AGRICULTURAL FINANCE, NON-FARM EMPLOYMENT, AND RURAL POVERTY: EVIDENCE FROM SUB-SAHARAN AFRICA By Serge Guigonan Adjognon A DISSERTATION Submitted to Michigan State University in partial fulfillment of the requirements for the degree of Agricultural, Food, and Resource Economics Ð Doctor of Philosophy 2016 !ABSTRACT AGRICULTURAL FINANCE, NON-FARM EMPLOYMENT, AND RURAL POVERTY: EVIDENCE FROM SUB-SAHARAN AFRICA By Serge Guigonan Adjognon Efforts to eradicate poverty in the world require a particular focus on agricultural households of SSA, where poverty remains ubiquitous. This dissertation, titled Agricultural Finance, Non-Farm Employment, and Rural Poverty, uses evidence from SSA to explore some of the constraints faced by farming households in Sub-Saharan Africa (SSA). First, while agriculture remains central in the economy of most SSA, yields are still relatively low compared to other parts of the world. Financial restrictions are cited amongst the main constraints to inputs use in SSA. Thus, the first essay of this dissertation titled Updating the Landscape on Farm Input Credit in SSA explores input financing and the role of credit therein. Our results consistently show that traditional credit use is extremely low (across credit type, country, crop and farm size categories) and farmers primarily finance modern input purchases with cash from nonfarm activities and crop sales. Second, the consistent lack of credit for agricultural inputs observed in the first essay, motivated the second essay titled Sustaining Input on Credit through Dynamic Incentives and Information Sharing. This essay uses a framed field experiment to explore conditions that can minimize strategic default, a key source of market failure in input credit markets in developing countries, where institutions for contract enforcement are weak or nonexistent. The results show that the existence of an information exchange system, amongst input sellers, which mimics the role of a Òcredit scoreÓ (with potential benefits from its informal nature), can effectively deter default behavior by farmers receiving inputs on credit. Moreover, productivity shocks that affect the return to the use of inputs also affect the opportunity cost of !repayment, and thus farmerÕs decision to repay. Third, the importance of non-farm activities (revealed in Essay 1), in addition to recent evidence from the literature, indicate increasing contribution of non-farm activities to householdsÕ income in SSA. Therefore, the third essay of this dissertation titled the Heterogeneous Welfare Effects of Non-Farm Employment explores the effects of rural non-farm activities (wage and self-employment) on rural household welfare. It also explores the heterogeneous effects of participating in non-farm activities across the welfare distribution. The results confirm that participation in non-farm activities is generally welfare improving and poverty reducing. However, households at the lower tail of the welfare distribution benefit significantly less from participation than the wealthiest. Low education, assets, and access to credit are important barriers that limit the participation of the poorest in lucrative non-farm employment opportunities. Together, these essays shed light on important policy considerations for improving the livelihood of poor households in developing countries. While access to credit remains extremely limited for farming households to finance agricultural intensification, the expansion of Information and Communication Technologies in SSA offers new hope for financial inclusion of those marginalized groups. Similarly, improving access to the non-farm sector by households will likely improve modern input use and agricultural productivity. Beyond just the effect on inputs use, participation in the non-farm sector significantly improves welfare and reduces poverty amongst rural households. However, it is important to address barriers that limit participation in more lucrative non-farm sectors by the poorest, who currently benefit less from participation compared to the non-poor. ! Copyright by SERGE GUIGONAN ADJOGNON 2016 !"! I dedicate this dissertation to my mother, Elisabeth Katary, and my late father, Innocent Adjognon. For all the sacrifice you have made to provide for your children, may this accomplishment be an expression of my eternal gratitude. Your boundless support, encouragement, and prayers have kept me going. Merci beaucoup. Je vous aime du fond de mon coeur.!"#!ACKNOWLEDGEMENTS I would like to express my profound gratitude for the incredible mentorship, guidance, inspiration, and motivation I received from my supervisor and major advisor Prof. Lenis Saweda Liverpool-Tasie who also funded my PhD study at Michigan State. Also, sincere thanks to the rest of my advisory committee: Prof. Mywish Maredia, Prof. Robert Shupp, Prof. Robert Myers, and Prof. Jeffrey Wooldridge, for precious comments on my research, teachings in class, as well as technical and personal advice throughout my PhD career. I also extend my gratitude to the faculty and staff of the Department of Agricultural, Food and Resource Economics (AFRE), Michigan State University, especially the Graduate Secretary Ms. Debbie Conway, and the Graduate Chair Prof. Scott Swinton for consistent support and assistance while navigating the PhD study. I am forever grateful for the support I received from my fellow student cohorts Jason Snyder, Joey Goeb, Sarah Kopper, Andrea Leschewski, Mukesh Ray, Piyayut Chitchimnong, Craig Carpenter, among others. I truly value the common experience and all the memories that we share. To all my East Lansing friends, many thanks for having supported me and given me the opportunity to keep a balanced lifestyle throughout my PhD program. Special thanks to my best friend, and soon-to-be-wife, Marie Steele, who has been truly supportive through this whole journey. Last but not least, I would like to recognize my family who encouraged me to accept my offer of admission into the PhD program and supported me throughout the whole program. I !"##!want to thank especially my mother Elisabeth Katary for her countless hours of prayers for my success, and for all the sacrifices made to ensure my education since my young age. I thank and pray for my late dad who instilled in me the values of education and good work ethic since my childhood and encouraged me to work hard and follow fearlessly my dreams in life. Thank you very much for all the sacrifices. !"###!TABLE OF CONTENTS LIST OF TABLES ........................................................................................................................ x LIST OF FIGURES ................................................................................................................... xiii KEY TO ABBREVIATIONS ................................................................................................... xiv 1. INTRODUCTION ................................................................................................................ 1 2. ESSAY 1: UPDATING THE LANDSCAPE ON FARM INPUT CREDIT IN SUB-SAHARAN AFRICA .................................................................................................................... 5 2.1. Introduction ................................................................................................................... 5 2.2. Data ................................................................................................................................ 8 2.3. Farm input purchases Ð abstracting for the moment from input finance ............. 10 2.4. Farm input finance for the overall sample and by farm size .................................. 16 2.5. Crop type and input purchase financing .................................................................. 21 2.6. Tied output input credit arrangements..................................................................... 25 2.7. HouseholdsÕ use of loans not specifically linked to input transactions .................. 29 2.8. Determinants of inputs purchase in Nigeria ............................................................. 33 2.8.1. Conceptual and empirical framework .............................................................. 34 2.8.2. Regression results ................................................................................................ 45 2.9. Conclusion ................................................................................................................... 50 REFERENCES ........................................................................................................................ 53 3. ESSAY 2: SUSTAINING INPUT ON CREDIT THROUGH DYNAMIC INCENTIVES AND INFORMATION SHARING: LESSONS FROM A FRAMED FIELD EXPERIMENT ........................................................................................................................... 57 3.1. Introduction ................................................................................................................. 57 3.2. Dealing with strategic default .................................................................................... 59 3.3. Theoretical framework and experimental hypotheses ............................................ 64 3.3.1. A simple model of input on credit ..................................................................... 64 3.3.2. The missing market problem in a single period game ..................................... 66 3.3.3. Enforcement of the input-on-credit contract using dynamic incentives ........ 67 3.4. Experimental design and procedures ........................................................................ 70 3.4.1. Decisions and Payoffs for Agro Brokers ........................................................... 73 3.4.2. Decisions and Payoffs for Farmers .................................................................... 75 3.4.3. Information Treatment Variation and General Implementation .................. 76 3.5. Results and discussion ................................................................................................ 78 3.5.1. General description of the data ......................................................................... 78 3.5.2. FarmersÕ behavior .............................................................................................. 79 3.5.2.1. Description of farmersÕ repayment behavior during the game .................. 79 !#$!3.5.2.2. Econometric model ......................................................................................... 81 3.5.2.3. Econometric results ........................................................................................ 82 3.5.3. BrokersÕ behavior ............................................................................................... 87 3.5.3.1. Proportion of farmers receiving input on credit throughout the rounds .. 87 3.5.3.2. BrokersÕ Punishment strategy ....................................................................... 89 3.6. Conclusion and Implications ...................................................................................... 93 APPENDICES ......................................................................................................................... 96 Appendix 1: BrokersÕ decision making sheet ................................................................... 97 Appendix 2: FarmersÕ decision making sheet .................................................................. 98 Appendix 3: Implementation and sequence of actions in each round of the game ....... 99 REFERENCES ...................................................................................................................... 101 4. ESSAY 3: THE HETEROGENEOUS WELFARE EFFECTS OF RURAL NON-FARM EMPLOYMENT: RECENT EVIDENCE FROM MALAWI ................................. 104 4.1. Introduction ............................................................................................................... 104 4.2. Conceptual framework for participation in non-farm activities .......................... 109 4.3. Data ............................................................................................................................ 112 4.3.1. Data source ........................................................................................................ 112 4.3.2. Patterns of non-farm activities in rural Malawi ............................................ 114 4.3.2.1. Non-farm self-employment in rural Malawi .............................................. 117 4.3.2.2. Non-farm wage employment in rural Malawi ............................................ 122 4.4. Empirical strategy ..................................................................................................... 124 4.4.1. Determinants of participation in non-farm employment .............................. 124 4.4.2. Impact of participation in non-farm activities ............................................... 128 4.4.2.1. Outcome variables ........................................................................................ 130 4.4.2.2. Distributional effects of non-farm employment participation: Quantile regression approach ...................................................................................................... 133 4.4.3. Explanatory variables ....................................................................................... 134 4.4.3.1. Determinants of participation in non-farm employment .......................... 134 4.4.3.2. Impact of participation in non-farm employment ..................................... 137 4.5. Results and discussion .............................................................................................. 138 4.5.1. Determinants of participation in non-farm activities .................................... 138 4.5.2. Impacts of participation in non-farm activities .............................................. 144 4.6. Conclusion ................................................................................................................. 152 APPENDICES ....................................................................................................................... 154 Appendix 1: A general framework for the identification of rural non-farm employment treatment effects .......................................................................................... 155 Appendix 2: Distributional effects of non-farm employment participation: A quantile effect model ........................................................................................................................ 157 Appendix 3: Full regression tables .................................................................................. 158 Appendix 4: Description and summary statistics of the variables used in this article, by treatment status, household level, 2010-2013, rural Malawi ........................................ 175 REFERENCES ...................................................................................................................... 184 !$!LIST OF TABLES Table 2-1: Share of households who purchase external inputs by country .......................... 12 Table 2-2: Household purchase of external inputs by farm size strata ................................. 14 Table 2-3: Share of households purchasing external inputs that finance the input purchase on credit ............................................................................................................................... 16 Table 2-4a: Shares of Strata in all credit-based expenditure on external inputs .................. 19 Table 2-4b: Share of credit-based input outlay for input i in total outlay for input i per stratum ÉÉÉÉÉÉÉÉÉÉÉÉÉÉÉÉÉÉÉÉÉÉÉÉÉÉÉÉÉÉÉÉÉ.É20 Table 2-5: Share of households producing key cash and food crops across farm size strata............................................................................................................................................... 22 Table 2-6: Cash crops versus food crops on which purchased external inputs are used that were financed on cash-credit for key cash and food crops (% of plots) ....................... 24 Table 2-7: Share of farmers using harvest to reimburse for inputs received on credit by farm size ............................................................................................................................... 26 Table 2-8: Financing inputs on credit with harvest across key cash and food crops. .......... 28 Table 2-9: Share of households with a member taking a financial loan ................................ 29 Table 2-10: Purpose for which loans are taken (by source of loan) in Malawi ..................... 30 Table 2-11: Purpose for which loans are taken (by source of loan) in Tanzania .................. 31 Table 2-12: Sources of Cash Income in Nigeria, North and South, 2010 and 2012 .............. 32 Table 2-13: Summary statistics of variables used in the regression analysis by region ....... 41 Table 2-14: Estimation results of determinants of fertilizer purchase and quantity purchased by farmers in Nigeria ....................................................................................... 48 Table 3-1: Experiment Villages in Kwara State, Nigeria ........................................................ 71 Table 3-2: BrokersÕ Commission/Penalty Schedule ................................................................. 74 Table 3-3: FarmersÕ payoff structure ...................................................................................... 76 !$#!Table 3-4: Public repayment records used in treatment villages ........................................... 77 Table 3-5: Statistics about the offers made and received through the game ......................... 78 Table 3-6: Estimation results for the determinants of farmersÕ repayment behavior ......... 85 Table 3-7: Determinants of receiving input loan offer as function of past repayment by communication treatment .................................................................................................. 90 Table 3-8: Determinants of receiving input loan offer as function of past repayment by communication treatment for round 2 only ..................................................................... 92 Table 4-1: Household participation in non-farm employment in rural Malawi, 2010/2013............................................................................................................................................. 116 Table 4-2: Selected characteristics of household enterprises in rural Malawi .................... 118 Table 4-3: Distribution of non-farm household enterprises and returns by Sector, by poverty status, in Rural Malawi, 2010/2013 ................................................................... 121 Table 4-4: Distribution of non-farm wage employment and returns by Sector, by poverty status, in Rural Malawi, 2010/2013 ................................................................................. 123 Table 4-5: Average partial effects estimates of determinants of non-farm employment participation ...................................................................................................................... 142 Table 4-6: Effect of participation in the non-farm activities on various outcomes in rural Malawi ................................................................................................................................ 145 Table 4-7: Distributional effects of participation in the non-farm activities on HHPCE in Malawi (Quantile regression results) .............................................................................. 147 Table 4-8: Seemingly unrelated system equation estimates of non-farm wage employment and non-farm self-employment participation model ..................................................... 158 Table 4-9: Effect of participation in the non-farm activities on HHPCE in Malawi, FE estimates ............................................................................................................................. 160 Table 4-10: Effects of participation in the RNFE on quintiles of HHPCE .......................... 162 Table 4-11: Effects of participation in the non-farm activities on poverty incidence, gap and severity in rural Malawi ................................................................................................... 164 Table 4-12: Effects of participation in the non-farm activities on food security and subjective well being ......................................................................................................... 166 !$##!Table 4-13: Effects of participation in non-farm activities on inputs purchases for farm households only ................................................................................................................. 168 Table 4-14: Multivariate recursive Probit estimation of effects of non-farm employment participation on activities on inputs purchases for farm households only. ................. 171 Table 4-15: Description and summary statistics of the variables used in this article, household level, 2010-2013, rural Malawi ...................................................................... 175 Table 4-16: Test of balancing of covariates between non-farm wage employment participants and non-participants households, 2010-2013, rural Malawi ................... 178 Table 4-17: Test of balancing of covariates between non-farm self-employment participants and non-participants, 2010-2013, rural Malawi ............................................................ 181 !$###!LIST OF FIGURES Figure 3.1: Extensive form representation of the farmer-trader theoretical game .............. 67 Figure 3.2: Histogram of repayment decisions by communication treatment status ........... 80 Figure 3.3: Histogram of repayment decision by weather state ............................................. 81 Figure 3.4: Patterns of offers throughout the rounds of the game ......................................... 88 Figure 4.1: Quantile effects of non-farm wage and self-employment on HHPCE in rural Malawi ................................................................................................................................ 147 !$#"!KEY TO ABBREVIATIONS SSA Sub-Saharan Africa FE Fixed effects FRM Fractional response Model ICT Information and Communication Technology LSMS Living Standards Measurement Surveys MKW Malawi Kwacha SD Standard Deviation SE Standard Error !%!1.!INTRODUCTION Poverty continues to affect millions of people globally, the majority of whom live in rural areas of Sub-Saharan Africa (SSA). The recent World Development Indicators1 published by the World Bank indicate that the share of total world poor living in SSA has practically tripled from 14.7 percent in 1990 to 43.4 percent in 2012. These households in poverty are exposed to a host of interrelated socio-economic constraints including hunger and malnutrition, low education, low assets, low levels of infrastructure, amongst others; which limit their productivity and trap them into a vicious cycle of poverty. This implies that efforts to eradicate poverty in the world require a particular focus on rural households of SSA, with interventions aiming to improve their productivity and income from the activities they are involved in. Consequently, this dissertation titled Agricultural Finance, Non-Farm Employment, and Rural Poverty uses evidence from SSA to explore some of the constraints faced by farming households in SSA2. Agriculture is the main activity for most people in rural areas of SSA, yet agricultural yields in SSA are considerably lower than in other parts of the world. There is widespread agreement that this is partly due to the significantly lower use of modern inputs in SSA compared to the rest of the world. Financial constraints have been mentioned amongst the main demand side barriers to agricultural input use in developing countries, including SSA. Therefore, improving householdsÕ access to financial solutions for inputs purchases is an important requirement for agricultural development and poverty reduction. While recent evidence shows that many Sub-Saharan African farmers use modern inputs, there is limited current and 1 http://data.worldbank.org/data-catalog/world-development-indicators 2 Though most farm activities occur in rural areas, there are some farm households in urban areas. Therefore farm households, in this essay, include households involved in agricultural activities in both urban and rural areas, unless otherwise specified. !&!comparable information on how these input purchases are financed. The first essay of this dissertation titled Updating the Landscape on Farm Input Credit in SSA uses recently available nationally representative data from four countries to explore input financing and the role of credit therein. Our results consistently show that traditional credit use is extremely low (across credit type, country, crop and farm size categories) and farmers primarily finance modern input purchases with cash from nonfarm activities and crop sales. Tied output-factor market arrangements (largely ignored in the literature) appear to be the only form of credit relatively widely used but it is mostly for labor and not for external inputs. These results motivated the second and third essays of this dissertation. The rational for the second essay comes from the consistent lack of credit for agricultural inputs observed in the first essay. Generally, rural credit markets in developing countries are characterized by market failures associated with imperfect information and risk. These failures persist due to weak contract enforcement institutions, thus increasing the potential for high strategic default rates. Knowing this, input suppliers are reluctant to provide inputs to farmers on credit. Therefore, the second essay of this dissertation, titled Sustaining Input on Credit through Dynamic Incentives and Information Sharing, uses a framed field experiment to explore conditions that can minimize default by farmers receiving input on credit from input sellers in developing countries where institutions for contract enforcement are weak or nonexistent. Using data collected through a framed field experiment that simulates a market for input on credit, the paper shows that the existence of an information exchange system, amongst input sellers, which mimics the role of a Òcredit scoreÓ (with potential benefits from its informal nature), can effectively deter default behavior by farmers receiving inputs on credit. Moreover, productivity !'!shocks that affect the return to the use of inputs also affect the opportunity cost of repayment, and thus farmerÕs decision to repay. The findings in Essay 1 also indicate that non-farm activities play an important role in agricultural finance in SSA. Recent evidence from the literature also shows that the contribution of non-farm activities (such as non-farm wage employment and non-farm enterprises) to household income in SSA is substantial and has increased over time. However, there are considerable entry barriers that restrict the poorest from participating in such activities, or limit them to low returns categories of employment with low potential for lifting them out if poverty. Therefore, the third essay of this dissertation titled the Heterogeneous Welfare Effects of Non-Farm Employment uses nationally representative panel data from Malawi and a combination of econometric approaches, to quantify the effects of rural non-farm activities (wage and self-employment) on rural household welfare. It also explores the heterogeneous effects of participating in non-farm activities across the welfare distribution. The results confirm that participation in non-farm activities is generally welfare improving and poverty reducing. However, households at the lower tail of the welfare distribution benefit significantly less from participation than the wealthiest. Low education, assets, and access to credit are important barriers that limit the participation of the poorest in lucrative non-farm employment opportunities. Together, these essays shed light on important policy considerations for improving the livelihood of poor households in developing countries. While access to credit remains extremely limited for farming households to finance agricultural intensification, the expansion of Information and Communication Technologies in SSA offers new hope for financial inclusion of those marginalized groups. Indeed, the establishment of information sharing systems amongst !(!input sellers, suggested in Essay 2, to enforce credit repayment and improve access to input credit, can be made easier and less costly, by leveraging on the opportunities presented by ICT technologies. Even in areas with limited access to digital technologies, the remarkable penetration of cell phones can supplement local social networks and facilitate the collection and sharing of credit repayment behavior, especially if community level retailers are integrated in the distribution of input credit. Meanwhile, farmers seem to rely increasingly on income from non-farm sources to address their farm inputs cash needs. This implies that, alongside strategies to provide better access to input credits, improving access to the non-farm sector by households will likely improve modern input use and agricultural productivity. More generally, the non-farm sector is shown in Essay 3 to significantly improve welfare and reduce poverty amongst rural households. This confirms that the benefits of developing the non-farm sector extend beyond just the financing of input purchases. However, Essay 3 also reveals an unequal distribution of welfare benefits from participation in non-farm employment which indicates that it is important to address barriers that limit participation in more lucrative non-farm sectors by the poorest, which currently limits their benefit from participation compared to the non-poor. !!)!2.!ESSAY 1: UPDATING THE LANDSCAPE ON FARM INPUT CREDIT IN SUB-SAHARAN AFRICA 2.1.!Introduction It is generally accepted that Sub-Saharan Africa (SSA) farmers often suffer low yields, and if they bought more modern inputs (e.g. fertilizer, pesticides, and seeds) they could increase yields, all else equal. It is a common hypothesis in the literature (e.g., Croppenstadt et al. (2003)) that one reason farmers in SSA do not buy enough of these inputs is that they are credit-constrained. Starting however from the knowledge that many SSA farmers in fact do purchase farm inputs (Sheahan and Barrett 2014), we pose a basic question concerning how they finance those inputs Ð that is, does credit play any role and if so which kinds of credit are important? Do the empirical facts from a systematic analysis of farm household surveys match with conventional wisdom about these issues? The three research questions we address are thus: (1) how do farmers finance input purchases? (2) Are there correlations with farm size and thus ÒinclusivenessÓ of the financial arrangement used and (3) is there a relation with crop type and thus relation to cash crop versus food crops? We derive the hypotheses to test from the literature, which we consider to be feeding and reflecting common wisdom about these questions. The key points are organized according to the set of potential finance arrangements available to smallholders. These include formal and informal tied and untied credit sources and then own retained earnings. First, in principal, formal credit from formal-sector banks could be a source of finance for inputs. But this source is in general depicted in the literature to be scant in SSA rural areas for two reasons. On one hand, parastatal agrarian banks and government credit for farmers input purchases have been largely dismantled during Structural Adjustment programs in the past !!*!several decades (Kherallah et al. 2002). On the other hand, it is generally held that private-sector banks lend little (and then mainly to larger farmers) to nothing to farmers for inputs or otherwise (Poulton et al., 2006). The reasons are diverse: as Dorward et al. (2009: 11) state, ÒCredit markets fail because of the inability to insure borrowers, lack of collateral, difficulties of recovering loans, and limited diversification of local economies, all of which impede the development of a sustainable model of rural financial services.Ó This evaluation of formal credit markets is echoed in other developing regions, such as South Asia (Binswanger and Khandker, 1995 and more generally (Besley, 1994, Conning and Udry, 2007). Our hypothesis to test is thus that farmers source little of their finance from banks, and only large farmers would get this. Second, informal credit from friends and family and local moneylenders appears to be in general deemed a major source of funds for farmers that do buy inputs and consumption items (Poulton et al., 2006, Binswanger and Khandker, 1995). Sometimes in the literature noting that farmers usually cannot get bank loans, it is implicit or explicitly noted that they must consequently resort to these local informal loans, or to trader credit noted below. Our hypothesis is thus that informal credit are important to all strata of farmers and all kinds of crop farmers would get them. Third, (informal sector) trader credit in Òtied output-creditÓ arrangements are deemed to be important and widespread (Bardhan, 1980, Dorward et al., 2009, Chao-B”roff, 2014). These can be crop traders or input traders. Traders and farmers enter these arrangements because local formal credit markets idiosyncratically fail for them, so in economics terms these are Òsecond bestÓ arrangements (Binswanger and Rosenzweig, 1986). We thus hypothesize that among farmers who buy inputs, obtaining these advances from output or input traders is important and not farm size or crop biased. !!+!Fourth, another kind of tied output-factor market arrangement is via Òlabor-outputÓ arrangements (Bardhan, 1984) where local farm workers advance labor in exchange for payment (typically in kind but can be in cash) at harvest. While this was researched in South Asia in the 1970s/1980s, it has not been examined empirically in SSA to our knowledge, and is an important gap in the literature we try to fill. Fifth, credit advances from (formal sector) processors in contract farming schemes (such as cotton or tobacco) are deemed important for the minority of farmers who participate in these arrangements (Fafchamps, 1999; Tschirley 2009; Key and Runsten 1999; Swinnen and Maertens, 2014). We thus hypothesize that for that subset of crops and farmers this will be found to serve many farmers. Sixth, household retained earnings are very rarely compared with credit sourcing by farmers. That seems to be because the credit literature focuses on patterns and determinants of sourcing credit rather than studying what are all the sources of funding inputs, per se. Non-credit related literature for example on rural nonfarm employment (RNFE) shows that source of cash to be the lead source in rural households in SSA (Haggblade et al. 2010). A few cases studies have compared RNFE with cash from credit and transfers and sales of crops and found the credit share (from any source) to be tiny, crops sales moderate, and RNFE as a cash source very important, predominant (for a case in the Sahel, see Reardon and Mercado-Peters 1994). We thus hypothesize that own cash sources will be important but possibly skewed toward larger farmers (due to skewed distribution of RNFE, see Reardon et al. 2000), with the smaller farmers relegated to funding inputs from the informal credit sources. This paper undertakes what to our knowledge is a unique cross-country empirical examination over various types of credit and different crops. We analyze recent LSMS farm !!,!household data sets with about 10,000 households from Malawi, Nigeria, Tanzania, and Uganda. We examine the purchase of Òexternal inputsÓ by which we mean non-labor variable inputs (fertilizer, pesticides, and seeds) and labor. We stratify by country, and also by farm size and crop type (primarily grains versus horticulture versus traditional Òcash cropsÓ like cotton. The determinants of fertilizer purchases are also explored in the specific case of Nigeria, to understand the importance of various sources of cash for agricultural inputs finances. The paper proceeds as follows. The next section discusses the data sources. Then we present descriptive results testing the hypotheses, and in the section thereafter, our econometrics results. Finally, we conclude with food and agricultural policy implications. 2.2.!Data We use survey data on farm household use of inputs and cash and in-kind arrangements to pay for them. The analysis is done by crop, household, and plot. The data also have characteristics of the farm households such as nonfarm income and farm size. The data come from four Living Standard Measurement Study (LSMS) surveys. The country surveys differ somewhat in the specific questions they use to elicit information on the variables of interest. We treat the survey datasets as uniformly as possible to ensure that the information is comparable over the sets. The details of the survey design and stratification methods are included in the documentation of the LSMS survey for each country. In general, the surveys used a two-stage sample design. In the first stage, enumeration areas were selected in each district of the country. Then, within each enumeration area a listing of households was conducted to provide the sample frame for the second stage selection of households. Then a random systematic sampling was used to select the households. !!-!In all countries, we extend the analysis to both urban and rural areas but select only households doing any farming. This is because we are interested in how households invested in agriculture generally finance input purchases. Also in all countries, there are a very tiny proportion of farm households in urban areas, and the separate analysis for urban and rural areas does not add any particular insight. Besides, in order to explore input finance arrangement, we focus on input purchases instead of input use, as some of the input use, though a very small part, come from government subsidies or from friends and relatives. Below we summarize the country data sets. First, we use the Malawi Integrated Household Panel Survey (IHPS) of 2012/2013, with 3219 households and 7705 plots. The data includes agriculture input credit use on fertilizer, pesticide and seed by season (rainy and dry). 3The seed data identify the purchase, crop type, whether credit was used and the type of credit. The fertilizer and pesticide data are by type of fertilizer and pesticide. There are also data on the use of harvest (in kind) to reimburse for inputs purchased on credit. The dataset also has information on the use of loans for purchase of farm inputs (among other uses). Second, we use the second wave of the Nigeria Living Standard Measurement Study ÐIntegrated Survey on Agriculture (LSMS-ISA) Panel for 2012/2013, covering 3000 farm households and 5819 plots. The data include agriculture input use/purchases and credit information for seeds, pesticides/herbicides, and fertilizer. The seed and fertilizer data indicate the crop type of the plot. There are also crop-level data on household use of in-kind or cash payments from the harvest to pay for labor, seed, and fertilizer received on credit. There are also data on households receiving and use of loans. 3 In Malawi we focus only on the rainy season information as agricultural activity is far less intense during dry season and inputs use and purchase decisions are considerably different. !!%.! Third, we use the United Republic of Tanzania National Panel Survey 2012/2013, covering 3047 households and 6165 plots. The data include (beside household and plot characteristics) input use by crop, use, purchase and finance method of organic fertilizer, inorganic fertilizer, and herbicides/pesticides. Credit use for these inputs was determined by reported purchase of input via credit. There are also data on loans and their uses (including for farming). Fourth, we use the Uganda National Panel Survey 2010/2011 covering 2109 farm households and 6003 plots across first and second season4. The data show per plot and by crop, for two seasons, the use and source of purchase for organic fertilizer, inorganic fertilizer, and pesticides/herbicides. We combine both types of fertilizers, and aggregate all the information to the household level. The data also show reimbursement for non-household labor used on the household plots and then the laborer paid with part of the householdÕs harvest. There is no loans (as contrasted with transaction related credit) section in the Uganda survey. 2.3.!Farm input purchases Ð abstracting for the moment from input finance In this section we examine farmersÕ purchase of what we call Òexternal inputs,Ó which are variable inputs apart from labor, and include inorganic fertilizer, seeds, and pesticides/herbicides. In this section we highlight the key patterns on how farmers financed these input purchases. In all the descriptive statistics we use sampling weights available in the dataset to account for the survey design and construct nationally representative statistics. The weight for each household is 4 Contrary to Malawi, we use information from first and second seasons in Uganda, as farmers are active in agricultural activities in both seasons. !!%%!the inverse of the probability of being selected based on the sample frame structure discussed above. Table 1 presents the shares of farmers purchasing what we call here Òexternal inputsÓ (variable inputs apart from labor). For the overall share of households buying external inputs, there is a marked contrast between Nigeria and Malawi, with a high share of farmers buying external inputs (71 and 70% respectively), compared to Uganda and Tanzania (16% and 18% respectively). The Malawi-Nigeria results are at odds with the traditional notion that very few farmers in SSA use external inputs but consistent with recent literature (Sheahan and Barrett, 2014). One might say that the Nigeria and Malawi results are driven by the fertilizer subsidy program. While this might be true in Malawi where about 60% of households receive subsidized fertilizer (IHS, 2013), this is not likely the case for Nigeria. In the Nigeria survey data, only about 5% of the households who purchase fertilizer bought it from government sources.5 The great majority of the householdsÕ external input purchases were with from the local market. 5 Note that there is no explicit question in the Nigeria LSMS that captures if a household participated in the government fertilizer subsidy program. However, until recently, subsidized fertilizer was distributed by the government. Since the government is not typically involved in the sale of non-subsidized fertilizer, we use the source of fertilizer purchase being government as a proxy for receiving subsidized fertilizer (Takeshima and Nkonya, 2014, Takeshima and Liverpool-Tasie, 2015). While this might be an underestimate in 2012 (since it was possible starting in 2012 for farmers to purchase subsidized fertilizer from dealers in the market with a coupon) we find the very low numbers in 2012 to be similar to those in 2010 (when the government was the sole distributor of fertilizer). !!%&!Table 2-1: Share of households who purchase external inputs by country Countries Share of households who purchase external inputs (%) Share of households (%) by type of inputs purchased Fertilizers Pesticides/ Herbicides Seeds Malawi 70 49 4 51 Nigeria 71 42 38 29 Tanzania 18 8 13 NA Uganda 16 5 14 NA Source: Generated by authors using LSMS data Note: NA implies that information is unavailable in the dataset used External inputs refer to fertilizer, seeds and agrochemicals (pesticides/herbicides) The shares are calculated amongst all households who farmed a plot in the year of the survey. Households with no agricultural plot farmed in that year are thus excluded. Moreover, for both Malawi and Nigeria, the data show the relative importance of fertilizer and seeds in terms of shares of households buying these. For Malawi, the incidence of purchase of pesticides/herbicides is much less. Interestingly, only about a half and a third of the farmers buying external inputs in Tanzania and Uganda buy fertilizer, yet a larger share buy pesticides and herbicides; this appears surprising, but is also consistent with Sheahan and Barrett (2014) for Uganda. Table 2 disaggregates input purchases by farm size. In each country, we stratified the farms by farm size strata: very small farmers (with less than 0.5 hectares) to larger (more than 5 hectares). Several points to note. First, farmland is very concentrated in the medium and large farm strata, while small farmers dominate in numbers of farms. Roughly 65-75% of the land is farmed by medium/large farmers, but 75-80% of the farms are small farmers and this is consistent across countries. Small farmers (less than 2 hectares, per Hazell and Rahman 2013) predominate in terms of shares of total numbers of farmers in the study countries: 96% in Malawi, 88% in Nigeria, 63% in !!%'!Tanzania, and 76% in Uganda, giving a simple average of 81% for all four countries, or 76% if one excludes Malawi. By contrast, the medium stratum (2 to 4.99 hectare) and larger farm stratum (5 and more hectares) have a total share of 38% (11 and 27% of farmland respectively) in Malawi, 65% (22 and 43% respectively) in Nigeria and 79% (32 and 47% respectively) in Tanzania, and 67% (30 and 37% respectively) in Uganda. For these two strata, the simple average over the study countries is 67%; excluding Malawi, it is 80%. Second, the shares of farmers buying external inputs are in a surprisingly tight distribution over farm size strata in all the countries. Again roughly grouping small farmers (up to 2 ha) and comparing with medium and larger farmers, one sees in Malawi that the shares (of farmers buying these inputs) is 71% for small farmers versus 88% for medium/large; for Nigeria, 78 versus 83%, for Tanzania, 15% versus 23%, and for Uganda, 14% versus 24%. However, despite farmland concentration in which the medium/large farmers have 67% of the land, they constitute a disproportionately lower share (36%) of the external input purchase ÒpieÓ. Thus the medium/large farmers are farming less intensively (in terms of applications of external inputs) than the small farmers and thus engaging less in the input market, and conversely, the small farmers are producing intensively with more external inputs per hectare, as farm technology intensification theory would predict (Binswanger-Mkhize and Ruttan, 1978, Boserup, 2005). This does not vary much over input types. But it does vary a lot over the countries: for Malawi, the medium/large group undertook 20% of the purchases in volume terms of all external inputs (versus its share in farmland of 38%); for Nigeria, the share of medium/large in external input purchases is 24%, versus its land share of 65%. For Tanzania, its share in inputs is 68%, near its share in land of 79%. In Uganda, its share is 42%, versus its share in land of 67%. !!%(!Table 2-2: Household purchase of external inputs by farm size strata Countries Farm size strata (ha) Share of farmers in this stratum (%) Share of national farmland in this stratum (%) Share of farmers who purchase external inputs (%) Share of total fertilizer demand bought by households in each stratum (%) Share of total pesticide/ herbicide demand bought by households in each stratum (%) Share of total seed demand bought by households in each stratum (%) Share of total inputs demand bought by households in each stratum (%) Malawi 0 -0.49 45 13 65 30 12 28 30 0.5 Ð 0.99 33 24 69 21 11 34 22 1 Ð 1.99 18 24 79 29 40 23 29 2 Ð 4.99 4 11 91 19 30 13 19 5+ 0 27 84 1 7 2 1 Nigeria 0 -0.49 53 8 62 30 19 55 30 0.5 Ð 0.99 20 12 78 25 20 17 23 1 Ð 1.99 15 16 83 23 24 13 22 2 Ð 4.99 9 22 82 16 21 8 16 5+ 3 43 85 5 16 7 8 Tanzania 0 -0.49 20 2 13 5 5 NA 5 0.5 Ð 0.99 19 5 14 9 7 NA 9 1 Ð 1.99 24 14 17 20 13 NA 19 2 Ð 4.99 26 32 22 41 46 NA 42 5+ 11 47 24 25 29 NA 26 !!%)!Table 2-2 (contÕd) Uganda 0 -0.49 26 4 6 6 5 NA 5 0.5 Ð 0.99 24 10 16 9 10 NA 10 1 Ð 1.99 26 20 20 35 48 NA 44 2 Ð 4.99 19 30 20 34 25 NA 28 5+ 6 37 28 16 12 NA 14 Source: Generated by authors using LSMS data Note: NA implies that information is unavailable in the dataset used External inputs refer to fertilizer, seeds and agrochemicals (pesticides/herbicides) !!%*!2.4.!Farm input finance for the overall sample and by farm size In this section, we explore the extent to which households use any credit arrangements for external input purchases and how that varies by farm size. We find consistent evidence across countries of very low use of any form of credit to buy these inputs (table 3). While Table 1 shows strong variation across countries in terms of shares of farmers buying external inputs, Table 3 shows only modest differences with respect to the very low shares (on average about 6%) of households that buy these inputs, using any form of credit. Table 2-3: Share of households purchasing external inputs that finance the input purchase on credit Countries Of those who bought external inputs, share of farmers buying on credit (%) Of those who bought seeds, fertilizers or pesticides/herbicides, share of farmers who bought on credit by input type Fertilizers Pesticides/ Herbicides Seeds Malawi 5 5 7 3 Nigeria 3 2 NA 3 Tanzania 11 14 7 3 Uganda 6 14 4 NA Source: Generated by authors using LSMS data Note: NA implies that information is unavailable in the dataset used External inputs refers to fertilizer, seeds and agrochemicals (pesticides/herbicides) Column 2 is the share amongst households who purchased at least one external input Column 3, 4, and 5 are shares amongst households who purchased fertilizers, agrochemicals, or seeds, respectively. The converse is that 94% only use their own cash to buy external inputs; this can be from non-credit resources such as cash sales of crops, and employment earnings (farm wage labor, migration, and rural nonfarm employment). In general for SSA, survey evidence has shown that !!%+!on average, among the employment earnings, rural nonfarm employment is a far greater source of income than migration or farm labor wage income (Haggblade et al. 2010). As noted in the introduction, there has been a presumption in the literature that to the extent farmers buy external inputs, they do it at least with informal credit or trader credit. But the analysis here shows that conventional wisdom is not supported empirically, and it is not just a lack of formal credit, but a near absence of the use of any credit, formal or informal, tied with input or output traders, in kind or in cash. Some other points to note. First, of the very small shares of farmers buying external inputs on credit, there is within those sets sharp variation over input types. There tends to be 2-3 times more households getting some kind of credit for fertilizer compared to seeds or pesticides/herbicides. Second, across all inputs, the limited credit based expenditures are for larger farmers. Table 4.a shows the shares of the landholding strata in all credit-based input outlays, so a sort of ÒpieÓ of strata shares in all credit transactions. The table shows that in Malawi, Tanzania, and Uganda, input credit is roughly farm size correlated - the great majority of the credit-based external input expenditures are concentrated outside the below-one-hectare group. Nigeria is a sharp outlier, with the great majority of the input credit taken by the Òunder 1 haÓ group. These results do not differ much over input type. This is largely confirmed by the distribution of shares, by stratum, of credit-based input outlay for each input i in total outlay for input i (table 4b). The importance of input credit tends to be concentrated in the middle to higher end of farm sizes, and be mainly in fertilizer and not very much in pesticides and seeds. In Malawi, Tanzania and Uganda, input credit is important only for fertilizer, averaging around 9% of fertilizer input outlay in Malawi but concentrated in the upper-small and medium farmers (1-5ha) where it averages a fifth of input expenditure. In !!%,!Tanzania, the share of input expenditure done on credit is correlated with land size, with about 10% for smaller farmers and about a quarter and a half for medium and larger farmers. For Uganda, it is only highly important for the 1-5ha group, where it reaches 40-50% of fertilizer expenditure. In Nigeria, the share is low for all, with about 3% on average and does not differ much over strata (except for small spikes to 11-12% among large farmers for fertilizer and seeds for 1-2 ha farmers). !!%-!Table 2-4a: Shares of Strata in all credit-based expenditure on external inputs Countries Farm size strata Share of stratum buying on credit (%) Share of stratum in all credit-based fertilizer outlay (%) Share of stratum in all credit-based pesticide/ herbicide outlay (%) Share of stratum in all credit-based seed outlay (%) Share of stratum in all credit-based input outlay (%) Malawi 0 -0.49 3 4 11 13 4 0.5 Ð 0.99 3 4 15 16 4 1 Ð 1.99 10 61 38 44 60 2 Ð 4.99 10 32 36 27 32 5+ 14 0 0 0 0 Nigeria 0 -0.49 3 49 NA 13 45 0.5 Ð 0.99 5 22 NA 22 22 1 Ð 1.99 4 11 NA 62 16 2 Ð 4.99 1 2 NA 0 2 5+ 6 16 NA 3 14 Tanzania 0 -0.49 2 0 0 NA 0 0.5 Ð 0.99 6 4 3 NA 4 1 Ð 1.99 8 10 15 NA 10 2 Ð 4.99 20 36 69 NA 38 5+ 24 50 12 NA 48 Uganda 0 -0.49 0 0 0 NA 0 0.5 Ð 0.99 2 3 17 NA 5 1 Ð 1.99 11 57 54 NA 56 2 Ð 4.99 11 40 28 NA 39 5+ 0 0 0 NA 0 Source: Generated by authors using LSMS data Note: NA implies that information is unavailable in the dataset used External inputs refers to fertilizer, seeds and agrochemicals (pesticides/herbicides) In Column 3 the share is amongst households who purchased at least one external input !!&.!Table 2-4b: Share of credit-based input outlay for input i in total outlay for input i per stratum Countries Farm size strata Share of credit-based input outlay for fertilizer in total outlay per stratum (%) Share of credit-based input outlay for pesticides/ herbicides in total outlay per stratum (%) Share of credit-based input outlay for seeds in total outlay for per stratum (%) Share of credit-based input outlay for all inputs in total outlay for per stratum (%) Malawi 0 -0.49 1 3 2 1 0.5 Ð 0.99 2 5 2 2 1 Ð 1.99 22 4 8 21 2 Ð 4.99 18 4 8 17 5+ 0 0 0 0 Nigeria 0 -0.49 6 NA 1 4 0.5 Ð 0.99 3 NA 3 3 1 Ð 1.99 2 NA 12 2 2 Ð 4.99 1 NA 0 0 5+ 11 NA 1 5 Tanzania 0 -0.49 2 0 NA 2 0.5 Ð 0.99 12 4 NA 11 1 Ð 1.99 15 10 NA 14 2 Ð 4.99 26 12 NA 23 5+ 58 3 NA 48 Uganda 0 -0.49 0 0 NA 0 0.5 Ð 0.99 12 3 NA 6 1 Ð 1.99 53 2 NA 17 2 Ð 4.99 40 2 NA 19 5+ 0 0 NA 0 Source: Generated by authors using LSMS data Note: NA implies that information is unavailable in the dataset used External inputs refers to fertilizer, seeds and agrochemicals (pesticides/herbicides) In Column 3 the share is amongst households who purchased at least one external input !!&%!2.5.!Crop type and input purchase financing In this section we explore the correlation between the type of crop grown by farmers and financing Òexternal inputsÓ on credit. We aggregate the purchase and the financing of external inputs over plots (to the household level) by crop. We classify crops into a set of what are traditionally called Òfood cropsÓ (although they are often also sold for cash), including grains, horticulture, legumes, and tubers (grown as a staple), and what are traditionally called Òcash cropsÓ, including tobacco, cotton, tea/coffee, and edible oil crops. Three points stand out from an analysis of the distribution of household producing at least some of the crop types of cash crop versus food crops, by country and by farm size strata (Table 5). First, as expected, grains dominate, but interestingly are not ubiquitous, reaching only about three-quarters of the farms in Nigeria, Tanzania, and Uganda, being near 100% only in Malawi. While there is a lot of variation, over the countries on average nearly a third of the farms grow horticultural crops, and on average a half grow beans/pulses, and a third grow tubers. The food cropping is thus fairly diversified on average. Second, by contrast, production of cash crops is much more concentrated over farms in every country. Overall, on average over countries only a fifth of farmers grow cash crops, and that is but a tenth if one excludes Uganda. The crop focus differs over countries, with tea/coffee and oil crops standing out in Uganda, cotton and oil crops in Tanzania, oil crops in Nigeria, and tobacco and cotton in Malawi. Third, there is little farm size bias in participation in all the food crops. The exception is that the smallest farms (below a half hectare) have a modestly lower participation rate than the other strata in all food crops but tubers, where they have higher or similar participation compared !!&&!with the other strata. By contrast, for cash crops, there is a marked correlation of the share of farms producing any cash crop and farm size. Table 2-5: Share of households producing key cash and food crops across farm size strata Crop types Farm size strata (hectares) Share of farmers producing each crop type (%) Malawi Nigeria Tanzania Uganda Cash crops Tobacco 0 -0.49 2 0 0 0 0.5 Ð 0.99 11 0 0 0 1 Ð 1.99 24 0 1 3 2 Ð 4.99 32 0 2 2 5+ 28 0 2 0 All 10 0 1 1 Cotton 0 -0.49 2 1 1 2 0.5 Ð 0.99 6 3 2 4 1 Ð 1.99 13 3 4 4 2 Ð 4.99 19 1 5 6 5+ 0 0 10 8 All 6 1 4 4 Tea/coffee 0 -0.49 0 0 2 22 0.5 Ð 0.99 0 0 2 21 1 Ð 1.99 0 0 2 22 2 Ð 4.99 0 0 3 22 5+ 0 0 0 36 All 0 0 2 23 Oil crops 0 -0.49 0 10 2 3 0.5 Ð 0.99 1 7 4 10 1 Ð 1.99 3 8 5 12 2 Ð 4.99 2 19 5 19 5+ 0 14 5 16 All 1 10 4 11 All cash crops 0 -0.49 4 10 4 26 0.5 Ð 0.99 18 10 8 34 1 Ð 1.99 39 11 11 39 2 Ð 4.99 49 20 14 46 5+ All 28 17 14 11 18 11 54 37 !!&'!Table 2-5 (contÕd) Food crops Grains 0 -0.49 98 69 61 70 0.5 Ð 0.99 99 87 74 83 1 Ð 1.99 99 86 79 86 2 Ð 4.99 99 84 83 81 5+ 100 88 85 81 All 99 77 76 80 Horticulture 0 -0.49 29 33 22 55 0.5 Ð 0.99 31 21 13 50 1 Ð 1.99 37 22 12 48 2 Ð 4.99 32 23 9 46 5+ 43 17 7 63 All 31 28 13 51 Legumes 0 -0.49 62 29 12 76 0.5 Ð 0.99 76 56 10 75 1 Ð 1.99 79 60 12 77 2 Ð 4.99 77 53 16 82 5+ 93 54 16 82 All 71 42 13 78 Tubers 0 -0.49 8 61 16 74 0.5 Ð 0.99 9 30 19 79 1 Ð 1.99 14 34 19 74 2 Ð 4.99 16 39 18 76 5+ 0 49 20 71 All 10 48 18 75 All food crops 0 -0.49 100 98 95 100 0.5 Ð 0.99 100 98 97 99 1 Ð 1.99 100 99 96 100 2 Ð 4.99 100 98 95 100 5+ 100 99 97 99 All 100 98 96 100 Source: Generated by authors using LSMS data Conventional wisdom suggests that farmers growing cash crops would commonly access external inputs on credit, in particular from processors, while food crop producers may not. To test this, we explore the shares (by crop type) of farm plots on which inputs purchased on credit !!&(!are used (Table 6). While there is a lot of variation over countries, the average over all cash crops is 13%, compared with 6% for food crops. First, this is surprising because the shares do not differ greatly, as we had expected. Second, this average figure masks higher ratios in two pairs of countries; Malawi and Tanzania who average 20% for cash crops and 7% for food, while Nigeria and Uganda average 6% for cash crops and 5% for food which is extremely similar. A closer look indicates that the main difference between cash and food crops in this respect is driven by tobacco in Tanzania and Uganda, where four-fifths of the plots are grown with inputs bought on credit from the processors. Removing the tobacco outlier (for just Tanzania and Uganda) puts the overall credit share for cash crops close to that of food; as observed in the other study countries. Also, that outlier is composed of a tiny group of tobacco farmers in the sample for each country, about 1% of the total sample. The very limited and ÒenclaveÓ nature of tobacco farming and its correlation with farm size (see Table 5) in those countries could explain why these are the main cases where the conventional image of contract-farming related credit is manifest. Table 2-6: Cash crops versus food crops on which purchased external inputs are used that were financed on cash-credit for key cash and food crops (% of plots) Malawi Nigeria Tanzania Uganda Cash crops Tobacco 16 NA 87 81 Cotton 11 8 11 0 Tea/coffee NA NA 22 1 Oil crops 6 3 4 11 All cash crops 14 4 26 8 Food crops Grains 5 3 11 7 Horticulture 4 3 0 4 Legumes 5 2 11 6 Tubers 7 3 4 5 All food crops 5 3 10 6 Source: Generated by authors using LSMS data Note: NA implies that information is unavailable in the dataset used The numbers are percentages amongst of households producing each type of crop in each country. !!&)!2.6.!Tied output input credit arrangements Tied output/input credit arrangements occur when repayment for inputs on credit (received at planting) are made at harvest time. The LSMS data for our four study countries include a section about the management of crop harvests. We use farmersÕ responses concerning use of part of their harvests to repay advances for inputs from input or output traders and processors (especially for cash crops) for external inputs, and labor from workers. Table 7 shows the share of farmers, overall and per stratum, using part of their harvests for these ends. The main finding is that such Òtied creditÓ is very rare for external inputs (fewer than 2% of the farmers) across all study countries. By contrast, and reported for the first time in the Sub-Saharan African literature using cross-country surveys for comparison, we find that labor-output tying is much more common, with as many as 42% of the Malawi, 26% of Nigerian, and 68% of Tanzanian farmers doing this practice. (The dataset for Uganda did not allow this calculation.) By contrast, and not reported in the table, tying the land and output markets was not found to be common; the land tenure section of the surveys showed that sharecropping was extremely limited. Moreover, the patterns of differentiation over strata differ by country so no single story emerges. For harvest payment to labor, in Uganda, the share rises with farm size, in Nigeria it slightly declines, and in Malawi it is in an inverted-U shape relation with farm size. Thus one cannot say that this traditional-tying of labor and harvest is more a phenomenon of the smallest farmers holding on to an old practice, as one might expect, given our hypothesis that larger farms are more apt to use monetized labor relations only. For harvest payment for external inputs, the shares are so small that there are no interesting inter-strata differences. !!&*!Table 2-7: Share of farmers using harvest to reimburse for inputs received on credit by farm size Countries Farm size strata Share of farmers using their harvest to repay labor received on credit (%) Share of farmers using their harvest to repay external inputs received on credit (%) Malawi 0 -0.49 37 1 0.5 Ð 0.99 45 3 1 Ð 1.99 50 2 2 Ð 4.99 47 1 5+ 24 0 All 42 1.8 Nigeria 0 -0.49 26 1 0.5 Ð 0.99 29 1 1 Ð 1.99 26 3 2 Ð 4.99 21 2 5+ 22 3 All 26 1.4 Tanzania 0 -0.49 NA 0 0.5 Ð 0.99 NA 1 1 Ð 1.99 NA 1 2 Ð 4.99 NA 4 5+ NA 5 All NA 1.9 Uganda 0 -0.49 54 NA 0.5 Ð 0.99 63 NA 1 Ð 1.99 74 NA 2 Ð 4.99 78 NA 5+ 81 NA All 68 NA Source: Generated by authors using LSMS data Notes: NA implies that information is unavailable in the dataset used !!&+!When we consider the Òreimbursement of credit with the harvestÓ by type of crop, it is very minor or zero for the other cash crops (except tobacco in Tanzania, discussed below), and all of the food crops (Table 8). By contrast, use of harvest repayment for labor is very minor for cash crops (except for oil crops in Uganda where it is a quarter of farmers using it), but is significant in food crops across the countries, such as about a third in horticulture and a quarter in grains. Interestingly, there is only a single situation (crop plus country) where this arrangement is important for external inputs, and that is for tobacco in Tanzania. We conjecture that this high prevalence of the use of harvest to reimburse for external inputs received on credit to produce tobacco in Tanzania is related to a widespread use of contract farming arrangement over tobacco production in Tanzania. If our conjecture is true, we should then expect to see a lot more contract farming arrangement over tobacco compared to cotton, tea/coffee, and oil crops. Therefore we investigated outgrower schemes in the Tanzania data set and found that a very small proportion of farmers (1.8%) are involved in outgrower schemes. Moreover, tobacco represents 78.1% of the crops grown as part of an outgrower scheme or contract farming system, followed by cotton (18.8%). Though this does not say how much of the tobacco produced is grown as part of outgrower scheme, at least it gives an indication of the dominance of tobacco amongst the crop grown as part of outgrower schemes, and therefore confirms our conjecture about the prevalence of tied output-input arrangements for tobacco in Tanzania. There is no information about contract farming or outgrower schemes in the other countries to allow us to compare this pattern across countries. !!&,!Table 2-8: Financing inputs on credit with harvest across key cash and food crops. Crops types Share of plots where harvest is used to repay (advanced) labor (%) Share of plots where harvest is used to repay external inputs (%) Nigeria Malawi Uganda Tanzania Nigeria Malawi Uganda Tanzania Cash crops Tobacco 0 2 0 NA 0 2 NA 79 Cotton 10 0 0 NA 0 1 NA 6 Tea/coffee NA NA 1 NA NA NA NA 3 Oil crops 8 0 25 NA 0 0 NA 0 Food crops Grains 17 22 27 NA 1 1 NA 1 Horticulture 18 32 36 NA 1 0 NA 0 Legumes 9 21 25 NA 1 1 NA 0 Tubers 5 29 30 NA 1 1 NA 0 Source: Generated by authors using LSMS data Notes: NA implies that information is unavailable in the dataset used !!&-!Overall our results indicate that there is much less tied credit arrangement to finance external input than expected. Even though those arrangements appear to be more formal (from contract farming arrangements) and more likely for cash crops, we still see far less than expected (except for tobacco), even though the literature indicates the contrary. 2.7.!HouseholdsÕ use of loans not specifically linked to input transactions We use the term ÒloansÓ for credit unconnected directly and specifically to transactions of outputs or inputs. Such loans can come from formal (banks), semi-formal (micro-finance), and informal sources (friends, relatives, cooperatives, etc.). The LSMS data show loan data for Malawi, Nigeria, and Tanzania, and Malawi and Tanzania show what the loans were used for. We find evidence that households in SSA do take loans however, this is rarely used for agricultural purposes. Nigeria had as much as 38% of farmers taking loans (Table 9). In Malawi, 23% of the households took a loan, but only 5% of them did so for farming; in Tanzania, it was but 11% taking loans of which 2% for farming purposes, hence a 5 to 1 ratio of overall loans to farm-destined loans in both (Table 9). Table 2-9: Share of households with a member taking a financial loan Country Share of HHs taking a loan (%) a Of those who took loan, share of HHs taking loans for farming (%) Malawi 23 5 Nigeria 38 NA Tanzania 11 2 Uganda NA NA Source: Generated by authors using LSMS data Notes: NA implies no data a captures whether any household member received a loan in the last 12 months !!"#!Table 2-10: Purpose for which loans are taken (by source of loan) in Malawi Source of loan Share of loans taken for land purchase (%) Share of loans taken for inputs purchase for food crops (%) Share of loans taken for input purchase for tobacco (%) Share of loans taken for input purchase for other cash crops (%) Share of loans taken for business start-up capital (%) Share of loans taken for purchase of non-farm inputs (%) Share of loans taken for consumption (%) Share of loans taken for other purposes (%) Relative 0 13 1 4 24 4 36 19 Neighbors 1 7 1 2 24 7 44 14 Grocery/local merchant 0 4 0 0 7 7 83 0 Money lender (katapila) 0 11 1 6 19 6 40 19 Employer 10 19 0 0 0 10 52 10 Religious institution 0 23 8 0 15 8 39 8 Mardef 0 0 0 0 71 0 29 0 Mrfc 0 0 14 0 43 29 14 0 Sacco 7 4 0 4 30 11 19 26 Loans from all sources 1 9 2 4 31 8 31 14 Source: Generated by authors using LSMS data Notes: MArdef and Mrfc are leading microfinance institutions in Malawi !!"$!Table 2-11: Purpose for which loans are taken (by source of loan) in Tanzania Source of loan Share of loans for subsistence needs Share of loans for medical cost Share of loans for school fees Share of loans for ceremony/ wedding Share of loans for land purchase Share of loans for purchase inputs Share of loans for other business inputs Share of loans for purchase Machinery Share of loans for buy/build dwelling Share of loans for other purposes Commercial banks 12 4 20 0 4 2 33 0 19 7 Microfinance institutions 8 0 15 0 5 8 42 0 15 8 Building society./mortgage 0 0 0 0 0 100 0 0 0 0 Other financial institutions 8 0 25 4 4 0 29 0 25 4 Neighbors / friends 48 12 4 3 2 5 14 1 3 10 Grocery/local merchant 62 6 2 0 0 5 19 0 2 4 Money lender 15 15 5 0 0 5 30 0 5 25 Employer 14 0 21 7 14 0 7 0 7 29 Religious institutions 0 0 0 0 17 17 33 0 0 33 NGO 0 0 13 0 0 25 25 13 13 13 Self-help groups 32 0 11 2 1 6 30 0 9 9 Others 23 7 5 2 5 23 11 5 7 14 Loans from all sources 31 6 10 2 3 6 24 1 9 9 Source: Generated by authors using LSMS data !!"%! Table 2-12: Sources of Cash Income in Nigeria, North and South, 2010 and 2012 INCOME SOURCES HOUSEHOLD CASH SOURCES (000 Naira) SHARE OF CASH FROM EACH SOURCE (%) NIGERIA SOUTH NORTH NIGERIA SOUTH NORTH 2010 2012 2010 2012 2010 2012 2010 2012 2010 2012 2010 2012 CASH INCOME Net profit from household enterprise 48.3 113.1 41.2 102.2 53.8 121.5 17 26 12 18 22 38 Wage income 193.2 261 249.7 406 149.9 151 67 60 75 70 60 47 Crop sales 42 56.7 38.2 66.2 44.9 49.4 15 13 11 11 18 15 Livestock net sales 0.7 0.9 0.4 1 0.9 0.9 0 0 0 0 0 0 Remittances 2.2 1 5 1.8 0.1 0.4 1 0 1 0 0 0 Total cash 286.4 432.7 334.5 577.2 249.6 323.1 100 100 100 100 100 100 Inputs credit transactions 0.2 0.3 0.1 0.1 0.3 0.5 0 0 0 0 0 0 Inputs non credit transactions 7.2 10.6 3.8 3.9 9.7 15.7 3 2 1 1 4 5 Total input purchase 7.3 11 3.8 4 10 16.3 3 3 1 1 4 5 Hired labor value for harvest only 18.9 12.4 8 7 27.2 16.5 7 3 2 1 11 5 Imputed value of own crop output 112.9 137.4 72.5 83.1 143.8 178.6 39 32 22 14 58 55 Source: Generated by authors using LSMS data Note: The numbers in the left panel are zero-in averages. The shares on the right are based on ration of number on the left to the total cash value. Inputs include fertilizer, seeds, and pesticides, except in Input credit transactions where it does not include pesticides credit because the information was not available in the data. For each value in the table, instead of deleting outliers we winsorized them i.e. replace top 10% values by the highest value within 90% of the distributions, thus creating a pile up at the top without changing the distribution (Cox, 2006). For Imputation of value of own crop output method, we estimate unit prices of crops for crops that were sold, and then we use the median price in the local governments and multiply by harvest quantities to get the value of crop sales. The harvest labor for planting activities is missing in the 2010 dataset, and therefore we focus on the harvest labor only in both years. !!""!Instead, the loans were taken for nonfarm business startup and for consumption (Tables 10 and 11). This is striking because one would expect credit-constrained farmers to use these loans to finance farm input purchases. As is shown below, a key factor that determines external input purchase is engaging in nonfarm enterprises and wage labor. Thus it appears that farmers prefer to use financial credit to finance the set up/expansion of their nonfarm enterprises but use the generated cash from these nonfarm enterprises to finance external input purchases. 2.8.! Determinants of inputs purchase in Nigeria Thus far, we find consistent evidence that the use of any form of credit to finance external input purchase is extremely low. This begs to question how households finance these purchases. While one might expect credit-constrained farmers to use loans to finance input purchase, we find that farmers tend to use loans to finance business startup and consumption. This could be driven by the risky nature of agricultural investments and/or low expected returns on investments in modern inputs relative to the cost of credit.6 In this section, we explore how farmers finance their input purchase by estimating the determinants of fertilizer purchases by Nigerian farmers. Our analysis lays emphasis on the role of non-farm employment (wage and self-employment) and agricultural productivity risks (captured by rainfall variability), as well as regional differences (north versus south) in fertilizer purchases decisions and intensity amongst Nigerian farm households. 6 We do not test this hypothesis as it is beyond the goal of this paper. However, we recognize that it is an important question that deserves further investigation. !!"#!2.8.1.!Conceptual and empirical framework The fertilizer purchase decision follows a standard input demand function derived from a constrained household utility maximization problem and presented in Sadoulet and de Janvry (1996). Fertilizer demand can be expressed as a function of a vector of prices, risk proxies, a vector of complementary and substitute farm capital, and relevant shifter variables such as crop type. We first consider the decision to purchase a particular input or not and then the extent of input purchase. We model the farmers fertilizer purchase decisions using the following unobserved effect binary dependent variable model (Green William, 2000, Wooldridge, 2010): !"#$%&'("#)*+,"#+-"./0 ; i=1, 2, É, N ; t=1, 2,É,T (1) In the model above, 123$ is the underlying latent variable which characterize farmer iÕs net benefit (or utility) from purchasing fertilizers, in period t. While this latent variable is unobserved, it determines the observed binary outcome variable 123 (fertilizer purchase) which takes value 1 if 123$.4 and a fertilizer is purchased, and 0 otherwise. 523 is the vector of explanatory variables included in the model. 6 is the vector of parameters of interest. 723 is the error term assumed to follow a standard normal distribution, leading to an unobservable effects Probit model for the fertilizer purchase equation (Green William, 2000, Wooldridge, 2010). 89:;<123%=>523?@2A%B<523)6+@2A (2) For the intensity of fertilizer purchase, we model it using the following unobserved effects Tobit model to account for zeros dues to corner solution in the dependent variable (Wooldridge, 2010): 1C23%DEFG<4?H23)6+@2+I23A ; i=1, 2, É, N ; t=1, 2,É,T (3) JI23H23?@2%K:9LMNG<4?OPQA !!"$!In this model, 1C is the dependent variable representing the number of kilograms of fertilizer purchased per unit of land cultivated. H23 is the vector of explanatory variables that potentially affect quantity of fertilizer purchased. I23 is the error term assumed to follow a normal distribution with mean 0 and standard deviation OP. In both models, @2 represents the unobserved effect parameter, modeled using the Mundlak (1978) special case of (Chamberlain, 1982) approach called correlated random effect (CRE): @2%GR+5ST+M2, M2>52GUGK:9LMN<4?OVQA where 5S represents time averages of the explanatory variables. Assuming conditional independence, the full model becomes: 89:;<123%=>523?5SA%B<523)6V+RV+5STVA, (4) with WV%WX<=+OVQACXQ , W%<6?R?TGA Average partial effects are identified and can be derived from the above model as: 8YZ[%\]^_`ebc?efA\Z[c%6Vghi523)6V+RV+5STV, with WV%WX<=+OVQACXQ, W%<6?R?TGA, for the Probit model; and 8YZ[%6VghBH23)6V+RV+HSTV with jV%jX