particularly good method of analysis for examining the profitability of fertilizer relative to other investments. Unfortunately, linear programming is a very data intensive method of analysis which requires detailed information on farm input/output relationships. If well done, however, a linear programming model which responds to fertilizer issues could provide 269 guidance in many other policy areas as well.1 Now that computers are readily available to ISRA researchers, we should be looking at the feasibility of doing more sophisticated types of analysis than have been conducted in the past. This will require greater coordination among the various departments of ISRA in the design of research pro- grams so that unnecessary and costly duplication of data collection efforts does not occur and, when appropriate, the data produced by one program are made available to others. 1Martin (1988) has developed a set of regional linear programming models for Senegalese agriculture which could serve as a base for developing models which look more specifically at fertilizer problems. APPENDICES APPENDIX I DATA FOR FERTILIZER DEMAND ANALYSIS Appendix 1* Date Used in Analyses of Sine Seloue Peanut Fertilizer Conewption for the “Program Agricole" Period 0 0 0 b b c c c c 1963 16 366 371 8052 8162 617 623 865 896 12 1O 10 23 1966 18 363 366 7889 8052 600 617 862 865 12 12 10 22 1965 20 656 363 10032 7566 698 600 1015 862 12 12 12 22 1966 28 680 656 10560 10032 526 698 1031 1015 13 12 12 21 1967 660 680 9660 10560 503 526 960 1031 16 13 12 18 1968 17 361 660 6698 9660 606 503 760 960 16 16 13 18 1969 6 327 361 5886 6698 377 606 723 760 12 16 16 19 1970 5 201 327 3819 5886 336 377 858 723 12 12 16 20 1971 6 216 201 6320 3819 290 336 786 858 12 12 12 26 1972 10 326 216 7776 6320 610 290 911 786 12 12 12 23 1973 9 279 326 6617 7776 339 610 803 911 16 12 12 30 1976 16 229 279 6870 6617 265 339 577 803 16 16 12 62 1975 16 388 229 16296 6870 398 265 926 577 20 16 16 62 1976 23 559 388 23678 16296 577 398 1159 926 25 20 16 62 1977 16 518 559 21756 23678 536 577 906 1159 25 25 20 62 1978 16 199 518 8358 21756 237 536 656 906 25 25 25 62 1979 1 367 199 16576 8358 379 237 786 656 25 25 25 66 1980 16 177 367 8162 16576 260 379 683 786 25 25 25 50 c c d d 1963 22 22 5 5.7 6 1.917 2.2 2.2 2.2 631 690 l 1966 23 22 6.8 5 5.7 1.833 1.917 2.2 2.058 689 631 1 1965 22 22 6.8 6.8 5 1.833 1.833 1.917 1.875 876 689 l 1966 22 22 6.2 6.8 6.8 1.615 1.833 1.833 1.833 679 876 1 1967 21 22 3.6 6.2 6.8 1.286 1.615 1.833 1.726 908 679 1 1968 18 21 3 3.6 6.2 1.125 1.286 1.615 1.651 955 908 0 1969 18 18 6.1 3 3.6 1.583 1.125 1.286 1.205 652 955 0 1970 19 18 6.3 6.1 3 1.667 1.583 1.125 1.356 862 652 0 1971 20 19 5.3 6.3 6.1 2 1.667 1.583 1.625 685 862 0 1972 26 20 5.1 5.3 6.3 1.917 2 1.667 1.833 773 685 0 1973 23 26 6.9 5.1 5.3 1.875 1.917 2 1.958 623 773 0 1976 30 23 6.9 6.9 5.1 2.625 1.875 1.917 1.896 505 623 0 1975 62 30 5.5 6.9 6.9 2.1 2.625 1.875 2.250 565 505 0 1976 62 62 6.6 5.5 6.9 1.680 2.1 2.625 2.363 761 565 0 1977 62 62 6.6 6.6 5.5 1.680 1.680 2.1 1.890 568 761 0 1978 62 62 6.6 6.6 6.6 1.680 1.680 1.680 1.680 639 568 0 1979 62 62 6.9 6.6 6.6 1.860 1.680 1.680 1.680 636 639 0 1980 66 62 5.1 6.9 6.6 2 1.860 1.680 1.760 666 636 0 *Hotes and sources on next page. 270 271 Appendix I, continued... NOTES: (a) Data in thousands of metric tons (b) Data in kilos/hectare (c) Data in FCFA/kilo (d) Data in millimeters/year SOURCES: Production, marketing, and rainfall data are from Sene (1981). Price data are from Cissé (1967), USAID (July 1983, Table 11),and USAID (1983, Annexe E. Table 3). Producer prices used are gross prices which include amounts withheld from farmers to cover various taxes, debts, etc. APPENDIX 11 CALCULATIONS FOR ESTIMATES OF TRACTION AND EQUIPMENT COSTS APPENDIX II CALCULATIONS FOR ESTIMATES OF TRACTION AND EQUIPMENT COSTS Depreciation Annual Costs Equipmenta Price Lifetime Themes Légers Themes Lourds Oxen Yoke 2,000 5 years 400 Seeder 84,780 15 years 5,652 5,652 Disks (2) 3,530 5 years 714 714 Hoe Sineb 60,330 15 years 5,170 "Ariana"c 180,310 15 years 12,020 PN Lifter 14,520 15 years 968 968 Subtotal Equipment 12,504 19,754 Maintenance (50% depreciation) 6,252 9,877 Depreciation on traction animals 5,000 - 0 - Maintenance feed traction animalse 21,900 29,200 Annual Fixed Costs 45,656 58,831 Maximum possible cultivated area 10 ha. 10.5 ha. Fixed Cost/year/hectare 4,566 5,603 Variable Cost/year/hectaref 1,094 3,500 Total Cost/year/hectare 5,660 9,103 (Cost SP2 - Cost SP1)/year/hectare 3,443 (Cost SP2 - Cost SP1/rotation/ha. 13,772 Source: Adapted from Tourte et al. (1971, p. 647) using informa- tion from Raymond, Monnier, and Cadot (1974) to estimate animal feed costs. Notes on next page... 272 273 Appendix 11, continued. aAll equipment costs are November 1987 prices quoted by SISMAR. bThe Hoe Sine costs include the frame, 3-toothed hoe, chaines, reins, etc. cThe Ariana costs include the frame, a simple plow, mounder, and 3- toothed hoe. due have not included a salvage value for the oxen, even though they are usually sold for more than the purchase price, in order to maintain continuity with the Tourte et a1. analysis; another problem was inadequate data on salvage value. eEstimates based on maintenance rations suggested by Raymond, Monnier, and Cadot (1974); only costs of peanut hay and sorghum grain were included as no price information is available on other feed (crop residues and field hay). fVariable feed costs are based on 350 grams of sorghum per horse-hour worked and 1 kilo of sorghum per oxen-team hour (Raymond, Monnier, and Cadot). APPENDIX III GENERAL SURVEY QUESTIONNAIRE APPENDIX III GENERAL SURVEY QUESTIONNAIRE F110: 1. RESEION BEN Inert: FICIE EXPLOITITIU 8010:1100: ”MINE” 00! ME 000 EXPLOITAIT mm» COIJUI. ll REPONIMT: SECTION VILLACEO 15E : EDUCATION CC: '0 LETTRE DANS EXP: POPULATION “[85: 36 ms: ms: “ICONS: FILLES: VIEUXIIEIKS: CIEF m DEREK]: COORCAIIAVETME: W11“ 85186: VILLAOE8 274 ETIIIE CIEF W CARIE: AUTRE ACTIVITE CC: 68E CHEF IU CAME: museum: COIIUNAUTE MALE: VILLAOE: Page 1 “TE: 275 1. cumin: 1. no: man mm m mm cm In to a warm sum as m mun Mum m: u "mm ummxtulim. m-mmmuumnmmmucmmmmuxtantalum 1mm um um I! nun man can In: A to mm: L: mm mm? 1. PI- “ rm In LA 11111 mum I11"! P01 ETE 3811“ "I LE mm. m 11 CE mm N'E1611 111 ”1 N 1.11 I m Ii 11 1'1"" 121 I mm ““1 PEI ram 11‘ 16 mu“. 11 PM! LE1 flC1Ell1 P61 ”Huh LE m 111 .-'1L 6 “IE a "C19. 1"“!1. 11. 1‘11 mm CE mm 1601 LA L111E C1°K11|11 A 10 PLACE MICE. mm m W100“!!! um mum: m in mm: mm rum: It tun mmxm «mama-tum mutt: mm mm It mu m room: an: at umxa. mm a mm mm mm 1. nmavutvmm:umm. MEIETOITLACNEEPIIKflm 11 MICE M11! 11 MIETE P11 m 11 1111,11“! 61 IE1“! 1“ LA ”1111”!“ 11 N1! 01 PM K PIOILEI mm c. m a mu u an: mm mm that: tx-mn 1. 11 m «a EU ”1 mm: H 861E11EL. an: El 11611 LA CAUSE H1K1PIE? 11 m 1! m PM 0 “TE! 11 H1013 m '61 "CPI”LE 11M1E11E1. I11 El I“ N VEEIO 11 m I mu: m LE1 ”3611.1 11 m I 1E1V1CE1 3 196161108 11 1EEV1CE1 I IEPWHU 11' ROI” 71 I115 11 '61 IE mun 1W1ll1 1. 11 m ”E! El ‘1 m1 I'M". KELLE El E1111 Ll CAOIE "1‘17“? [1'611112181‘121'11118'13‘ ”NICE!" 116111! OIPIEIEPIOILE‘ 7. “111. E11 LA CITIAIITE "1‘17“ '1 VII CHECK I’MITEI LE1 W1C!“ CILT1VEE1? 11 CEIEKE1 21 111E113 11 m I 11611 61 M11 I'm 11 m I m 61 m 71 P61 I CflTIAIflE 1. “LL! E11 Ll mm man 1981! COKI CI-IECCUfl 276 11. M111" flICflE 1. PI” 1W1 LE1 mm 1M1 LA M111“ AIlCIE ILIT 11W ”111 1110. “L E11 CELOI I11 A El L'1'AC1 LE PLOI m 1“ I15 P109511” 1. A mu: A111. at 1‘11'1L 111‘ m m CE mum 1. 11 LA “FUN m L'111L11A11I I1 “111. m PEOT'I mm LE WENT? 199'PA1 OE 1111111 1.IELITLEWIHAEOLEPLII'EPPETAPIRCELIIKIUHCI'IW 1. A VOFIE A111. II 1811-1L 11111 PM um CE PIOILE‘? 1. 11 1W1 I'AV1E1 I'I m1! PAII1 1'11 LE1 W1 I P111111! A1111” M11111 1M1 LA LISIE 1011M“. LEIIL CN11111EZ'W 11 M111 LE P111 I'M!“ 1W1LEI1E IIIIAIAEI A 121 CPA!" 11 19111111 L’AI1EI mm 11 100 11LOI111E m LA 111111I111I K1 mu AVEC LE W1 A 140% 11 CIT1IJEZ LE 111111 11 LA 1111!! PM LEI MES AVEC a! 111E111 I 21 CPAIIO CIAEACIAUEE 1PM m 10“ W1M1EE I IECEKIA 11.”. C11 PA1EEI1 11 111 11LO 11 115““ EA town. CE 1111 1191111 A 110 CPAIKO 7. 11 1W1 I'A'IEZ I'I cum PAII1 1W1 LE1 mu I “1111! A1111“ 11511111 DANE LA L111! 101VAITE. LEILEL C1011111EI'VW17 11 AUIEITEI LE P111 I'AIACNII 11111LEA1E IIIIAIIEI A 121 CPAIII 21 1911111 LE P111 I L'EIIA11 A 11 CPAIKO C'1I1 ET LE KTTIE El VEITE IAIS cm COINIAOIE MAL 11 mm LE C1E111 PM LE1 M11 AVEC I P111 K W A 121 CPA!" 11. 1AC I'EAC1A11 CWTEAA 1211 CPA. 11111 N 11E111 M111 1. 11 M1 I’AV1EI I'I CI“ PIII 1’11 LE1 W! I M111“ AIICIE m: IAI1 LA L11TE 1111VAI1E. LEIIL 0011111114”? 11 AOIEITEI LE P111 I'AAACIII 1N1L111E “11111111 A 121 CPAIIO 11 mm I mm I C111" P” LE1 ACAAT I1 AATEE1EL 11111111 Cl 111‘ 11 ELA1111 LE m I name: PIOEAI 11 ETAI11 I 1111“ K C1011 PM LE1 ACNAT1 IO IA1EliEL PICEAOI 1. IAIITEIAIT. PAII1 LE1 11111 C1011 II 1W1 AVE! 1111 El “111" 1 A 1. L111“ PIEPEA1EI'VWS 11111 1W1 LE1 mum 1C1” 0 1. 7. 11 11. LEIEL I1111EZ'VI11 El mm PLACE? ICO‘1 ' 1. 7. 11 277 111. L’m11111' E1 L'111L11A11M ‘1 11111111. LE IA1E11EL E1 LE 11E111 A. mum - 1. w I: ma» mm um um am 1:! $0111 um I; u 111111I111I Imcmu an an :1qu - I .......... I I Imam Io Immune I: III II 1! MI I II I I. was: minim. “mm mun mus-II 10!! II II III": I I I. II mm: ICC"!!! ms m was, mm mm manor: (KG «mt/«3 I I. commun:! II mp: III A u mm: II II 11960 it“ “L55 1f Or I I. II IquoIIm m mm um I. It III: III n W 1111 I. I: nut Isl: mm 1111 mama: mum: In IoIm III mums mm IIuoIIz-III-IIIIHIII-tmcmcwuu 1 11 I 2. nun 1110111111" K1 1111A11 I mm H mm E1 A 111111 1111111 . . . . - . 2 2 48" .. Z 3 cm: ACK1E1 1 ms 111M1 1 1AA1£1 ECNHOLEES 1111111 11101111111 Aw I mm mm. N111 1 P111 1 ”All 2 M11 M11 2 P111 quu 111111 11 11“ ”MENMA 11 S CFA 1111M!NMA ME 3 11 1 11 5 film 1 11 EIEIIMSEI ' T l I .' J . 1 ‘ 1 I" N 1.1. | A l .I -- j I 1 - - ' 1 ‘ I ‘ I i I i IE1 P111111 1 J 1 l 1. 1m ‘1 111L11A11~ 1111111 1! W 1m in IE 1101 11111 1101 FOAAE 1E 11111! 1“ “1‘1 mamas: 1111! 11111“ IRWIN!“ k1lm Emsum IMWIIEAIF 278 1. m1 ““1 PI LE “P 1'E1PL111111I 1111111 11K1“ 111111“ K SEIEICE1 mu: CE11E All“ 1W1 PM K 1“ 1N1 PM K 111118 KEN1110EE1 mnmm rum cmmm 1 1. 11 m I'AIEI P11 IEK K1 1111111K1 1E11E A“ W'ALLEZM P111E L'1IKE "RINK? 11 “1E1 K1 and: 11 I P11! I! K1 111.1.“ mums 11 1111E 11 P11 1'1KE ’11 1'11 1E! K1 ““1 I.I'ILrummummmsumummzmxmnmmmxnmmmmr 1W1. m. 1'PA1 1'1KE 7. 11 a. K'KLEZ-m PAIR? 11 K 111K! IE LE1 1111A“! K 11A11 11 m 1W1 LE1 11111111- 11 VEIIIE 1K P1111E 1E1 M11111 1E 11A" 11 111L11E1 LA PA1LLE E1 LE1 11K1 11 11111E W 1‘01“: 1'01". 1'10” PM P111E PACE ”1 m K PAK. KP111 L'1l11111K11fl K LA 11 11. 111L11E1'M1LE11EKRE171fl11PflE1111117 11 m I'IIEE m mum am It mu 1 1. MHZ-1W1 PLUS K PAK N'AVM1 m ”11 LE1 ”1111117 1 11. “Elm 1111!! LE “IE 1'11111017 I 11. 111L11EI'11U1 1111111K LE1 11K1 K 11L N K 1111 m L1 “11” K1 111111117 1 11. 11 max-m 1111111K LA PA1LLE 1111‘ PARKE 1111 LA ms: PM L1 ”HIRE 1E1 11111111? 1 11. Kml'm PL11 1E '11 “111111 1111 II MW 11’11111? A 1 11. mum-ms PL11 K 11EK PM 11111 11 PMAEE7 279 1. m1 L1 man 1111” m 1111111 1111.111 P11 '1 1111111111. PIIKIP“ 111111111- 111.1111! 11K 1131111 1111 1'1K1111 ”111111 11111111 11K I IN UK 1'11“]1111. P11! 111‘ 111111 N K1111 111 111.111 11K 11 L1 P11111111! 1111 PM“ 1111 1’11111111111 1111 1‘U11L11111111 1111 1'111L1111111 u mum. m u 11L H mm 1: 19011111 Lox-nut u 11L 2: mm was 2: man: 21 mm «I cum: 2: w 1: cm: 2: meme u m a mm: 3: ms 3: mm can! 3: can I'm 3) mean: rmxum u u mm 0 cm: 0 mm mm 4: an IE m 4: 1mm»: tamer: s: was new a: In a: comm a» scum a) 3mm 5: cmnm rm: 5: um a non .1 meme mm 7: am mm on n __ 7: mm 7: PEIIE 7: com I: can no can: u __ u n u m 9: mm: mm n __ n 9» moo m um I: m m m m m m rum scomm m w tram nu ma mm m 1 mm m m ——d 280 1.N1K1'MK11N11I1111KL1K11IK111111K£1 11 K111 11 K1. 11L. 11 can 111'. M. 11 M1K 11 P11K11 71 P11K11111111K1111 1. 1111'!” K11 L11 M11 111 Il111'1L PM" II 111 11 K1111 111 P11 K11 11 P11 11111 111 1111111 1. P. W! 1 L'A‘! K1111K. 111‘“ I! W K1 M111 [1111 111 K1LLEII 111 P11! 111 P11 K 11PPEKK1 1. 1'1L 1 11111 “ 11PM. I‘LL! 11111 LA 1111“ P11K1P“ 119-P11 K 1131111111 1. mmuuxmmummnummmxsmmnmmum? 7. 1.1111 K 1111 1K1°1fl1 am 1. can!" 1111-1” K11K II 11111 1111111111 111 L1 [1111111 1 111 1.111. 11“” 1111-1” K1111 sun L1 ”111111 my 1. KP111 1911. PM P111! P111 1111 P111LEK1 1'1“”11. 111111114” m K P111111 W "11111! K M111 N'111111 II'W1. mm: 11. 11 fl. mum (”'111L111 PL11 11'111111 11. K'L11 L1 mum 1111111: “11 I'M" 11 111011 K L1 m (11 1'11" 11 1 1111111 ' 11I1 11I1 11!! “R m K 1111 1'1'111 111(111 PK L! 1.11. PM1KII 11'1U1K ”MI ““1111" 10111911111111 «111K - ”1 P11 1'11111 P11! P11 "U 111 R1111“. 11111111 1111" 11 1111” PM 111111! mm “11 «um P1IAK1 111 111111? IIOKVEIJ "4111“” 1.11111 1'M111111 1'1111111 mm 1111“ 1‘1111! ”1 P11 1'11111 11. town-ms K m “ 1111 K L1 11PPEKKE 111K L1 KIKKIT K111 '1'“ P111 11111 11 N KK 11 1K11£1 1'11111111 1111 1111111 11 KLU1 ‘1'" P111 11111 11 11 11K 11 1111111 1111 11111111 1PM ”1111 PK111£1 L1 V111£1E.1 11. 111K 191111 11111 1111 1' L'111L11111I K M111 K 1111 1’EK1111 PM L11 11 111111 1'111111K1? 11. PINK-1W1 K nun K 1K1 K LA 11m 111K L1 K1KK11 U111 11'“ P111 1V1" 11 N 11K 1 111.1 K 11L1111.1 M1 W111 11 mu: Il'fl K11 11111 11 I 11K L1 KK 111M111! 1M1 (”11117 11. N1K KP1111 11111 11K 1. L'1111L11111fl K M111 K 1111 I'M" PM L11 1 11L1 K KKK“? 281 11. L'111L11111fl K L'1K1111. L'11111111 K 11.111. L1 1111111. L1 L1” 111 K 111L1 11 L1 1111 11 ‘111KK 1N1 1W1 K111K1 1 11111L1111 L1 11111 11 M 1’1111111 K 1K1 11111K111. PINK-1W1 1L11111 1I1111 111 K1101“ K L1 K1LL11I1 1 L1 N111 11111111. K1LL11I ”111 11111111 17. 11 L1 PLW1W1111 1'111 111 m. 111111K1 K 111 K1111K1 11.111111 nus K 111111 11! L11 1111111 P11! L11 11111.11. I‘LL! 111 L1 KflflK L1 11.111 11111111111. m L'111111111 m L1 11L? 11. I‘LL1 111 11LL1 1111 11111 11111? M L'111111K? PM L1 11L? 11. 11 L11 1111111 m 11 V1111 1 11 111I11L1 111 m K L1 11.11111L1111111 11111 11111 11 L1 1111111 111 11111. M111 K 111 115111124”? 111'?“ 1'11111 11. 11 L'1I11 111 11 11111 P" 111 111II1L1. 1K1111 K 111 “11112-1”? 111-P11 1'1K11 1. 11111111. W 1. 1112-1” 110 K 11111111. m L11 ”11111111 1111111111 111.11W11 1. 1912-1” V1111 K 1111111L K1111 1111 1111 L'1V111 mm 111 W1. L111!“ 1. 111111 1111 1112'“ 111311 N M11111L I1 K1 111111 K1111K117 111 I11. L111£L1 1. “1°11 II 11111 1111111L 1 K1111 1'1111 111111 PLU1 W111 K1111 1111 II 1111 L11 111111 1111111141! 1. 1111-1” K 1111111L ”411L111”? 111 I11. L1IEL1 1. K1111 1111K1111 ° 111111 1. K1111 11” 11111 P11Ifl11K ML! 1'1-1LL1 11mm 11111 PM M11 V1111 F111LL11 1'I11. 2W 2. K1111 1111 V1111 "OM11" M11 1°1°1LL1 1W1”1 11511 m "111 91111 F111LL1 11 M 1 ”I111 L'11K KM111117 “W1. m 282 3. PM u mm: X! 3 “HES “I! mum “I m CM 32 2.! KW I 222 “I“ I! "M! L! cum. (22 DE!“ II" "8 now: 2.! cum. N 232 PM can: LE CID" MI 2 cm 2 Ali!" 2 I... I!" 2 m: 2 mm mm 2 2 2 2 2 0 W20 2 2 2 2 : mum 2 2 2 2 2 M28 2 2 2 2 2 ”I!“ 2 2 2 2 : MI! 2 2 2 2 2 2. m u PERIOD! K8 5 III!!!“ “I! “I"!!! “I! am can: 8! L! m A W W I" El “'5 I.“ I]!!! EMS M LI um I m Ill mum “WW: II) ' m. (22 l I" El “5‘. 232 I! WW, I! I“ ll WE. It“!!! cm 2 mu 2 menu. 2 All“! 2 0‘2“.“th 2 mm 2 mum 2 MI“ 2 I!“ 2 2 2 2 2 2 2 2 “ll! 2 2 2 2 2 2 2 2 m 2 2 2 2 2 2 2 22 mm 2 2 2 2 2 2 2 2 mum . 2 ’2 2 2 2 2 2 2 mm IMHIEL 2 2 2 2 2 2 2 =2 3 NM. Mil: 2 2 2 2 2 2 2 2 m 2 2 2 2 2 2 2 -2 WM 2 ‘ I 2 2 2 2 ’ 2 2 .2 m I "A" 2 2 2 2 2 2 2 ,2 W K norm 2 2 2 2 2 2 2 ' g' 8. SIWIKIWIIMIIINIS IMWflOEYEUfllm 22 I!“ W rm 22 “I"! m PU"! 32 “I" K MN"! 0 “M! K! ms 8| MI! ”I mam P“ m I. 82 M "II «no KI “It“ 2”" 6mm 30" I LA mmmm WICIELLEI W! "ll. mu II | m U mm II It" M MD 12 mm 2" PH"! 3| KIM K M"!!! 22 WW"! M!!! 32 I!!!” IS 22.0125 62 Mill! ”2 ms: "8 VIII!!! 283 IV. IIVBHWI I! nu me I" 23.0“ C" "MILE! m N xmxm II m I) III I I“. com: LE8 mam-m "12.2853? mu .2 2.3 mm (”2‘ M mu I!!!“ m um I“ MUCH". m LE8 Mill-m "In!!!” 2|.l.. 8! u m CPI!” um I [MSTIMI "4.2:”. mm: COME" LE mm UTILISERMI lEI 23,000 “A 3'“ “I" ”L"! I'lmfll M L'MICIIJIIIJ mu I! 2.! ISO“ mm M '0"! KW!“ um EU" I!” “"8"". COME" L8 MIH'VWS 012L256? mum 82 I! m WM! m 23.00. C" III um I! N28 K I“ m “RYE! DE mu: EWIPEIEIIS. IE5 PIECES “TACK“. 00 PM “PM“ '08 INNER!" ”182M". '0'“le "I" m1? 8! 1! was ”A” m 23..“ I" W um N ”23 K M! PM mm K8 W28. MIEZ'VWS “I! L’EMIS IIL. PM" “III. um. II a: CWIWSW mun 284 I. AEUVUEE "4'16“ 2. AV“! 2’" (LA "I I2 m AERICIE2 WEE ACIWITEE EXERCIEZ'VWS El m I LA CIIIIE? “MENU!“ mama. MAE MHMH Em ”VII ”VIE ELEVACE "AVAIL “HIE IAIA EMA I2 EMILE m 1. mu 2’” Efl-CE I! m m K CE! ACTIVITES M 1mm. 2'22”“. EIEOIT RES!!! A22 IRE IWEAU MAE I'AflWflE 3. 3! LEE man: an "Allis. I‘LLE El E82 LA IA!“ HIKIPALE‘I 22 N2” l’llVESHEEEKIY N’AVMT 22 mm IE 1mm PM LE P3022222 I2 “NICE 32 AUTIEI ”2 PA! I'ACUVITE I2 IEVEII PAS "III! 4. awn-ms "All" I m ACUVUES M'AHXCRES DEN" "IO "WI. mm 32 I22 1mm LEEUSLLES 5. 32 I22. MINI AVEPVM m LEE ACYIVHEE? 299‘?“ DE WE “2292252 APPENDIX IV DESCRIPTION OF VARIABLES EXAMINED IN LOGIT ANALYSES APPENDIX IV DESCRIPTION OF VARIABLES EXAMINED IN LOGIT ANALYSES* VARIABLES REFLECTING FARMERS’ CIRCUMSTANCES Afifi AVPQP FOODPROB N N GACT OTHREV 1 if principal farmer < 50 years 0 if principal farmer 2 50 years 1 if farmer owned herd of z 15 ruminants or 2 5 beef cattle during most of 1981-85 period 0 otherwise AVAREA/AVPOP Average cultivated area during 1981-85 period Average farm population during 1981-85 1 if farmer mentioned lack of food as one of three major constraints during 1981-85 period 0 if food not one of three biggest problems if average farm size 2 6 hectares if average farm size < 6 hectares OH if < 6 hectares and < 1.5 active workers per hectare if < 6 hectares and 2 1.5 active workers per hectare if 2 6 but < 17 hectares and < 2 active workers per hectare 4 if 2 6 but < 17 hectares and 2 2 active workers per hectare 5 if 2 17 hectares wk).— 1 if farmer participated in non-agricultural activities 0 otherwise 1 if farmer obtained regular, reliable non-crop revenues during 1981-85 0 otherwise *Variables that are underlined are available for both large and small samples; variables without underline are available for small sample only. 285 286 PARC 1 if farmer had access to 'parcage” 0 otherwise EQPAQI Average active workforce for 1981-85 (ISRA norms used) EBQBMAI 1 if farmer mentioned lack of agricultural equipment as one of three principal constraints 0 if equipment not one of three principal constraints ZQNE 1 if farmer in Gossas 2 if farmer in Nioro VARIABLES REFLECTING FARMERS' ATTITUDES ABOUT FERTILIZER FAB 1 if farmer believes a fertilized field always achieves better yields than an unfertilized field, even if rains are very poor 0 otherwise FBAM 1 if farmer believes chemical fertilizer increases both millet and peanut yields more than organic fertilizers 0 otherwise FBMP 1 if farmer believes 3,000 FCFA spent for one sack of fertilizer is a better investment than 3,000 spend for either the manure of one horse during one year, or one month of "parcage' during the rainy season 0 otherwise FBS 1 if farmers believes that a farmer who spends 6,000 FCFA for two sacks of fertilizer to use on an existing peanut field would obtain a larger harvest than a farmer who spent the 6,000 on an additional 30 kilos of peanut seed 0 otherwise INVFER 1 if fertilizer mentioned as one of three priority uses which the farmer would have considered if he had been given 15,000 FCFA just before the rains in 1985 0 otherwise RISK 1 if risk preferring - 2 if inconsistent 3 if risk averse SA 1 if farmer had enough food and peanut seed, but no cash, and he would be willing to sell some small ruminants to purchase fertilizer 0 otherwise SONLY 287 1 if farmer had enough food and 20,000 FCFA cash but no seed or fertilizer and he would be willing to spend some of 20,000 on fertilizer 0 otherwise 1 if farmer believes that solution to seed, equipment, and fertilizer distribution problems is to raise producer price of peanuts 0 if farmer believes that price policy is not best sglution for at least one of the three inputs mentioned a ove VARIABLES DESCRIBING FARMERS’ PAST FERTILIZER PURCHASING BEHAVIOR CASHPA CALPA CONT if farmer bought fertilizer for cash during PA otherwise 1 0 0 if farmer bought no fertilizer through PA 1 if farmer bought average of < 5 sacks/year during PA 2 if farmer bought average of 5-10 sacks/year during PA 3 if farmer bought average of > 10 sacks/year during PA 1 if farmer has obtained fertilizer through contracts during 1981-85 period 0 otherwise VARIABLES REFLECTING FARMERS’ KNOWLEDGE OF FERTILIZER TECHNOLOGY DONTKNQ TIME 1 if farmer unable to quantify fertilizer response for either millet or peanuts 0 if able to quantify for both crops 1 if farmer states that the best time to apply fertilizer is early (before the first weeding) 0 if farmer states that the best time is after the first weeding LIST OF REFERENCES LIST OF REFERENCES Abt Associates, Inc. .§gnggg1gsg__flgy1ggltggal Policy Analysis. Cambridge, MA: Report prepared for USAID, Senegal, April 1985. Aldrich, John, and Forrest Nelson. Lingg: Probability, ngjt, and Ergbig_flggg1s. Beverly Hills: Sage Publications, 1984. Amin, Samir. d s f l . Paris: Les Editions de Minuit, 1969. Andersen, Arthur, and Gaye Associes. a i n stabilisation (1.95 prjx: Etggg diagnostic, Vol. 3: "Filiére arachide- huile." Dakar: Date and publisher not indicated, but about 1982 or 1983. Anderson, J. R., J. L. Dillon, and B. E. Hardaker. Agricultural Decision Analysis. Ames, Iowa: Iowa State University Press, 1977. Antel, John. "Econometric Estimation of Producers’ Risk Attitudes", Amgr, Q. Agr, Egon. 69(1967):509-522. Barnett, Douglas. A Stugy gf Eggmgrs’ gggls ang anstrgjgts; Their Effggts 1n: the Cultivation of Crops in Sine Saloum, Senegal Lafayette, IN: Unpublished Master’ 5 Thesis, Purdue University, 1979. Belloncle, G. "Peut-on Sauver les Coopératives Sénégalaises?" Paris: Unpublished paper, June 1980. Benoit-Cattin, Michel. Les Unités Expgrimgntglgs du Sénégal. Paris Dakar: Joint publication by ISRA, CIRAD, and FAC, 1986. Benoit-Cattin, M., and J. Faye. L’Explgitgtion agrigole familiale gn Afrigge soudggg-sahglignng. Paris: PUF, 1982. Binswanger, H. ”Attitudes toward Risk: Experimental Measurement in Rural India,” Amer. J. Agr, Eggn. 62(1980): 395-407. and Binswanger, H. et al. R ra Household St die in Ai . Singapore: Singapore University Press (1980) 288 289 Bockelee- -Morvan, A. and R. Vaillant. l'L’ Evolution de l’ utilisation d’engrais sur l’ arachide au Sénégal, ” in vLa o to ‘ _ f a“ 0 on W. Pars: 111111111957 Bray, M. 'Essai de Determination de l’ Effet de la Fertilisation Annuelle de l’Arachide dans la Region du Siné- Saloum". L’Agrgngm jg Irgpjgalg, 24(1969): 1098- 1108. 'Essai d’Evaluation de l’Effet des Techniques Vulgarisées au Sine-Saloum sur la Culture de l’Arachide". L’Agronomja Tropjgala, 25(1970): 192-204. Brochier. J. La_d1IIus1on_du_nr29res_1echoinue_en_milieu_nunal sagggalais. Paris: Presses Universitaires de France, 1968. Casswell, N. "Autopsie de l’ONCAD: La politique arachidiére au Sénégal, 1966-1980," £91111gga_Afriga1na, 14(1984): 39-73. Cissé, Ben Madi, R. Billaz, A. Bockelee-Morvan, M. Bray, R. Dorsemaine, G. Pocthier, and Young. 'Bilan des réalisations sénégalaises en matiere de vulgarisation d’engrais," in Le galloggg sgr la farailjgé gas sgls trgpjgagr, Tananariva br 7. Paris: IRAT, 1967a. Cissé, Ben Madi, R. Billaz, A. Bockelee-Morvan, M. Bray, R. Dorsemaine, G. Pocthier, and Young. "Motivations de la reponse a la vulgarisation: L’application des techniques a l’incidence pluriannuelle au Sénégal,” in Le collggue sur la f ' ité ‘ s tr i u ananariv Mad a car 9 a 25 govambrg 1262. Paris: IRAT, 1967b. Crawford, Eric, Curtis Jolly, Valerie Kelly, Philippe Lambrecht, Makhona Mbaye, and Matar Gaye. A Field Study of Fertilizer Distribution and Use in Senegal, 1984: Final ReggrL. Dakar: ISRA/BANE, Travaux et Documents, No. l, 1985. Also available as a Michigan State University Reprint Paper in English and French. Crawford, Eric, and Valerie. Kelly. Enggata sgr la gjstribgtjgn t ’uti is tion l’en rais au Séné al 1984: Résum Analysigua. Dakar: ISRA/BANE, Working Paper No. 1984-3, 1984. Also available as a Michigan State University Reprint Paper in English. David, Christina. of r i i m n th A i Ri e n ' r n ' . Stanford: Unpublished Ph. D. Dissertation (1975). Eicher, C. "Facing Up to Africa’s Food Crisis". Foreign Affairs, 61(1982): 151-174. 290 Evenson, R. "Food Policy and the New Home Economics,“ Eond Enlicy, (August, 1981). Fall, Mankeur. ’ u 9 - . Bambey: ISRA/CNRA, May 1981. FAO Fertilizer Program. [he first Qeeede. Rome: FAO, 1974. Faruquee, Rashid. Segrees nf Eertiligy Deelin ne; Feetnr Analysis of -C . Washington, D. C: World Bank Staff Working Paper No. 318, Februray 1979. Faye, A., and E. Landais. ’ m he v ann da le eentre-nord du bessin ereehidier en §énéQ§|. Dakar: ISRA/Departement Systemes, 1984. Forest, Francis. ”Evolution de la pluviometrie en zone Soudano- Sahelienne au cours de la periode 1940-1979,’I L’Agrenomie ‘Iremigele, 37(1982): 17-23. Franzel, Steven. i r d for Small Farmers: A Case Study of Farmingn Systems Reseereh in Kirinyaga District, Kenya E. Lansing: Department of Agricultural Economics, Michigan State University, Unpublished Ph.D. dissertation, 1983. Frélastre, Georges. 'L’évolution de la politique agricole du Sénégal," t de 0 i i i e & soci l africeines. 17(1982): 49-62 plus. Gaye, Matar. a u i s m 'er dan l c dr la ouv l Pelitigue Agricole. Dakar: ISRA/BANE, Memoire de Confirmation, 1986. Gazagnes, J.M., and B. D’Hiver. Typologie des exploitatiens agricoles en Sine Saleem. Dakar: SODEVA, 1978. Gellar, Sheldon. ”Circulaire 32 Revisited: Prospects for Revitalizing the Senegalese Cooperative Movement in the 19805,” in Ine Enliticel Eennnmy of Risk eng gnniee in Senegel, edited by John Waterbury and Mark Gersovitz. London: Frank Cass & Co. Ltd. (1987) Gersovits, Mark and John Waterbury (eds. ). The Political Economy of Risk end Cheiee in Senegal. London: Frank Cass & Co. Ltd., 1987. Gillier, P., and P. Prévot. "Fumures minérales de l’arachide au Sénégal". Belletin of the Reseereh Council of Isreel, 80(1960): 131-150. Gladwin, C. ”A View of the Plan Puebla: An Application of Hierarchical Decision Models, Amer. J, Agr. Eeon,, 58(1976): 881-887. 291 Government of Senegal, Ministry of Rural Development. Neuyelle Eelitigue_Agrienle. Dakar: MDR, March-April, 1984. Ministry of Rural Development. 511an_gleba1_des :. .111 o o "1 l_f. ‘ 011‘ ._ 0‘1 ’ n n a . Dakar: MDR, February 1982. Groupe de Réflexion sur l’Engrais. - Bilan gamnagne SSZSZ e; ijeetifs §7Z§§. Dakar: Mimeo, November 1986. Rannnrt Final: Pr i ns our r n d la c9ns2mmat19n_d_engrais_au_§énégal- Dakar: Mimeo, December 1986. Hopkins, E. W0 5 ° n ° f Agrieultural_£rtensign_Seneme. Sussex: Unpublished Ph. D. dissertation, University of Sussex, 1975. IFDC. s n o G n n Mill i " n in" S . Muscle Shoals, Alabama: IFDC, 1980. . W Afric n r ili r n al. Florence, Alabama: IFDC, 1977. IRAT. Nesess1té_agrQn9m1gue_et_intér§t_ecgn2migue_dLune int ' a ion d ric n . Bambey: IRAT, 1972. ISRA. La nelitigued e fer tili satign gu miletn de l’ ar raenide gans l C tr Sud d ri e e; perspectives. Bambey: ISRA/CNRA,1980. Reflerinn d’ensemble sur le nrgnleme de l’engrais n b n hidi . Bambey: ISRA/CNRA, 1975. l v F el Que_gues_2bser_at1Qns_sur;le_r§nnont_1_QC_n__ati£_a l’etu de de la reponse de l’ arachide et du mil aux engrais NPK l b n idier é . Bambey: ISRA/Service de Fertilisation et de Chimie des Sols, December 1979. Kelly, V. F ers’ D ma d for Fert i r in the Context of Senegal’s New Agricultural Policy: A Study of Factors Influencing Farmers’ Fertilizer Purchasing Decision. Dakar: ISRA/BANE, Document du Travail 86-4, 1986a. Also available as a Michigan State University Reprint Paper in English and renc . 292 h gentext of Senegal’ s New Agrieultur al Policv: The Im mnl ieatjnn ns gf Farmers’ Attitudes ang In nnut Purehasing eh or h i ri l P lie and R . Dakar: ISRA/BANE, Document du Travail 86-5, 1986b. Also available as a Michigan State University Reprint Paper in English and French. Kelly, V. and M. Gaye. c i tion Use the_Denartments_gf;figssas_and_fliorn Dakar: ISRA/BANE. Report prepared for USAID, 1985. Kingsbury, David. t l l ev vel t r m 5- 9 .0" 0‘ I ' I: ‘ “i '1 ‘10 0 LI -‘:c._] July 1985. Lipton, M. “The Theory of the Optimising Peasant." Jeurna1_gf Dexelenment_§tudies. 4(1958)= 327-351- Luning, H. n ec f w u - c Wageningen: Centre for Agricultural Publications and Documentation, Agricultural Research Report No. 699, 1967. Ly, Abdoulaye. ’ d i ’ t l n l: 7- . Paris: Présence Africaine, 1958. Martin, Fredéric. La Befurme de la Pelitigue cerealiere au SaheI: . A report prepared by Elliot Berg Associates for CILSS/Club du Sahel. Paris: OCDE, March 1986. feud Seeurity and Cnmnarative Agvantage in Senegal: A Micro- ac E. Lansing: Ph. D. Dissertation, Michigan State N_sr2_A22r9__h University, 1988. McPherson, M. Why an Res eareners Continue tn Igngre Risk in Tests of farmer Efficieney? Cambridge: Harvard Institute for International Development, Harvard University, Development Discussion Paper No. 160,1983. Nguyen-van-chi-Bonnardel, R. Vie ge relations au Senegal; La eireulatien_ges_niens. Dakar: Institut Fondamental d’Afrique Noire, Memoire No. 90, 1978. Niang. L.. and D. Sarr. Enguétes_sur_les_Leastigns_et_strategies n ve ’ u em n e e gtengrais, Kaolack: ISRA/Département Systemes, February 1986. Norman. D. ta_mgth2de_de_recbenches_sur_les_sxst§mes r' 0 ° ° b'l‘ i 293 . E. Lansing: Department of Agricultural Economics, Michigan State University, Cahiers MSU sur le Développement Rural, No. 5, 1980. __ A] 71101 I 1 1 - .OS 1 .’ 01 , Zaria: Ahmadu Bello University, IRA, Samaru Miscellaneous Paper No. 37, 1972. Ogunfowora. 01 abisi . W W i . Ames, Iowa: Unpublished Ph. D. Dissertation (I972) Pieri, Co '_ T 1 0100 a 0 ‘ 0-111 "ll f ' -f- WW. Bambey and Montpellier: ISRA/CNRA and IRAT, November 1979. "Bilans minéraux des systémes de cultures pluviales en zones arides et semi-arides.' L’Agrpngmie Trppieale, 40(1985): 1219. Pieri, C., F. GANRY, and P. SIBAND. "Proposition pour une interpretation agro-économique des essais d’engrais.” Wale 330978): 32-39- Pocthier, G. Bappert Annuel, Campaign IQSZ-S; (II AmeIipratipn Epneiere, (III Systeme ge CuIture. Dakar: ISRA, June 1983. Pocthier, G.; C. Pieri, and M. Mara. Test de régénératipn des lo m 976- 7. Bambey: ISRA, May 1977. Poli, A. "Les nouvelles structures de l’économie rurale du Sénégal." L’Agrongmje IrgpiCaIe 25(1970): 772-778. Raymond, C., J. Monnier, and R. Cadot. Etude de systémes teennigues de production pour Ie Sine Saloum sud et est. Bambey: ISRA/CNRA, (published jointly with IRAT and the Institut d’ élevage et de médecine vétérinaire des pays tropicaux), June 1974. Richard, J. F., Moussa Fall, and J. M. Attonaty. Le Mgdéle ”4S": Prpgramme Linéaire pour les explpitations agrienles du Sine m Sud éné l. Bambey and Montpellier: ISRA and IRAT, September 1976. Sarr, P. L. An des ff ' d ' "nt ns fi a eultures sur guelgues caractéristigues physico-chimigues d’un sel ferrugineux tropical du Sénégal. Montpellier: Thése de 3e cycle, agronomie-pédologie, Université des Sciences et Techniques du Languedoc, 1981. 294 Sarr, Désiré. Mise en pIaee ges seetions inIagepises: Etuge de ._ ‘ 'to‘ 111.0‘0' 111 . IL ._té'_ ._‘c' kaymer. Kaolack: ISRA draft paper, 1985. Schultz, T. Transforming [ragitipnal Agrieulture. New Haven: Yale University Press, 1964. Schumacher. E. W Senegal. Berkeley: University of California Press, 1975. Sene. 1. WWW araenidiere au Sine Saloum. Kaolack: Inspection Regionale de la Production Agricole, mimeo, 1981. Shalit, H., and H. Binswanger. Fertilizer Subsidies: A Beyiew of Policy Issues with Special Emphasis on Western Africa. Washington, D.C.: World Bank (Research Unit, Agriculture and Rural Development, Operational Policy Staff), Report ARU 27, November 1984. Shaner, W., P. Philipp, and W: Schmehl. Earming_Systems_Besearen n ev lo m nt: G l1 v o 1 0 tries. Boulder: Westview Press, 1982. Shapiro, K. “Efficiency Differentials in Peasant Agriculture and Their Implications for Development Policies," [ne Jgurnal pf , 19(1983): 179-190. Sheldon, V., and N. Le. Senegal Fertilizer Study fer Eegnpmie Suppprt Grant. Dakar: Unpublished Report Prepared for USAID, Senegal, 1982. SYSTAT, Inc. SYSTAI: Ine System for Statisties. Evanston: SYSTAT, Inc., 1986. Thénevin, P., and J. M. Yung. Evaluation de la filiere arachide au_Senegal. Paris: Ministere des Relations extérieures, Coopération et Développement, I982. Thiam, Abdoulaye. Personal Communication, December 1988. Thiam has several draft papers in process summarizing the results of various trials conducted during 1986, 1987, and 1988. Tignor, Robert. ”Senegal’s Cooperative Experience, 1907-1960," in Ine P i ic l no of R k and ho' e ' Sene l, edited by John Waterbury and Mark Gersovitz. London: Frank Cass & Co. Ltd., 1987. 295 Tomek, William, and Kenneth Robinson. ”Agricultural Price Analysis and Outlook,” in A Suryey pf Agrieultural Eepnpmies Literature, Volume I, edited by Lee R. Martin. Minneapolis: University of Minnesota Press for the American Agricultural Economics Association, 1977. Tourte, R. with the collaboration of G. Pocthier, C. Ramond, J. Monnier, R. Nicou, J. F. Poulain, R. Hamon, C. Charreau. "Themes Légers - Themes Lourds - Systémes Intensifs: Voies Différentes Ouvertes au Développement Agricole du Sénégal". Week 260971): 632- 645. Tuck, L. d f n R Dakar: Report prepared for USAID, Senegal, 1983. “Financial Markets in Rural Senegal," in [ne EgliticaI Ecgnomy pf Risk and Chgiee in Senegal, edited by John Waterbury and Mark Gersovitz. London: Frank Cass & Co. Ltd., 1987. USAID. P A nc v l l r n men nd 6 5-0 4 . Dakar: USAID, July 28, 1983. Venema, L. Th W f lo : So i . Wageningen: Centre for Agricultural Publishing and Documentation, 1978. Waterbury, John. "Dimensions of State Intervention in the Groundnut Basin,” in T 'tic nom i n e in Sen al, edited by John Waterbury and Mark Gersovitz. London: Frank Cass & Co. Ltd., 1987. Welsch, D. "Response to Economic Incentives by Abakaliki Rice Farmers in Eastern Nigeria," Journal of Farm Eepnomies, 47(1965): 900-914. Wolgin, J. "Resource Allocation and Risk: A Case Study of Smallholder Agriculture in Kenya," Amer, J. Agr. Ecpn., 57(1975): 622-630. mm 1 AAAAA IIIIIIIIIIIIIII’IIIIII/IIII 2921