Economic analyses of maize storage innovations in southern Benin

Economic Analyses of Maize Storage Innovations in Southern Benin Patrice Ygué Adegbola Thesis committee Thesis supervisor Prof. dr. ir. A.J. Oskam Professor Emeritus of Agricultural Economics and Rural Policy Wageningen University Thesis co-supervisor Dr. ir. C. Gardebroek Assistant professor, Agricultural Economics and Rural Policy Group Wageningen University Other members Prof. dr. ir. E.H. Bulte, Wageningen University Prof. dr. ir. P.C. Struik, Wageningen University Dr. A. van Tilburg, Wageningen University Dr. O.N. Coulibaly, International Institute of Tropical Agriculture, Cotonou, Benin This research was conducted under the auspices of Mansholt Graduate School of Social Sciences. Economic Analyses of Maize Storage Innovations in Southern Benin Patrice Ygué Adegbola Thesis Submitted in fulfillment of the requirements for the degree of doctor at Wageningen University by the authority of the Rector Magnificus Prof. dr. M.J. Kropff, in the presence of the Thesis Committee appointed by the Academic Board to be defended in public on Tuesday 25 May 2010 at 1.30 p.m. in the Aula. Patrice Ygué Adegbola Economic Analyses of Maize Storage Innovations in Southern Benin, 191 pages. Thesis, Wageningen University, Wageningen, NL (2010) With references, with summaries in Dutch and English ISBN 978-90-8585-639-9 This thesis is dedicated to my wife Elisabeth, to my kids, to my mother Micheline and the memory of my late father Adegbola ABSTRACT Maize is a staple food and an important source of income for farmers in southern Benin. It is stored at village level in traditional storage structures and treated with conservation products. To improve control pest damage in stored maize, improved wooden granaries and a new product, Sofagrain®, were introduced in 1992. On-farm trials indicated that after six months of storage, the losses were reduced from 30% to only 5% for maize treated with Sofagrain® stored in an improved wooden granary. Although the effectiveness of storage innovations against pests is well documented, little is known about the socioeconomic aspects of promotion of these innovations in southern Benin. Using appropriate econometric models, this study investigates the perceptions of farmers regarding the characteristics of storage innovations and the causal effect of participation in extension on their formation, the adoption of storage innovations and effect of sources of information on the determinants of adoption, the impact of adopting storage innovation on schooling expenditure and the factors that affect the abandonment of storage innovations. First, the empirical results show that the effectiveness against pests and the length of the storage are the most important preferred characteristics and are provided by the storage innovation. Second, farmer’s participation in an extension program on these storage technologies has an important effect on the probability that positive perceptions of the quality of effectiveness against insects are provided by the improved wooden granary and the Sofagrain®. Third, there are differences in adoption and modification decisions between farmers who are informed by extension agents and those informed by other farmers. Fourth, adoption of a storage innovation increases the schooling expenditure of adopters. Finally, the study highlights the effect of road conditions, availability of family labor and availability of the protection measure Sofagrain® on the probability of abandonment of storage innovations. Key words: Storage innovations, maize, information sources, farmers’ perceptions, adoption and modification, treatment effects, sample selection bias, correction function approach, technology abandonment, cross-sectional and panel data, Benin. i ACKNOWLEDGEMENTS Several institutions and individuals have contributed to the successful completion of this thesis. I am grateful to the Programme Analyse de la Politique Agricole (PAPA) for providing most of the data used in this dissertation. I would like to express my deep gratitude to my promoter Prof. Arie Oskam for accepting my application as a PhD candidate at Mansholt Graduate School of Social Sciences of Wageningen University and giving me critical comments starting from the draft research proposal till the final thesis script. My copromoter and daily supervisor Dr. Ir. (Koos) Cornelis Gardebroek is greatly appreciated for his professional guidance, constructive criticism on all of the chapters in this work and for translating the summary of this thesis into Dutch. Without him this thesis would not have been completed in this way. I am also grateful to Dr. Doortje Wartena for her constructive comments and editing the Chapters 3 and 5 of this thesis. I would also like to thank Dr. ir. Rob Nout from Food Technology at Wageningen University who recommended me to Prof. Arie Oskam. My deepest gratitude to Prof. Joseph Hounhouigan, Abomey-Calavi University (Benin) Food Sciences and Nutrition Department who introduced me to Rob Nout. I am particularly indebted to Dr. Delphin Olorunto Koudandé, Research Director of Institut National des Recherches Agricoles du Bénin (INRAB). Dr. Delphin Olorunto Koudandé submitted my application as a PhD candidate and recommended me to Prof. Arie Oskam in 1997. Since 2005 as Research Director of INRAB, he played a great role in my application to NFP scholarship and in the completion of this thesis. I am also sincerely grateful to Dr. Victor Manyong for his help in applications to PhD programs and scholarships and his constant encouragement. Many thanks to Dr. Ousmane Coulibaly and Dr. Aliou Diagne for the training opportunities they gave me on methods of adoption analysis and impact assessment and for their encouragement. I am indebted to the members of Agricultural Economics and Rural Policy Group for the good working climate. I highly appreciate Dineke Wemmenhove for the secretariat assistance and for organizing my travels, living rooms and other administrative services. Thanks Karen van der Heide for providing useful support in solving financial and logistic problems. I would also like to thank Ineke van Driel for her cooperation related to my PhD research project. My sincere thanks also go to Mirjam Oskam (with Arie) for the parties you ii have been organizing. It was a great opportunity for us, as international PhD students, to socialize outside the academic world. The contribution of colleagues in the Programme Analyse de la Politique Agricole was very helpful. I highly appreciate Souleïmane Adekambi, Alphonse Singbo, Aminou Arouna, Léonard Hinnou, Epiphane Sodjinou and Jacques Zinsou for their assistance at various stages of this thesis. I am also indebted to the staff of the administration of PAPA. Cécile Odouyomi and Alice Gnangnon provided useful support in solving administration problems and facilitated all my travels. I also wish to acknowledge enumerators who helped me in data collection and farmers who devoted their time in giving answers to the surveys questionnaires and discussions during focus-groups even though they did not know how they would benefit from this work. I am also greatly indebted to my parents, Micheline Olou and late Adegbola Biaou, for sending me to school. I see this academic achievement as a great success to you. The contribution of my family has been enormous. My sincere thanks go to my wife Elisabeth for your understanding and bearing the family responsibilities during the course of this work. To my kids, please follow this achievement as a model for your life. It is my hope that you will probably note that three things are important in life: Faith, Courage and Humility. The list of people whose kindness positively influenced my work is too long. I say thank you all from the bottom of my heart. Patrice Ygue Adegbola Wageningen, December 2009 iii TABLE OF CONTENTS ABSTRACT ............................................................................................................................................................ I ACKNOWLEDGEMENTS ................................................................................................................................. II TABLE OF CONTENTS.................................................................................................................................... IV LIST OF TABLES .............................................................................................................................................. VI CHAPTER 1 INTRODUCTION ......................................................................................................................... 1 1.1 BACKGROUND ........................................................................................................................................ 1 1.2 PROBLEM STATEMENT ............................................................................................................................ 2 1.3 OBJECTIVES OF THE THESIS .................................................................................................................... 6 1.4 METHODOLOGICAL APPROACH AND DATA ............................................................................................. 7 1.4.1 General analytical framework........................................................................................................... 7 1.4.2 Specific theoretical background and empirical approaches ............................................................. 9 1.5 OUTLINE OF THE THESIS ....................................................................................................................... 11 CHAPTER 2 DESCRIPTION OF THE STUDY AREA, POST HARVEST SYSTEMS AND SURVEY DATA ................................................................................................................................................................... 13 2.1 2.2 2.3 2.4 2.5 2.6 INTRODUCTION ..................................................................................................................................... 13 DESCRIPTION OF THE STUDY AREA ....................................................................................................... 13 OVERVIEW OF MAIZE POST-HARVEST SYSTEM AT THE FARM LEVEL ..................................................... 15 MAIZE STORAGE TECHNOLOGIES .......................................................................................................... 16 DEVELOPMENT AND PROMOTION OF STORAGE INNOVATIONS ............................................................... 17 SURVEYS’ DESCRIPTION ....................................................................................................................... 21 CHAPTER 3 ANALYSING FARMERS’ PERCEPTIONS OF MAIZE STORAGE INNOVATIONS IN SOUTHERN BENIN........................................................................................................................................... 24 3.1 3.2 INTRODUCTION ..................................................................................................................................... 25 ASSESSING FARMERS’ PERCEPTIONS ON CHARACTERISTICS OF GRANARIES AND CONSERVATION MEASURES ......................................................................................................................................................... 28 3.2.1 Evaluation method .......................................................................................................................... 28 3.2.2 Data................................................................................................................................................. 30 3.2.3 Estimation results of farmers’ perceptions of storage innovations characteristics ........................ 31 3.3 PROJECT PARTICIPATION EFFECTS ON FARMERS’ PERCEPTIONS ON STORAGE INNOVATIONS CHARACTERISTICS ............................................................................................................................................. 34 3.3.1 The modeling framework................................................................................................................. 34 3.3.2 Description of variables included in the model ............................................................................... 38 3.3.3 Data................................................................................................................................................. 43 3.3.4 Results of impact assessment of farmers’ participation in extension program ............................... 46 3.4 CONCLUSIONS AND IMPLICATIONS ....................................................................................................... 53 CHAPTER 4 THE EFFECT OF INFORMATION SOURCES ON TECHNOLOGY ADOPTION AND MODIFICATION DECISIONS ........................................................................................................................ 58 4.1 INTRODUCTION ..................................................................................................................................... 59 4.2 MAIZE STORAGE AND CONSERVATION SYSTEMS................................................................................... 61 4.3 A MODEL FOR ANALYZING ADOPTION AND MODIFICATION DECISIONS ................................................. 61 4.4 DATA.................................................................................................................................................... 70 4.5 EMPIRICAL RESULTS AND DISCUSSION .................................................................................................. 72 4.5.1 Awareness of improved maize storage technologies ....................................................................... 73 iv 4.5.2 Adoption decisions and the effect of different information sources................................................. 74 4.5.3 Modification of improved maize storage technologies .................................................................... 77 4.6 CONCLUSION AND IMPLICATIONS ......................................................................................................... 78 CHAPTER 5 IMPACT OF MAIZE STORAGE INNOVATIONS ADOPTION ON SCHOOLING EXPENDITURE IN SOUTHERN BENIN ....................................................................................................... 81 5.1 INTRODUCTION ..................................................................................................................................... 82 5.2 DEVELOPMENT AND PROMOTION OF STORAGE INNOVATIONS IN SOUTHERN BENIN .............................. 85 5.3 GENERAL FEATURES OF PRIMARY SCHOOLING SYSTEM ........................................................................ 86 5.4 CONCEPTUAL FRAMEWORK .................................................................................................................. 87 5.5 MODEL ESTIMATION ............................................................................................................................. 92 5.6 DATA.................................................................................................................................................... 98 5.7 ESTIMATION RESULTS AND DISCUSSION ............................................................................................. 100 5.7.1 Goodness-of-fit of the GMM model and determinants of schooling expenditures ........................ 100 5.7.2 Results of GMM estimation for the storage adoption impact on schooling expenditures ............. 103 5.8 CONCLUSIONS .................................................................................................................................... 105 CHAPTER 6 ONE STEP FORWARD, ONE STEP BACK? WHAT DROVE ABANDONMENT OF MAIZE STORAGE INNOVATIONS IN SOUTHERN BENIN? ................................................................. 110 6.1 6.2 6.3 6.4 6.5 6.6 6.7 INTRODUCTION ................................................................................................................................... 111 REASONS FOR TECHNOLOGY ABANDONMENT: A REVIEW OF EXISTING LITERATURE ........................... 113 CONCEPTUAL FRAMEWORK ................................................................................................................ 115 EMPIRICAL FRAMEWORK .................................................................................................................... 116 DATA AND DESCRIPTIVE STATISTICS .................................................................................................. 123 ESTIMATION RESULTS......................................................................................................................... 127 CONCLUSIONS .................................................................................................................................... 130 CHAPTER 7 CONCLUSIONS AND DISCUSSION ..................................................................................... 132 7.1 7.2 7.3 7.4 INTRODUCTION ................................................................................................................................... 132 SUMMARY AND DISCUSSION OF MAIN FINDINGS ................................................................................. 133 POLICY IMPLICATIONS ........................................................................................................................ 138 FUTURE RESEARCH ............................................................................................................................. 140 REFERENCES .................................................................................................................................................. 142 SUMMARY ....................................................................................................................................................... 156 SAMENVATTING............................................................................................................................................ 160 TRAINING AND SUPERVISION PLAN....................................................................................................... 165 CURRICULUM VITAE ................................................................................................................................... 166 v LIST OF TABLES Table 2. 1 Average construction costs of maize storage wooden granaries (FCFA) ........................................ 19 Table 2. 2 Net margin of maize storage systems (FCFA/ton of stored maize) ................................................. 19 Table 2. 3 Evolution of percent of users of storage innovations within farmers who are aware of these innovations ................................................................................................................................................... 20 Table 2. 4 Percent of respondents by main reason for changing adoption status ............................................. 21 Table 2. 5 Major storage problems and applied solutions ................................................................................ 22 Table 2. 6 Numbers of farmers aware of storage innovations divided into their adoption status in 2002 and 2008 ......................................................................................................................................................... 23 Table 3. 1 Farmers’ perceptions of storage structure attributes ................................................................... 32 Table 3. 2 Farmers’ perceptions of conservation measures .............................................................................. 33 Table 3. 3 Variables used in equations of participation in the project.............................................................. 44 Table 3. 4 Variables used in equations of causal effect of project participation on farmers’ perceptions (standard deviations in parentheses). ............................................................................................................ 45 Table 3. 5 Estimation results for the bivariate probit models on project participation and perceptions of characteristics of improved wooden granary (standard errors in parentheses) ............................................. 49 Table 3. 6 Estimation results for the bivariate probit models on project participation and perceptions of characteristics of Sofagrain® (standard errors in parentheses) ..................................................................... 50 Table 3. 7 Major storage problems and applied solutions ................................................................................ 52 Table 3A. 1 Demand and supply weights used for the robustness evaluation ................................................ 55 Table 3A. 2 Correlations between Attainment scores given by different weighting formulae for granaries .. 56 Table 3A. 3 Correlations between Attainment scores given by different weighting formulae for conservation measures .................................................................................................................................................... 57 Table 4. 1 Summary of sample means of model variables (standard deviation in parentheses) .................. 72 Table 4. 2 Estimation results for information, adoption and modification of improved granaries (t-statistics in parentheses) .................................................................................................................................................. 73 Table 4. 3 Estimation results for information, adoption and modification of Sofagrain® (t-statistics in parentheses) .................................................................................................................................................. 77 Table 5. 1 Evolution of percent of users of storage innovations within farmers who are aware of these innovations ................................................................................................................................................... 86 Table 5. 2 Definition and sample means for model variables of schooling expenditure equation (standard deviation in parentheses). ............................................................................................................................. 99 Table 5. 3 Estimation results for the GMM model of average effect on children schooling expenditure of improved maize storage innovations adoption. .......................................................................................... 102 Table 5A. 1 Definition and sample means for model variables of adoption of maize storage innovations equation (standard deviation in parentheses).............................................................................................. 108 Table 5A. 2 Estimation results for the reduced probit model of maize storage innovations adoption. ......... 109 Table 6. 1 Numbers of farmers aware of storage innovations divided into their adoption status in 2002 and 2008 124 Table 6. 2 Percent of respondents by main reason for changing adoption status ........................................... 125 Table 6. 3 Definition and sample means of variables used in the models of changes in adoption decisions of storage innovations (Standard deviation in parentheses)............................................................................ 126 Table 6. 4 Fixed effects logit estimates of changes in adoption status of maize storage innovations. ........... 128 Table 6. 5 Marginal effects of fixed effects logit models for adoption status of maize storage innovations. 130 vi LIST OF FIGURES Figure 1.1 A Conceptual framework of agricultural innovations adoption process and the pathway of adopting impact at the household level. ......................................................................................................... 8 Figure 2.1 Southern Benin and the geographical locations of surveyed villages ......................................... 14 Figure 2.2 Maize post-harvest system on the farms ..................................................................................... 15 vii CHAPTER 1 INTRODUCTION 1.1 Background Population increases are at a high rate in Sub-Saharan Africa where poverty and hunger are widespread. Diffusion of improved agricultural production technologies and policies aiming to improve transportation, storage and information infrastructure and/or regulatory frameworks are mostly applied to increase food production and improve food security. However, in many developing countries crop production and harvesting are carried out during the wet season, when it is difficult to dry and store grain properly. Traditional post harvest systems are often not equipped to dry and store such large quantities properly, and therefore post harvest losses are often aggravated during storage (Goletti and Wolff, 1999). At household level, food security may thus be affected by the magnitude of the physical grain loss. Moreover, household incomes may be affected by lower grain prices due to quality losses. These losses can be reduced through new drying and pest management systems (Goletti and Wolff, 1999). Maize is a major staple food and an important source of income and employment for many farmers in southern Benin. It accounted for 34% of total cereal area and for 47% of cereal production in 2000. The estimated per capita annual consumption was 114kg in 2002 (Arouna, 2002). Almost all maize produced in southern Benin is from the first rainy season and harvested in the wet season. Drying and storage of the grain is therefore difficult because the moist grain attracts more insects than properly dried grain. Maize is stored in traditional storage structures at village level and treated with protectant products. Although these structures and conservation products in some cases seem to be adapted to the prevailing environmental conditions, they are not always effective in protecting maize against pest infestation leading to storage losses. Estimates of losses after six months of storage range from 17% to 40% of the total maize production (Kossou and Aho, 1993; Affognon et al. 2000). Such losses seriously affect food security and household income. To reduce pest damage in stored maize, several projects have been implemented in southern Benin since 1960. Most of the projects implemented up to 1990 failed because of the Introduction lack of adoption of the storage innovations whilst pest attacks remain an important storage constraint for maize producers. Since 1991 projects dealing with post harvest losses use participatory approaches to develop appropriate storage innovations. A package of complementary innovations including an improved wooden granary and Sofagrain®1 was designed and promoted in southern Benin. On-farm trials indicated that after six months of storage, losses were reduced from 30% to only 5% for maize that was treated with Sofagrain® and stored in the improved wooden granary (PADSA, 2000). Although these storage innovations were shown to be effective against pests, their adoption by farmers and the persistence of the adoption process provide challenging questions for scientists, policy makers and donors. Despite several years of storage innovations research and diffusion in southern Benin, there remains a dearth of empirical information on their behavioral impacts. It is important to understand by whom, how and when the storage innovations are used and what their impact is (Doss, 2006). Such information is essential to researchers and the national agricultural extension service to measure the persistence of the adoption process and the social relevance of the storage innovations. 1.2 Problem statement Technological change in agriculture is considered as an important means to foster economic growth at the early stage of development and to improve the well being of poor households (Doss et al. 2003; Self and Grabowski, 2007; Tiffin and Irz 2006). Adoption of agricultural innovations has been studied intensively since Griliches’ (1957) pioneering work on adoption of hybrid corn in the United States. Feder et al. (1985), Feder and Umali (1993) and Sunding and Zilberman (2001) provide excellent surveys on the technology adoption literature. To improve agricultural productivity and the welfare of farmers, policy makers need information on the adoption pattern of agricultural innovations to formulate policies for their dissemination and diffusion (Dimara and Skuras, 2003). Furthermore, since the declaration of the Millennium Development Goals in 2002, policy makers and donors have increased their interest in the impact of agricultural innovations on the livelihood of poor peoples in subSaharan Africa (Alwang and Siegel, 2003). Understanding the adoption process of 1 The symbol  stands for ‘Registered trade mark’. Sofagrain is an insecticide protectant constituted of 0.2% Delmethrin and 1.5% Pyrimiphos-Methyl. It’s used to control pests in stored grains, notably cereals and leguminous. 2 Introduction agricultural innovations and their impact on the welfare of farmers is therefore a challenge for social scientists (Doss, 2006). Following an expected profit maximization framework, most of empirical economic studies on technology adoption use a probit, logit or tobit model to identify the specific factors that affect adoption or intensity of adoption of an agricultural innovation at a point in time. However, the expected profit maximization framework does not condition the adoption decisions on the information acquired by the producer (Dimara and Skuras, 2003). Moreover, Rogers (2003) argues that the adoption-decision process starts when an individual is exposed to an innovation’s existence and collects information necessary to use it properly (Rogers, 2003; p. 172). Accordingly, a farmer who is not exposed to the existence of an agricultural innovation is excluded from the subsequent adoption decisions. Hence, adoption studies not controlling for such exposure yield non-consistent estimates of the effects that explanatory variables have on adoption (Saha, et al., 1994; Dimara and Skuras, 2003; Diagne and Demont, 2007). Most farmers rely on information from the near peer adopters, because their (subjective) opinions of the innovation are accessible and convincing to them (Rogers, 2003). The role of early adopters in information dissemination on new technologies is recognized in the literature on copying behavior in technology diffusion (Bevan et al., 1989:109-122; Pomp and Burger, 1995). However, it is often not explicitly known whether farmers adopted after contacts with extension agents or whether they copied adoption decisions from others. This makes it difficult to assess the role of copying in the diffusion of innovations. Another neglected aspect is that copying may involve modifications of the technology. Farmers that imitate their neighbors in adopting a certain technology may modify it, thereby adapting the innovation to their circumstances. At a given point in time, the decision to adopt or reject an innovation or to defer this decision is postulated to be influenced by the attitude towards the technology and the beliefs about that technology (Fishbein, 1967). In addition, the intensity of having a certain attitude is a major determinant of anticipated behavior (Lemon, 1973). These insights led to an increased attention towards the role of farmers’ perceptions on innovation characteristics in recent work on adoption of agricultural innovations. Recent studies show that perceptions of various attributes of an innovation influence the expected relative value of the innovation and subsequent adoption decisions (Adesina and Zinnah, 1993; Batz et al. 2003; Negatu and Parikh, 1999; Llewellyn, et al., 2004). However, a drawback of these studies is that they 3 Introduction neglect the factors that determine the formation of perceptions among farmers. Few studies dealt with the influence of the source of information on the farmers’ perceptions of the characteristics of agricultural innovations (Guerin, 1999). Awareness of the factors that influence perceptions would contribute to development and transfer of appropriate innovations. Assessing the impact of adopting agricultural innovations on farmer welfare is a major concern to policy-makers and donors. The surplus method in a partial equilibrium framework is often used to assess the economic impact of agricultural research (Marasas et al., 2003; Mather et al., 2003). This method requires additional assumptions on the prices of output and consumed commodities, and income of the farm household to compute the measures of changes in a target outcome. To avoid these assumptions and extend the impact assessment of adopting new technologies on behavioral, efficiency and well-being outcomes, recent works have used the treatment effect approach (Vella and Verbeek, 1999). Assessing the impact of agricultural innovations is complicated by the lack of experimental design in this field. One of the challenges in impact assessment of agricultural innovations is therefore how the changes in peoples’ welfare can be attributed to a specific new technology. Changes in poverty indicators such as income, expenditure, nutrition and health may arise from changes in the external environment that have nothing to do with the new technology. Treatment effects estimators are based on the counterfactual (potential outcomes) framework in which each individual has an outcome with and without treatment (Wooldridge, 2002, p. 603). This framework also underlies the standard methods for establishing causal treatment effects on observed outcomes in natural experiments. Recently developed econometric treatment effect estimation strategies can help to distinguish impacts of single interventions and thus bring a solution to the evaluation and attribution problems encountered in assessment of the impact of an intervention. Nevertheless the major drawback of the applications of this method is the assumption of homogeneity of the impact of the “treatment” being evaluated (Blundell and Dias, 2002). Therefore the major focus in this recent microeconomic policy evaluation literature is to design and estimate models in which the heterogeneity in responses to treatment among observationally identical people is assumed. Since the change of government objectives toward poverty reduction, there has been growing concern to assess the adoption impact of agricultural innovations on poverty indicators such as income, expenditure, food, nutrition and health at farmer’s level. Expenditure levels are 4 Introduction generally recognized as a better measure of economic status than income, since income does not reflect permanent wealth and can be seasonally variable (Waters, 2000). Many of the agricultural innovations impact assessment studies have been concerned with the effect on the income of adopting farmers (Bravo-Ureta et al., 2006; McBride and El-Osta, 2002). According to our best knowledge, to date very few studies on the effect of adopting agricultural innovations on expenditures on food staples and non-food items such as children schooling have been performed (Adekambi et al., 2009). Recent literature on development shows that schooling expenditure is an effective means to increase labor productivity and to improve the income distribution and individual well-being (Groot and Maassen van den Brink, 2007). Therefore schooling expenditure is an important outcome on which the impact of adopting agricultural innovations could be assessed. Most empirical studies on agricultural technology adoption only use cross-section data to analyze adoption decisions. They divide a population into adopters and non-adopters, and analyze the reasons for adoption or non-adoption at a point in time. However, a simple classification of farmers as adopters and non-adopters may not be adequate to understand the adoption process fully. Besides, adoption of agricultural innovations is a dynamic decisionmaking process in which farmers move from learning to adoption to continued use or abandonment of the technology over time. In addition, decisions in one period may depend on decisions made in previous periods. This dynamic decision-making process is characterized by the time pattern of factors such as information gathering and updating, learning by doing, or accumulating resources that may affect the farmer’s decision (Feder et al., 1985; Sunding and Zilberman 2001). Accordingly, changes in these variables would help in explaining why individuals choose to adopt an agricultural innovation at different periods (Koundouri et al. 2006). The dynamic aspect of agricultural innovation adoption decisions has been recognized in the theoretical literature (Cameron, 1999). To understand the dynamics of adoption, the decisions of the farmers and the factors related to these decisions need to be followed over a period of time. This is best done using panel data. Because of commitment of time and resources that are required to develop panel data sets, few economic adoption studies estimate dynamic models to analyze the adoption decisions of agricultural innovations. Exceptions to this are Cameron (1999) and Moser and Barrett (2006). Moser and Barrett analyze the factors affecting the decisions of rice producers to adopt, expand and abandon a system of rice intensification (SRI) in Madagascar. One of their main finding is that learning effects play a 5 Introduction major role in the adoption decisions of the SRI. Moreover, they analyze the factors that affect farmers’ decisions to discontinue the use of the SRI, one of the aspects of the adoption process that is rarely studied. However, due to a lack of panel data, Moser and Barrett (2006) used data based on a recall procedure on each farmer’s adoption history as proposed by Besley and Case (1993). Cameron studied the dynamic process of learning in the adoption of a new high-yielding variety cotton seed in India. Using a panel data set of households in India, she estimated a fixed effects model including the one-period lagged profit differential to reflect the farmer's knowledge on the seeds. She applied a linear probability model to avoid estimation problems related to the use of fixed and random effects probit, logit and tobit models (Maddala, 1987; Greene, 2002; Greene, 2008: 796-806). However, the linear probability model also has serious shortcomings such as heteroskedasticity, constant marginal effects and the fact that predicted probabilities are not always between 0 and 1 (Cameron, 1999). 1.3 Objectives of the thesis The general objective of this thesis is to analyze the adoption patterns of the maize storage innovations promoted in southern Benin since 1992 and assess the impact of their adoption on the well-being of adopters. This objective is based on the issues raised in the previous section. In more detail, the specific objectives of this thesis are to: i. Examine the extent to which the storage innovations have characteristics that match the needs of the maize producers; ii. Evaluate the impact of farmers’ participation in the extension program on their perceptions of the quality of characteristics provided by the storage innovations; iii. Determine the factors influencing the adoption and modification of maize storage innovations and to assess the role of the sources of information; iv. Assess the impact of adopting storage innovations on schooling expenditures in rural areas of southern of Benin; and v. Analyze changes in adoption status and factors that affect the discontinuation of storage innovations use. 6 Introduction 1.4 Methodological approach and data This section presents the general analytical framework which guides the whole study. In addition, the specific theoretical background and empirical approaches used to achieve each specific objective of this thesis are introduced. 1.4.1 General analytical framework The theoretical framework that underlies this study is the agricultural household model as illustrated in Taylor and Adelman (2003). This framework provides a behavioral model of the farm household acting simultaneously as producer and consumer. The agricultural household model assumes that the farm household makes its consumption and production choices to maximize the utility of consumption subject to a set of constraints including the production technologies and full income constraints. Adoption of new agricultural technologies allows the household to alleviate the constraints related to the production technologies. The decision to adopt or reject versus the decision to adopt with or without modifying a new agricultural technology is based on a comparison of expected utility. The expected utility maximization framework makes explicit the role of the information in the adoption decisions making process (Dimara and Skuras, 2003). Moreover, the budget of the farm household is in part determined by the profit realized from the production activities and can be increased with the adoption of new agricultural technologies. An increased budget leads to changes in the demand for food and non-food commodities as given by the consumer maximization problem and thereby in its welfare outcome. This study is based on a complete analytical framework presented in Fig. 1.1. The conceptual representation describes the adoption of decision-making process for agricultural innovations and related factors. In addition, the conceptual framework shows the impact of technological change on the poverty indicators such as (schooling) expenditure at the farm household level and the exogenous factors that may affect the (schooling) expenditure. At each period of time, the decision to adopt or reject an agricultural technology versus the decision to adopt it with or without modifications can be viewed as one element in the total decision making process of the farm household. Similarly, following the agricultural 7 Introduction household model, the impact of technological changes on schooling expenditure can be assessed at each point in time. Policies and institutions Credit Extension Resource Endowment Profile Market Infrastructures Natural resources Agro-climatic conditions Political and institutional constraints Resource constraints and conditioning environment Own experiences Near peers experiences Farmer characteristics Technology characteristics Community characteristics Perception Knowledge FIRST PERIOD Innovation existence information Beliefs & Expectations Preferences Value formation Impact Income Adoption/ Modification Non Adoption Nutrition Health status SECOND PERIOD Expenditures on food staples Increase Intensity / Modification Disadoption Adoption/ Modification Non adoption Expenditures non-food items (schooling expenditures, etc.) Figure 1.1 A Conceptual framework of agricultural innovations adoption process and the pathway of adopting impact at the household level. When a farmer is exposed to the existence of a new technology and collects information to use it properly, he develops a perception or belief about it and decides whether to adopt or reject the technology or defer it for a decision to be taken on it later. The decision to reject or to defer it may result from an inadequate level of information to use the new technology properly. Whatever decision is taken, in the next period the farmer gathers new knowledge and experience from learning-by-doing as well as observing performance of near peer adopters. According to the decision to adopt or reject, the subsequent decisions may follow two pathways. First, a decision to adopt may be followed by a decision to increase 8 Introduction intensity and/or modify the new technology or to discontinue the use of the new technology. Second, following the decision to reject the farmer modifies his initial perception or belief about the new technology based on his new knowledge and/or observed performance from the adopters. Accordingly he takes a new decision about the adoption of the new technology. At each time-period, based on his knowledge and experience the farmer forms an attitude towards the technological innovation. An individual's decision in a time-period is a joint function of his attitude towards the new technology and his beliefs about what is expected for that situation (Fishbein, 1967). The attitude of a decision-maker towards an innovation depends on his valuations of the set of characteristics of that innovation (Wossink et al., 1997). Beliefs of the farmer about his adoption behavior in a given period are influenced by his socio-economic characteristics such as resources endowment and characteristics of his community at that time (Feder et al., 1985; Sunding and Zilberman, 2001). Impact of adoption of a new technology on outcomes such as income, expenditures, food security at the farm household level can be assessed at each time-period. Figure 1.1 shows that the same factors can determine both adoption and the outcome leading to selection bias in estimating impact of adoption of a new technology. Agricultural household models provide useful guide to researchers for deciding which variables should be treated as endogenous and which are to be held exogenous in impact evaluation of new technologies. Drawing from the agricultural models, adoption variable is assumed endogenous in equations of factors related to individual and household behaviors such as production (farm and nonfarm) decisions, consumption decisions, investment and saving decisions. On the other hand, human resources such as labor, health, knowledge are exogenous factors. 1.4.2 Specific theoretical background and empirical approaches This subsection presents an overview of the specific theoretical frameworks and related estimation methods used to achieve each specific objective of this study. Analysis of farmers’ perceptions To achieve the first two objectives of this thesis, farmers’ perceptions of the characteristics of the improved wooden granary and Sofagrain® are analyzed in two ways. Firstly, an empirical analysis adapted from Reed et al. (1991) is used to evaluate the extent to which, each of the storage innovations provides characteristics that meet the expectations of the maize producers. 9 Introduction The approach is based on the calculation of demand, supply and attainment indices, respectively. The demand index measures the importance that farmers give to each characteristic of the storage innovation they desire. The supply index assesses the farmers’ perception on the achieved level of each desired characteristic in the storage innovation. The attainment index evaluates how the perceived importance of a characteristic matches with the extent to which this characteristic is supplied in the storage innovation. Secondly, following Wooldridge (2007a) a bivariate probit model is applied to correct for endogeneity of farmer participation in the extension program and evaluate its impact on their perceptions of the quality of characteristics provided by the storage innovations. The total marginal effect of participation in extension program is obtained following a method developed by Bartus (2005). Data used in this analysis is obtained from an exploratory survey followed by crosssection data collection carried out in 2002. Analysis of adoption and modifications of maize storage innovations and effect of sources of information To achieve the third objective of this study, the adoption decision on maize storage systems by individual farmers is modeled following Saha et al. (1994) and Dimara and Skuras (2003). These authors recognized that farmers can only adopt a new technology if they are sufficiently informed about it. Moreover, after having decided to adopt an innovation or not, adopters also decide whether to modify the innovation or not. In the empirical analysis probit specifications are used to estimate equations for information, adoption and modification decisions, respectively. The probit specification for adoption is corrected for sample selection since the adoption decision is only relevant for those who heard about the innovation component (Saha et al. 1994). Similarly, the probit equation for modification is corrected for sample selection bias since this decision is only relevant for the non-random subsample of farmers who decided to adopt the innovation. Maximum likelihood is used for model estimation. The models are also estimated according to the sources of information and the results are compared to see whether having information from the extension service that promotes the storage innovations affects the determinants of adoption and modification differently than having information from peers. Cross-section data collected in 2002 as indicated earlier is used