Agriculture global practice technical assistance Paper 94883 tanzania Agricultural Sector Risk Assessment Carlos E. Arce and Jorge Caballero WORLD BANK GROUP REPORT NUMBER 94883-TZ June 2015 Agriculture Global Practice Technical Assistance Paper TANZANIA Agricultural Sector Risk Assessment Carlos E. Arce and Jorge Caballero © 2015 World Bank Group 1818 H Street NW Washington, DC 20433 Telephone: 202-473-1000 Internet: www.worldbank.org Email: feedback@worldbank.org All rights reserved This volume is a product of the staff of the World Bank Group. The findings, interpretations, and conclusions expressed in this volume do not necessarily reflect the views of the Executive Directors of World Bank Group or the governments they represent. The World Bank Group does not guarantee the accuracy of the data included in this work. The boundaries, colors, denominations, and other information shown on any map in this work do not imply any judgment on the part of World Bank Group concerning the legal status of any territory or the endorsement or acceptance of such boundaries. Rights and Permissions The material in this publication is copyrighted. Copying and/or transmitting portions or all of this work without permission may be a violation of applicable law. World Bank Group encourages dissemination of its work and will normally grant permission to reproduce portions of the work promptly. For permission to photocopy or reprint any part of this work, please send a request with complete information to the Copyright ­ Clearance Center, Inc., 222 Rosewood Drive, Danvers, MA 01923, USA, telephone: 978-750-8400, fax: 978-750-4470, http://www.copyright.com/. All other queries on rights and licenses, including subsidiary rights, should be addressed to the Office of the Publisher, World Bank Group, 1818 H Street NW, Washington, DC 20433, USA, fax: 202-522-2422, e-mail: pubrights@worldbank.org. Photo credits from left to right: Anne Wangalachi/CIMMYT, Cecilia Schubert/CCAFS, Anne Wangalachi/CIMMYT, and Scott Wallace/World Bank. Contents Acknowledgments vii Abbreviations and Acronyms ix Executive Summary xi Chapter One: Introduction and Context 1 The Scope of the Study 1 The Risk Assessment Process 1 Contents of the Report 2 Chapter Two: Agricultural System 3 Agricultural Sector Overview and Performance 3 Agro-Climatic Conditions 4 Rainfall Trends 4 Crop Production System and Exports 5 Yield, Production, Acreage, and Market Trends 6 Principal Constraints in the Agricultural System and Policy Reform 6 Chapter Three: Agricultural Sector Risks 9 Empirical evidence17 Chapter Four: Adverse Impact of Agricultural Risk 21 Quantification of losses21 Production volatility21 Chapter Five: Stakeholders’ Assessment 25 Impact of Risks at Individual Stakeholder Level 25 Vulnerable hotspots25 Chapter Six: Risk Prioritization and Management 29 Risk prioritization29 Priority Risk Management Measures 30 References 37 Appendix A: Weather Analysis 41 Appendix B: Impact of Climate Change on Agricultural Sector 63 Appendix C: Vulnerability Analysis 73 BOXES Box 2.1: Commodity Boards Reform during the 1990s and Early 2000s 7 Box 3.1: Regional Vulnerability 10 Box 3.2: Examples of Changing Environment 16 Agricultural Sector Risk Assessment iii FIGURES Figure 1.1: The Risk Assessment Process 2 Figure 1.2: Risk Identification and Prioritization 2 Figure 2.1: Tanzania Real GDP Growth by Sector 4 Figure 2.2: Land Resources Zones 5 Figure 2.3: Maize Production by Region 6 Figure 2.4: Traditional Export Crops 6 Figure 2.5: Planted Area 6 Figure 3.1: International Coffee Price Change 12 Figure 3.2: Coffee International-Domestic Price Comparison 12 Figure 3.3: International Cotton Price Change 13 Figure 3.4: Cashew Nut Export Price 15 Figure 3.5: Maize Price 17 Figure 3.6: Regression Chart for Arusha 17 Figure 3.7: Regression Chart for Dodoma 18 Figure 3.8: Coffee: Occurrence of Risk Events 18 Figure 3.9: Maize and Rice: Occurrence of Risk Events 18 Figure 3.10: Cotton: Occurrence of Risk Event 19 Figure 4.1: Maize Production Volatility 23 Figure 5.1: Food Balances 27 Figure A.1: Monthly Rainfall Pattern for Pixel #84 42 Figure A.2: Monthly Rainfall Pattern for Pixel #77 42 Figure A.3: Tanzania Region Centroids 42 Figure A.4: Maize Surface Sowed and Production Volume, Time Series 43 Figure A.5: Maize Yield, Time Series 43 Figure A.6: Average Maize Surface by Region in Thousand Hectares 44 Figure A.7: Maize Yield Histogram for All Regions 44 Figure A.8: Sowing Season Rainfall and Yield Time Series for Arusha and Manyara Regions 46 Figure A.9: Linear Regression Models for Arusha and Manyara Regions (stage 1) 46 Figure A.10. Linear Regression Models for Arusha and Manyara Regions (stage 2) 46 Figure A.11: Linear Regression Models for Arusha and Manyara Regions (stage 3) 47 Figure A.12: Lindi and Kagera Regions Mid-Season Rainfall Models 48 Figure A.13: Tabora Region Harvest Season Model 49 Figure A.14: Yearly National Paddy Rice Surface Area Sowed and Production 49 Figure A.15: Yearly National Paddy Rice Yield 49 Figure A.16: Average Distribution of Surface Sown per Region 49 iv Tanzania Figure A.17: Relationship between Rice Yield and Rainfall Variability in Dar es Salaam 50 Figure A.18: Relationship between Rice Yield and Rainfall Variability in Kilimanjaro  50 Figure A.19: Relationship between Rice Yield and Rainfall Variability in Rukwa and Manyara  51 Figure A.20: Cotton Area and Production  51 Figure A.21: Cotton Yield 51 Figure A.22: Linear Regression Models for Iringa and Mwanza Regions 53 Figure A.23: Linear Regression Models for Iringa and Manyara Regions 53 Figure A.24: Sorghum Surface Sown and Production Volume 54 Figure A.25: Sorghum Yield, 1981–82 and 2009–10 54 Figure A.26: Linear Regression Models for Manyara and Ruvuma Regions 55 Figure A.27: Millet Surface Sowed and Production Volume  55 Figure A.28: Millet Yield 1981–82 and 2009–10 55 Figure A.29: Linear Regression Model for the Kagera Region 56 Figure A.30: Linear Regression Model for the Tabora Region 56 Figure A.31: Tobacco Surface Sowed and Production Volume 56 Figure A.32: Linear Regression Models for Mbeya and Rukwa Regions 57 Figure A.33: Linear Regression Model for the Ruvuma Region 58 Figure A.34: Linear Regression Models for the Kigoma Region 58 Figure A.35: Linear Regression Model for the Arusha Region 59 Figure A.36: Linear Regression Model for the Manyara Region 59 Figure A.37: Linear Regression Model for the Tanga Region 59 Figure A.38: Linear Regression Model for the Dodoma Region 60 Figure A.39: Linear Regression Model for the Manyara Region 60 Figure A.40: Linear Regression Model for the Mbeya Region 61 Figure A.41: Linear Regression Model for the Ruvuma Region 61 Figure B.1: Mean Annual Dry-Land Maize Yield Changes under HOT, COOL, WET, and DRY Scenarios, 2041–50 68 Figure C.1: Distribution of Poor and Borderline Food Consumption Household 75 Figure C.2: Hazard Occurrence in the Agro-Ecological Zones 77 TABLES Table 2.1: Tanzania at a Glance 4 Table 2.2: GDP Composition by Sectors 4 Table 2.3: Average 2007/08–2009/10, Except for Coffee and Cashew Nut, Which Is 2008/09–2009/10 7 Table 4.1: Value of the Average Annual Losses (at 2010 prices) 22 Table 4.2: Maize Production Variability by Region 23 Table 5.1: Summary of Stakeholder Risk Profiles 27 Table 6.1: Risk Prioritization—Food Crops 30 Agricultural Sector Risk Assessment v Table 6.2: Risk Prioritization—Export Crops 31 Table 6.3: Risk Solutions: The Long List 32 Table 6.4: Gap Analysis 33 Table A.1: The Five Nearest Pixels to Each Region’s Centroid 43 Table A.2: D  etermination Coefficient (R2) of the Linear Regression Models Applied to The First Rainfall Pattern on All Regions 45 Table A.3: Determination Coefficients for Each Stage and Region 47 Table A.4: Summary of Rice Regression Analysis Results 50 Table A.5: Cotton Regression Analysis Results 52 Table A.6: Sorghum Regression Analysis Results 54 Table A.7: Millet Regression Analysis Results 56 Table A.8. Tobacco Linear Regression Results 57 Table A.9: Maize Sowing Calendar 58 Table A.10: Maize Regression Analysis Results  59 Table C.1: Factors Associated with Food Security by Region  76 vi Tanzania ACKNOWLEDGMENTS This report was developed by a team led by Carlos E. Arce, Senior Economist from the Agricultural Risk Management Team at the World Bank. The activities were sup- ported by the following consultants: Jorge Caballero, Aira Htenas, Michael Westlake, Hussein Nassoro, Cyril Chimilila, and Nicholai Chiweka. David Rohrbach and Vikas Choudhary from the World Bank participated at various stages during the process. The team is grateful to the leadership and coordination received from Janet Simu- kanga, Director of Policy and Planning at the Ministry of Agriculture, Food Security and Cooperatives, as well as to her technical team for the fruitful and productive dis- cussions over the findings of this assessment. The team would like to extend its appreciation to the stakeholders from major agri- cultural supply chains that participated at various moments during the fieldwork and during the workshops to discuss the findings. Their participation obliged the team to be realistic and practical. Valuable comments were received from Sergiy Zorya, Senior Economist, World Bank, Andrew Temu, Professor of Agricultural Economics at Sokoine University, and from Gungu Mivabu, Agricultural Sector Policy Analyst from the Ministry of Agriculture, both acting as external peer reviewers. The team also appreciated the constructive comments received from Kevin McCown, Senior Agricultural Economist from U.S. Agency for International Development (USAID) Tanzania Office, and is grateful also for the work of consulting editor Damian Milverton. This activity would not have been possible without the generous contributions from USAID, Ministry of Foreign Affairs of the Government of the Netherlands and State Secretariat for Economic Affairs (SECO) of the Government of Switzerland. Agricultural Sector Risk Assessment vii ABBREVIATIONS AND ACRONYMS ASDP Agricultural Sector Development Program FEWS NET Famine Early Warning System Network ASDS Agricultural Sector Development Strategy GCM Global Climate Model CEEST Centre for Energy, Environment, Science and GDP Gross Domestic Product Technology GIEWS Global Information and Early Warning System CFVSA Comprehensive Food Security and GPCP Global Precipitation Climate Project Vulnerability Analysis IFPRI International Food Policy Research Institute CGIAR Consultative Group on International INC Initial National Communication Agricultural Research IPCC International Panel for Climate Change CSIRO Commonwealth Scientific and Industrial MAFC Ministry of Agriculture, Food Security and Research Organisation Cooperatives DCGE Dynamic Computable General Equilibrium MIROC Model for Interdisciplinary Research on Model Climate FAO Food and Agricultural Organization of the NAPA National Adaptation Program of Action United Nations TACRI Tanzania Coffee Research Institute FAOSTAT Food and Agriculture Organization Statistics TTC Tanzania Tobacco Council Division Agricultural Sector Risk Assessment ix EXECUTIVE SUMMARY In spite of Tanzania’s comparative advantage in the production of many crops (cashew nuts, coffee, cotton, tea, tobacco, maize, and rice, for example) and the rela- tive abundance of natural resources, 38 percent of the rural population, or 13 million rural inhabitants, live below the poverty line. The agricultural gross domestic product (GDP) grew at an annual average rate of 4 percent between 2005 and 2012, which is significant but below the 6 percent considered necessary for reducing poverty. Most small-scale farmers in Tanzania still continue to use low, purchased-input technologies that result in poor yields and scanty economic returns while facing high production price volatility and limited incentives to invest. In 2001, the government established the Agricultural Sector Development Strategy (ASDS). The ASDS highlights the key constraints to achieving agricultural growth targets, among them “un-managed risks with significant exposure to variability in weather patterns with periodic droughts.” The Agricultural Sector Development Program (ASDP) Framework and Process Document (2005) provides the overall framework and processes for implementing the ASDS. Development activities at the national level are to be based on the strategic plans of the line ministries while activi- ties at the district level are to be implemented by local government authorities. The ASDP components are to be financed through a basket fund. Currently, there is an attempt to link risk management interventions to the new ASDP. This study was undertaken by the Agricultural Risk Management Team (ARMT) of the Agriculture and Environment Services Department of the World Bank under the leadership and coordination of the Directorate of Policy and Planning from the Min- istry of Agriculture, Food Security and Cooperatives (MAFC). This volume comprises the first phase of the Agricultural Risk Assessment for Tanzania related to identi- fication and prioritization of agricultural risks. The Second Phase will address risk management solutions and will be developed as a separate volume. The findings of this assessment aim at informing the Tanzania’s Agricultural Sector Development Program, currently in preparation. ­ Tanzanian agriculture has not suffered natural or artificial events at a catastrophic level during the past 20 years, and that is reflected in the agricultural GDP growth Agricultural Sector Risk Assessment xi rate, which has never been negative during that period. In calculation involves the following crops: tobacco, coffee, effect, it is fortunate that its abundance of natural endow- cotton, cashew nuts, sesame, maize, rice, beans, and cas- ments to date have not been impacted by catastrophic sava, which in aggregate make up more than 80 percent shocks. However, aggregated figures at the sector level of agricultural GDP. Drought was the main cause of these tend to mask volatility at crop and regional levels, which in shocks, sometimes in combination with other events. With turn hide fundamental vulnerabilities. As was highlighted regard to maize, more than 40 percent of losses over a by the ASDS, such volatility represents an important con- 30-year period are concentrated in Mbeya, Manyara, straint to growth and poverty reduction. Shinyanga, and Iringa. Kilimanjaro and Arusha have also been adversely affected by production volatility. Alto- Unreliable rainfall in terms of intensity and distribution gether, the six regions account for 61 percent of all losses. has been identified as one of the most likely and dam- aging production risks by most stakeholders. Drought is How the losses are distributed among stakeholders within also recognized as a severe risk that occurs with lower the supply chain is to a great extent a function of value frequency while retaining the potential to severely affect chain governance and the actors’ capabilities and oppor- agriculture. Pests and diseases are also important produc- tunities for risk management. Some exporters, millers, tion risks that cause yield volatility and, occasionally, when and large trading companies are able to hedge price risks outbreaks occur, can result in severe and extensive dam- globally through the practice of standard futures risk age to agriculture. However, their damage potential varies management strategies. The great majority of farmers, greatly among crops and is highly correlated to any risk traders, and cooperatives are highly exposed to price risk, management actions in place. largely through a lack of risk management practices and knowledge. Primary cooperative societies, ginners, and Price volatility is a key market risk in Tanzania and is par- other procurement agents involved in export crops take ticularly present in coffee and export crops, where inter- significant risks when they make advance payments to annual domestic price changes are very much in line with farmers or keep the products in storage until delivery in the high international price volatility of these commodi- the auction (coffee) or to the exporting companies (cot- ties. Sudden fluctuations in prices are negatively affecting ton). Small-scale farmers’ capacity to protect themselves the segments of the supply chain with little capacity to against price risk is extremely limited. Primary cooper- manage volatility, being for the most part farmers. The ative societies are also the weakest segment in the sup- enabling environment is another source of risk. For the ply chain. Product price variations within the marketing purpose of this report, enabling environment risk refers year can expose primary cooperative societies to financial to the set of conditions that facilitate the efficient perfor- losses when practice multipayment systems. mance of business along the supply chain, among which public policy and regulation are the most prominent. The All actors along the supply chains are exposed to the vari- most prevalent enabling environment risks identified are ability in primary farming production. However, small- changes in regulation regarding the marketing system and holder farmers are particularly vulnerable to production the role of stakeholders in the supply chains; decision- and yield variability. Their family food security and mon- making processes of primary societies in their intermedi- etary income are extensively dependent on the crop har- ary roles; and logistic disruptions in the supply, access, and vest. Thus, to mitigate weather and pest and diseases risks availability of inputs to agriculture. The relative impor- at the farm level, many producers adopt low-risk and low- tance with respect to each supply chain is discussed in the yield crop and production patterns to ensure minimum body of this document. volumes for food security purposes. The value of the average annual production losses in The identified risks were prioritized according to the the agricultural sector as a result of unmanaged pro- frequency of realized risk events, their capacity to cause duction risks has been estimated at approximately losses, and the ability shown by the different stakeholders US$203 million, or 3.5 percent of agricultural GDP. The to manage the risks. The prioritization exercise indicated xii Tanzania that the following were the major risks causing losses to »» Strengthening the agricultural technology inno- the agricultural sector: drought events mainly for maize, vation system to mitigate agricultural production rice, and cotton; widespread outbreaks of pest and dis- losses. Agricultural risk mitigation practices can eases especially for cotton, maize, and coffee; price vola- have very significant impacts on reducing risks tility for cotton and coffee; and regulatory risks, mostly derived from irregular or insufficient rainfall as within the trade policy framework, for various cash crops well as from diseases and pests. and for maize. Although these risks do not necessarily »» Current maize trade policy adds market volatility manifest themselves in the form of catastrophic shocks to the normal production (climatic and sanitary) to agriculture as mentioned above, they are identified as risks because of the variability and unpredictabil- the main drivers of agricultural GDP volatility that cause ity of the norms restricting trade and the way they stakeholders income instability and recurrent food secu- are enforced. rity problems. »» Risk management strategies for high-priced, vola- tile export crops (principally coffee and cotton) Field interviews identified a number of potential solutions are needed to reduce exposure to risk of the most related to a combination of risk mitigation, risk transfer, vulnerable stakeholders in the respective supply and risk coping instruments. chains. The areas of focus for risk management solutions in the These solutions will be addressed in a Risk Management second phase have been identified as the following: Solutions mission (the second phase) and will be put within »» Strengthening seed supply chains for producing the framework of an action plan that addresses the most and delivering drought tolerant seeds, disease resis- relevant risks with appropriate investments, programs, tant seeds, and planting material, as well as inef- and policy measures. ficient seed markets that need to be addressed to reduce risks in agriculture. Agricultural Sector Risk Assessment xiii CHAPTER ONE INTRODUCTION AND CONTEXT The Scope of the Study This study aims to achieve a better understanding of the agricultural risk and risk management situation in Tanzania with a view to identifying key solutions to reduce current gross domestic product (GDP) growth volatility. For the purpose of this assessment, risk is defined as the probability that an uncertain event will occur that could potentially produce losses to participants along the supply chain. Persistence of unmanaged risks in agriculture is a cause of great economic losses for farmers and other actors along the supply chains (for example, traders, pro- cessors, exporters), affecting export earnings and food security. The Risk Assessment Process The Agricultural Risk Management Team of the Agriculture and Environment Services Department of the World Bank is conducting this Agricultural Sector Risk Assessment under the leadership and coordination of the Directorate of Policy and Planning from the Ministry of Agriculture, Food Security and Cooperatives (MAFC). The Agricultural Sector Risk Assessment is a straightforward methodology based on a three-phase sequential process. Phase one begins by analyzing the chronological occurrence of inter-seasonal agricultural risks with a view to identify and prioritize the risks that are the drivers of agricultural GDP volatility. A short list of potential risk management solutions is also identified during this process. Those solutions are then assessed in a second phase that details the gaps that need attention to reduce risks. Finally, in the third phase, the solutions are placed within the framework of an action plan that addresses the most relevant risks with appropriate investments, programs, and policies that involve the participation of the concerned government and private sector stakeholders. Figure 1.1 shows the process. This report contains the findings and recommendations of the first phase and includes the identification, analysis, and prioritization of major risks facing the agricultural sec- tor in Tanzania, as well as recommendations regarding key solutions (see figure 1.2). Agricultural Sector Risk Assessment 1 A combination of quantitative and qualitative techniques was used to generate the findings of this assessment, Contents of the Report include an existing secondary analysis as well as com- Chapter 2 of this report contains an overview of the agri- ments and analyses from experts from MAFC at the cen- cultural sector and its performance, as well as a discus- tral and regional levels. sion of key agro-climatic, weather, and policy restrictions and opportunities. Chapter 3 includes an assessment of Figure 1.1. The Risk Assessment major risks (that is, production, market, and enabling environment risks) facing key export and food crops. Process Chapter 4 presents an estimate of historical losses due to First phase: Second Third phase: realized production risks and a correlation of such losses risk phase: design of identification risk action plan with production volatility. Chapter 5 provides insights & management (projects, into the exposure to risks by different stakeholders and prioritization solutions programs, policies) their actual capacities, vulnerabilities, and potential to manage agricultural risks. Finally, chapter 6 presents a Implementation risk prioritization by different supply chains and discusses the possible solutions, as well as specific recommenda- Monitoring and tions for the Agricultural Sector Development Program evaluation (ASDP). Figure 1.2.  Risk Identification and Prioritization Risks Risk Stocktaking Focus on (production, Risk analysis prioritization Long list of of current solutions that market, and and (based on possible interventions address main enabling underlying probability solutions and analysis underlying environment) causes and impact) of gaps causes of risk identification 2 Tanzania CHAPTER TWO AGRICULTURAL SYSTEM Agricultural Sector Overview and Performance Tanzania is endowed with 44 million hectares suitable for agriculture, representing 46 percent of its territory. However, part of this arable land is currently only margin- ally suitable for agricultural production owing to, for example, soil leaching, drought proneness, and tsetse fly infestation. According to the Agricultural Sector Development Strategy (2001), only 10.1 million hectares (23 percent of arable land) are cultivated. This includes around 2.2 million to 3.0 million hectares of annual crops, fallow for up to five years, permanent crops, and pasture. It is also estimated that out of 50 million hectares suitable for livestock production, only half are currently being used, mainly owing to tsetse fly infestation. Agriculture (which also encompasses livestock, forestry, hunting, and fishing) is an important pillar of the Tanzanian economy, accounting for 28 percent of GDP (2010) and 25 percent of export earnings (2011) and it provides a livelihood to more than 75 percent of the population (National Bureau of Statistics 2011). While the agricultural sector’s contribution to national GDP declined significantly from 1990 (46 percent) as other sectors like industry and manufacturing began to grow across the country (see table 2.2), it is still an important sector that serves as one of the main activities and income sources for rural households (National Sample Census of Agriculture 2012).1 According to the Household Budget Survey (2007), 38 percent of the rural population lived below the poverty line (basic needs). Poverty is highest among those living in arid and semi-arid regions that depend entirely on livestock and food crop production for their livelihood. Although there is not one significantly worse- or better-off region in 1 Sale of food crops was the main cash income earning activity (61.6 percent in 2008 compared with 37.4 percent in 2003 of all the rural agricultural households), followed by sale of cash crops (9.9 percent compared with 17 percent in 2003) and other casual cash earnings (7.8 percent in 2008 compared with 15.1 percent in 2003.). Agricultural Sector Risk Assessment 3 Table 2.1. Tanzania at a Glance Figure 2.1.  Tanzania Real GDP Growth Area (km2) 948,087 by Sector Total population (millions) (2010) 44.8 Agriculture and fishing Services 11 Industry and construction Overall GDP Rural population (% of total population) (2010) 73.7 Rural population growth (annual %) (2010) 2.4 10 Economy and poverty 9 GDP (US$ billions) (2010) 22.9 8 Annual % Agriculture GDP (% of total GDP) (2010) 28.1 7 Agriculture growth rate (2010) 4.1 6 GNI per capita (2010) 520 5 4 Poverty gap at national poverty line (%) (2007) 9.9 3 Poverty gap at rural poverty line (%) (2007) 11 2 Poverty head count ratio at rural poverty line 37.4 2005 2006 2007 2008 2009 2010 2011 2012 (% of rural population) (2007) Prel Proj Source: NBS, IMF, WB. Source: World Development Indicators 2013. Table 2.2. GDP Composition by Sectors GDP Composition 1990 2000 2009 2010 Agro-Climatic Conditions Agriculture (% of GDP) 46 33.5 28.8 28.1 Located in East Africa, Tanzania is composed of seven Industry (% of GDP) 17.7 19.2 24.3 24.5 agro-ecological (land resource) zones (see figure 2.2) Manufacturing (% of GDP) 9.3 9.4 9.5 9.6 having different soils and topography, altitude, rain- Services (% of GDP) 36.4 47.3 46.9 47.3 fall regimes, and growing seasons, with dry periods and Source: World Bank 2010. extreme rainfall during the two rainy seasons that are prevalent in some zones (National Sample Census of Agriculture 2012, xv). Tanzania, the most severe poverty can be found near the coast and in the southern highlands. The combination of dry periods and heavy rainfalls, along with an inadequate land maintenance system, aggravates Agriculture performance. The agricultural GDP continu- the land degradation process and makes the country’s ously increased during the past eight years at an annual agricultural production highly vulnerable to weather- average rate of 4 percent, which is acceptable but well related shocks (Enfors and Gordon 2007). below the almost 7 percent rate of the overall economy (see figure 2.1). Rainfall Trends The agricultural GDP growth rate improved to 4.6 per- Rainfall follows two different patterns in Tanzania. In the cent in 2008 from 4.0 percent in 2007 and 3.8 percent in northeast and coastal areas there is a bimodal rainfall pat- 2006, largely reflecting favorable weather experienced in tern with short rains (Vuli, in local parlance) from Octo- the 2007–08 agricultural season, improved irrigation and ber to December and a long period of rains (Masika) from rural road infrastructure, and increased use of fertilizers. March to May. A different rainfall pattern is observed In 2009, the agricultural GDP growth rate registered the across the south and west. A unimodal pattern (Musumi) lowest value (3.3 percent) in the period as a consequence occurs with rainfall from December to April. Appendix A of a drought in 2008–09, especially in the northern part provides detailed information on the cumulative monthly of the country. Growth continued at about the average rainfall in both zones. 4 percent rate in the following years with the exception of 2011, when preliminary data show a slight dip to Annual crop production takes place during the two 3.6 percent. rainfall patterns. The Vuli planting season is around­ 4 Tanzania Figure 2.2.  Land Resource Zones Source: World Bank 1994. October–November, with harvesting in late January–­ have very little access to modern farm technologies and February; the Masika, or main planting seasons, starts inputs. Productivity and profits are low. in late February–March, with harvesting in July–August. Most of the country’s crop production occcurs in the Over the decades, a Tanzanian farmer’s choice of crop Masika season with around 80 percent of total planted production has been influenced by environmental (soil area, compared with 20 percent of the total planted area quality, water accessibility, pest resistance), resource during the Vuli period (National Sample Census of Agri- (fertilizers, quality seeds, machinery) and economic culture 2012, 30). factors (such as marketability and seed prices) (Greig 2009). Farming preferences in the past have been largely Crop Production System focused on millet, cotton, sugarcane, and banana-based systems that have been shifted around greater maize, and Exports cassava, and rice production for the past few decades The crop subsectors account for the highest contribution (Fermont 2009). Generally, mixed maize production is to Tanzania’s agricultural GDP, followed by livestock, for- common in central semi-arid regions whereas the north- estry, and fishing (Ministry of Agriculture 2011). As such, ern zones provide better conditions for coffee, maize, crop production is the main rural smallholder household and tea. Coffee and tobacco production is predominant activity (National Sample Census of Agriculture 2012).2 in the southern and western zones, and the Lake Victoria Food crops are grown for both family consumption and area is suitable for cotton (Ponte 2002, 38–39; Kimaro for sale. Pure cash crops (coffee, tobacco, cotton, rice, and others 2009, 115). peas, and so on) are key commodities in the country. Crop smallholder farming is labor intensive and those farmers According to the latest crop census data, maize is produced across the country, with a relative concentration in some regions, and is the main crop for the majority of house- 2 At the national level, crop production was the dominant agricultural activ- holds (more than 5.1 million) (National Sample Census of ity, which engaged 3,508,581 households (60.1 percent), followed by 2,268,255 Agriculture 2012, 30). Figure 2.3 shows the geographical (38.8 percent) households engaged in mixed crop and livestock, 57,770 distribution of producing regions. The largest producing (1 percent) households engaged in livestock only, and only 3,917 (0.1 percent) regions (Shinyanga, Mbeya, Iringa, Rukwa, Tanga, Man- households engaged in pastoralism (chart 2.12). Of the total crop-growing households, 3,422,072 (98 percent) were on the mainland and 86,509 (2 per- yara) and Ruvuma are also surplus areas, with per capita cent) were in Zanzibar. production 20 percent above the national average. Agricultural Sector Risk Assessment 5 Figure 2.3.  Maize Production by Figure 2.5. Planted Area Planted area (000 ha) by main crops for 2010 Region Maize production by region: Maize Percentage of total Beans national production - Avg. 2007/08–2009/10 Rice Cassava Cotton 2% 1% 1% 1% Cashew nuts 3% Sesame 3% Shinyanga Tobacco 3% 11% Coffee 3% Mbeya 0 500 1000 1500 2000 2500 3000 3500 11% 4% Source: MAFC. Iringa 4% 9% Kilimanjaro 4% Rukwa that period. The export volume of coffee remains low, but 8% Ruvuma given the recovery of the international prices during the 5% Morogoro 2000s, export earnings are increasing. 5% Tabora Manyara Tanga 6% 6% Yield, Production, 6% Source: MAFC. Acreage, and Market Figure 2.4.  Traditional Export Crops Trends Traditional export crops (million US$) 800 Maize dominates according to the planted area, followed 700 by beans, rice, cassava, and cotton (see figure 2.5). Sisal 600 Cloves 500 Tea Yields for maize are low and decreasing, averaging about Cotton 1.3 tons per hectare. This is a very low level compared 400 Cashew nuts 300 Coffee with what has been achieved elsewhere. Rice productivity 200 Tobacco in irrigated areas is variable depending on location but, 100 on average, is higher in irrigated than nonirrigated areas. 0 The average yield in nonirrigated areas is less than 2 tons 2007/08 2008/09 2009/10 2010/11 2011/12 Source: Bank of Tanzania. per hectares for paddy. The ASDP—Joint Implementa- tion Review (2012) revealed that the yield per hectare for irrigated paddy can be as high as 8 tons per hectare in Agricultural exports. Tanzania’s major exports include gold, some places, such as KPL-Mngeta, where farmers receive tobacco, coffee, tea, cashews, cotton, gemstones, and some some extension services as out-growers.3 Table 2.3 shows manufactured goods. Based on Bank of Tanzania data, the average area, yields, and production for the main food the value of traditional commodity exports has increased security and export crops. during the past few years, and tobacco, coffee, and cashew nuts have been the leading commodity exports (see figure 2.4). Principal Constraints in the Agricultural System International Trade Centre (INTRACEN) data show that and Policy Reform the agricultural sector (including livestock, forestry, hunt- Tanzania’s agricultural sector has a comparative advan- ing, and fishing) accounted for 25 percent of export earn- agricultural tage in the production of a diversified set of ­ ings in 2011, down from 42 percent in 2007. The total export crops (cashews, coffee, cotton, tea, and tobacco) as value of agricultural exports increased between 2007 and well as those for food consumption (rice and maize), and 2011 from US$908 million to US$1.18 billion, but over- all exports increased much more, from US$2.14 billion to US$4.73 billion. Tobacco export earnings grew steadily in 3 Information provided by Professor Gungu M. Mibavu. 6 Tanzania Table 2.3. Average 2007/08–2009/10, Box 2.1. C  ommodity Boards Reform Except for Coffee and during the 1990s and Early Cashew Nut, Which Is 2000s 2008/09–2009/10 The major cotton reform began with the Cotton Act of Area Yields Production 1994, which allowed competition in both marketing and (000 ha) (tons/ha) (000 tons) ginning of cotton. The Cotton Industry Act of 2001 specified that the Tanzania Cotton Board would officially Maize 3,328 1.30 4,313 regulate the cotton sector. Reform opponents argue that Rice 943 1.90 1,795 taxation remains excessive and the sector is still overly regu- Tobacco 65 0.96 63 lated. The quality of cotton also has been declining (Baffes Coffee 65 0.79 51 2004). Cotton 505 0.62 313 In the early 1990s, the Tanzanian government privatized Cashew nut 320 0.28 90 previously nationalized coffee estates and abandoned Source: Based on data from MAFC. controls on coffee prices. Today, the private sector has emerged in the coffee business and the overall processing capacity has increased. However, the export volume of there is a relative abundance of natural resources (includ- coffee remains low and arguably the quality of coffee has ing arable land and range land) that can be used for pro- decreased. Additionally, declining world coffee prices have ductive purposes. Market opportunities are expanding in cut real producer prices. During the 2000s, international prices recovered, giving producers an incentive to expand domestic markets for food, especially for livestock prod- production. However, there are still high marketing costs ucts and crops with high-income elasticity of demand. that depress the ratio of farm gate price to auction price. Moreover, Tanzania’s membership in regional trade groups (East African Community and the Southern Afri- The marketing and pricing strategy for tobacco was over- hauled with the Tobacco Industry Act of 2001. Reforms can Development Community) and signatory status to included assigning price according to tobacco demand international trade protocols improves its market oppor- worldwide and negotiations between tobacco farmers and tunities both within the region and globally. buyers in U.S. dollars. Additionally, tobacco producers became free to take agricultural inputs on loans that helped Starting in the 1980s, the Tanzanian government imple- productivity growth. mented a series of agricultural reforms to support market liberalization, remove state monopolies, and encourage private sector development.4 This policy process envi- sioned the transition of the agricultural sector from sub- included competitive prices and free entry of marketing sistence to export agriculture. The policy instruments actors (producers, traders, processors, and exporters) at included demand-driven and market-led technology all levels of the marketing channel. See box 2.1 for brief development and encouraged private sector involvement. details on commodity boards reform. These reforms included the liberalization of marketing of nontraditional export crops in 1986, which was followed It should be noted, however, that these policy reforms have by liberalization of marketing of food crops in 1989 and had implementation weaknesses that prevented maxi- finally liberalization of marketing of traditional export mization of available market opportunities domestically, crops in the 1993/94 marketing season. The ­ liberalization regionally, and internationally. Most rural small-scale of agricultural marketing was expected to pave the way farmers in Tanzania continue to use low-purchased-input for the participation of cooperatives and private trad- technologies that result in poor yields, weak economic ers in crop marketing in a competitive environment that returns, and high production volatility. The Agricultural Sector Development Strategy and Program. In 4 This section was extracted from Ministry of Industry, Trade and Marketing, 2001, the government established the Agricultural Sector Agricultural Marketing Policy, December 2008. Development Strategy (ASDS). The ASDS was created Agricultural Sector Risk Assessment 7 as an integral component of macroeconomic adjustment ii.  Underinvestment in productivity-enhancing tech- and structural reforms that were supported by Tanzania’s nologies. Although recent progress has been made development partners. Its primary objective was to create in increasing land productivity, progress has been an enabling and conducive environment for improving hampered by the relative under-investment in profitability in the sector as the basis for increasing farm research. incomes and reducing rural poverty in the medium and iii.  Limited access to technology demand and deliv- long term. The Agricultural Sector Development Pro- ery channels—with 60 to 75 percent of house- gram (ASDP) Framework and Process Document (2005) holds estimated to have no contact with research (United Republic of Tanzania 2006), developed jointly and extension services. by the lead ministries for the agricultural sector, provides iv.  Limited access to financing for the uptake of tech- the overall framework and processes for implementing nologies. the ASDS. Development activities at the national level v.  Unmanaged risks with significant exposure to vari- are to be based on the strategic plans of the line minis- ability in weather patterns with periodic droughts. tries, whereas activities at the district level are to be imple- The impact of these events is amplified by the mented by local government authorities, based on District dependency on rain-fed agriculture and the lim- Agricultural Development Plans (DADPs).5 The ASDP ited capacity to manage land and water resources. national and local components are to be financed through vi.  Weak coordination and capacity in policy, and an ASDP Basket Fund. the formulation and implementation of public intervention among the various actors in the sec- The ASDP highlights the key constraints to achieving tor (including the multiplicity of ministries dealing agricultural growth targets: with agriculture). (United Republic of Tanzania i.  High transaction costs due to the poor state or lack 2006, 7) of infrastructure and the overall policy and regu- latory environment governing market transactions This risk assessment aims at complementing current agri- (including tax regimes and licensing requirements cultural risk management analysis and practices within and costs). ASDP. 5 Tanzania mainland is divided into 25 administrative regions: Dodoma, Aru- sha, Kilimanjaro, Tanga, Morogoro, Pwani, Dar es Salaam, Lindi, Mtwara, Ruvuma, Iringa, Mbeya, Singida, Tabora, Rukwa, Kigoma, Shinyanga, Kag- era, Mwanza, Mara, and Manyara. Furthermore, the regions are divided into urban and rural districts, totaling up to 119 administrative districts (National Sample Census of Agriculture 2012). 8 Tanzania CHAPTER THREE AGRICULTURAL SECTOR RISKS Tanzania has not suffered catastrophic natural or artificial events during the past 20 years, good fortune that is reflected in a positive agricultural gross domestic product (GDP) growth rate throughout those two decades. In effect, Tanzania’s natural endow- ments reduce its exposure to systemic agricultural risks. However, aggregated figures at the sector level tend to mask volatility at crop and regional levels, which in turn hide fundamental vulnerabilities (see box 3.1 on regional vulnerabilities). The analysis focused on critical food security crops: tobacco, coffee, cotton, and cashew nuts (representing 88 percent of the total value of the agricultural exports in 2011/12) as well as rice and maize, the main food staple (Bank of Tanzania). Risks are highly concentrated in these supply chains, which also show great, largely unmanaged volatility. It is recognized, however, that there are other crops, such as roots and tubers (for example, cassava) and legumes (for example, beans), that are food security coping crops in specific regions or grown under intercropping systems, with risk profiles not radically different from the crops that are actually being studied. Tanzania’s dependency on rain-fed agriculture makes it acutely vulnerable to weather changes. Most stakeholders have cited unreliable rainfall—in terms of intensity and distribution—as one of the most likely and damaging production risks. Drought is also recognized as a severe risk that occurs with lower frequency but with great potential to severely affect agriculture. Climate change may be exac- erbating the typical inter-seasonal weather risks facing the agricultural sector. A couple of climate change assessments (see appendix B) predict impact variations across geographical areas and among agricultural subsectors. There might even be gains in some regions while losses might occur in others. This study focused spe- cifically on short-term interannual risks rather than long-term changes in climate. However, the recommendations provided in this report will need to be linked to the findings contributed by climate change studies regarding mitigation and adaptation to climate change. Pests and diseases are also important production risks that cause yield volatility and, occasionally when outbreaks occur, can result in severe and extensive damage to Agricultural Sector Risk Assessment 9 Box 3.1. Regional Vulnerability Finally, there are long-term threats that if not properly addressed may become actual risks. These are challenges The World Food Program’s 2010 Comprehensive Food over the medium to longer term because of structural Security and Vulnerability Analysis reports the top three most frequent shocks to household food security as drought tendencies that signal sustainability issues for the supply (58 percent of surveyed households), high food prices chain. Three different types of long-term threats were (53 percent), and plant disease/animal pests (35 percent). identified: environmental, financial, and price related. Drought was most frequently reported in the northern regions (Arusha, Tanga, Manyara, Kilimanjaro, and Mara), This section presents findings regarding the production, central regions (Dodoma and Morogoro), and southeastern market, and enabling environment risks in the most rel- regions (Mtwara and Lindi). The “increasingly bimodal” evant supply chains as well as discussions on the impact of tendencies and rainfall patterns in the north correspond with this finding. High food prices were cited across Tan- the adverse events on the different stakeholders. zania, but particularly so in Kilimanjaro, Mara, Dodoma, Singida, Lindi, and Mtwara; western regions reported this Tobacco shock less frequently. Shocks related to plant disease and Tobacco is Tanzania’s largest agricultural export crop animal pests were more prevalent in regions close to the water, specifically, Lindi, Kigoma, Mara, Mtwara, and (US$272 million in 2011/12) and a major cash crop Mwanza. The Shinyanga, Ruvuma, and Arusha regions for many smallholders. It is an important source of for- were least affected. eign exchange, tax revenue, and income for stakeholders along its supply chain. The National Sample Census of Agriculture (2012) estimated that about 64,572 hectares agriculture. However, their damage potential varies very at the end of the 2007/08 agricultural year were under much among crops and is highly correlated to the actual tobacco cultivation on the mainland.6 The Tabora region risk management actions in place. Other risks identified accounted for the largest tobacco production, equivalent include fake and expired chemicals and strong winds to 51.1 percent of total harvested quantity (36,056 tons; (strong winds were recorded in 2009 and 2010, affecting National Sample Census of Agriculture 2012). More than banana and coffee trees). The greatest impact occurred 85 percent of people in Tabora depend on tobacco for when strong winds coincided with drought and therefore their livelihood.7 exacerbated the damage to already weakened trees. The tobacco supply chain is regulated by the Tanzania Price volatility is a key market risk in Tanzania and is Tobacco Board (TTB). Tobacco growers organize them- particularly present in export crops, for which interan- selves in primary cooperative societies. These primary nual domestic price changes are very much in line with societies form a cooperative union at the regional level the international and regional market variations. Sudden (such as the Western Zone Tobacco Growers Cooperative fluctuations in prices negatively affect the segments of the Union, or WETCU) and these cooperative unions give supply chain with little capacity to manage volatility. rise to a tertiary cooperative organization, the Tanzania Tobacco Co-operative Apex, which represents farmers at Enabling environment, another source of risk, for the the Tanzania Tobacco Council (TTC).8 TTC is a body purpose of this report refers to the set of conditions that that comprises all stakeholders of the tobacco industry facilitate the efficient performance of business along the and is the institutional forum in which important issues supply chain, among which public policy and regulation related to the industry are discussed. Lastly, there is the are the most prominent. The prevalent enabling environ- Association of Tanzanian Tobacco Traders (ATTT), ment risks identified are changing regulation regarding the marketing system and the role of stakeholders in the 6 There was no tobacco production in Zanzibar. supply chains; decision-making processes of primary 7 Personal communication Mr. Yobu Kiungo, Regional Forestry Officer, Tab- ora, January 14, 2013. societies in their intermediary roles; and, logistical dis- 8 The Tanzania Tobacco Co-operative Apex unites approximately 300 tobacco ruptions in the supply, access, and availability of inputs primary cooperative societies, representing more than 100,000 small holder to agriculture. tobacco growers. Source: IPP Media 2012. 10 Tanzania owned by the two largest tobacco processors in Tanzania, Coffee and practically acting as a service provider to farmers on Today, coffee is Tanzania’s second-largest export crop behalf of the two largest exporters. after tobacco. It accounted for 20 percent of agricultural export proceeds10 and 3.6 percent of all export proceeds Production risks. In tobacco production, weather risks tend in 2012. More than 400,000 households with an average to come in the form of droughts or heavy rainfall with area of 0.5 to 1 hectare are responsible for most coffee strong winds and hail storms that can either hurt the production. Participants in the supply chain are farmers, quality of tobacco or completely destroy the crop. Wet cooperatives, farmer groups, traders, exporters, and dry soil after heavy rain causes the tobacco plant to become mills. stunted. Such weather occurrences are common at least once a year, but they are localized. Small-scale farmers trade their produce through traders and primary cooperative societies. Traders and primary Risks posed by pests and diseases are negligible, according cooperatives collect roughly 75 to 80 percent of the mar- to the ATTT, as crops are kept as clean as possible. The ket, and estates account for the rest. Coffee harvested by Tobacco Research Institute of Tanzania also claimed that cooperative member farmers is processed in the coop- diseases do not affect tobacco very much, as nematodes are erative centers using the wet method (arabica coffee) to left in the tobacco field after harvesting. Standard tobacco obtain parchment coffee. Cooperative unions buy from farming practice to avoid contamination prescribes mov- the associated primary cooperative societies, arranging for ing plants after harvesting and planting another crop such hulling and grading in private mills, and then sell beans at as maize rotating with tobacco. auction or export them directly (if authorized). The auc- tion is an efficient pricing mechanism, in the sense that Market risks. Production of tobacco is price driven, and realized prices move in accordance with the New York lags a season behind. Although farmers base their plant- Board of Trade futures prices but, it is argued, the man- ing decisions on the previous year’s price, exposing datory nature of the auction increases marketing costs.11 themselves to the risk of fluctuation of both agricultural The second grade associations (cooperative unions) pro- inputs and tobacco purchase price, the product price is vide bank-financed credit resources to primary coopera- reported to be more or less steady, with the exception of tive societies to enable them to afford the processing costs some years such as 2011, which witnessed the entrance and to prefinance farmers. Some exporters also have pro- of a new tobacco purchaser—a development that drove duction promotion support programs to assist farmers to prices higher.9 expand and improve production. Long-term threat (environmental risk). Tobacco farmers tend to Production risks. Coffee is exposed to erratic rains in all engage in a farming practice whereby after harvesting one agro-ecological zones where it is produced, although rain plot, they leave it fallow, move to the next plot, and thus irregularities are more pronounced in the north. It has advance into forestland. In addition, flue-cured tobacco been reported by the local stakeholders in Arusha that farming requires the use of woodland. It is estimated that short rains (November–December) seem to have been approximately one hectare of woodland is required to diminishing during the past few years. For instance, the flue-cure one hectare of planted tobacco. In the absence Meru Rural Cooperative Society Ltd. (Singesi-Arusha) of reforestation programs, this represents a great environ- identified 4 years of drought during the past 10 years, mental constraint and may exacerbate production risks in and in 2010/11 coffee processed by cooperatives dropped the future. to 12,000 kg against an expected 24,000 kg. The Tanza- nia Coffee Research Institute (TACRI) is working on the 9 Agricultural input decisions are taken at the primary society level in January, a few months before the purchase price for tobacco is negotiated at the Tanzania Tobacco Council in May–June. In June–July, seeds are issued to farmers who 10 Source: Bank of Tanzania. prepare the seedlings to be planted between October and December. 11 See John Baffes, “Tanzania’s Coffee Sector: Constraints and Challenges.” Agricultural Sector Risk Assessment 11 Figure 3.1. I nternational Coffee Price Figure 3.2. C  offee International- Change Domestic Price Comparison Coffee price change, U.S. cents per pound Coffee price index 50% Rural Meru Cooperative Society, Singisi, Tanzania 40% 250 International price 1/ 30% 200 20% 10% 150 0% –0% 100 –20% 50 –30% Jan-88 Nov-88 Sep-89 Jul-90 May-91 Mar-92 Jan-93 Nov-93 Sep-94 Jul-95 May-96 Mar-97 Jan-98 Nov-98 Sep-99 Jul-00 May-01 Mar-02 Jan-03 Nov-03 Sep-04 Jul-05 May-06 Mar-07 Jan-08 Nov-08 Sep-09 Jul-10 May-11 Mar-12 0 92 /92 95 /95 97/97 98/98 99/99 00/00 01/01 02/02 03/03 04/04 05/05 06 06 07/07 08/08 09/09 10/10 11/11 12/12 3 93/93 94/94 96 96 /1 / / 91 19 Source: International Coffee Organization: Coffee, Other Mild Arabicas, New York cash price, ex-dock New York. Source: International Coffee Organization—Coffee, Other Mild Arabicas, New York cash price, ex-dock New York, U.S. cents per pound. development of drought-resistant varieties. The program actors in the value chain vulnerable to the volatility in is now at the initial stage. the international markets. Figure 3.2 illustrates the quasi-­ perfect price transmission effect on the prices received by Coffee growers interviewed in focus groups also identified the Rural Meru Cooperative Society from Singisi, Aru- diseases as a considerable risk for coffee plants, whereas sha, as compared with the New York cash price over the pests are considered less important (thrips in 2011 and past 20 years. 2012, with 20 percent local losses in 2012). The Tanzania Coffee Research Institute (TACRI) undertakes extensive Value chain actors relying on the multipayment system for research in coffee plant diseases and pests, including cof- settling payments to farmers experience different degrees fee bean diseases, coffee wilt disease, and coffee leaf rust. of impact resulting from price volatility. Farmers receive Disease research has been successful, mainly in leaf rust, an initial payment before the auction from the primary and the results are being transferred to farmers. TACRI is cooperative societies and a second payment afterward. promoting the replacement of old varieties with the new The multipayment system allows farmers to benefit from disease resistant varieties, the introduction of improved any price increase between the two payment moments but management practices, and the use of pesticides suitable it introduces considerable price risk for the cooperatives for each agro-ecological zone. Regarding pest research, given the long period between the delivery to the primary TACRI is still at the experimental stage to develop resis- society and the auction. If the first payment made to the tant varieties and is for the moment recommending coffee growers was higher than the auction realization appropriate practices. The adoption rate is low, however, plus other costs, the cooperatives would operate at a loss. and much effort is needed at different levels to ensure the Because CRDB Bank Plc. has been involved in financing acceptance of the new practices and varieties. the coffee subsector through the cooperatives, it has been Market risks (price volatility). Coffee prices in the interna- sharing the risk with farmers and farmers’ organizations. tional markets are subject to great variability. Monthly Meanwhile, exporters operate in the local market pro- New York price changes between January 1988 and tected by futures operations in the international markets December 2012 are shown in figure 3.1. The series stan- and therefore are less exposed to price volatility. dard deviation is more than 8 percent. Long-term threats (price related). Coffee farmers perceive the Transmission of the interannual international price price drop risk very strongly after the international price changes to domestic producer prices is high, making all crisis strongly hit their economies. Coffee area expanded 12 Tanzania significantly during the 1970s and 1980s when prices were Figure 3.3. I nternational Cotton favourable and declined thereafter during the world cof- Price Change fee price crisis—from 1980/81 to 1998/99 coffee sales 30% declined from 61,000 tons to 41,500 tons. Production in 20% 2010/11 increased again to 60,500 tons but small farmers 10% are extremely cautious about investing in improving and expanding coffee cultivation. 0% –10% Cotton –30% Cotton is a key cash crop in Tanzania not only in terms –30% of foreign exchange earnings (US$88 million in 2011/12) Mar-88 May-89 Jul-90 Sep-91 Nov-92 Jan-94 Mar-95 May-96 Jul-97 Sep-98 Nov-99 Jan-01 Mar-02 May-03 Jul-04 Sep-05 Nov-06 Jan-08 Mar-09 May-10 Jul-11 Sep-12 but also in terms of provision of direct employment in primary production and in marketing (transport) and Source: Cotlook via IMF, Cotton, Cotlook ‘A Index,’ Middling 1–3/32 inch processing. It provides livelihoods to more than 1 mil- staple, CFR Far Eastern ports, U.S. cents per Pound. lion people. Farm production of cotton is predominantly undertaken by smallholder farmers within an average area ranging from 0.5 to 1 hectare per household. Farmers reported that insect pest and diseases together represent another production risk, which is captured in The Cotton Development Trust Fund procures inputs the annual reports by the Cotton Board. Farmers are and distributes through private ginners on a credit basis, fairly vulnerable to production shocks partly because of with payments deducted from the cotton seed price paid the low margins of profitability that constrain them from by the ginners to the farmers. Farmers are organized in adopting more effective agricultural practices. They suf- local primary cooperative societies and they have formed fer most from low volumes of production, followed by a regional body known as the Nyanza Cooperative Union. other actors along the value chain such the ginners; tex- The Cotton Board plays a key role regulating the subsec- tile industries may suffer, but only marginally. Promotion tor and advising the MAFC on cotton-related policies. of newly released agricultural technologies is not effec- This organ is vested with the responsibility of engaging tively undertaken owing to institutional problems related key stakeholders in establishing indicative prices through to lack of functioning linkages between extension services a farm gate price-setting forum. The price determined in within the local government authorities and the agricul- this forum becomes a floor price that will be applicable in tural research centers. In addition, the existence of fake a specific cotton-buying season. seeds or agrochemicals in the market is an additional risk for farmers. Production risks. Cotton farm production relies on rainfall and, as a result, it succumbs to sporadic adverse weather Market risks. The price announced by the cotton forum is conditions. The Tanzania Cotton Board (2011) annual supposed to be indicative but, in reality, it tends to become report of 2010/11 reports a 39 percent decline in produc- the actual buying price for all practical purposes. Given tion with respect to the previous year, which is attributed that world cotton prices play a significant role in setting to localized drought, more than average rainfall in some the indicative price, significant volatility is transmitted to areas, and failure of the voucher inputs system.The lower domestic prices. In effect, cotton international prices are cotton production in the 2010/2011 marketing season very volatile, as can be observed in figure 3.3 (standard resulted in a failure for most local ginners to fulfill their deviation is 6 percent). contractual obligation of supplying bales to external buy- ers and a poor supply of lint to the local textile indus- Unexpected losses can occur when the world price falls try. Nonfulfillment of the contracts eventually led to the below the corresponding indicative seed cotton price. The blacklisting of some local cotton companies by the Inter- capacity of ginners and traders to manage such price risks national Cotton Association in Liverpool. varies markedly depending on their expertise, size, and Agricultural Sector Risk Assessment 13 scale. Price risk is currently borne by ginners, particularly tenders for these lots, which are opened at an auction small operations (totaling around 34) that basically sell to attended by union staff and staff of the societies whose textile industries. Considerable side-selling is present given nuts are being sold. that farmers sell to different primary societies when there is an opportunity for better margins or less uncertainty about Production risks. The main risk that cashew farmers face payment for the seed cotton they deliver (counterparty risk). is too much rain at the wrong time, leading to outbreaks Some primary societies may also be at risk if they are not of fungal diseases on the leaves, flowers, premature nuts, able to recover enough seed cotton to cover the amounts of and apples. Fungal diseases can be controlled relatively credit they provide to farmers in terms of inputs. easily through the application of sulfur. If these diseases are not treated, virtually the entire crop can be lost. The Enabling environment risks. Major risks in this perspective principal problem stems from the possibility that farmers include port delays due to inefficiencies caused by mul- may be unable to acquire the necessary chemicals, either tiple factors, including poor technology. The most affected because they are physically unavailable at the right time stakeholders are the cotton exporters who end up in dis- or because they cannot afford to purchase them. The putes with their customers aboard. Naliendele Agricultural Research Institute, which has national responsibility for cashew nut research, has devel- oped recommendations as to the appropriate timing and Cashew nuts quantity of sprays that extension staff can pass on to farm- Cashew nuts are grown along the coastal lowlands, with ers. Currently, this advice has yet to be transmitted to all the more productive areas in the south, close to the border farmers, so some individual farmers risk losing a part of with Mozambique. Nationally, a total of some 400,000 their crop through poor spraying practices. households produce cashew nuts. The majority of farms are small but there are also some relatively large farms, Market risks (price volatility). International prices for exceeding 100 hectares. There are a number of small fac- cashew nut kernels change markedly from day to day, tories, as well as one large and one medium-scale facility. month to month, and from year to year. This volatil- About 70 percent of national exports are made in raw ity is reflected in the international prices negotiated for form, with the cashew kernel still within the shell of the raw cashew nuts. There are no long-term time series of nut. Virtually all raw nut exports are destined for India, international cashew prices.12 The only available long- where they are processed and sold in the domestic market term price series for valuing Tanzanian cashew produc- or reexported. tion is the Food and Agriculture Organization Statistics Division (FAO-STAT) data on the national unit value The sector is supervised and regulated by the Cashew of raw nut exports (see figure 3.4). Because Tanzania Board of Tanzania and supported by a Cashew Indus- exports virtually all its cashew output, price volatility in try Development Trust Fund, established in April 2011, the international market is reflected in export prices and which is financed partly through the export levy and in the prices received by growers. partly by government financial contributions. There is currently a single-channel marketing system referred to as Enabling environment risks. The changing regulatory frame- the warehouse receipt system, under which farmers must work has been a major cause of dysfunction in the cashew deliver their entire crop to their local cooperative society supply chain over the past 40 years and the origin of great for acceptance without grading. There is thought to be production volatility (see box 3.2). Similarly, the sesame a significant unrecorded amount of side-selling outside supply chain has been subject to changing regulations that this system. The society transports accumulated deliveries affected the marketing system and the roles of primary to the store of its parent cooperative union, where they are held separately from deliveries from other societies. 12 There are no readily available time series indicative of the international prices At intervals during the buying season, the union prepares of raw cashew nuts given that worldwide there are no formal markets in which a sales catalogue listing as separate lots the stocks that it prices are formed. (A futures market for 320-count cashew nut kernels operated holds from each society. Licensed buyers submit sealed at the Kolam exchange in India from 2005 to 2009.) 14 Tanzania Figure 3.4. C  ashew Nut Export Price trading intermediaries and is less rules based. Wholesale Price of Cashew W-320 count kernals traders and millers in main urban centers similarly tend to delivered Kollam, India* 7500 specialize in trading either maize or rice. Regarding inter- Indian rupees per carton 7000 national trade, protection barriers for rice are high and 6500 constant and for maize are lower but variable. Exports are 6000 regulated with periodic bans depending on the season’s 5500 domestic supply. 5000 4500 4000 These policies are focused on the short term, mostly 0 3 0 6 0 9 06 2 0 3 0 6 0 9 0 2 0 3 0 6 0 9 0 2 0 3 0 6 0 9 0 2 0 3 09 6 9 directed to guarantee national and regional (subnational) 20 5-0 20 5-0 20 5-0 20 5-1 20 -0 20 6-0 20 6-0 20 6-1 20 7-0 20 7-0 20 7-0 20 7-1 20 8-0 20 8-0 20 8-0 20 8-1 20 9-0 20 9-0 -0 0 20 food security, but deny market opportunities that may be *Price for the nearest expiration futures contract at the National Commod- ity and Derivatives Exchange Ltd., India. (There are no data for March to available in neighboring countries. September 2007.) Production risks. Rainfall shortages and drought pose societies. Further examples of disruption in the supply critical risks for maize and upland rice. During the past chain are the risk of the government withholding loan 30 years, yield has oscillated at an average of 1.4 tons per guarantees to tobacco primary societies, and the unreli- hectare, with many years above the average in the 1980s able supply of agricultural inputs by ginneries to farmers and 1990s, but with worsening yield in recent years. Dur- in cotton production. ing 2000–07, maize production increased at a slower rate (2 percent) than the overall population growth rate Long-term threat (financial). A sustainable single-channel (3 percent). Such poor performance has been attributed marketing system, as for most export crops, must either to erratic rainfall and low application of fertilizers and lead to payment of a total price to farmers that reflects improved seeds. Currently, the country has introduced a the net farm-gate value of their output or it must provide farm input subsidy (voucher system) program covering a mechanism for systematically subsidizing farmers when fertilizers and improved seeds to address the low adoption world prices fall. In the absence of this, a single-channel rates (MAFC 2011a). system will simply stagger from crisis to crisis, creating continued uncertainty within the industry. Farmers have already developed practical mitiga- tion strategies. The most common drought mitigation Maize and rice strategy is to mix farming and intercropping in small Maize and rice are the main staple food crops in Tan- plots (for instance, maize-beans, maize-peas, beans-­ zania. Maize is the traditional food in both rural and sunflower, coffee-bananas-beans). In addition, the urban areas and rice is increasingly becoming more government has assisted farmer households by guar- important in towns as family income tends to rise. anteeing food security and rehabilitating agriculture Maize is also the most widespread crop among small- when severe droughts occurred in the past as a way to holders, and production surpluses are traded to the cope with the losses. extent that in good years it may become a relevant cash crop. Agro-ecological conditions for growing maize are Pests and diseases. Armyworms and rodents are relatively good in Tanzania, and normal conditions are better moderate risks provided that they are controlled in a than in neighboring countries. In spite of this, in gen- timely manner with chemicals. Rice yellow mottle virus eral productivity is low. is also a moderate risk (disease) if controlled adequately. Birds feeding on rice can be devastating if not prevented Large traders that have developed broad buying networks from doing so; sometimes entire farmer families have dominate maize trade and their purchases from local been forced to spend all day in the fields chasing birds. farmers, middlemen, and farmer associations are rules Armyworm damage has also been very severe when out- based (quality, and so on). Rice trade involves few large breaks spread. Agricultural Sector Risk Assessment 15 Box 3.2. Examples of Changing MARKET Environment Cashew nuts. The cashew nut sector performed strongly after tion plummeted by 45 percent in 2001/02, representing the independence, with production growing strongly. Market- biggest fall yet recorded. It would appear that, this time, it was ing was through a single-channel system using cooperative the result of a tightening of marketing regulations that com- societies and unions. Production fell abruptly in 1973/74 and plicated trading, as well as the impact of the 1999 Local Gov- again, by an even greater percentage, in 1974/75. After a ernment Act. This law added local authorities as collectors of short recovery, the decline continued with further massive revenues and led to an increase in the tax burden of farmers. falls between 1976/77 and 1979/80. Production continued Production has since grown strongly with the reintroduction to fall by large percentages during the next decade but from a of single-channel marketing using the cooperative societies much smaller base, as illustrated in the accompanying figure. and unions, and the sale of raw nuts to exporters by tender. National cashew nut production (1961/62/2011/12) Sesame. Single-channel marketing systems for sesame under (MT raw nuts) 180,000 which all seed was to be channeled through cooperative Single-year unions were introduced by regional councils in 2008, includ- 160,000 decline: ing in the Lindi and Mtwara regions. Different systems were 140,000 2-year decline 2000/01 established by the regional councils in Lindi and Mtwara. 1973/74–1975/76 2001/02 120,000 =45% These single-channel systems were poorly organized and =42% 100,000 inadequately managed, controlled, and monitored. In the 3-year decline key Lindi producing region, farmers were paid an admin- 80,000 1977/78–1979/80 istratively determined price by their primary cooperative 60,000 =58% society that was well above the export parity producer price, 40,000 meaning that one or more of the institutions involved in the value chain would necessarily end up losing money. In 20,000 Mtwara region, a more rational two-payment system was 0 introduced with a relatively low first payment. However, 1961/1962 1963/1964 1965/1966 1967/1968 1969/1970 1971/1972 1973/1974 1975/1976 1977/1978 1979/1980 1981/1982 1983/1984 1985/1986 1987/1988 1989/1990 1991/1992 1993/1994 1995/1996 1997/1998 1999/2000 2001/2002 2003/2004 2005/2006 2007/2008 2009/2010 2011/2012 a part of the revenue from sales to exporters was diverted by at least one cooperative union to fund investments not specifically devoted to sesame marketing, including the con- Source: MAFC. struction of a new warehouse and the acquisition of trucks. What was the cause of these production falls and why did The large 2011/12 cashew harvest, coupled with problems they extend over such a long period? If we look back, we encountered by cooperatives in selling their stocks of raw see that the successful marketing system was disrupted, first, cashew nuts, resulted in union warehouses being full at the in 1974 with the creation of a crop-specific marketing board time of the local 2012 sesame harvest. Consequently, the that took away much of the power and influence of the coop- government directed that sesame buying should be opened eratives. Second, in 1976, primary societies were abandoned up to private traders resulting in the reestablishment of a and replaced by village agents. Many farmers were also relo- free market. Farmers now sell directly to small-scale private cated under “villagezation” policies away from their trees. traders operating on their own account or to agents of the This institutional disruption was the main cause of these out- main exporting companies, OLAM, Export Trading, and put declines. After more than a decade of recovery, produc- Mohammed Enterprises. Market and enabling environment risks. Assessing the role of in “normal years” (or when there are no droughts) the price volatility in food crops is more complex than it is for domestic price tends to align to the longer-term trend of export crops, in particular maize. Maize’s domestic price the international price. for the most part reflects crop availability in the domestic market and is less correlated to the short-term oscillations In effect, maize trade is somehow dominated by a policy of the price in international markets (see figure 3.5, with that establishes export bans and import permits when monthly prices). The revision of the past six years shows the government deems it necessary to stabilize prices and that the higher peaks in the domestic price (wholesale) are guarantee food security. These interventions have not been reached after a drop in domestic production. However, successful in stabilizing prices, and because there is much 16 Tanzania Figure 3.5. M  aize Price Figure 3.6. R  egression Chart for Production (000 tons) SAFEX Arusha Dar es Salaam–Wholesale 450 Drought 5000 3 400 4500 y = 0.009x + 0.2306 2.5 4000 R 2 = 0.5353 Yield (tons per hectare) 350 3500 3000 2 US$/Ton 000 tons 300 2500 250 2000 1.5 200 1500 1000 1 150 500 100 0 0.5 2006M01 2006M05 2006M09 2007M01 2007M05 2007M09 2008M01 2008M05 2008M09 2009M01 2009M05 2009M09 2010M01 2010M05 2010M09 2011M01 2011M05 2011M09 2012M01 2012M05 2012M09 0 0 50 100 150 200 250 Cumulative rainfall (mm.) Source: FAO—Global Information and Early Warning System (GIEWS) and Source: Global Precipitation Climate Project (GPCP), author’s analysis. South African Futures Exchange (SAFEX). uncertainty about when and for how long the restrictions Empirical evidence will operate (generating a separate and distinct risk), they Maize yield—cumulative rainfall relationship. An attempt tend to create confusion and obscure the functioning of was made to correlate yields and cumulative rainfall to the market. This policy also results in additional transac- find statistical evidence of yield variations by crops and tion costs that are transferred to farmers through lower regions.14 Some results for maize are shown in this sec- costs. tion for illustrative purposes; the entire study is included in appendix A. Moreover, in the long run this policy tends to curb the trade opportunities offered by markets in neighboring The best results were found for the Manyara region, where countries (mostly but not exclusively Kenya), thus reduc- both the sowing and the mid-season periods explain a ing the investment incentives to increase production and significant amount of the variability in yield (72 percent reduce volatility, with the side effect of encouraging infor- and 75 percent). For the sowing season alone, cumulative mal cross border trade.13 Price volatility impacts are great- rainfall is significant for three regions: Arusha, Manyara, est for the most vulnerable segment of the supply chain and Tanga. Figure 3.6 offers an example of a regression (small-scale farmers and traders) that cannot profit from analysis for Arusha alone. high prices determined either by the international market or domestic scarcity. For Arusha, the relationship is quite clear, with a deter- mination coefficient (R2) of 54 percent, meaning that For rice, domestic price oscillations tend to be less dra- 54 ­ percent of the variability in yield can be explained by matic than they are for maize. This can be explained by the cumulative rainfall of the sowing season alone. The the fact that part of the crop is cultivated under irrigation slope is positive, which means that more rain results in a and because rice is a less sensitive commodity in terms of higher yield, signaling that drought is the main threat here. food security in Tanzania and is more likely to be a cash It is also clear that the worst years in terms of rainfall (1997 crop for farmers. 14 Linear regression models were built for each region to establish the relation- ship between maize yield, expressed in tons per hectare, and the cumulative rainfall of each of the crop seasons (sowing, mid-season, harvest). The model is 13 Tanzania’s recorded trade in maize is modest; during the period 2005–07, expressed as: Yield = b0 + b1 Raini The determination coefficient (R2) was calcu- imports averaged about 3 percent of apparent consumption and exports repre- lated for each model. The R2 is a measure of the proportion of the variance in sented just 2 percent of maize production. yield that can be explained by the cumulative rainfall in each season. Agricultural Sector Risk Assessment 17 Figure 3.7.  Regression Chart for Figure 3.8. C  offee: Occurrence of Dodoma Risk Events 1.8 Coffee production and price 1.6 y = 0.0035x – 0.3913 R 2 = 0.3294 Production ICO (tons) Lineal (Production ICO (tons)) Yield (tons per hectare) 1.4 International price 1.2 (U.S. cents/pound) 75,000 250 1 70,000 230 0.8 65,000 210 0.6 U.S. cents/pound Drought 190 0.4 60,000 Drought and secular and pests 170 55,000 low export Tons 0.2 150 prices 0 50,000 130 0 100 200 300 400 500 600 45,000 Cumulative rainfall (mm.) 110 40,000 90 Source: GPCP, author’s analysis. 35,000 70 Drought 30,000 50 0 1 0 2 0 3 0 4 0 5 0 6 0 7 0 8 0 9 1 0 1 1 12 2 3 20 0/0 20 1/0 20 2/0 20 3/0 20 4/0 20 5/0 20 6/0 20 7/0 20 8/0 20 9/1 20 0/1 20 1/1 /1 0 20 Source: TCB and ICO. Chart 3.9. Maize and Rice: Occurrence of Risk Events Maize production and yields Rice production and yields 4500 4.00 1200 4.00 2004/05 3.50 4000 3.50 1000 Drought in 3500 northern 3.00 3.00 800 region 3000 2.50 Tons/ha 000 tons 000 tons 2.50 2005 Tons/ha 2500 2011 2009 Drought 600 Drought 2.00 2000 Drought 2.00 2003 1.50 1500 Drought 400 1.50 1.00 1000 1.00 200 0.50 500 0 0.50 0 0.00 0 1 0 2 0 3 0 4 0 5 0 6 0 7 0 8 0 9 1 0 1 1 12 2 3 20 /01 20 /02 20 /03 20 /04 20 /05 20 /06 20 /07 20 /08 20 /09 20 /10 20 /11 20 /12 3 20 /0 20 1/0 20 2/0 20 3/0 20 4/0 20 5/0 20 6/0 20 7/0 20 8/0 20 9/1 20 0/1 20 1/1 /1 /1 00 00 01 02 03 04 05 06 07 08 09 10 11 12 20 20 Production (000 tons) Yield (tons/ha) Production (000 tons) Yield (kg/ha) Lineal (yield (tons/ha)) Lineal (yield (kg/ha)) Source: USDA. and 2000, when only 44 mm and 55 mm fell through each yield, and the opposite. Rainfall wasn’t significant for any season, respectively) were also the worst years in terms of region in the harvest season. yield (with 129 kg and 300 kg per hectare, respectively). Therefore, it is clear that drought in the sowing season has Risk event occurrence. Many of the production declines can an important effect on maize yield in the Arusha region. be easily explained by natural hazards, mostly weather events, as is reported in different technical reports and As for mid-season, rainfall explains the variability of yield other publications. Figure 3.8 shows the evolution of cof- for the following regions: Dodoma, Manyara, Mbeya, and fee production at the national level, since coffee prices Ruvuma. Figure 3.7 illustrates the relationships in Dodoma. began to recover in the international market around 2002. Even though the determination coefficient for Dodoma is not very high (33 percent), it is clear that there are two Coffee output steadily increased between 2003/04 and different groups of points: those with low rainfall and low 2008/09 followed by two pronounced drops in 2009/10 18 Tanzania Figure 3.10. C  otton: Occurrence of Risk Event 2009/10 (27% decline of production) Likely causes: Tanzanian seed cotton production 1. Fear of anticipated fall in cotton (MT) prices (10% drop in acreage) 2. Low and poorly distributed rainfall 400,000 3. Suspension of passbook system (input supply) 4. Unchecked insect a attack 350,000 300,000 250,000 2003/04 Delayed rainfall (and other causes??) 200,000 150,000 2006/07 (Over 60% drop in production) 100,000 2010/11 (39% drop in production) Causes: 1. Devastating drought Likely causes: 2. Doubling of input prices 1. Localized droughts 50,000 3. Problems with the input system. 2. More than average rainfall in some 4. Other? areas 3. Failure of voucher scheme for inputs 0 0 /1 2 3 4 5 6 7 8 9 0 1 /0 /0 /0 /0 /0 /0 /0 /0 /0 /1 /1 00 99 01 02 03 04 05 06 07 08 09 10 20 19 20 20 20 20 20 20 20 20 20 20 Source: Tanzania Cotton Board (TCB) annual reports. and 2011/12. These declines are attributed to a combina- There have been three years for maize and two years for tion of production (yield) problems: drought and/or pest rice when production has dropped, coinciding with drops attacks. in yields below the trend line. Droughts were reported during those years. Previous drops in production are most likely associated with the extended depressed market period, character- Another example is cotton. Figure 3.10 shows a sequence ized by low and unstable prices. In 2003/04, however, the of events that caused production to drop in recent years, fall in production was also driven by drought problems. and identifies a combination of causes related to weather, Figure 3.9 illustrates production performances for maize ­ pests, and regulatory risks. and rice during the last decade. Agricultural Sector Risk Assessment 19 CHAPTER FOUR ADVERSE IMPACT OF AGRICULTURAL RISK Quantification of losses The quantification of losses presented in this report refers largely to production risks, such as drought and pest attacks. In this section, the indicative value of agricultural output lost for a particular year is calculated as the deviation of the actual annual yield from a historic yield trend value multiplied by the actual area that year, valued at 2008–10 average producer prices and converted into U.S. dollars at the 2010 exchange rate. Indicative loss values are also compared with agricultural gross domestic product (GDP) in the relevant year to provide a relative measure of the loss. Approximately US$203 million or 3.5 percent of the agricultural GDP was esti- mated as the value of the average production loss annually in the agricultural sector as a result of unmanaged production risks. The calculation involves the following crops: tobacco, coffee, cotton, cashew nuts, sesame, maize, rice, beans, and cassava, which together are responsible for more than 80 percent of agricultural GDP and as such as representative of sector risks. Drought was the main cause of these shocks, sometimes in combination with other events. See table 4.1 for detailed information by crop. In terms of the regional distribution of the losses with regard to maize, more than 40 percent of the 30-year losses are concentrated in Mbeya, Manyara, Shinyanga, and Iringa. Kilimanjaro and Arusha have also been badly affected by production volatility. Altogether, the six regions account for 61 percent of all losses. Production volatility We should expect that an agricultural system (for example, a country or a particular region) that is intrinsically exposed to high production volatility would be more prone to suffer greater economic losses from natural hazards (a drought, for instance) than one that is more stable. Higher volatility means that the production system (yields) Agricultural Sector Risk Assessment 21 Table 4.1. Value of the Average Annual Losses (at 2010 prices) Average Annual Average Annual Losses as % of Crop Period Losses (tons) Losses (US$) Agricultural GDP Export Tobacco 1982–2007 2,697 10,511,265 0.18% Cotton 1981–2010 14,676 21,506,786 0.37% Coffee 1981–2011 1,539 5,148,697 0.09% Sesame 2001–2012 9,210 2,267,709 0.04% Cashew nuts 2003–2010 733 348,738 0.01% Food security Maize 1981–2010 246,823 55,767,795 0.96% Rice 1981–2010 58,044 39,246,307 0.68% Cassava 1981–2010 177,139 28,805,572 0.50% Beans 2003–2010 59,982 39,924,138 0.69% TOTAL 203,527,008 3.50% Source: Author’s calculations. can change significantly from cycle to cycle in either direc- minimum of 14.2 percent in Ruvuma (see table 4.2). The tion; and lower volatility means that production (or yields) highly productive regions of Mbeya, Iringa, and Rukwa do not fluctuate dramatically, but change at a steady pace exhibit moderate volatility (23 to 25 percent). The national over time. In this context, agricultural volatility is closely coefficient is relatively low (20 percent) as this aggregate linked to the natural resources base, the predominant value masks variability within regions. Table 4.2 shows the technology and skill, and the market development and coefficient of variation of yields and the annual average regulations. production loss that results from unmanaged production risks, by region. A study was performed to measure the relative volatility of the different regions and at a national level using the The different levels of production volatility among agro- coefficient of variation of yields.15 For illustrative pur- ecological regions reflect the great diversity in terms of poses, maize was the focus of this analysis, given that there weather patterns and natural resource endowments. Fur- is available a relatively extensive database for all regions ther analysis would be required to examine the impor- from the Ministry of Agriculture, Food Security and tance of developing technology and market strategies for Cooperatives (MAFC) and because maize is in practice smoothing those differences and reducing overall agricul- the only crop cultivated throughout Tanzania and traded tural volatility throughout the country. This is an issue for extensively within the country. the risk solution stage of this study. The regions with larger maize production are certainly Maize production volatility is very different among the those suffering greater losses when adverse natural events regions as measured by the coefficient of variation of occur, simply due to the volume of production, as shown yields, with a maximum of 56.3 percent in Coast and a in table 4.2. That is true, for instance, for Mbeya, Shin- yanga, and Iringa. Questions arise, however, about whether lower volatility would necessarily result in fewer 15 Calculated as the standard deviation divided by the series arithmetic media. It shows the extent of variability in relation to mean of the population: the higher losses, and whether risk-related losses and production vol- the number is, the worse the situation is. atility are linked. A strong correlation would support the 22 Tanzania Table 4.2. M  aize Production Figure 4.1. M  aize Production Variability by Region Volatility 60,000 Average annual loss (US$) / per caput Maize production avg. 2007/08–2009/10 Production Coefficient 50,000 by Region— Average of Avg. 2007/08 Annual Variation 40,000 R 2 = 0.3511 (kg/person) and 2009/10 Loss of Yields 30,000 Regions (000 tons) (US$) (%) 20,000 Mbeya 503 4,844,386 23.3 Manyara 253 4,799,232 26.7 10,000 Shinyanga 508 4,623,777 43.3 0 Iringa 407 4,490,286 25.5 0% 10% 20% 30% 40% 50% 60% Coefficient of variation of yields Kilimanjaro 199 3,972,870 45.1 Arusha 152 3,972,284 45.3 Source: Author’s calculations. Dodoma 158 2,858,077 41.1 Singida 110 2,254,946 40.4 Rukwa 367 1,903,361 25.6 understanding that it is meaningful to put in place specific Tabora 251 1,711,112 33.8 policies to reduce volatility. Kagera 151 1,642,120 37.9 Such correlation is attempted below through a regression Tanga 283 1,374,540 24.7 of the monetary losses weighted by per capita (2007/08– Morogoro 240 1,169,570 32.3 2009/10) average production (as a way to isolate the Ruvuma 234 1,051,896 14.2 regions’ size effect) with the coefficient of variation of yields Coast 66 835,494 56.3 in a cross section regression among regions. Figure 4.1 Mtwara 50 788,085 51.0 illustrates this relation. Kigoma 149 705,524 22.9 Mwanza 194 351,611 31.6 Although the correlation coefficient is not high (35 per- Lindi 64 219,752 36.4 cent), the points are well aligned, corresponding to what Mara 173 44,386 45.3 would be expected if volatility and average annual losses Total National 20.0 were connected positively.16 Source: Based on data from MAFC. 16 In this cross-section analysis, the number of observations is limited by the number of regions. Agricultural Sector Risk Assessment 23 CHAPTER FIVE STAKEHOLDERS’ ASSESSMENT Impact of risks at individual stakeholder level How the losses are distributed among stakeholders within the supply chain is to a great extent a function of value chain governance and the actors’ capabilities and opportu- nities for risk management. Price risk. Exporters, millers, and large trading companies are capable of managing price risks globally through the practice of standard futures risk management strat- egies. CRDB Bank, the main agricultural financing bank, has unsuccessfully tested innovative ways to manage price risk for coffee and cotton, such as gaining access to international markets for price hedging. Normally, banks manage lending risk through regular banking risk management procedures, such as collateral management, due diligence, and the maintenance of loan loss provisions (as their loan recovery prospect is particularly related to short-term commodity price variation). Traders, middlemen, and small storage and processing companies can manage price risk via keeping/releas- ing physical stocks, at the same time assuming the additional risk of accumulating higher losses if prices decline. However, those involved in export crops take important risks when they make advance payments to farmers or are required to keep the prod- ucts in storage until delivery in the auction or to the exporting companies. Small-scale farmer capacity to manage price risk is extremely limited. Primary coop- erative societies and, to some extent, second-level farmers’ associations are also the weakest segment in the supply chain. Product price variations within the marketing year can expose farmers to financial losses when they practice multipayment systems. Primary cooperative societies tend to have fragile financial structures and rely on bank credits to support farmers in marketing their products and paying for marketing and processing costs. Price risk management strategies are needed at the farm level (farmers and coopera- tives) to reduce the exposure to price risks without having to resort to financial price hedging instruments, because only a small group of stakeholders can benefit from Agricultural Sector Risk Assessment 25 these market-based instruments. In any case, primary cooperative societies could be the target for such policies, Vulnerable hotspots and appropriate training should be a central component. During an average year, Tanzania has enough food for Timing is of the essence in price risk management, and its population. Data available from Food and Agriculture there is plenty of room to reduce price volatility exposure Organization-Global Information and Early Warning for the various participants along coffee and cotton supply System (FAO-GIEWS) show that total domestic cereals chains. availability has surpassed requirements for domestic uti- lization since at least 2002/03.17 The same holds true for Production risks. All actors along supply chains are exposed maize (see figure 5.1). to the variability in primary farming production. How- However, the Ministry of Agriculture, Food Security and ever, smallholder farmers are particularly exposed to pro- Cooperatives (MAFC 2010) reports that for 2010/11 duction and yield variability. Their family food security “pockets of vulnerable areas” or “vulnerable hotspots” and monetary income are dependent on the crop harvest. had been identified in 45 districts in 11 regions, namely, Thus, to mitigate weather and pest and disease risks at Arusha, Tanga, Shinyanga, Mwanza, Kilimanjaro, Coast, the farm level, many producers adopt low-risk and low- Tabora, Mara, Manyara, Kagera, and Mtwara. Of yield crop and production patterns to ensure that they end these, four are definitely food deficit regions, five are self-­ up with at least a minimum quantity of food available. sufficient, and two are surplus regions. With the exception These production patterns come at the expense of high- of Tanga and Manyara, all the other regions have high risk, high-return production that could create income production volatility (coefficient of variation higher than growth and the buildup of capital. More cost-effective 30 percent). technologies and agricultural practices can provide better protection against production risks but that would imply Moreover, the 2009/10 Comprehensive Food Security that improved research and extension services are avail- and Vulnerability Analysis reported that, at the time of able for smallholder farmers. In export crops, such as cof- the survey, 4.1 percent of households in rural mainland fee or cotton, exporters and processors tend to provide Tanzania had poor food consumption, and 18.9 percent marketing and productive service assistance to farmers to had borderline consumption.18,19 In terms of undernu- increase and stabilize supplies. trition, 5.7 percent of children under five years of age were wasted, 36.6 percent were stunted, and 14.3 percent The government role. Most of the specific risk mitigation were underweight. Regions of Mtwara, Manyara, Aru- actions implemented by the government are directed at sha, Singinda, and Lindi had the highest prevalence of coping with the impact of natural hazards (food aid, seed food-poor consumption households, whereas Dodoma, distribution, and so on). The government aims at main- Morogoro, and Manyara reported the highest prevalence taining social, economic, and political stability as well of households with borderline consumption. Child wast- as assuring food security. Government expenses to cope ing rates were highest in Arusha, Manyara, and Mtwara. with agricultural risks are usually met through budget The latter two also had the highest prevalence of under- resources when they are not required in response to cata- weight rates. Stunting prevalence, by contrast, was high- strophic events. est in regions such as Iringa, Rukwa, and Kigoma, which did not report poor or borderline food consumption Summary. Table 5.1 summarizes the stakeholders risk profile, which is defined by the following variables: the sources of risks that are most common for each stakeholder; the signifi- 17 Further details regarding vulnerability analysis are contained in appendix C. cance of the perceived damage expected from the realized 18 Poor food consumption households have a diet based mainly on cereals with risk events; and, finally, the stakeholders’ current capacities almost no animal protein and very little of any other food item (three days per week vegetables, and two days per week pulses). to manage those risks. Smallholder farmers and their fam- 19 Borderline food consumption households have a marginally better diet than ilies are the weakest segment in the supply chain and the poor food consumption households as they consume approximately one day prevalence of risks contribute to the vicious cycle of poverty. more per week pulses, vegetables, and fruits. 26 Tanzania Table 5.1. Summary of Stakeholder Risk Profiles Most Common Current Risk Stakeholders Sources of Risk Perceived Risk Management Capability Smallholder Natural hazards (climate and Significant production losses, No drought risk mitigation. Inefficient farmers pest and diseases). reduced family income, and food diseases and pest risk mitigation. No Price drop and exchange rate insecurity. price and exchange rate risk transfer. variation. Minor to medium income losses. Indebtedness. Sell assets. Primary Climate and pest and Reduced procurement and higher Delay payments to farmers. Fall cooperative diseases. unit costs. into arrears with creditors. Increase societies/union Product and input short-term Potential important financial losses. indebtedness. cooperatives price variation. Traders/processors Climate and pests and Reduced procurement and higher Diversification. Price bargaining. diseases. unit costs. Stockpiling. Increase indebtedness. Short-term price variations. Revenue variation and possibility of breakdown when price drop is pronounced. Exporters Short-term price variations. Revenue variation. Hedging. Banks Financial losses accrued by Credits in arrears and economic Collateral management, due diligence, clients owing to production losses. maintenance of loan loss provisions. and market risks. Government Natural hazards. Social instability. Budget Budget resources for risk coping Internationally soaring food implications. programs (food aid, seed distribution, prices. cash transfers, and so on). Source: Authors. Figure 5.1. Food Balances Total cereals balance Maize balance 6500 5000 6000 4500 5500 4000 000 tons 000 tons 5000 3500 4500 4000 3000 3500 2500 3000 2000 2002/03 2003/04 2004/05 2005/06 2006/07 2007/08 2008/09 2009/10 2010/11 2011/12 2013/14 2002/03 2003/04 2004/05 2005/06 2006/07 2007/08 2008/09 2009/10 2010/11 2011/12 2013/14 2012/13* 2012/13* domestic availability domestic utilization Source: FAO, GIEWS. Notes: Domestic availability is the sum of opening stocks and production; domestic utilization is the sum of food use, feed use, and other uses. * = 2012/13 is current forecast ­ ouseholds.20 Food consumption was lowest among the h The coincidence of certain regions as hosts of both poorest households, whereas households with the poorest ­ “pockets of vulnerable areas” and poor and borderline food consumption tended to have access to less livestock, food consumption households, as well as areas specially cultivate less diverse crops, cultivate less than one ha of exposed to agricultural risks (such as Arusha, Manyara, land, and be less likely to use chemical fertilizers. and Mtwara) is especially worrisome. These regions should be specially targeted with interventions aimed at 20 The regional distribution of child malnutrition simply confirmed that food guaranteeing food security under the Agricultural Sector availability and consumption do not translate necessarily into adequate nutrition. Development Program (ASDP). Agricultural Sector Risk Assessment 27 CHAPTER SIX RISK PRIORITIZATION AND MANAGEMENT To better utilize scarce resources, it is important to understand which risks are caus- ing major shocks to the sector in terms of losses and observe at what frequency they occur. The sections below summarize the risks facing the agricultural sector and pos- sible solutions. The latter were identified by the World Bank mission team and then validated and prioritized with stakeholders at different levels and at a workshop on January 28, 2013, in Dar es Salaam. This workshop was organized by the Ministry of Agriculture, Food Security and Cooperatives (MAFC) to present the findings of the mission and to reach consensus on the identified risks and key risk solution areas. Risk prioritization The following tables, arrived at through a consensus process with stakeholders, provide a summary of agricultural risks aggregated on the basis of the probability that risk events occur and the expected impact (losses) for food and for cash crops. The identified risks located in the darkest shaded area (upper right corner) represent the most signifi- cant risks owing to their potential to cause the greatest losses and the frequency of their occurrence. The second level of importance is represented by the lighter shaded boxes, whereas the unshaded boxes (on the left side of table 6.1) represent identified risks that either have low potential to cause damages or their frequency of occurrence is also low. In summary, the exercise of risk prioritization (based on the frequency of realized risk events, their capacity to cause losses, and the ability to manage the risks shown by the different stakeholders) identified the most significant risks, listed below: »» Drought events, especially for maize, rice, and cotton »» Widespread outbreaks of pests and diseases, especially for cotton, maize, and coffee »» Price volatility for cotton and coffee »» Regulatory risks, mostly linked to the trade policy framework, for various cash crops and maize Although these risks do not necessarily manifest themselves in the form of catastrophic shocks to agriculture (as shown in table 6.1), they are identified as the main drivers Agricultural Sector Risk Assessment 29 Table 6.1. Risk Prioritization—Food Crops Probability of event Negligible Moderate Considerable Critical Catastrophic Highly •  Diseases (rice yellow •  Drought probable mottle virus) (R) (M) Probable •  Aflatoxins (M) •  Pests (rodents, •  Price volatility (R) •  Droughts •  Pests (wild armyworms, quealea •  Unpredictable (R) animals) (Ca) birds) (R) trade policy (M) •  Diseases (for •  Erratic rainfall (R) example, •  Cassava mosaic maize streak disease) (CMV) (Ca) diseases) (M) •  Diseases (B) •  Pests (for example, rodents, armyworms, stock borer) (M) •  Cassava brown stick diseases (CBSD) (C) •  Excess water (Ca) •  Insects and pests (for example, beetle, armyworm) (B) •  Food deficits/surplus in neighboring countries (M) Occasional Remote •  Flood (R) Source: Authors. Key: R=Rice, M=Maize, B=Beans, Ca=Cassava, Cot=Cotton, Tob=Tobacco, Co=Coffee, C=Cashew Nuts, S=Sesame. of agricultural volatility that cause stakeholders income strategies are a combination of risk mitigation, risk trans- instability and recurrent food security problems. Whereas fer, and risk coping instruments. Risk mitigation refers to implementation of the solutions will certainly entail actions taken to eliminate or reduce events from occur- regional specificities, an appropriate national institu- ring, or reduce the severity of losses (for example, water- tional and policy framework must first be identified. The draining infrastructure, crop diversification, extension, assessment of regional risk dimensions will be part of the and so on); risk transfers are mechanisms to shift the risk detailed solutions definition and the program and proj- to a willing third party, at a cost (for example, insurance, ect design that will follow. This will be part of the second reinsurance, financial hedging tools, and so on); and risk assessment mission. coping makes up actions that will help cope with the losses caused by a risk event (for example, government assistance Priority risk to farmers, debt restructuring, and so on). management measures Filtering risk management Risk solutions: The long list measures Below is the long list of risk solutions discussed during the Many of the actions included in the Agricultural Sector risk assessment mission with various stakeholders (table Development Program (ASDP) and other specific proj- 6.3). These potential solutions were identified during field ects and programs for the agricultural sector are already interviews and were previously suggested in various gov- tackling some of the risk solutions identified in the long ernment and nongovernmental documents. Usually, risk list in table 6.3. 30 Tanzania Table 6.2. Risk Prioritization—Export Crops Probability of event Negligible Moderate Considerable Critical Catastrophic Highly •  Counterparty •  Sesame flea beetle •  Diseases (for •  Cotton probable risk (farmers)/ infestation example, CBD, price Side-selling (Cot (insects) (S) CWD, CLR) (CO) volatility and Tob) •  Fungal diseases (for •  Insects/pests (Cot) example, powdery (cotton bull worm mildew) (C) and so on) (Cot) •  Drought (Cot) Probable •  Pests and •  Pests (for example, •  Erratic •  Regulatory diseases (Tob) thrips) (Co) rainfall (CO) risk •  Incidence of •  Price volatility (CO) •  Price volatility diseases (for •  Counter party (unstable world example, leaf (Ginners)/ prices) (C) spot, bacterial International blight, stem buyers (Cot) rot) (S) •  Excess rainfall (Tob) Occasional •  Occurrence of severe drought (S) Remote Source: Authors. Key: R=Rice, M=Maize, B=Beans, Ca=Cassava, Cot=Cotton, Tob=Tobacco, Co=Coffee, C=Cashew Nuts, S=Sesame. Moreover, the government of Tanzania is now working The shortlist areas for deepening the risk solutions are, in on a new agriculture policy that is expected to be final- brief, the following: ized and approved soon and may include changes to the ASDP. Highly drought- and pest-tolerant seeds. There are weaknesses in the supply chains for delivering drought Table 6.4 contains a number of these projects and pro- tolerant seeds, disease resistant seeds, and planting mate- grams, as identified by the mission, indicating their con- rial, and inefficiencies in seed markets that should be nection with the risk assessment results and the potential addressed. In principle, this encompasses food crops such gaps to be covered with specific risk management actions as maize and rice, and export crops such as cotton and additional to existing measures. coffee. This would imply the need to effectively intervene in the short to medium term to make the seed supply Risk Solutions: The short list chains work more effectively along the range of stakehold- The long list of general solutions in table 6.3 and the gap ers involved (from breeders to seed producers to farmers) analysis presented in table 6.4 were used to start narrow- as well as to clearly define the roles of public and private ing down to specific areas of solutions that tackle the key sectors in developing this market. risk issues. The final result will be a package of interven- tions that could effectively lower volatility and increase Good agricultural practices to address drought, resilience in agriculture. The identified interventions to pests, and diseases. Widespread, improved agricul- reduce agricultural risks will also have the added benefit tural risk mitigation practices can have significant impacts of contributing to higher productivity and a direct posi- in reducing risks derived from irregular or insufficient tive impact on the reduction of poverty. rainfall, as well as from diseases and pests. This implies a Agricultural Sector Risk Assessment 31 Table 6.3. Risk Solutions: The Long List Risk Mitigation Transfer Coping Drought •  Drought tolerant seed varieties •  Insurance •  Food reserves •  Water harvesting and irrigation •  Food imports •  Improving early warning systems •  Social safety net •  Reforestation/afforestation programs •  Contour farming/Soil and water conservation programs/ •  Risk financing Assisted natural regeneration/Land and water management •  Agronomic practices for on-farm drought management •  Crop diversification Price volatility •  Improved understanding of price risk management •  Hedging •  Imports •  Managing food stocks •  Trade policies •  Trade policies •  Social safety net •  Increased domestic processing programs •  Improved quality to access stable niche markets •  Improved market information systems and transparency •  Contract farming •  Improved storage •  Infrastructure development •  Foster competition in markets Diseases •  Scale-up disease tolerant varieties •  Quarantine measures •  On-farm agronomic practices •  On-farm agronomic •  Early warning systems practices •  Integrated pest management •  Integrated pest •  Quarantines measures management •  Improved phytosanitary laboratory systems •  Improved extension services Pests •  On-farm agronomic practices •  Quarantine measures •  Early warning systems •  On-farm agronomic •  Integrated pest management practices •  Quarantines measures •  Integrated pest •  Improved phytosanitary laboratory systems management •  Improved extension services Regulatory risks •  Improved efficacy of commodity councils •  Promote proactive rather than reactive policies •  Develop clear, long-term, efficient, and transparent policies for commodities and sector development •  Improved transparency in policy decision making Source: Authors. need to strengthen the existing disconnected technology time allow for a transparent market. Policy predictability, systems through effective coordination among research, market transparency, and fewer nontrade barriers would extension, and training, including the effectiveness of result in greater incentives for farmers to invest in tech- information and communication outreach to farmers. nology that increases productivity and reduces production volatility in a sustainable way. This would create a better Balanced maize trade policy. The export and import balance between the short-term food security goal and the policy has to be predictable and stable and at the same long-term productivity growth aim. 32 Tanzania Table 6.4. Gap Analysis (Continued ) Statement on Short List Risk and Current Projects Risk Solving Solution Rating Solution or Programs Perspective Gap Proposal Drought— Irrigation ASDP: A total of 353 Prospective Expansion of None critical or systems, land irrigation schemes were sector risk coverage to considerable and water upgraded, rehabilitated, or reduction build up from and probable or management. newly developed. impact is small current projects’ highly probable TAFSIP: Irrigation because of experiences. risk development, sustainable nonmassive type water, and land use of investment. management. Feed the Future Program: Increase area under irrigation by 15.5% through the development of seven smallholder irrigation schemes in Morogoro and Zanzibar. More extensive There are available drought Reduction of Planting Specific program use of drought yield variability tolerant seeds in Tanzania (for materials and to be included resistant seeds. example, maize) and there and crop losses research results in set of specific is research under way (for but learning are available; proposals. example, coffee) but there is required for information and low adoption. optimal balance promotion are between risk missing. reduction and high productivity. Pests and Good agricultural ASDP: Strengthening Existing research To build up Redesign current diseases—critical practices to agricultural research and results need to be from the ASDP programs or considerable address drought training. disseminated to subprograms. and expand and probable or and pest and Feed the Future Program: farmers. Specific geographically to highly probable diseases. Agricultural support services knowledge is cover the entire risk and capacity building including required to country and/or research and development and address pest new more specific financial services. and disease technology transfer prevention and program. control. Introduce disease No comprehensive program Reduction of Progress in Specific program resistant seeds in place. There are available yield variability coffee, for to be included and planting disease resistant seeds (coffee and crop losses, instance, but no in set of specific material. and so on). depending on program covering proposals. the crop. most crops. (Continued ) Agricultural Sector Risk Assessment 33 Table 6.4. Gap Analysis (Continued ) Statement on Short List Risk and Current Projects Risk Solving Solution Rating Solution or Programs Perspective Gap Proposal Price volatility— Improved Agricultural Marketing Project-related No practical Reforms needed moderate to understanding Systems Development interventions strategies in place to deal with critical and of price risk Program: (a) agricultural can have good to reduce price price volatility. probable and management, marketing policy development; results in terms risk exposure To deepen into highly probable, market (b) small producers’ of targeted to vulnerable institutional mostly cotton, information, and empowerment by building stakeholders stakeholders. arrangements and cashew nuts, and hedging. their entrepreneurial and but massive current roles of coffee organizational capacity and and sustainable public and private Trade policies. improving their links to achievements sector. markets; (c) introducing a require Contract warehouse receipt system, nationwide farming. allowing the small farmers policies and using the warehouses to institutional Improved obtain loans for the period buildup. storage. between harvest and sale; and (d) the development of Infrastructure rural marketing infrastructure, development. including storage facilities, market places, and roads. Improved efficacy ASDP: Marketing and Private of commodity Sector Development Improving councils. overall sector policy, regulatory and legal framework. Maize short– Develop clear, Several programs to support Difficult to have Policy framework Find an adequate term policy long-term, maize production at farming impact from still suffering equilibrium variability— efficient, and level. single projects if of great between the short- considerable and transparent policy framework variability and term food security probable risk commodities is weak. discretionarily goal and the long- and sector (because of food term productivity development security goal) growth aim. policies. and therefore provides poor incentives to invest in production. Source: Authors. Risk management strategies for key export crops would imply a need to deepen institutional arrange- with high price volatility (in principle, coffee, and ments and clarify current roles of the public and private cotton). The way these supply chains are ­ organized sectors. depend on which stakeholder is exposed to price risk. A set of options on how to reduce exposure to risk can be Whereas there are already interventions of various tem- explored by analyzing the physical and financial flows poral and spatial natures in Tanzania on these shortlist on current transaction arrangements for exports. This solutions, the key issue is to identify the gaps among cur- 34 Tanzania rent interventions and design a package of solutions that assistance solutions) for better managing the risks identi- addresses the main underlying causes of risk. A Risk Man- fied in the short list. In particular, the mission will: agement Solutions Assessment will be planned as a follow- »» (i) Identify the risk management gaps in existing up to current risk identification. The coming assessment interventions; and, will have the task of linking the risk management inter- »» (ii) Propose a set of interventions for incorporating ventions to the ASDP by developing concrete propos- them into the medium-term ASDP, which could be als (policy solutions, investment solutions, and technical financed by the public sector and/or donors. Agricultural Sector Risk Assessment 35 REFERENCES Agricultural Council of Tanzania. 2010. Value Chain Analysis of Rice and Maize in Selected Districts in Tanzania. Volume I: Introduction, Context Analysis and Recommended Way For- ward. November. Study commissioned by ACT-Tanzania Agricultural Partnership funded by EU-EC-Food Facility Grant 2009/213-569. Ahmed, Syud Amer, Noah S. Diffenbaugh, Thomas W. Hertel, and William J. Mar- tin. 2012. “Agriculture and Trade Opportunities for Tanzania: Past Volatility and Future Climate Change.” Review of Development Economics 16 (3): 429–447. Arndt, Channing, Will Farmer, Ken Strzepek, and James Thurlow. 2011. “Brief 107: Climate Change, Agriculture, Food Security in Tanzania.” Working Paper No. 2011/52, UNU-WIDER. September. Arndt, Channing, William Farmer, Kenneth Strzepek, and James Thurlow. 2012. “Climate Change, Agriculture and Food Security in Tanzania.” Policy Research Working Paper 6188, World Bank, Washington, DC. Atkinson, A. B., and M. A. Lugo. 2010. “Growth, Poverty and Distribution in Tanza- nia.” Working Paper 10/0831. International Growth Center, London, November. Baffes, J. 2003. “Tanzania’s Coffee Sector: Constraints and Challenges in a Global Environment.” World Bank. May 29. ———. 2004. “Tanzania’s Cotton Sector: Reforms, Constraints and Challenges.” Development Policy Review 22(1): 75–96. Beegle, Kathleen, Juachim De Weerdt, and Stefan Dercon. 2006. “Adult Mortality and Consumption Growth in the Age of HIV/AIDS.” Policy Research Working Paper 4082, World Bank, Washington, DC. Web. Birkmann, Jörn. 2006. Measuring Vulnerability to Natural Hazards: Towards Disaster Resilient Societies. United Nations University Press. eBook Collection (EBSCOhost). Web. Cass Business School. 2005. “Agricultural Reforms and the Use of Market Mecha- nisms for Risk Management.” A study commissioned by the Futures and Options Association. March. Christiaensen, L., V. Hoffmann, and A. Sarris. 2006. “Coffee Price Risk in Perspective. Household Vulnerability among Rural Coffee Growing Smallholders in Rural Tanzania.” Paper presented at the CSAE Conference 2006, “Reducing Poverty and Inequality: How Can Africa Be Included?” Oxford, England, March 19–21. ———. 2007. “Gauging the Welfare Effects of Shocks in Rural Tanzania.” Policy Research Working Paper, World Bank. November. Christiansen, L., and A. Sarris, eds. 2007. “Rural Household Vulnerability and Insurance against Commodity Risks.” Food and Agriculture Organization (FAO) of the United Nations Commodities and Trade Technical Paper No. 10, Trade and Markets Divi- sion, Food and Agriculture Organization (FAO) of the United Nations, Rome, Italy. CIA Fact Book. 2013. Tanzania. CRDB. 2005. “Case Study—Managing Risk in the Cotton and Coffee Sectors: The Case of CRDB Bank Tanzania.” Agricultural Sector Risk Assessment 37 CRDB. 2005. “Managing Price Risk for Cotton & Coffee Borrowers: The CRDB Bank Experience.” Paper presented at the Regional Workshop on Agricultural Price Risk Management, San Salvador, El Salvador, March 27–28. Directorate International Trade: Africa Desk, Department of Agriculture, Pretoria. 2006. Country Profile of the United Republic of Tanzania (Agricultural Trade). December. Economic and Social Research Foundation. 2008. “Survey 2: Agricultural Trade Poli- cies Tanzania.” “Enabling Poor Rural People to Overcome Poverty in the United Republic of Tan- zania.” 2012. Country Factsheets: Tanzania. UN International Fund for Agricultural Development. Web. Enfors, E. I., and L. J. Gordon. 2007. “Analysing Resilience in Dryland Agro-Ecosys- tems: A Case Study of the Makanya Catchment in Tanzania Over the Past 50 Years.” Land Degradation and Development 18: 681–688. FAO/GIEWS-WFP. 1999. “Special Report FAO/WFP Crop and Food Supply Assess- ment Mission to Tanzania.” February 15. Fermont, A. V., A. Babirye, H. M. Obiero, S. Abele, and K. E. Giller. 2010. “False Beliefs on the Socio-Economic Drivers of Cassava Cropping.” Agronomy for Sustain- able Development 30 (2): 433–44. The Fourth IPCC Assessment. 2009. Gongwe Mhando, D., and G. Mbeyale. 2010. “An Analysis of the Coffee Value Chain in the Kilimanjaro Region, Tanzania.” NCCR North-South Dialogue 27. Greig, L. 2009. “An Analysis of the Key Factors Influencing Farmers’ Choice of Crop, Kibamba Ward, Tanzania. Journal of Agricultural Economics 60 (3): 703. Haggar, Jeremy, and Kathleen Schepp. 2009. Desk Study: Impacts of Climate Change in the Pilot Country Tanzania of the Coffee & Climate Initiative. Heitzmann, K., R. S. Canagarajah, and P. Siegel, 2002. “Guidelines for Assessing the Sources of Risk and Vulnerability.” Social Protection Discussion Paper Series No. 0218. Social Protection Unit, Human Development Network, World Bank, Washington, DC. Human Development Index. Tanzania. UNDP. Accessed 6 Jan 2013. IFPRI (International Food Policy Research Institute). 2006. “Poverty and malnutrition in Tanzania: New approaches for examining trends and spatial patterns.” January 2006. Tanzania. IFPRI. 2012. East African Agriculture and Climate Change: A Comprehensive Analysis—­ CGIAR. December. (Short summary only—main document to be published shortly) International Coffee Organization. Statistics on Coffee. Web page. IPP Media. 2012. “Farmers Opposed to WHO Blocks in Tobacco Farming.” http:// www.ippmedia.com/frontend/index.php?l=45913. Kijazi1, A. L., and C. J. C. Reason. 2009. “Analysis of the 1998 to 2005 Drought over the Northeastern Highlands of Tanzania.” Climate Research 38: 209–223. doi:10.3354/cr00784. Kimaro, A. A., V. R. Timmer, S. A. O. Chamshama, Y. N. Ngaga, and D. A. Kimaro. 2009. “Competition between Maize and Pigeonpea in Semi-arid Tanzania: Effect on Yields and Nutrition of Crops.” Agriculture, Ecosystems & Environment. MAFC. 2001. “Agriculture Sector Development Strategy.” http://www.tzonline.org /pdf/agriculturalsectordevelopmentstrategy.pdf. MAFC (Ministry of Agriculture, Food Security and Cooperatives). 2010. 2010/11 Final Food Crop Production Forecast. 38 Tanzania MAFC. 2011a. “Trends and Outlook Report on Key Agriculture and Rural Develop- ment Indicators in Tanzania.” Prepared by the Agricultural Sector Development Program Monitoring and Evaluation Thematic Working Group. May, pp. 16–17. MAFC, 2011b. AGSTATS for Food Security, Volume 1: The 2010/11 Final Food Crop Produc- tion Forecast for 2011/12 Food Security. December. Executive Summary. MAFC. 2010. Statistical data. Meena, Hubert, and Paul O’Keefe. 2007. “Sustainable Livelihoods in the Context of Vulnerability and Adaptation to Climate Change Impacts in Tanzania: A Case Study of Kilimanjaro Region.” The Netherlands Climate Assistance Program. CEEST Foundation, June. Web. Ministry of Finance. 2011. “The Economic Survey 2010.” Ministry of Finance and Economic Affairs. 2009. “Poverty and Human Development Report 2009,” Research and Analysis Working Group. MKUKUTA Monitoring System. December. Ministry of Industry, Trade, and Marketing of Tanzania. 2008. “Agricultural Market- ing Policy.” Ministry of Industry and Trade. 2011 “Tanzania Textiles and Garment Development Strategy,” Interim Report for the Ministry of Industry and Trade. A study funded by the Tanzania Gatsby Trust. December. Minot, N. 2010. “Staple Food Prices in Tanzania.” Prepared for the Comesa Policy Seminar under the African Agricultural Marketing Project (AAMP), “Variation in Staple Food Prices: Causes, Consequence, and Policy Options,” Maputo, Mozam- bique, January 25–26. National Bureau of Statistics, Ministry of Finance. 2011. “Tanzania in Figures 2010.” June. National Sample Census of Agriculture. 2012. Small Holder Agriculture, Volume II, Crop Sector, national report. April, p. 13. OECD (Organization for Economic Co-operation and Development). 2009. “Risk Management in Agriculture: A Holistic Approach; Highlights.” ODXF Food Security Analysis Service. 2010. “United Republic of Tanzania Compre- hensive Food Security and Vulnerability Analysis.” VAM Food Security Analysis. World Food Program. September. Ponte, S. 2002. Farmers and Markets in Tanzania: How Policy Reforms Affect Rural Livelihoods in Africa. Portsmouth, NH: Heinemann. Population and Housing Census 2002. Tanzania. REPOA (Research on Poverty Alleviation). (2007). “Children and Vulnerability In Tan- zania: A Brief Synthesis.” Special Paper 07.25, Dar es Salaam, REPOA. UNICEF. Republic of Tanzania. 2003. Initial National Communication under United Nations Framework Convention on Climate Change. Vice President’s Office. Republic of Tanzania. 2007. National Adaptation Programme of Action. Vice Presi- dent’s Office. Romer Løvendal, C., M. Knowles, and N. Horii. 2004. “Understanding Vulnerability to Food Insecurity Lessons from Vulnerable Livelihood Profiling,” ESA Working Paper No. 04-18, October. Agricultural and Development Economics Division, the Food and Agriculture Organization of the United Nations. Sarris, Alexander, and Panayiotis Karfakis. 2006. “Household Vulnerability in Rural Tanzania.” Paper presented at the CSAE conference “Reducing Poverty and Inequality: How Can Africa Be Included?” March. Web. Agricultural Sector Risk Assessment 39 Schlenker, W., and D. B. Lobell. 2010. “Robust Negative Impacts of Climate Change on African Agriculture.” Environmental Research Letters 5. Southern Agricultural Growth Corridor of Tanzania. “Appendix IV: Value Chain and Market Analysis.” “Tackling Vulnerability: An Approach to Poverty Reduction.” 2004. A Plain Language Guide to the 2002/2003 Tanzania Participatory Poverty Assessment. Research and Analy- sis Working Group. Government of Tanzania. October. Tanzania. 2013. Statistics. http://www.bot-tz.org/publications/PublicationsAndStatistics .asp. Tanzania Coffee Board. 2010. “Annual Report 2009–2010.” Tanzania Cotton Board. 2011. Annual Report and Accounts for the year ending on 30th June 2011. TaCRI (Tanzania Coffee Research Institute). 2005. “Coffee Baseline Report.” Sokoine University of Agriculture—Bureau for Agricultural Consultancy and Advisory Service. August. TechnoServe. 2012. “Weather Risk Management Pilot Program Tanzania.” Report compiled by Nargis Suleman with contributions by Jonathan Edmonds (Tech- noServe Consultants). Tumushabe, Joseph. 2005. “HIV/AIDs and Changing Vulnerability to Crisis in Tan- zania: Implications for Food Security and Poverty Reduction.” Presented at the International Conference on HIV/AIDS, Food and Nutrition Security, Makere University, South Africa, April 14–16. United Republic of Tanzania. 2006. Agricultural Sector Development Programme (ASDP). Support Through Basket Fund, government programme document, p. 7. USAID. 2010. “Staple Foods Value Chain Analysis. Country Report—Tanzania.” Prepared by Chemonics International Inc. June. USAID. 2010. “Tanzania Coffee Industry Value Chain Analysis. Profiling the Actors, Their Interactions, Costs, Constraints and Opportunities.” Prepared by Chemon- ics International Inc. 28 May. Wheeler, David. 2011. “Quantifying Vulnerability to Climate Change: Implications for Adaptations Assistance.” CGD Working Paper 240. Washington, DC: Center for Clobal Development. (Accompanying data set and maps) World Bank. 1994. “Tanzania Agriculture.” World Bank. 2009a. “Tanzania: Country Brief.” World Bank. 2009b. “Eastern Africa: A Study of the Regional Maize Market and Marketing Costs.” Report 49831-AFR, World Bank, Washington, DC. World Bank. 2010. “Tanzania at a Glance.” World Bank. 2012. “Tanzania Economic Update, Spreading the Wings: From Growth to Shared Prosperity.” October. World Bank Data Bank. 2013. Tanzania. The World Bank. Accessed 6 Jan 2013. World Development Indicators. 2013. http://data.worldbank.org. World Food Program. 2010. “Comprehensive Food Security and Vulnerability Analysis.” September. Wolter, D. 2008. “Tanzania—Why a Potential Food Exporter Is Still Importing Food.” OECD Development Centre. 40 Tanzania APPENDIX A WEATHER ANALYSIS Tanzania comprises 26 regions and each crop is sown is some regions so some of them may not have available data for all crops. Agricultural information is provided on a regional basis, made up of two variables: sowed area in thousand hectares and produc- tion in thousand tons. Yield is not provided, but can be estimated as follows: Production Yield Area Rainfall patterns in Tanzania Rainfall data were available through a gridded database from the Global Precipita- tion Climate Project (GPCP, http://precip.gsfc.nasa.gov/). The resolution of the grid is 1 degree so there is a pixel point with data from January 1, 1997, to August 31, 2009, for the whole country. Rainfall follows two different patterns in the country. In the northeast and coastal regions, a bimodal rainfall regime with short (vuli) rains from October–December and a long (masika) period of rains from March–May. The following chart has the mean cumulative rainfall per month for pixel #84 on the east coast. In the rest of the country (south and west), a different rainfall pattern is observed. A unimodal (musumi) regime occurs with rainfall from December to April. Figure A.2 illustrates this pattern. It is worth noting that geographical resolution of data is not the same. Rainfall data are available on point estimates whereas yield data are available regionwide, making up the whole political region as described above. Therefore, the geographical reso- lution of both data sets must be made equivalent. Because there is no information regarding the sowing zones within each region, the centroid of each region was con- sidered as the coordinates to relate to the rainfall grid. Figure A.3 shows the centroid for each region. Agricultural Sector Risk Assessment 41 Figure A.1.  Monthly Rainfall Pattern Figure A.3.  Tanzania Region Centroids 0 for Pixel #84 30 32 34 36 38 40 180 Kagera Mara –2 Average cumulative rainfall (mm.) Mwanza 160 Arusha Shinyanga Kilimanjaro 140 Latitude (South) –4 Kigoma Manyara 120 Tabora Tanga Pemba Singida Dodoma Zanzibar 100 –6 Rukva Pwani Dar es Salaam 80 Morogoro Mbeya –8 Iringa 60 Lindi 40 –10 Ruvuma Mtwara 20 0 –12 Longitude (East) ry ry ch ril ay ne Au ly pt st O er r ec er r N obe be Ju Ap Se gu ua ua b b ar M Ju em em em Source: GPCP. n br M ct Ja Fe ov D Month Source: GPCP. Thus, the average of the five nearest pixels can be used as Figure A.2. M  onthly Rainfall Pattern a proxy of region’s rainfall. for Pixel #77 180 Maize Average cumulative rainfall (mm.) 160 140 Maize is grown in most of Tanzania with an average of 120 approximately 2 million hectares sown countrywide. But 100 the area grown was previously less than 2 million hect- 80 60 ares prior to the 2000–01 cycle, when the surface was 40 increased up to a maximum of 5.8 million hectares in the 20 2002–03 cycle. It then decreased to a steady amount of 0 about 3 million hectares a year after that cycle. Figure A.4 y ry ch ril ay ne Au y pt st O er em r ec er r shows the total area and production for each cycle. N obe be ar l Ju Ap Se gu ua b b ar M Ju nu em em br M ct Ja Fe ov D Month National production follows a similar pattern, with an Source: GPCP. increase in the new century, and an average production of 3.8 million tons after year 2000. Figure A.5 shows the To assign rainfall pixels to political regions the distance time series of yield on a national basis. between each pair of centroid (i) and pixel (j) was calcu- lated using the Euclidean Distance Formula, as follows: Yield has oscillated around an average of 1.368 tons per 2 2 hectare, with many years above the average in the 1980s Dist xi xj yi yj and 1990s, but declining yield recently. Year 2002/03 stands out as the worst when yield reached its lowest point where: of 0.593 tons per hectare, representing 43 percent of the Dist = Euclidean Distance mean yield. xi = longitude from region’s i centroid xj = longitude from pixel j Figure A.6 shows the distribution of surface by region yi = latitude from region’s i centroid color coded to identify nearby regions. yj = latitude from pixel j By using this formula and comparing each region’s cen- The regions in which the most maize is sown are Mbeya, troid to all pixels, we can get the five nearest pixels to each Shinyanga, and Iringa, representing 13.6 percent, 12.7 per- region’s centroid (table A.1). cent, and 11 percent of the national surface, respectively. 42 Tanzania Table A.1.  The Five Nearest Pixels to Each Region’s Centroid Region Number Pixel 1 Pixel 2 Pixel 3 Pixel 4 Pixel 5 Arusha 1 135 136 121 149 134 Dar es Salaam 2 82 83 96 97 68 Dodoma 3 93 107 92 94 79 Iringa 4 64 50 65 51 63 Kagera 5 144 145 130 131 143 Kigoma 6 102 116 101 115 103 Kilimanjaro 7 123 122 137 136 109 Lindi 8 53 54 39 40 67 Manyara 9 108 122 107 121 109 Mara 10 147 148 146 133 134 Mbeya 11 62 63 48 49 76 Morogoro 12 66 67 80 81 52 Mtwara 13 26 27 40 41 12 Mwanza 14 132 146 131 145 133 Pemba 15 111 110 97 96 125 Pwani 16 82 81 68 67 96 Rukwa 17 74 75 60 61 88 Ruvuma 18 23 37 24 38 22 Shinyanga 19 118 132 119 133 117 Singida 20 92 91 106 105 78 Tabora 21 104 90 103 89 105 Tanga 22 109 110 95 96 123 Zanzibar 23 96 97 82 83 110 Source: Author. Figure A.4. M  aize Surface Sowed and Figure A.5. M  aize Yield, Time Series Maize Production Volume, Time 2 Series 1.8 Yield (Tons per hectane) Maize 1.6 7,000 Area in 000 hectares / production 1.4 6,000 Surface sowed 1.2 Average yield = 1.368 ton / ha Production 1 5,000 0.8 in 000 tons 4,000 0.6 3,000 0.4 2,000 0.2 0 1,000 81–82 82–83 83–84 84–85 85–86 86–87 87–88 88–89 89–90 90–91 91–92 92–93 93–94 94–95 95–96 96–97 97–98 98–99 99–00 00–01 01–02 02–03 03–04 04–05 05–06 06–07 07–08 08–09 09–10 0 Source: MAFC, author’s calculations. 81-82 82-83 83-84 84-85 85-86 86-87 87-88 88-89 89-90 90-91 91-92 92-93 93-94 94-95 95-96 96-97 97-98 98-99 99-00 00-01 01-02 02-03 03-04 04-05 05-06 06-07 07-08 08-09 09-10 Year / Cycle Source: MAFC. Agricultural Sector Risk Assessment 43 Figure A.6.  Average Maize Surface three stages: a sowing stage from February–March; a by Region in Thousand mid-season stage from April–June; and the harvest stage from July–August. In the south, the calendar follows this Hectares 129 118 pattern: a sowing stage from December–January; a mid- 125 3 86 season stage from February–April; and the harvest stage 59 229 from June–July. 266 64 73 Cumulative rainfall was calculated for each pixel for each 73 stage to determine the relationship between rainfall and 115 44 yield. Although it is known that the first calendar applies 116 190 only to the north-coast regions, it is not clear precisely 127 45 which regions follow it, so regressions were run with both 44 108 286 rainfall patterns against all regions to determine which Arusha Dar es Salaam Dodoma Iringa Kagera regions follow which pattern (table A.2). Kigoma Kilimanjaro Lindi Manyara Mara Mbeya Morogoro Mtwara Mwanza Rukva Ruvuma Shinyanga Singida Taboro Tanga As table A.2 shows, Arusha and Manyara in the northeast Source: MAFC. were the regions in which the first rainfall calendar related more closely. Figure A.8 shows the time series of cumula- tive rainfall for the first stage (February–March) and yield The histogram in figure A.7 shows the distribution of yield for the Arusha and Manyara regions. in all regions. Figure A.8 shows that poor harvest years such as 1997, The regional mean is very similar to the national mean at 2000, 2003, and 2009 also have low cumulative rainfall 1.341 tons per hectare, but the histogram shows that the for the sowing stage, meaning that February–March rain- fifth percentile is 0.49 tons per hectare, meaning that in fall is a good indicator of the yield that can be obtained in 5 percent of the regional cases yield has been even lower the cycle. The charts in figure A.9 show the linear regres- than half a ton per hectare. sion models for each region. Following the rainfall pattern and seasonality. the sow- The charts in figure A.9 confirm that stage 1 (February– ing calendar in the northeast and coastal regions has March) cumulative rainfall explains yield in both regions. Figure A.7.  Maize Yield Histogram for All Regions Maize yield 0.490 2.300 5.0% 90.0% 5.0% 0.8 0.7 0.6 Input 0.5 Minimum 0.0123 Maximum 4.7965 0.4 Mean 1.3413 0.5797 Median = 1.3001 Std dev Values 556 0.3 0.2 0.1 0.0 0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5 Yield (tons/ha) Source: MAFC. 44 Tanzania Table A.2.  Determination Coefficient (R2) of the Linear Regression Models Applied to the First Rainfall Pattern on All Regions Determination Coefficient (R2) Number Region Sowing (%) Mid-Season (%) Harvest (%) 1 Arusha 53 1 1 2 Dar es Salaam 7 1 30 3 Dodoma 6 1 1 4 Iringa 4 3 0 5 Kagera 7 15 6 6 Kigoma 7 0 3 7 Kilimanjaro 2 1 1 8 Lindi 6 25 2 9 Manyara 65 76 17 10 Mara 7 1 2 11 Mbeya 2 3 0 12 Morogoro 5 3 19 13 Mtwara 5 5 1 14 Mwanza 7 0 3 17 Rukwa 0 5 1 18 Ruvuma 0 25 6 19 Shinyanga 1 3 4 20 Singida 10 1 1 21 Tabora 4 3 0 22 Tanga 0 1 14 Source: Author’s calculations. The determination coefficient (R2) is significant, meaning Lastly, for stage 3 the relationship between the harvest that approximately 60 percent of the variation in yield is stage cumulative rainfall (July–August) and yield are not being explained by stage 1 cumulative rainfall. For both significant for both regions because the determination cases, it is clear that drought is the main threat to maize coefficient is very low in both cases (see figure A.11). yield. Years with low cumulative rainfall, such as 1997 and 2000 when 48 mm and 56 mm on average fell over the For the second rainfall pattern, table A.3 summarizes the Arusha region, showed the lowest yield records (129 kg determination coefficient of each stage and region. and 300 kg per hectare, respectively). Table A.3 shows that cumulative rainfall for the sow- The charts in figure A.10 show the relationship between ing stage (December–January) is significant only for the stage 2 cumulative rainfall (April–June) and yield for both Morogoro and Kagera regions but with a very low deter- regions. mination coefficient, meaning that less than 20 percent of variation in yield can be explained by sowing season It can be seen that stage 2 cumulative rainfall is not sig- rainfall. nificant for the Arusha region as the determination coef- ficient is very low; for the Manyara region it is significant, For the mid-season stage, cumulative rainfall is though lower than for stage 1. significant for Manyara and Arusha. This can be Agricultural Sector Risk Assessment 45 Figure A.8. S  owing Season Rainfall and Yield Time Series for Arusha and Manyara Regions 3 Arusha region 250 2 Manyara region 300 1.8 2.5 250 200 1.6 Cumulative rainfall (mm.) Cumulative rainfall (mm.) 1.4 2 200 Yield (Tons/Ha.) Yield (Tons/Ha.) 150 1.2 1.5 1 150 100 0.8 1 100 0.6 50 0.4 0.5 Yield Yield 50 Cumulative rain 0.2 Cumulative rain 0 0 0 0 97 98 99 00 01 02 03 04 05 06 07 08 09 03 04 05 06 07 08 09 19 19 19 20 20 20 20 20 20 20 20 20 20 20 20 20 20 20 20 20 Year Year Source: MAFC, GPCP. Figure A.9.  Linear Regression Models for Arusha and Manyara Regions (stage 1) 3 Yield - Rainfall relationship for Arusha region 2 Yield - Rainfall relationship for Manyara region 1.8 2.5 y = 0.0064x + 0.051 1.6 R 2 = 0.6554 y = 0.0076x + 0.3196 1.4 Yield (Tons/Ha.) Yield (Tons/Ha.) 2 R 2 = 0.5393 1.2 1.5 1 0.8 1 0.6 0.4 0.5 0.2 0 0 0 50 100 150 200 250 0 50 100 150 200 250 300 Stage 1 rainfall (mm.) Stage 1 rainfall (mm.) Source: MAFC, GPCP, author’s calculations. Figure A.10.  Linear Regression Models for Arusha and Manyara Regions (stage 2) Arusha region Manyara region 3 2 y = 0.0004x + 1.3193 1.8 y = 0.0093x + 0.025 2.5 2 R = 0.0024 2 R = 0.5104 1.6 2 1.4 Yield (tons/ha.) Yield (tons/ha.) 1.2 1.5 1 0.8 1 0.6 0.4 0.5 0.2 0 0 0 50 100 150 200 250 300 350 400 0 20 40 60 80 100 120 140 160 180 200 Stage 2 rainfall (mm.) Stage 2 rainfall (mm.) Source: MAFC, GPCP, author’s calculations. 46 Tanzania Figure A.11. L  inear Regression Models for Arusha and Manyara Regions (stage 3) Arusha region Manyara region 3 2 y = 0.0041x + 1.2912 y = 0.0176x + 0.9676 1.8 R 2 = 0.0112 R 2 = 0.2745 2.5 1.6 1.4 2 Yield (tons/ha.) Yield (tons/ha.) 1.2 1.5 1 0.8 1 0.6 0.4 0.5 0.2 0 0 0 10 20 30 40 50 60 70 80 0 5 10 15 20 25 30 35 40 Stage 3 rainfall (mm.) Stage 3 rainfall (mm.) Source: MAFC, GPCP, author’s calculations. Table A.3.  Determination Coefficients for Each Stage and Region Determination Coefficient R2 Number Region Sowing (%) Mid-Season (%) Harvest (%) 1 Arusha 0 40 0 2 Dar es Salaam 5 5 12 3 Dodoma 5 4 6 4 Iringa 8 0 0 5 Kagera 16 23 3 6 Kigoma 9 3 1 7 Kilimanjaro 0 8 2 8 Lindi 2 31 0 9 Manyara 0 53 27 10 Mara 2 18 21 11 Mbeya 4 7 1 12 Morogoro 19 9 4 13 Mtwara 3 1 5 14 Mwanza 5 0 0 17 Rukwa 1 4 0 18 Ruvuma 4 10 0 19 Shinyanga 3 1 11 20 Singida 1 2 0 21 Tabora 7 5 23 22 Tanga 10 0 1 Source: Author’s calculations. explained because the stage runs from February to For the Lindi region, figure A.12 shows significance and two April, almost matching the sowing season of the first data points with low rainfall and low yield corresponding exercise (February–March). But the Lindi, Kagera, to the 2003 and the 2005 cycles in which rainfall was 382 and Mara regions show a significant relationship mm and 380 mm, respectively, while yield was 324 kg and as well. 510 kg, respectively (though a level of 380 mm is hardly an Agricultural Sector Risk Assessment 47 Figure A.12.  Lindi and Kagera Regions Mid-Season Rainfall Models 1.6 2.5 Kagera region Lindi region 1.4 y = 0.0017x + 0.0818 2 1.2 R 2 = 0.3052 y = 0.005x + 2.9155 R 2 = 0.2311 Yield (tons/ha.) Yield (tons/ha.) 1 1.5 0.8 0.6 1 0.4 0.5 0.2 0 0 0 100 200 300 400 500 600 700 0 50 100 150 200 250 300 350 400 450 Stage 2 rainfall (mm.) Stage 2 rainfall (mm.) Source: MAFC, GPCP, author’s calculations. indication of drought. For the Kagera region, the slope of Rice yield has been quite steady, oscillating around the line seems to indicate that excess rainfall affects yield. In an average of 1.65 tons per hectare, the worst seasons 2006, less than 100 kg were obtained with 420 mm of rain- being the 1990–91 cycle when 1.12 tons per hectare fall, the highest amount of rainfall in the region. Still, the were obtained and year 2000–01 when yield was exactly determination coefficient is not very high, indicating that 1 ton per hectare, because 323.500 tons were produced other factors could be the cause of the low yield. in 323,500 hectares (figure A.15 illustrates the yearly national yield). The harvest season has a lower significance, and is only relevant for the Manyara, Tabora, and Mara regions. The main regions where rice is sown are Shinyanga, Figure A.13 shows the linear regression model for yield Morogoro, and Mwanza with 91,500, 74,200, and 62,500 in the Tabora region against harvest season cumulative hectares, respectively, sown on average each year (see rainfall. ­ figure A.16). The chart shows that even though the determination To establish the rainfall-yield relationship, the same algo- coefficient is low, yet significant (23 percent), the slope rithm used for maize will be used for all crops. Rice has is negative indicating that the more rain, the less yield. the following sowing schedule: sowing stage from Janu- Particularly notice year 2005, during which cumulative ary 15 to March 15; mid-season stage from March 16 to rainfall is 56 mm but yield was 466 kg per hectare. This June 30; and a harvest stage from July–August. Table A.4 would seem to indicate that excess rainfall is the main summarizes the regression analysis using the cumulative threat in this region, but the determination coefficient is rainfall of all three stages as regressors against yield. not high enough nor does 56 mm seem to be an indication of excess rainfall. The sowing stage cumulative rainfall explains 72 percent of the variability in the Dar es Salaam region. Figure A.17 Paddy Rice illustrates this relationship. Paddy rice is sown throughout the whole country but the surface has been increasing steadily. The whole country The relationship is quite clear, with a positive slope, mean- sowed less than 500,000 hectares per year in the 1980s. ing that the more rainfall, the better yield. It is worth not- Since the turn of the century, the surface sown has ing that the worst yield was in year 2003 when 396 kg per increased up to a maximum of 1,136,000 hectares in year hectare were obtained; cumulative rainfall for that stage 2010, the most recent data available. Figure A.14 illus- was a mere 29 mm, a clear indication that drought in the trates the yearly increase: sowing season affected yield that year. 48 Tanzania Figure A.13.  Tabora Region Harvest Figure A.15.  Yearly National Paddy Season Model Rice Yield 2 Yield - Rainfall relationship for Tabora region 3 1.8 y = –0.0076x + 1.2984 2.5 1.6 R 2 = 0.2279 Yield (tons per hectane) 1.4 2 Yield (tons/ha.) 1.2 1 1.5 Mean yield = 1.65 0.8 1 0.6 0.4 0.5 0.2 0 0 0 10 20 30 40 50 60 70 80 81–82 82–83 83–84 84–85 85–86 86–87 87–88 88–89 89–90 90–91 91–92 92–93 93–94 94–95 95–96 96–97 97–98 98–99 99–00 00–01 01–02 02–03 03–04 04–05 05–06 06–07 07–08 08–09 09–10 Stage 3 rainfall (mm.) Year Source: MAFC, GPCP, author’s calculations. Source: MAFC, author’s calculations. Figure A.14.  Yearly National Paddy Figure A.16.  Average Distribution of Rice Surface Area Sowed Surface Sown per Region 3.9 and Production 4.0 4.6 5.3 3000 Area 8.0 3.6 7.6 Production 42.9 7.8 2500 8.1 15.0 4.5 4.3 Area / Production 2000 42.6 1500 91.5 1000 500 0 74.2 81–82 82–83 83–84 84–85 85–86 86–87 87–88 88–89 89–90 90–91 91–92 92–93 93–94 94–95 95–96 96–97 97–98 98–99 99–00 00–01 01–02 02–03 03–04 04–05 05–06 06–07 07–08 08–09 09–10 Year 25.2 Source: MAFC. 24.3 23.3 62.5 Arusha Dar es Salaam Dodoma Iringa Kagera For the mid-season stage, only the Kilimanjaro region Kigoma Kilimanjaro Lindi Manyara Mara shows significance, with a much lower determination Mbeya Morogoro Mtwara Mwanza Rukva coefficient of 38 percent (figure A.18). Ruvuma Shinyanga Singida Taboro Tanga Source: MAFC. In this case, the slope is negative; the more cumulative rainfall, the worse the yield, suggesting that excess rainfall is the main threat in this stage. The worst yield year, 2001, coefficients higher than 50 percent. The charts in when only 1 ton per hectare was obtained, was not the figure A.19 illustrate the relationship for these regions. year with most rainfall. In 1998, 383 mm fell and yield was also low at 1.81 tons per hectare, representing about In both cases, the slope is positive, meaning that drought is half the mean yield of the region (3.93 tons per hectare). the main threat. But this stage is the dry season, as shown by the low amounts of rainfall that accumulate in two For the harvest stage, many regions show significance, months (July and August); so rain is hardly expected in but Rukwa and Manyara stand out with a determination this time of the year. Agricultural Sector Risk Assessment 49 Table A.4.  Summary of Rice Regression Analysis Results Determination Coefficient (R2) Number Region Sowing (%) Mid-Season (%) Harvest (%) 1 Arusha 4 8 42 2 Dar es Salaam 72 4 0 3 Dodoma 1 4 0 4 Iringa 11 2 9 5 Kagera 7 3 26 6 Kigoma 4 15 0 7 Kilimanjaro 0 38 0 8 Lindi 13 0 10 9 Manyara 4 17 51 10 Mara 21 5 43 11 Mbeya 0 0 11 12 Morogoro 5 19 19 13 Mtwara 0 18 0 14 Mwanza 25 0 5 17 Rukwa 9 10 57 18 Ruvuma 0 8 0 19 Shinyanga 1 0 0 20 Singida 33 12 29 21 Tabora 18 1 32 22 Tanga 5 1 4 Source: Author’s calculations. Figure A.17. R  elationship between Figure A.18.  Relationship between Rice Yield and Rainfall Rice Yield and Rainfall Variability in Dar es Variability in Kilimanjaro 7 Kilimanjaro rice yield-rainfall relationship Salaam Dar es salaam rice yield-rainfall model 6 1.8 Yield (tons per hectare) 1.6 5 y = 0.0055x + 0.2023 1.4 Yield (tons per hectare) 2 R = 0.7245 4 1.2 1 3 y = –0.012x + 6.3565 2 R = 0.3793 0.8 2 0.6 1 0.4 0 0.2 0 50 100 150 200 250 300 350 400 0 Stage 2 cumulative rainfall (mm.) 0 50 100 150 200 250 Stage 1 cumulative rainfall (mm.) Source: MAFC, GCPC, author’s calculations. Source: MAFC, GCPC, author’s calculations. 50 Tanzania Figure A.19.  Relationship between Rice Yield and Rainfall Variability in Rukwa and Manyara 6 Rukwa rice yield-rainfall model 8 Manyara rice yield-rainfall model 7 y = 0.1078x + 2.2141 5 y = 0.2001x + 1.5147 R 2 = 0.5145 R 2 = 0.574 Yield (tons per hectane) Yield (tons per hectane) 6 4 5 3 4 3 2 2 1 1 0 0 0 2 4 6 8 10 12 14 16 18 0 5 10 15 20 25 30 35 40 45 Stage 3 Cumulative rainfall (mm.) Stage 3 Cumulative rainfall (mm.) Source: MAFC, GCPC, author’s calculations. Cotton Figure A.20.  Cotton Area and Production Cotton is sown in 14 regions of the country. Unfortunately, 700 Area surface sown data are not available from the 1992–93 Area and production in thousand 600 Production cycle to the 2000–01 cycle, although production data are 500 hectares and tons available those years. Figure A.20 shows the time series of area sowed and production for the available years. 400 300 On average, 419,000 hectares are sowed nationally, 200 steadily oscillating between 300,000 and 500,000 hectares. 100 But because sowed area is not available from the years in 0 blue, the mean area was used to estimate the yield. Figure 81–82 82–83 83–84 84–85 85–86 86–87 87–88 88–89 89–90 90–91 91–92 92–93 93–94 94–95 95–96 96–97 97–98 98–99 99–00 00–01 01–02 02–03 03–04 04–05 05–06 06–07 07–08 08–09 09–10 A.21 shows the yield, with the blue line indicating esti- Year mated figures using mean area. Source: MAFC. The chart in figure A.21 shows that yield has not been steady; it was approximately 400 kg per hectare in the 1980s but it almost tripled to 1.5 tons per hectare in 1991– Figure A.21. Cotton Yield 1.6 92. Since 2000, the mean yield is 671 kg per hectare, with 1.4 year 2005–06 almost doubling that mean with 1.18 tons Yield (tons per hectare) 1.2 per hectare. 1 0.8 Cotton areas are concentrated mainly in two regions, Mean yield Estimated figures = 581 kg per ha. 0.6 with Shinyanga (220,000 hectares mean surface sowed) and Mwanza (132,000 hectares mean surface sowed) 0.4 accounting for approximately 85 percent of the total 0.2 surface. 0 81–82 82–83 83–84 84–85 85–86 86–87 87–88 88–89 89–90 90–91 91–92 92–93 93–94 94–95 95–96 96–97 97–98 98–99 99–00 00–01 01–02 02–03 03–04 04–05 05–06 06–07 07–08 08–09 09–10 The sowing schedule for cotton has the following Year stages: sowing season from November 15 to January 31; Source: MAFC, author’s calculations. Agricultural Sector Risk Assessment 51 ­ id-season from February to June 15; and harvest sea- m lower, several data points align perfectly within the line of son from June 16 to August 31. Regression models were the model. The worst yield year (2000) with 239 kilograms run using cumulative rainfall in these stages against yield. per hectare is an estimated figure (area data are not avail- Table A.5 summarizes the results. able for that year), but cumulative rainfall was the lowest for the stage at 356 mm; again, hardly a drought event. It is worth noting that all regions are systematically lack- ing data for surface in the same years as the national data For the harvest season, Iringa and Manyara have high above, so some of the yield values were estimated using significance. The charts in figure A.23 illustrate the the mean surface for the region. ­relationship. For the sowing season, Arusha and Mbeya regions have a Determination coefficients are remarkably high, espe- significant coefficient higher than 40 percent, but surface cially for Iringa, mainly due to the high point in 2004 sowed in those regions is of much significance. when yield was 1.66 tons per hectare, three times as high as the other cycles, whereas rainfall was 32 mm, the high- For the mid-season stage, Iringa has a high coefficient est of all points. A similar pattern occurred in Manyara, of 60 percent, but Mwanza, one of the most important where the outstanding cycle in 2006 (when almost 4 tons regions, also has a significant coefficient. The charts in per hectare were obtained) correlates with a relatively figure A.22 illustrate the model for these two regions. high amount of precipitation (70 mm). Iringa has few data points but rainfall in the low yield years (2001 and 2005, when loss in yield was less than Sorghum 100 kg per hectare cumulative rainfall) was relatively low Sorghum is sown in most of the country. Total surface at about 400 mm for the whole stage, although 400 mm sown increased during the 1990s and has been steady can hardly be considered a catastrophic drought event. since then, up to an average of about 600,000 hectares For Mwanza, even though the determination coefficient is sown nationwide (figure A.24). Table A.5. Cotton Regression Analysis Results Determination Coefficient R2 Number Region Sowing (%) Mid-Season (%) Harvest (%) 1 Arusha 44 0 19 4 Iringa 6 60 84 5 Kagera 3 7 4 6 Kigoma 4 7 0 7 Kilimanjaro 1 1 45 9 Manyara 13 37 74 10 Mara 3 5 14 11 Mbeya 40 36 9 12 Morogoro 6 9 18 14 Mwanza 7 34 4 19 Shinyanga 9 14 0 20 Singida 5 4 3 21 Tabora 8 34 9 22 Tanga 22 1 9 Source: Author’s calculations. 52 Tanzania Figure A.22. L  inear Regression Models for Iringa and Mwanza Regions 1.8 Iringa rainfall - Yield relationship 1.4 Mwanza rainfall - Yield relationship 1.6 y = 0.0049x – 1.9502 1.2 R 2 = 0.5966 y = 0.0021x – 0.497 1.4 Yield (tons per hectare) Yield (tons per hectare) 1 R 2 = 0.3371 1.2 1 0.8 0.8 0.6 0.6 0.4 0.4 0.2 0.2 0 0 0 100 200 300 400 500 600 700 0 100 200 300 400 500 600 700 Stage 2 cumulative rainfall (mm.) Stage 2 cumulative rainfall (mm.) Source: MAFC, GCPC, author’s calculations. Figure A.23.  Linear Regression Models for Iringa and Manyara Regions 1.8 Iringa rainfall - Yield relationship 4.5 Mwanza rainfall - Yield relationship 1.6 4 y = 0.0818x – 1.0921 R 2 = 0.8364 Yield (tons per hectare) 1.4 Yield (tons per hectare) 3.5 y = 0.0608x – 0.9582 1.2 3 R 2 = 0.739 1 2.5 0.8 2 0.6 1.5 0.4 1 0.2 0.5 0 0 0 5 10 15 20 25 30 35 0 10 20 30 40 50 60 70 80 Stage 3 cumulative rainfall (mm.) Stage 3 cumulative rainfall (mm.) Source: MAFC, GCPC, author’s calculations. Production figures have been very similar to surface January; a mid-season from February–April; and a sown; thus, yield has been very steady for an average of harvest season from May–June. The same analysis was 1.02 tons per hectare (figure A.25). performed to establish the rainfall-yield relationship. Table A6 summarizes the determination coefficients As illustrated in figure A.25, yield on a national basis has of each stage. never been too low, as the lowest it has been was in the 1983–84 cycle when yield was 840 kg per hectare, just As shown in table A.6, neither the sowing nor the mid- barely below the mean of 1,020 kg per hectare. But nei- season cumulative rainfall explains a significant amount ther has yield been too high. The best year in terms of of yield; only the harvest season rainfall explains a sig- yield was 2010, when it reached 1.86 tons per hectare. nificant variation of yield for the Manyara and Ruvuma regions. The charts in figure A.26 illustrate these Although sorghum is sown in most of the country, the fol- ­relationships. lowing regions stand out regarding the average surface of sorghum sown: Shinyanga, Dodoma, and Singida with In both cases, the slope is positive, indicating that the more 107,000, 85,800, and 69,900 hectares, respectively. rainfall, the more yield while also showing that drought in this stage could be harmful for the crop. For Manyara, The sowing calendar for sorghum is very similar to the linear relationship is quite clear, although it has fewer that for maize: A sowing season from December–­ data points. However, the worst yield year (2009, with Agricultural Sector Risk Assessment 53 Figure A.24. S  orghum Surface Sown Figure A.25. S  orghum Yield, 1981–82 and Production Volume and 2009–10 Sorghum Sorghum 1,000 2 Area in 000 hectares / Production Surface sowed 900 Production 1.8 Yield (tons per hectare) 800 1.6 700 1.4 Mean yield in 000 tons 600 1.2 = 1.02 ton / ha. 500 1 400 0.8 300 0.6 200 0.4 100 0.2 0 0 81–82 82–83 83–84 84–85 85–86 86–87 87–88 88–89 89–90 90–91 91–92 92–93 93–94 94–95 95–96 96–97 97–98 98–99 99–00 00–01 01–02 02–03 03–04 04–05 05–06 06–07 07–08 08–09 09–10 81–82 82–83 83–84 84–85 85–86 86–87 87–88 88–89 89–90 90–91 91–92 92–93 93–94 94–95 95–96 96–97 97–98 98–99 99–00 00–01 01–02 02–03 03–04 04–05 05–06 06–07 07–08 08–09 09–10 Year / Cycle Year / Cycle Source: MAFC. Source: MAFC, author’s calculations. 765 kg per hectare) matches the lowest amount of rainfall, with only 18 mm. Table A.6. S  orghum Regression For the Ruvuma region, the relationship is not as clear. It Analysis Results is highly influenced by one data point, which is suspect. The average yield in the region is 1.26 tons per hectare Sowing Mid- Harvest Number Region (%) Season (%) (%) but during 2006 the yield was 5.5 tons per hectare, match- ing the year with the most cumulative rainfall in the har- 1 Arusha 2 1 3 vest stage with 71 mm. But it is questionable that rainfall 2 Dar es 3 4 0 in the harvest season can produce such a high yield rela- Salaam tive to other years. 3 Dodoma 23 2 8 4 Iringa 9 11 0 5 Kagera 22 12 0 Millet 6 Kigoma 0 1 4 Millet is also sown in many regions of the country, with 7 Kilimanjaro 3 5 0 an average surface sown of 271,000 hectares. Figure A.27 8 Lindi 1 3 0 illustrates the surface and total production for each year. 9 Manyara 6 2 80 10 Mara 20 6 12 Apparently there are data missing for year 1990–91, 11 Mbeya 19 0 9 whereas in the mid-1990 decade, surface reached its max- 12 Morogoro 1 10 18 imum of 473,000 hectares sown, but has not reached that 13 Mtwara 3 4 0 level again. 14 Mwanza 0 8 18 17 Rukwa 1 6 0 Yield has been steady as well as for sorghum, oscillating 18 Ruvuma 21 1 50 around 600 kg and 1,200 kg per hectare, with a mean 19 Shinyanga 4 14 0 yield of 874 kg per hectare (see figure A.28). 20 Singida 2 6 6 21 Tabora 2 11 13 One region stands out as the most important in terms 22 Tanga 2 1 16 of surface. Dodoma has recently sown the most surface Source: Author’s calculations. in the country with a yearly average of 90,000 hectares, 54 Tanzania Figure A.26. L  inear Regression Models for Manyara and Ruvuma Regions 6 Ruvuma region yield-Rainfall relationship 2 Manyara region yield-Rainfall relationship 1.8 y = 0.0085x – 0.6923 y = 0.0548x – 0.5543 5 1.6 R 2 = 0.8005 R 2 = 0.5005 Yield (tons per hectare) Yield (tons per hectare) 1.4 4 1.2 1 3 0.8 2 0.6 0.4 1 0.2 0 0 0 20 40 60 80 100 120 0 10 20 30 40 50 60 70 80 Stage 3 cumulative rainfall (mm.) Stage 3 cumulative rainfall (mm.) Source: MAFC, GCPC, author’s calculations. Figure A.27.  Millet Surface Sowed Figure A.28. M  illet Yield 1981–2010 Millet and Production Volume 1.4 Millet 700 Surface sowed Area in 000 hectares / Production 1.2 Production Yield (tons per hectare) 600 1 500 0.8 Mean yield in 000 tons 400 = 0.874 ton / ha. 0.6 300 0.4 200 0.2 100 0 81–82 82–83 83–84 84–85 85–86 86–87 87–88 88–89 89–90 90–91 91–92 92–93 93–94 94–95 95–96 96–97 97–98 98–99 99–00 00–01 01–02 02–03 03–04 04–05 05–06 06–07 07–08 08–09 09–10 0 81–82 82–83 83–84 84–85 85–86 86–87 87–88 88–89 89–90 90–91 91–92 92–93 93–94 94–95 95–96 96–97 97–98 98–99 99–00 00–01 01–02 02–03 03–04 04–05 05–06 06–07 07–08 08–09 09–10 Year / Cycle Year / Cycle Source: MAFC, author’s calculations. Source: MAFC. roughly representing 33 percent of the total surface sown. The chart in figure A.29 shows a negative slope, which Rukwa and Shinyanga follow Dodoma with approxi- indicates that the more rainfall, the less yield; it also sig- mately 30,000 hectares sown on average each. nals that excess rainfall during this stage affects yield. Two particularly poor years stand out: 2006, with zero yield The sowing calendar for millet follows the masika rainfall and 251 mm rainfall, and 2009, with 150 kg per hectare pattern, with a sowing season from February–March; a and 248 mm rainfall. But it is also worth noting that the mid-season from April–June; and the harvest season of year with the highest cumulative rainfall (2007) with 269 July–August. A similar regression analysis was performed mm does not have such a low yield. using cumulative rainfall during these stages. Table A.7 summarizes the determination coefficient of each stage. For the mid-season stage, rain only shows significance for the Tabora region (figure A.30). Apparently Lindi has a very good fit (100 percent), but only because there are just two available years of data in As with the chart for Kagera, the slope is negative, signaling this region, thus forcing a line with perfect fit. As such, the an excess rainfall problem. The worst years in terms of yield result should be disregarded. For the sowing season, Kag- (2003 and 1997, with less than 100 kg per hectare) match era has a significant relationship (figure A.29). with high precipitation (148 mm and 194 mm, respectively). Agricultural Sector Risk Assessment 55 Table A.7.  Millet Regression Analysis Figure A.30. L  inear Regression Model Results for the Tabora Region 1.4 Tabora millet yield-rainfall relationship Sowing Mid- Harvest Number Region (%) Season (%) (%) 1.2 y = –0.0048x + 1.2503 Yield (tons per hectare) 1 Arusha 8 0 20 1 R 2 = 0.351 3 Dodoma 5 9 14 0.8 4 Iringa 9 1 4 0.6 5 Kagera 45 12 6 6 Kigoma 24 1 9 0.4 7 Kilimanjaro 3 17 0 0.2 8 Lindi 100 100 100 0 10 Mara 3 7 25 0 50 100 150 200 250 Stage 1 cumulative rainfall (mm.) 11 Mbeya 25 0 1 Source: MAFC, GCPC, author’s calculations. 12 Morogoro 1 6 10 13 Mtwara 8 12 0 14 Mwanza 1 0 14 Figure A.31.  Tobacco Surface Sowed 17 Rukwa 5 6 0 18 Ruvuma 10 0 9 and Production Volume 140 Surface sowed 19 Shinyanga 8 0 0 Production 120 20 Singida 3 0 18 Production in 000 tons Area in 000 hectares / 100 21 Tabora 12 35 2 80 Source: Author’s calculations. 60 40 Figure A.29.  Linear Regression Model 20 for the Kagera Region 1.4 Kagera millet yield-rainfall relationship 0 05–06 06–07 07–08 08–09 09–10 1.2 Year / Cycle Yield (tons per hectare) Source: MAFC. 1 y = –0.0068x + 2.1245 R 2 = 0.4486 0.8 was sown with 116,000 hectares. Three regions make up 0.6 72 percent of the surface sown: Tabora, Ruvuma, and 0.4 Shinyanga, with 30,000, 13,000 and 10,000 hectares 0.2 sown on average, respectively, each year. 0 0 50 100 150 200 250 300 The sowing calendar for tobacco runs December–­ January Stage 1 cummulative rainfall (mm.) for the sowing season, February–April for the mid-season, Source: MAFC, GCPC, author’s calculations. and May–June for the harvest season, which will be con- sidered as stages one through three for the rainfall analy- Tobacco sis. Table A.8 summarizes the determination coefficient of the regression analysis performed. Tobacco is sown mostly in the southern highlands with an average of 74,700 hectares sown each year nationwide As a general note, coefficients were higher because of the (figure A.31). low number of observations available; there were only Unfortunately, data are only available for five cycles, from four data points with which to run regressions, so the 2005 to 2010. In 2006–07, the highest amount of surface coefficients should be viewed with care. For the sowing 56 Tanzania Table A.8.  Tobacco Linear Regression For the mid-season rainfall Iringa, Kagera, and Ruvuma Results have a high coefficient. Figure A.33 shows the relation- ship for Ruvuma, because it is one of the most important Sowing Mid- Harvest Number Region (%) Season (%) (%) tobacco regions. 1 Arusha 11 3 91 Again, the slope is negative, clearly indicating that the 4 Iringa 77 77 3 higher the rainfall, the lower the yield. But rainfall in this 5 Kagera 49 76 0 stage had low variability in the four years considered, 6 Kigoma 53 38 70 varying from 535 mm to 600 mm. Coincidentally, in 2007 11 Mbeya 74 28 63 when 600 mm fell, the lowest yield was observed at 309 kg 17 Rukwa 95 40 67 per hectare. In 2006, 580 mm fell and almost 500 kg per 18 Ruvuma 8 67 21 hectare were recorded. This explains why the determina- 19 Shinyanga 21 37 21 tion coefficient is so high, but again, only four data points 20 Singida 14 8 20 were considered. A larger sample size should be used to 21 Tabora 26 21 7 draw more solid conclusions. Source: Author’s calculations. For the harvest stage, Arusha and Kigoma were signifi- cant. Figure A.34 shows the relationship for Kigoma. season, Rukwa, Iringa, and Mbeya stand out with high coefficients. Figure A.32 illustrates the model for Mbeya Again, a negative slope indicates excess rainfall. In par- and Rukwa. ticular, the worst yield year—2006, when only 650 kg per hectare was recorded—matches the year with most rain- The slope is negative in both charts, indicating that fall, with 83 mm. excess rainfall is the main threat. For Mbeya, however, the slope is not significantly different from zero, mean- ing that despite the fact that the R2 is high, rainfall does Maize (based on a specific not influence yield but for Rukwa the model resembles the relationship quite well. It should be noted that only calendar for every four observations are considered. The lowest produc- region) tion year—2007, the only one with less than 1 ton per Because planting and harvesting take place in different hectare—matches the highest accumulated rainfall with times throughout the year, the specific calendar for each 469 mm. region was used (table A.9). Figure A.32. L  inear Regression Models for Mbeya and Rukwa Regions 1.6 Mbeya tobacco yield-Rainfall relationship 1.8 Rukwa tobacco yield-Rainfall relationship 1.4 1.6 1.4 Yield (tons per hectare) Yield (tons per hectare) 1.2 1.2 1 1 0.8 y = –0.0007x + 1.4216 R 2 = 0.7443 0.8 0.6 0.6 y = –0.0033x + 2.411 2 0.4 R = 0.9517 0.4 0.2 0.2 0 0 0 100 200 300 400 500 600 0 50 100 150 200 250 300 350 400 450 500 Stage 1 cumulative rainfall (mm.) Stage 1 cumulative rainfall (mm.) Source: MAFC, GCPC, author’s calculations. Agricultural Sector Risk Assessment 57 Figure A.33. L  inear Regression Model Figure A.34. L  inear Regression Models for the Ruvuma Region for the Kigoma Region 1.2 Ruvuma tobacco yield-rainfall relationship 1.2 Kigoma tobacco yield-rainfall relationship 1 1 y = –0.0088x + 5.6268 Yield (tons per hectane) Yield (tons per hectare) 2 R = 0.6694 0.8 0.8 0.6 0.6 y = –0.0054x + 1.1847 R 2 = 0.7003 0.4 0.4 0.2 0.2 0 0 0 100 200 300 400 500 600 0 10 20 30 40 50 60 70 80 90 Stage 2 cumulative rainfall (mm.) Stage 3 cumulative rainfall (mm.) Source: MAFC, GCPC, author’s calculations. Source: MAFC, GCPC, author’s calculations. Table A.9. Maize Sowing Calendar Region Sowing Mid-Season Harvesting Kilimanjaro February–March April–July August Arusha February–March April–July August Manyara February–March April–July August Kagera November–December January–February March Mwanza November–December January–March April Kigoma November–December January–March April Tabora November–December January–March April Shinyanga November–December January–March April Singida November–December January–March April Dodoma November–December January–March April Morogoro February–March April–June July–August Rukwa November–December January–June July–August Mbeya November–December January–June July–August Iringa November–December January–June July–August Pwani March April–June July–August Dar es Salaam March April–June July–August Lindi March April–June July–August Mtwara March April–June July–August Ruvuma March April–June July–August Tanga March April–June July–August Source: MAFC. Linear regression models were built to establish the rela- Table A.10 summarizes the determination coefficient (R2) tionship between yield, expressed in tons per hectare, and obtained for each region. the cumulative rainfall of each of the crop seasons above. The model can be expressed as follows: The determination coefficient is a measure of the pro- portion of the variance in yield that can be explained by Yield = b0 + b1Raini the cumulative rainfall in each season. So, for instance, in 58 Tanzania Table A.10. M  aize Regression Analysis Figure A.35. L  inear Regression Model Results for the Arusha Region 3 Arusha region Sowing Mid-Season Harvest Region (%) (%) (%) 2.5 y = 0.009x + 0.2306 R 2 = 0.5353 Yield (tons per hectare) Arusha 54 1 1 2 Dar es Salaam 12 2 15 Dodoma 0 33 0 1.5 Iringa 6 14 0 1 Kagera 9 6 10 Kigoma 2 0 2 0.5 Kilimanjaro 2 1 1 0 Lindi 2 23 10 0 50 100 150 200 250 Cumulative rainfall (mm.) Manyara 72 75 8 Source: MAFC, GCPC, author’s calculations. Mbeya 1 29 0 Morogoro 6 3 17 Mtwara 7 6 0 Mwanza 2 0 0 Figure A.36. L  inear Regression Model Coast 2 3 8 for the Manyara Region Rukwa 1 9 1 2 Manyara region Ruvuma 3 24 12 1.8 1.6 y = 0.0071x + 0.0482 Shinyanga 1 7 0 R 2 = 0.723 Yield (tons per hectare) 1.4 Singida 3 14 1 1.2 Tabora 15 2 1 1 Tanga 48 1 14 0.8 Source: Author’s calculations. 0.6 0.4 0.2 0 the Manyara region, both the sowing and the mid-seasons 0 50 100 150 200 250 300 Cumulative rainfall (mm.) explain a significant amount of the variability in yield. Source: MAFC, GCPC, author’s calculations. For the sowing season, cumulative rainfall shows signifi- cance for three regions: Arusha, Manyara, and Tanga. The regression charts for these regions follow in figures Figure A.37.  Linear Regression Model A.35, A.36, and A.37. for the Tanga Region 2.5 Tanga region For the Arusha region, the relationship is quite clear, with 2 a determination coefficient (R2) of 54 percent, meaning Yield (tons per hectare) that 54 percent of the variability in yield can be explained y = 0.0098x + 0.4722 1.5 by the cumulative rainfall of the sowing season alone. The R 2 = 0.4833 slope is positive; the more rain, the higher the yield, sig- 1 naling that drought is the main threat here. 0.5 It is also clear that the worst years in terms of rainfall, which were 1997 and 2000, when only 44 mm and 55 mm 0 0 20 40 60 80 100 120 140 160 fell through each entire season, were also the worst years Cumulative rainfall (mm.) in terms of yield with 129 kg and 300 kg per ­ hectare, Source: MAFC, GCPC, author’s calculations. Agricultural Sector Risk Assessment 59 Figure A.38. L  inear Regression Model Figure A.39.  Linear Regression Model for the Dodoma Region for the Manyara Region 1.8 Dodoma region 2.5 Manyara region 1.6 y = 0.0035 x – 0.3913 R 2 = 0.3294 y = 0.0103x – 0.3004 1.4 R 2 = 0.7477 Yield (tons per hectare) 2 Yield (tons per hectare) 1.2 1 1.5 0.8 1 0.6 0.4 0.5 0.2 0 0 0 100 200 300 400 500 600 0 50 100 150 200 250 Cumulative rainfall (mm.) Cumulative rainfall (mm.) Source: MAFC, GCPC, author’s calculations. Source: MAFC, GCPC, author’s calculations. respectively. So it is clear that drought in the sowing sea- As for the midseason, rainfall explained the variability in son has an important effect in maize yield in the Arusha yield only in Dodoma, Manyara, Mbeya, and Ruvuma region. (figures A.38, A.39, A.40, and A.41). For the Manyara region, the determination coefficient is For Dodoma, even though the determination coefficient even higher, which means that roughly 72 percent of the is not very high (33 percent), it is clear that there are two variance in yield can be explained by cumulative rainfall different groups of points: those with low rainfall and low in the sowing season alone. yield, and the opposite group. The slope is also positive, which signals that drought is But in this case, the worst rainy year (1997 at 223 mm) is also the main threat here. It is also clear that the worst not the worst yield year (642 kg/ha). There are observa- year in terms of rainfall, which was 2009 when 113 mm tions with lower yield, for example, 2003 and 2005 with fell, was also the worst year in terms of yield, with only approximately 400 kg per hectare, respectively, but with 626 kg per hectare. But it seems that drought has not been cumulative rainfall of 297 mm and 335 mm. This season very frequent in this region, because the lowest rainfall is long for Dodoma (at three months) but 300 mm would in this region was 113 mm. This also explains why yield not be considered a drought event. has not been lower than 626 kg per hectare in the years observed. For the Manyara region, the mid-season cumulative rain- fall was also significant in explaining variability in yield. The same pattern is observed in the Tanga region. Almost It was the only region in which both the sowing and the 50 percent of the variability in yield can be explained by mid-season were important, although the high R2 can be cumulative rainfall in the sowing season alone. The slope explained because this is also the region with fewest obser- is also positive, signaling a drought effect. vations available. The year with least rain, 2004, when only 57 mm fell, matches the lowest yield with 674 kg per hectare. Again, the slope is positive, implying that drought is the main threat. Clearly the worst year (2009) when only 626 It can be concluded that for Arusha, Manyara, and kilograms per hectare were recorded, was the year with Tanga, cumulative rainfall in the sowing season is a very the lowest rainfall (114 mm). Hence, Manyara is suscep- important driver for yield, and that drought can threaten tible to drought not only in the sowing season, but also in the yield obtained. the mid-season. 60 Tanzania Figure A.40.  Linear Regression Model Figure A.41.  Linear Regression Model for the Mbeya Region for the Ruvuma Region 2.5 Mbeya region 2.5 Ruvuma region y = 0.0036x – 0.6268 y = 0.0024x – 1.3251 2 R 2 = 0.2853 2 R 2 = 0.2449 Yield (tons per hectare) Yield (tons per hectare) 1.5 1.5 1 1 0.5 0.5 0 0 0 100 200 300 400 500 600 700 800 0 50 100 150 200 250 300 Cumulative rainfall (mm.) Cumulative rainfall (mm.) Source: MAFC, GCPC, author’s calculations. Source: MAFC, GCPC, author’s calculations. In Mbyea, the determination coefficient is not very high For Ruvuma, not only is the determination coefficient (29 percent), but the relationship is skewed by one extreme low (24 percent) but also it is clear that the slope is almost observation: during 2003, the yield was only 181 kg per equal to zero, signaling that rainfall is not significant in hectare. Although rain was relatively low, compared with this region. Rainfall was not significant for any region in how much rain falls regularly in this region, 472 mm the harvest season, because the highest determination would not be considered the cause for such a low yield. coefficient was 17 percent for Morogoro. There were other years with similar amounts of rain, but yields were higher. Agricultural Sector Risk Assessment 61 APPENDIX B IMPACT OF CLIMATE CHANGE ON AGRICULTURAL SECTOR Introduction Agriculture is highly vulnerable to climate change in Tanzania, although the effects are heterogeneous across regions and crops/livestock. Some 80 percent of the popula- tion is involved in agriculture (CIA Fact Book 2013), and the majority of those on an informal, small-scale nature without many chemicals or mechanizations. Agriculture composes almost 28 percent of Tanzania’s gross domestic product (GDP) (CIA Fact Book 2013). If climate change is left unaddressed, progress in agricultural develop- ment, food security, and poverty alleviation in general will be reversed. In the Mapping the Impacts of Climate Change index under “Agricultural Productivity Loss,” the Center for Global Development ranks Tanzania 68 out of 233 countries globally for “direct risks” due to “physical climate impacts” and 33 out of 233 for “overall vulnerability” due to “physical impacts adjusted for coping ability” (Wheeler 2011). However, the impacts of climate change vary widely based on what assumptions are made, and which scenarios are played out. There are direct impacts, such as changes in crop yields due to precipitation changes, and indirect impacts, such as rising food prices due to production changes and conflict over land tenure based on shifting agro- climatic zones. The newest installment of the International Panel for Climate Change (IPCC) did not narrow expected results from climate change, but rather widened the frame of variability. This, in combination with various approaches to impact studies, makes it difficult to generalize regarding the effects of climate change on agriculture in Tanzania. This appendix discusses the various possible outcomes. Principal Findings As Tanzania is highly dependent on rain-fed agriculture, shifts in precipitation and temperature patterns due to climate change will have significant impacts on the sector. Agricultural Sector Risk Assessment 63 A high level of variability both geographically and Three other assessments have been completed recently. between scenarios makes it difficult to generalize about In 2009, Review of Development Economics published the impact of climate change. Because of high levels of an article, “Agriculture and Trade Opportunities for variation, there is a need for subnational assessments, Tanzania: Past Volatility and Future Climate Change,” particularly for the design of climate change policy supported by Stanford University. In September 2012, responses. the World Bank published the working paper, “Climate Change, Agriculture and Food Security in Tanzania,” The literature generally agrees that climate change will which was a continuation of the 2009 United Nations have negative impacts on key food crops for domestic con- University Working Paper. Finally, in December 2012, the sumption, such as maize, while it may not have significant International Food Policy Research Institute (in collabo- impact or possibly have positive impacts on cash crops ration with the Association for Strengthening Agriculture such as coffee and cotton. Research in Eastern and Central Africa, and the Con- sultative Group on International Agricultural Research Although there is regional variability (and possible [CGIAR] Research Program on Climate Change, Agri- increases in some areas), on a national average, maize culture and Food Security) released a summary of their yields are likely to decrease. upcoming publication, “East African Agriculture and Cli- mate Change: A Comprehensive Analysis—Tanzania.” Cotton will not be affected by changes in temperature, but changes in precipitation may affect yields. Throughout all of these studies, most data and models came from the United Kingdom, the United States, or Agriculture and food security, including livestock, is listed Canada. The older assessments rely primarily on agri- in the National Adaptation Program of Action (NAPA) cultural crop models, while some of the newer assess- as the most important sector to address in climate change ments seek to quantify impacts through economic growth adaptation. accounting or other various models. In most analyses, projections are made through the mid-21st century. Brief History of Climate Change Impact Methodologies Assessments The INC and NAPA used the assessments supported by the United States Country Studies Program, global envi- Many climate change agricultural impact assessments ronmental facility (GEF), and United Nations Environ- have been done at the regional (eastern and Sub-Saharan ment Program (UNEP) that were carried out by CEEST. Africa) and global levels; however, there are few specific The scenarios used were developed from the General analyses of Tanzania at a national or subnational level. In Circulations Models (GCMs).21 The base climate data 1994, the United States Country Studies Program (in part- came from 1951–1980, and were used to create 30-year nership with of the Global Environmental Facility and the climate scenarios. The scenarios projected an increase United Nations Environmental Programme) supported in the mean daily temperatures by 3.5°C. One scenario a vulnerability and climate change impact study (called doubles CO2 concentration, resulting in an annual tem- The National Vulnerability and Adaptation Assessment perature increase of 2.1°C in the northeast to 4°C in of Tanzania). It was conducted by the Centre for Energy, the central and western areas. Under these scenarios, the Environment, Science and Technology (CEEST), and bimodal rainfall areas (the northeast and northwest, the results were published in 1997. These studies were the Lake Victoria basin, and the northern part of the coastal basis for the government of Tanzania’s Initial National belt) would see rainfall increases for both seasons from Communication (INC) to the United Nations Framework 5 to 45 percent. The unimodal rainfall areas (the south, Convention on Climate Change (UNFCCC) in 2003 and the National Adaptation Programme of Action to the UNFCC in 2007. 21 In particular from UK 89, CCCM, GFD3, GFDLOI, and GISS. 64 Tanzania southwest, west, central, and east) would see decreases in patterns, food prices, calorie consumption, and child mal- annual rainfall of 5 to 15 percent with greater volume or nutrition.” They used an optimistic scenario with high per rain during the long rains, and less during the short rains. capita income growth and low population growth, and Rain is expected to increase by 5 to 45 percent annually corresponding pessimistic and intermediate scenarios. in the southeast. In looking at particular crops, relevant regression models were used for cotton and coffee, but the The World Bank analysis found four limitations in pre- Crop Environment Resource Synthesis (CERES) model vious studies: (1) Most assessments are conducted at a and GCMs were used for maize. global/regional level, and more information is needed at national and subnational levels because impacts vary In “Agriculture and Trade Opportunities for Tanza- widely geographically; (2) many assessments rely on only nia: Past Volatility and Future Climate Change,” the a few projections despite the great uncertainty in climate authors used the Coupled Model Intercomparison Project change; (3) autonomous adaptation, which may offset (CMIP3). This is the same model which Working Group I damage due to climate change, is not included in cali- of the Fourth Assessment Report of the IPCC used for the brated agronomic crop models; and (4) assessments may 2007 publication (Ahmed and others 2012). exclude indirect and general equilibrium effects (such as household income changes, price, and inter-sector link- The International Food Policy Research Institute (IFPRI) ages) (Arndt and others 2012). analysis took four downscaled global climate models (GCMs) from the IPCC AR4 and projected agricultural In the World Bank study, agricultural production changes yields out to 2050 (IFPRI 2012). The scenarios modeled brought about because of climate change on a subna- include changes in precipitation and temperature. The tional level are projected using four GCMs through 2050. Model for Interdisciplinary Research on Climate (MIROC) The assessment takes the climate projections and inserts model is the wettest, with increased rainfall across the them into calibrated crop models that predict changes in country, and a median 200 mm increase in precipitation yield, and then imposes them on a “highly-disaggregated, per year (with some areas seeing a 300 mm increase.) The recursive dynamic economywide model of Tanzania.” Commonwealth Scientific and Industrial Research Organ- The economic model thereby evaluates both availability isation (CSIRO) model did not note significant precipita- (production) and accessibility (income) as crucial com- tion change over the majority of Tanzania (60 percent), but ponents of food security. All four scenarios assume an projected increases over the east of 50 to 100 mm (IFPRI increase in temperature between 1 and 2 percent. HOT 2012). All four models projected higher temperatures by projects a 5.67 percent increase in precipitation; COOL, 2050, with the lowest median temperature increases of a 5.37 percent increase; WET, a 13.3 percent increase; around 1°C in the CSIRO and MIROC models, and 2°C and DRY, an 11.14 percent decrease (Arndt and others or higher in the other models (MIROC showed spatial vari- 2012). The model captures indirect effects and allows for ability.) The report notes that temperature increases could some autonomous adaptation (Arndt and others 2012). have negative consequences for agricultural productivity owing to the spread of diseases and crop pests. The generic crop model the World Bank study uses is called CLICROP. It simulates the impacts of climate The IFPRI analysis then used the Decision Support Sys- change on rain-fed and irrigated crops as well as on tem for Agrotechnology Transfer (DSSAT) crop model- demand for irrigation water. CLICROP has a daily time ing software projections for rain-fed maize and compared scale and includes both water-logging and crop-specific 2000 crop yields with projected 2050 crop yields that parameters. The FAO model CROPWAT is a simpler resulted from climate change (IFPRI 2012). IFPRI also (earlier) such model. CLICROP indirectly measures the ran the IMPACT global model for food and agriculture effects of the atmosphere through evatranspiration and “to estimate the impact of future GDP and population sce- infiltration to the soil layers. Fertilization via CO2 is not narios on crop production and staple consumption, which considered in this analysis and therefore yield losses may can be used to derive commodity prices, agricultural trade be overestimated. Within CLICROP, they used four Agricultural Sector Risk Assessment 65 ­cenarios/projections: COOL, WET, HOT, and DRY. s base, and environmental degradation (Republic of Tan- The CLICROP analysis was performed for nine crops zania 2007). (cassava, groundnuts, maize, millet, potatoes, sorghum, soybeans, sweet potatoes, and wheat), with a focus on The Tanzania NAPA listed the following additional maize as a principle food crop (Arndt and others 2012). vulnerabilities in the agricultural sector due to climate Similar to other models, there is a considerable degree of change: (1) unpredictable rainfall, resulting in cropping variability both between the four climate scenarios, and pattern uncertainty; (2) prolonged dry spells leading to across the subnational regions. drought; (3) increased competition between weeds and crops for moisture, light, and nutrients; (4) ecological Finally, the Haggar and Schepp desk study of coffee changes in pests and diseases; and (5) vulnerability in the builds upon the INC, but is a valuable addition because agriculture/livestock sector (Republic of Tanzania 2007). it takes into account results from farmer surveys and also addresses the impact of El Niño and La Niña The World Bank study found that, relative to a no climate cycles. change baseline with the principal impact channel being domestic agricultural production, “food security in Tan- General Findings zania appears likely to deteriorate as a consequence of climate change.” It also found significant impact differ- There were several confluences of results throughout ences by region, income category, and across households. these assessments. All of the assessments agreed that dependency on rain-fed agriculture in Tanzania made The Review of Development Economics article found that it acutely vulnerable to climate change. The assessments more than 50 percent of Tanzania’s dry years might coin- all reiterated impact variations across geographical areas, cide with nondry years in selected African trading part- between models/scenarios, and among agricultural sec- ners between the early 2000s through the 2050s (Ahmed tors. There was a wide consensus that maize yields would and others 2012). The article goes on to suggest that there generally decline. is great potential for Tanzania to benefit from the hetero- geneous climate impacts on agriculture. It notes that these Several generalizations are made about climate change benefits can only be realized through a removal of export impacts. First, increased rainfall leads to nutrient leach- restrictions or movement to a rules-based policy mecha- ing, topsoil erosion, and water logging, thereby affecting nism. These steps will remove policy uncertainty and the plant growth and yield. Second, climate change will favor resultant price instability (Ahmed and others 2012). pests and diseases because of increased temperature and rainfall. Farmers will therefore be inclined to use costly agrochemicals and disease resistant cultivars, placing vul- Cotton nerable and poor small-scale farmers at a disadvantage. The only study that directly addresses cotton is the Third, agro-climatic zones will shift, and areas with less National Vulnerability and Adaptation Assessment of rainfall will require irrigation (which is costly because of Tanzania, which is referenced in the INC, the NAPA, reduced river runoff and shallow well vulnerability) and and various other climate change impact-related docu- drought resistant plant varieties (Republic of Tanzania ments. The study assessed the impacts of climate change 2003). on cotton, using relevant regression models, finding no significant impact on cotton growth due to temperature. The NAPA notes that in Tanzania there would be a However, with increased rainfall, yield will rise by 17 shift from perennial crops to annual crops, and global percent whereas decreased precipitation will result in a warming that accelerated plant growth would reduce the 17 percent yield drop. In the studied areas (Mwanza and length of growing seasons (Republic of Tanzania 2007). Morogoro regions), rainfall is projected to increase by Agricultural vulnerabilities include: decreased crop pro- 37 percent and 7 percent respectively. With a doubling of duction exacerbated by climatic variability and unpre- the CO2 levels, the average temperature increase would dictability of seasonality, erosion of natural resource be 2.7°C, which still falls in optimal cotton conditions 66 Tanzania of 18°C to 30°C. Pests and disease are a side effect of 15 percent decreases) (Haggar and Schepp 2009). As with increased rainfall that may adversely affect production cotton and coffee production, increased temperatures and (Republic of Tanzania 2003). rainfall would increase pest and disease incidence, nega- tively affecting production as well. For both cotton and coffee, the NAPA notes a projected increase by 18 percent in bimodal rainfall areas and In the article “Robust Negative Impacts of Climate 16 percent in unimodal rainfall areas due to a 2°C to 4°C Change on African Agriculture,” Schlenker and Lobell increase in temperature (Republic of Tanzania 2007). are cited as suggesting that there will be a 22 percent The NAPA also suggested that cotton yields could be neg- decline in average maize productivity across Sub-Saharan atively affected by pests and diseases, resulting in a 10 to Africa by 2050 (Schlenker and Lobell 2010). The article 20 percent loss. goes on to argue that if global and regional maize produc- tion and supply are low, Tanzania can take advantage of Maize high prices, even if producing at a rate below trend. A his- Maize is the most important staple food in East Africa, torical analysis suggests that Tanzania may be only mildly and the most widely-traded agricultural commod- affected by dry conditions while its major trading partners ity (World Bank 2009b). Similarly, maize is the primary are severely affected, giving them a comparative advan- staple crop in Tanzania, and is greatly important to food tage in exports. As noted previously, these advantages can- security. There is broad agreement that maize production not be realized without a removal of export restrictions will be adversely affected by climate change. Further, it and other policy measures (Ahmed and others 2012). appears that poor producers will be particularly affected, as they may not be able to afford the required cost of irri- The World Bank models found that there were heteroge- gation, varieties, or chemicals needed to adapt. As noted neous impacts geographically and between scenarios, but in the World Bank assessment, in regards to food avail- there will be regional correlations; essentially, favorable cli- ability, yield impacts on the major producing areas should mate outcomes for maize farmers in a specific region will be examined. The future of maize as a staple crop and likely favor farmers in neighboring regions. The same can continued reliance upon it may be at risk (Haggar and be said of unfavorable impacts. In general, yield declines Schepp 2009). are more prevalent than yield increases throughout the scenarios and across regions. Also the coastal islands The INC and NAPA (as a result of the CEEST study) generally appear to remain fairly unaffected (Arndt and reported that increases in temperature and reduced rain- ­others 2012). fall would lead to increased moisture loss and a reduced growth period thereby affecting maize growth and yields. In the WET scenario, maize yields in the Northern Using the CERES model and GCMs, the projections Zone increase substantially. Maize yields are projected suggest that farmers may move away from corn produc- to increase by 15 percent in Manayara in the North- tion because of lack of control over temperature, and the ern Zone, but to decline by 12 percent in Tabora in added cost of irrigation to supplement rainfall. Using the the Central Zone. The WET scenario also saw mean CERES maize model, maize yields will be lower than increased yields near Mount Kilimanjaro and its under a baseline climate projection by about 33 percent southern slope, with very significant decreases in the across the country. This varies across regions; the cen- west near Lake Tanganyika. In the COOL scenario, tral regions (Dodoma and Tabora) (Republic of Tanza- yield increases in the Northern Zone (around Mount nia 2007) would see a projected 84 percent production Kilimanjaro), but results in slight yield declines in the decrease, with a 22 percent decrease in the northeastern southern coast and southern highlands (Arndt and highlands, a 17 percent decrease in the Lake Victoria others 2012). basin, and a 12 percent decrease in the southern highlands (Republic of Tanzania 2003) (or rather, in the southern Maize yields are generally more favorable under WET highlands, Mbeya and Songea were estimated to see 10 to and COOL, than HOT and DRY scenarios. Maize yields Agricultural Sector Risk Assessment 67 Figure B.1.  Mean Annual Dry-Land Maize Yield Changes under HOT, COOL, WET, and DRY Scenarios, 2041–50 HOT scenario WET scenario –27% to –9% –19% to –13% –8% to –3% –12% to –3% –2% to +2% –2% to +2% +3% to +12% +3% to +7% +13% to +20% +8% to +12% COOL scenario DRY scenario –25% to –11% –37% to –10% –10% to –3% –9% to –3% –2% to +2% –2% to +2% +3% to +6% +3% to +4% +7% to +10% Source: World Bank 2012. decrease in the Northern Zone under the HOT and DRY The INC also assessed the impacts of climate change scenarios. Yield increases in very few areas and only by on coffee using relevant regression models, looking at small percentages. Under HOT, there is damage to the the major producing areas of Lyamuno in the northeast yields in the vast majority of Tanzania, particularly in the and Mbozi-Mbeya in the south. The INC assumes that north and in the Lake Victoria region (Arndt and others the rainfall increase is 37 percent in the northeast and 2012). there is a rainfall decrease of 10 percent in the south. An increase of 2°C in both areas would put coffee pro- Coffee duction within the optimal values, and changes in rain- fall would determine production. An increase in rainfall Outside of climate change, the coffee sector in particu- would correspond with increased yield. The model shows lar has been historically rather unstable owing to global only a minimal decrease in rainfall in the southern areas prices and climate. A few characteristics exacerbate cof- and yield would not be affected. As such, the INC finds fee production vulnerability to climate change impacts: that yield will increase by an average of 17 percent in (1) intercropping with bananas in the north and west with each area (taking into account and increase in pests and low plant densities, productivity, and replanting rates; diseases that would reduce yield by 20 percent on aver- (2) coffee is a major crop in the southern highlands with age). In Lyamungo, rainfall is bimodal and yields are high density and replanting rates; (3) minimal manage- expected to see an 18 percent increase, whereas Mbozi ment of coffee trees and shrubs (agrochemicals only used has unimodal rainfall and is expected to see a 16 percent in the southern highlands, and by less than half of pro- increase. ducers); (4) pre-existing vulnerability to variability in the El Niño/La Niña cycles. Further, the National Coffee If, however, there was a 4°C increase in temperature, Development Strategy does not address climate change coffee production would be “significantly reduced” and risks, but aims to double coffee production by 2020 (Hag- particularly limited in the southern highlands. Irrigation, gar and Schepp 2009). training, and drought/disease resistant coffee varieties 68 Tanzania would be needed to keep coffee as a major cash crop. compounded by pests and diseases, forcing farmers Generally, coffee may be more successful in areas with to adjust grazing habits and rangeland management increases in rainfall, such as the northern, northeastern, (Republic of Tanzania 2003). These problems would be and southeastern areas (Republic of Tanzania 2003). multiplied as farmers employ various strategies which may cause further environmental degradation or have In a comparative analysis with findings from neighbor- large economic losses. ing Kenya and Uganda, the Haggar and Schepp (2009) desk study found that climate change would result in a Climate change is already shrinking rangelands vital to “significant redistribution” of viable coffee-growing land. livestock producers and communities. The loss of range- For example, the study concluded that the minimum alti- lands will be aggravated because around 60 percent is tude for arabica production would increase by as much infested by tsetse fly, making it unsuitable. As a result, as 400 m, and robusta cultivation would need to shift to Tanzania may see increased conflicts between livestock areas with higher rainfall (most likely in the north). Coffee producers and farmers (Republic of Tanzania 2007). growing may become unviable in lower altitudes or lose quality (Haggar and Schepp 2009). The study also suggests In the NAPA, vulnerability in the livestock sector is that the robusta growing region in Tanzania would move projected to increase owing to the effects of increased toward the Rwandan border, away from Lake Victoria. temperature and rainfall: changes in plant species compo- sitions affecting grazing; a general increase in dry matter The desk study also does some qualitative analysis as well yields, a favorable condition for pests and disease; long of the potential impact of climate change, with an analy- droughts and disease outbreaks limiting pasture size; and sis of coffee farmer surveys. The farmers generally agreed heat waves directly leading to livestock deaths (Arndt and that the climate is changing, particularly with irregular others 2012). rainfall patterns and less rain in turn resulting in lower productivity. Other Crops The desk study warns of potential environmental impacts The IFPRI crop modeling projected that rice production as coffee production expands at higher altitudes and com- would be geographically variable, making it hard to gen- petes with forestry and natural ecosystems. There is par- eralize. There might be gains in some regions while losses ticular concern over the Mount Kilimanjaro region. may occur in others. Under the IMPACT model, rice yields would “roughly double between 2010 and 2050” Livestock (IFPRI 2012). The INC also conducted a climate change vulnerability Using the IMPACT model, IFPRI found that cassava assessment for grasslands and livestock, finding changes yields will remain largely unchanged between 2010 and in foliage associations and a shift in foliage species as the 2050, but with population growth, demand will greatly “most palatable species” in semiarid areas are grazed exceed supply. The same model found yields tripling for out and replaced with more climate-tolerant species. It sorghum, factoring in both climate change and techno- also found that the rangeland carrying capacity would logical improvements. If it is assumed that the area under be low, but that the carrying capacity for areas with production expands by 40 percent, allowing total produc- increased rainfall as CO2 doubles will rise (the north- tion to increase fourfold, 70 percent of sorghum produc- ern, northwestern, and northeastern regions of Kigoma, tion in 2050 could be exported (IFPRI 2012). Mwanza, Musoma, and Same, and some southern areas such as Iringa). In areas with increased precipitation, there would be surplus foliage, but crude protein content Beyond Crop Impact would be lower. As a result, grazing animals would have Studies poor performance, and there would be negative impacts Several of the recent climate change impact assessments on milk and meat production. These problems would be seek to quantify the economic impact of agricultural Agricultural Sector Risk Assessment 69 changes in the broader economy. Generally, they find Changes in food consumption are less pronounced than that climate change has a negative impact on agriculture, changes in agricultural GDP. The paper accounts for this which results in a negative impact on the economy and a with assumptions of ability to import food and developed deterioration of food security. transport systems by 2050. For example in the DRY sce- nario there is an 11.5 percent decline in national agricul- The World Bank study uses a dynamic computable gen- tural production offset by a 37.1 percent increase in net eral equilibrium model (DCGE) of mainland Tanza- food imports, and food consumption falls only 8 percent. nia to project economywide effects (including indirect Outcomes are also variable due to region-specific impacts effects and economywide linkages) of the agricultural of climate change, crop-specific impacts (and thereby impact channel and potential indirect impact channels incomes and ability to reallocate farm resources), and the such as agro-­processing. In the DCGE model, predicted percentage of household income composed of agriculture annual yield deviations for rain-fed crops estimated by and a consumption basket composed of food. For further CLICROP affects domestic agricultural production, eco- details, and region specific numbers, refer to the World nomic growth, and household incomes. The net effect of Bank Report (Arndt and others 2012). climate change in this model is a significant reduction in national GDP in the HOT and DRY scenarios, with a slight decrease in COOL, and a slight increase in WET Conclusion (Arndt and others 2012). On a general level, a review of the literature suggests that there will be a decline in agricultural production because In the DRY scenario, agricultural GDP is 11.5 percent of climate change that in turn will affect various com- below the baseline by the end of the 2040s. This con- ponents of the national GDP. The production declines tracts the supply of raw inputs such as grain for the will occur in food production principally, while there agro-­processing sectors (for example, milling). The agro- are opportunities for increases in some production (such processing GDP is then 7.8 percent below the baseline. as coffee). These changes may limit export growth and Food imports, however, are expected to increase, offsetting household income, which in turn reduces Tanzania’s abil- declined domestic production and potentially benefiting ity to import food. some traders. The HOT and DRY scenarios project large agricultural GDP reductions in the Northern and Cen- Climate change is likely to alter the makeup of Tanza- tral Zones around Lake Victoria. These areas currently nia agriculture. Shifts in production and cropping will account for a large portion of Tanzanian agriculture; also have large socioeconomic impacts due to changes in therefore, future changes have implications nationwide. livelihoods. In particular, there is widespread reliance on In the WET scenario, there is significant variation on the corn among subsistence farmers, who may not have the regional level (increases in the northern coast and North- resources available to invest in different crops to feed their ern Zone with falls in other areas including the Lake Vic- families. Further, the crops with the greatest potential for toria region) and within agro-climatic zones, but overall increased favorable conditions such as coffee and cotton agricultural production rises (Arndt and others 2012). are export-oriented cash crops. This may contribute to the overall economy, but not to the increasing food inse- In the DCGE models, households are affected by climate curity due to climate change. change both through consumer prices and agricultural incomes. Household adaptation decisions are based on In conclusion, because of the impacts of climate change, both supply and demand. They might adapt by reallo- Tanzania may see problems related to land tenure, agri- cating resources and changing livelihoods. Or, because cultural incomes, food availability, food prices, and food of rising consumer prices (from falls in agricultural pro- security, among others. These changes demand better duction), some resources may be reallocated to affected crop (Republic of Tanzania 2003) and land management agricultural sectors in hopes of benefiting from the high strategies, and their incorporation into agricultural devel- prices. opment approaches is crucial. 70 Tanzania Major Stakeholders of concern. Also omitted are summaries of various vul- nerability studies and poverty analyses as to the effects of This group includes the Tanzania Meteorology Agency, climate change. There are several such studies, including the MAFC, regional institutions (river basin manage- a joint study from CEEST and The Netherlands Climate ment offices, and regional and district government Assistance Program using the United Kingdom Depart- offices in charge of land use planning and investment ment for International Development (DFID) Sustainable promotion), the Tanzania Coffee Research Institute, Livelihood Framework. The IMPACT model used by farmers and producers, and the University of Dar es IFPRI also accounts for projections in international prices Salaam. of crops, which is not mentioned in this review, but could provide important insights for food security. This assess- Limitations ment could benefit from crop-specific analysis. In particu- This literature review does not consider impact studies lar, there appear to be no studies on the impact of climate focusing on the minor islands or coastline of Tanzania change on cashew nuts in Tanzania, a major commodity. where a rising sea level and resulting coastal erosion are Further research should be done on these topics. Agricultural Sector Risk Assessment 71 APPENDIX C VULNERABILITY ANALYSIS Introduction The World Bank defines vulnerability as exposure to uninsured risk, leading to a socially unacceptable level of well-being. An individual or household is vulnerable if they lack the capacity and/or resources to deal with a realized risk. It is generally accepted that in low-income countries, rural populations are both poor and vulner- able, and that primary risks to these populations may include climate and market shocks (Sarris and Karfakis 2006). In Tanzania, shocks and stresses that will trigger a decline or drop in well-being may be on the household/micro-level (crop disruption, malaria, HIV/AIDs), at the community/meso-level (refugee populations competing for resources, food price shocks), and at the national/macro-level (climate change, natural disasters). Vulnerability is discussed here particularly in the context of food security. Major findings: »» Demography: The primary vulnerable populations in Tanzania are women, children, widows, and the elderly, the disabled, poor, and ill. »» Location/Livelihoods: The primary vulnerable areas are rural subsistence- based agricultural communities. Other particularly vulnerable rural groups include those dependent on aid, daily workers, and those with little access to assets. »» The major shocks to these vulnerable groups include »» Climate and other natural disasters (particularly drought and pests) »» High food prices (international commodity price shocks) »» Pests and plant disease »» Human illness (HIV/AIDS, malaria, and so on) »» Other shocks and stresses to these vulnerable groups might include changes in aid flows, refugee populations competing for resources (particularly in the northwest regions), governance changes, and others (to be discussed). »» The population can be vulnerable on an individual level as well as meso- and macro-levels. Agricultural Sector Risk Assessment 73 »» Vulnerability is context specific and difficult to agriculture, and agricultural production accounts for 27.8 measure in Tanzania, but addressing vulnerability percent of gross domestic product (GDP) (CIA Factbook and poverty appears to be a high priority of gov- 2013). Agriculture is extremely important to the Tanza- ernment and donor agencies. nian economy and is the primary income source for the »» Many valuable impact studies and vulnerability poor. The poor are composed primarily of the rural pop- assessments are out of date. ulations and are small-scale or subsistence farmers. Literature In 2010, 74 percent of the population was rural, with a The government of Tanzania, in partnership with inter- rate of 5 to 10 percent annual rate of change (urbaniza- national development organizations such as the World tion), depending upon the source. Food crop producers Bank, has been focusing on reducing vulnerability over are generally poorer than cash crop producers (“Enabling the past decade. The Tanzania government (the Ministry Poor Rural People” 2012). These poor rural households of Finance, the President’s Office of Planning and Priva- are particularly vulnerable to extreme weather shocks tization, and the National Bureau of Statistics) has been (such as drought and flood), and price shocks in inter- deeply involved in vulnerability and poverty assessments, national commodity markets. These groups lack links to the foremost of which are the participatory poverty markets, inputs, credit, and irrigation, making them less assessments. Other assessments include the Food Crop resilient and more vulnerable to shocks. Production Forecast and Vulnerability Assessments. The 2002–03 Participatory Poverty Assessment in particular, According to International Fund for Agricultural performed by the United Republic of Tanzania, resulted Development (IFAD), approximately 90 percent of the in a comprehensive qualitative assessment of households’ poor live in rural areas, and poverty is highest among risk environments, coping strategies, and vulnerabilities. those living in arid and semiarid regions that depend Surveys make up the substance of these assessments. entirely on livestock and food crop production for sur- vival. Although there is not one significantly worse off The 2009/10 United Republic of Tanzania Comprehen- or better region in Tanzania, generally, the most severe sive Food Security and Vulnerability Analysis (CFVSA/ poverty can be found near the coast and in the south- MKUKUTA), published by the World Food Program, gives ern highlands, and the most poorly nourished people an in-depth assessment of vulnerability as it relates to good live in the central and northern highlands (“Enabling security. Their analysis was based on data obtained through Poor Rural People” 2012). Dependence on rain-fed surveys conducted during a “relatively lean period” in both agriculture makes households in the semiarid areas unimodal and bimodal regions—capturing food consump- (the central and northern regions) particularly vulner- tion patterns while food was less available. For more recent able to weather shocks because it affects access to food assessments of vulnerability, particularly as it relates to (Enabling Poor Rural People” 2012). Food insecurity in food security, the Famine Early Warning System Network turn leads to further vulnerability to disease, livelihood (FEWS NET) and the Food and Agricultural Organization loss, and so on. According to the most recent World of the United Nations (FAO) provide remote monitoring.22 Bank data, the poverty head count ratio at US$1.25 a day (PPP) is 68 percent (World Bank 2007), the poverty Dimensions of Poverty head count ratio at the rural poverty line (percentage of and Vulnerability in rural population) is 37%, and the poverty head count ratio at the urban poverty line (percentage of urban Tanzania population) is 22%. Life expectancy at birth in Tan- In 2011, Tanzania’s GDP per capita (PPP) was US$1,600 zania is 57, malnutrition in terms of height-for-age of and the country was ranked 199 out of 228 in terms of children younger than five years is 17 percent, there are wealth. Around 80 percent of the labor force works in approximately 230,000 children (0–14) living with HIV, and countless others orphaned by HIV/AIDS (World 22 For the FEWSNet updates: http://www.fews.net/east-africa/tanzania. Bank Data Bank 2013). 74 Tanzania Figure C.1. D  istribution of Poor and Borderline Food Consumption Households Source: CFSVA 2009/10. Tanzania is ranked 152 on the Human Development The CFSVA produced the two maps above (­figure C.1), Index (HDI, and has a 0.332 inequality adjusted HDI diagraming the frequency of poor food consumption value), the mean years of schooling (of adults) is 5.1 years, and borderline food consumption regionally. Poor and and the country scored a 0,606 on the Gender Inequality borderline consumption centered in a “band of vul- Index. Inequality also contributes to vulnerability in that nerability” running from the central northern regions unequal access to productive assets such as land, finance, down to the southeast. As expected, acceptable con- livestock, and education affects the ability to cope with sumption prevailed along the coast and in the west shocks and stresses. Overall, the UN Development Pro- (ODXF 2013). gramme places Tanzania in the “low” human develop- ment category (although above the average for countries There is some geographical overlap between regional in Sub-Saharan Africa), with significant levels of gender distribution of maternal and child malnutrition rates inequality (Human Development Index 2013). and poor food consumption households. But it should be noted that several regions, such as Kigoma, reported Food Security and Vulnerability elevated wasting and underweight prevalence and yet had The 2009/10 CFVSA found 4.1 percent of households a high level of acceptable consumption. Stunting was also in rural mainland Tanzania with poor food consumption, not correlated regionally with food consumption patterns meaning diets primarily are cereal based with almost no (ODXF 2013). animal protein and little else. It also found 18.9 percent of households with borderline food consumption (mean- Table C.1 lists the factors that the CFSVA found to be ing a marginally better diet including pulses, vegetables, associated with food security, both positively and nega- and fruits at least one more day a week than poor con- tively. The regions indicated are most affected by the sumption households), and 77 percent of households variables, resulting from interactive models. Based on with acceptable food consumption (a threefold increase in multivariate analyses, after controlling for the variables pulse and fruit consumption, larger increases in milk and below, the CFSVA found that small subsistence farmers animal protein) (World Food Programme Food Security were “significantly worse off” than the most food secure Analysis Service [ODXF] 2010). individuals (salaried workers) (ODXF 2013). Agricultural Sector Risk Assessment 75 Table C.1. Factors Associated with Food Security by Region Factors Significantly Associated With Regions Where Factors Show a Strong Regions Where Factors Show a Strong Food Security Positive Association with Food Security Negative Association with Food Security Illiteracy of household head Mwanza and Mara Access to livestock Tanga, Mtwara, and Ruvuma Kagera Cultivating four or more crops Dodoma, Arusha, Kilimanjaro, Singida, Rukwa, Shinyanga, Kagera, and Mara Using chemical fertilizers Arusha and Shinyanga Asset wealth Arusha Source: CFSVA 2009/10. and Arusha. Women affected are generally the poor and The Vulnerable elderly (Research on Poverty Alleviation [REPOA] 2007). Populations In 2010, Tanzania had the 23rd highest maternal mortal- According to the 2009/10 CFVSA, the food insecure ity rate globally: 460 deaths to 100,000 live births (CIA (poor food consumption households) and thereby those Factbook 2013). Malnutrition has effects that are cumula- vulnerable to shocks, had the following characteristics: tive and intergenerational. Maternal health is intrinsically (1) dependent on aid, daily work, small subsistence farm- linked with child health, and maternal mortality rates ing, and agro-pastoralism for their livelihoods; (2) female- have not significantly improved over the last few decades. headed households and illiterate households; and (3) poor, with the least access to assets. Specifically, poor consump- Children tion households have access to fewer livestock, cultivate Children are a particularly vulnerable group in Tanzania less diverse crops, cultivate less than one hectare of land, (over 18 million Tanzanians are under 18 years old). The and are less likely to use chemical fertilizers (ODXF 2010). 2009/10 CFVSA found that nationally 5.7 percent of children 0 to 59 months old were wasted, 36.6 percent The Tanzania Participatory Poverty Assessment found stunted, and 14.3 percent underweight (ODXF 2010). the following social groups to be the most vulnerable The depth of vulnerability for children is weighted heav- because of having the “least freedom of response” to ily toward rural populations. According to the National shocks and stresses: children (especially orphans), child- Bureau of Statistics, Population Census in 2002 and the bearing women and women with young children, widows, Tanzania Demographic and Health Survey (TDHS) the elderly, people with disabilities, people with chronic 2004/05, around 41 percent of children are stunted illnesses, people in HIV/AIDS-affected households, and under-five in rural areas, whereas only 26 percent are in destitute persons. Other studies agree with this assess- urban areas. These trends are similarly mirrored in mal- ment, and other social groups might include drug addicts, nutrition and mortality rates. More than 1 in 10 children unemployed youths, and alcoholics. These groups all have die before they turn five years old. Many children live in low access to assets, which limits their capacity to cope households that do not have income sufficient to provide (Sarris and Karfakis 2006). minimum nutritional requirements, resulting in physi- cal and mental problems, serious economic and social Women well-being consequences, and distortions of their poten- In Tanzania, women are particularly vulnerable owing to tial contribution to national development. Particularly a lack of rights and various physical, social, and finan- vulnerable groups of children are those with disabilities, cial inequalities. For example, female genital mutilation orphans (especially those orphaned due to HIV/AIDs affects 15 percent of women in Tanzania and is partic- and subsequently stigmatized), and others (such as child ularly common in regions such as Manyara, Dodoma, laborers and street children) (REPOA 2007). 76 Tanzania Major Shocks Figure C.2. Hazard Occurrence in the and Stresses Agro-Ecological Zones 90 Drought HIV/AIDS Based on the 2009/10 CFSVA study, the top three shocks Disease outbreak Pests 80 to household consumption were drought (at 58.4 percent), 70 high food prices (at 53.4 percent), and plant disease/ani- % of respondants 60 mal pests (at 34.7 percent). Drought was most frequently 50 reported in the northern regions (Arusha, Tanga, Man- 40 yara, Kilimanjaro, and Mara) central regions (Dodoma 30 and Morogoro), and southeastern regions (Mtwara and 20 Lindi). The “increasingly bimodal” tendencies and rain- 10 fall patterns in the north corresponds with this finding 0 (ODXF 2010). Across the varying shocks, invariably, the Zone 1 Zone 2 Zone 3 Zone 4 Zone 5 Zone 6 Zone 7 Agro - ecological zones most vulnerable were the most affected. Source: United Nations University. Legend: Zone 1 = Coastal; 2 = Eastern plateau and mountain blocks; High food prices as a shock were most reported across 3 = Southern highlands; 4 = Northen rift valley and volcanic highlands; Tanzania, but with particularly high percentages in Kili- 5 = Central plateau; 6 = Rukwa-Ruaha rift zone; 7 = Inland Sedimentary; manjaro, Mara, Dodoma, Singida, Lindi, and Mtwara Ufipa plateau and western highlands. (western regions reported this shock with less frequency). Groups disproportionately affected by the high food price Other recent vulnerability studies cite environmental shocks included daily workers, fishermen/hunters, house- and macroeconomic conditions, governance, ill health, holds reliant on aid, and “others.” Those least affected life-cycle conditions and cultural beliefs and practices as were large producers of both food and cash crops (ODXF being important impoverishing forces (Sarris and Kar- 2010). fakis 2006). Ill health makes populations vulnerable as it reduces the capacity to work, resulting in a loss of pro- Plant diseases and animal pests acted as shocks most fre- duction and income as well as generating treatment costs, quently in regions adjacent to bodies of water, specifically, which reduces their ability to cope with further shocks Lindi, Kigoma, Mara, Mtwara, and Mwanza. The Shin- (that is, a poverty trap). Vulnerability also increases as yanga, Ruvuma, and Arusha regions were least affected. populations sustain successive shocks. Looking through the lens of livelihoods, the households most affected were large subsistence farmers and “others” (ODXF 2010). Natural Disasters Tanzania has a long history of natural disasters through- Participants in the 2002–03 Participatory Poverty Assess- out its seven differing agro-ecological zones (see figure C.2), ment cited having vulnerabilities to material well-being diverse and varied as the geography, physical, social, and (such as money, land, farming gear, and so on) and physical economic factors throughout the country. A disaster vulner- well-being (health, security, dignity and freedom of choice ability assessment carried out in 2006 used both perceptions and action, and so on). According to this and subsequent and a regression analysis wherein the 1992 United Nations studies, the significant shocks and impoverishing forces Development Programme (UNDP) formula for vulnerabil- include drought and other natural disasters, environmen- hazard Risk ity was used v . tal degradation, worsening terms of trade, corruption, manageability and coping strategies inappropriate taxation, lack of physical security, HIV/ AIDs, malaria, and aging. The most significant category The resulting vulnerability index suggests that the Rukwa- of shock/stress varied from community to community, Ruaha rift zone was most vulnerable to disease outbreak; but three cases emerged as having the greatest impact: the central plateau to drought; and the southern high- governance, macroeconomic influences, and environmen- lands, eastern plateau, and mountain blocks to pests (Birk- tal forces (“Tackling Vulnerabilty” 2004). mann 2006). Agricultural Sector Risk Assessment 77 Commonly occurring disasters occur as a result of epi- communities are disproportionately affected by the epi- demics, pests, flood, and drought leading to famine, fire, demic, with livelihoods unsustainable in sickness, ill adults accidents, cyclones and strong winds, refugees, conflicts, relocated, and orphaned children sent to villages to be landslides, explosions, earthquakes, and technological cared for by relatives (Tumushabe 2005). hazards. The disaster vulnerability assessment identified 15 hazards. The most commonly occurring were pests, A 2006 World Bank report studied the effect of HIV/ drought, and disease outbreaks. At the household and vil- AIDS as a shock on short- and long-term consumption lage levels, pests received the highest scores, whereas at among surviving households. Over a 13-year period, the the district level disease outbreaks (including HIV/AIDS) study found that affected households saw a 7 percent con- were most common (followed by pests, drought, and sumption drop within the first five years after an adult strong winds) (Birkmann 2006). All of these disasters can death, and had a 19 percent growth gap with unaffected lead to food crises, livelihood failures, and deeply negative households. The effects of shocks may last for 13 years, impacts for vulnerable populations. and adult female death has a particularly severe impact on a household (Beegle, De Weerdt, and Dercon 2006). Climate Change HIV/AIDS has become a long-term stress in Tanzania, There is both individual and collective vulnerability to and the interaction effect means that concurrent shocks climate change across Tanzania. The Centre for Energy, (such as price shocks) will have a greater impact. The epi- Environment, Science and Technology (CEEST) pro- demic has affected vulnerability in Tanzania by creating vides some indicators of vulnerability to climate change. a new underclass of highly vulnerable and disadvantaged For individuals, useful indicators include poverty indexes, people (the majority of whom are children, women, and the proportion of income dependent on risky resources, the elderly who fell into poverty because of the impact of dependency, and stability. Collective indicators might HIV/AIDS at the household level) and devastating par- include GDP per capita, relative inequality, qualitative ticular local economies (Tumushabe 2005). indicators of institutional arrangements, levels of infra- structure, availability of insurance, and formal or infor- mal social security (Meena and O’Keefe 2007). Coastal Children The major shocks for children include being orphaned, communities are particularly vulnerable to sea rise and encountering natural disasters and other disruptions, and flooding. Increased pests and diseases are likely the result illness. Children are particularly susceptible to malaria, of increased temperature and moisture in some areas. other diarrheal diseases, and respiratory infections. All of (See appendix B on climate change.) these illnesses affect appetite, which in turn affects nutrition and may affect their physical and mental development, HIV/AIDS thereby increasing their future vulnerability. Children, HIV/AIDS has been considered by international organi- particularly those who have been orphaned, are vulner- zations and the government of Tanzania to be the primary able to a lack of education and exploitation, including threat to human development in Tanzania. Estimates in child labor in mining, sex work, commercial agriculture, 2009 ranked Tanzania 12th in global prevalence of HIV, and domestic work. Those orphaned as a result of HIV/ with a 5.6 percent rate (CIA Factbook 2013). HIV/AIDs AIDS also are vulnerable to being ostracized because of can create localized crises, such as in the Makete district, social stigma. Studies have suggested there are geographi- Iringa region, where there was a livelihood collapse owing cal area–specific factors that play a role beyond common to the high prevalence of AIDS. Estimates range up to a determinants (education, income, and risk of malaria). potential 20 percent negative impact on GDP. Small-scale Under-five mortality rates are four times higher in Lindi studies across Sub-Saharan Africa and Tanzania have and Mtwara than they are in Kilimanjaro and Arusha. found that HIV/AIDS also causes serious losses at the Higher percentages of children with fever are reported household level, including lower income, decreased food along the coast and in Mara and Kigoma (both on large cultivation, and depletion of assets. Rural households and lakes) (REPOA 2007). 78 Tanzania Existing Coping Methods men learning to cook, women collecting firewood, and both sexes participating in decision making) (Tumushabe There are various coping methods depending on the type 2005). of shock, but one of the first reactions is to either sell or use assets, whether they are human, social, political, natu- ral, physical, or financial (Sarris and Karfakis 2006). A Natural Disasters presentation by the Tanzania MUCHALI23 team to the Coping strategies for natural disasters and climate change Southern Africa Vulnerability Initiative in July 2010 iden- are similar and reflect the capacities of vulnerable groups. tified the following various coping strategies in response They include selling assets, migration, reduction in con- natural disasters and food insecurity: reduction in the num- sumption, income and crop diversification, and other ber and size of meals; increased livestock sales; increased various strategies listed here. One study listed coping sale of charcoal, handcrafts, and firewood; increased con- strategies at the zonal level for drought and pests. Listed sumption of wild-food; sales of household assets; sending in order of decreasing frequency, these are pesticides, sell- children to relatives; urban migration; government collab- ing assets, employment elsewhere, and drought resistant oration with development partners (free food aid, school crops (Birkmann 2006). feeding programs, nutrition and food, cash transfers, food fortification, and input subsidies). Conclusion The various studies reviewed here have found varying HIV/AIDS types of vulnerability and vulnerable groups across Tan- Coping strategies to deal with the socioeconomic impacts zania, as well as a plethora of coping strategies. Factors of HIV/AIDS include selling assets such as livestock, which contribute to vulnerability, but have not been dis- drafting in new adults to the household, strong social cussed here include larger distances from medical services cohesion for the transfer of assets, assistance from non- (lack of access), lack of access to finance, and other assets governmental organizations and government interven- that would foster resiliency. Understanding local and cir- tions, and burden-shifting (such as moving the dying to cumstantial vulnerabilities is imperative in designing pol- rest in better-off households). In relation to food security, icy and agricultural development strategies so that they farming systems with low labor requirements are less vul- may be more effectively targeted. nerable (particularly if there is good rain and a reliance on tree crops, which does not work in unimodal rainfall areas). Other food security coping strategies include cut- Limitations ting the number of meals consumed and cultivating short Although a primary focus of the government, many of season crops such as cassava, sweet potatoes, cabbage, these vulnerability assessments were conducted several beans, and groundnuts for both small-scale consumption years ago, and updating is recommended. Further, this and sale. Coping strategies have resulted in casual labor appendix does not discuss coping strategies that have been by surviving adults and orphans, changes in gender roles put in place by the government, such as safety nets that and the division of labor (for example, one study found reduce vulnerability. 23 MUCHALI is the Swahili abbreviation for the Food Security and Nutrition Information System implementation framework. MUCHALI team members are analysts from government ministries, Sokoine University, FAO, and FEWS NET. Agricultural Sector Risk Assessment 79 A g r i c u l t u r e g l o b a l p r a c t i c e t e c h n i c a l a s s i s t a n c e Pa p e r W ORL D B A NK GROUP REPORT NUMBER 94883-TZ 1818 H Street, NW Washington, D.C. 20433 USA Telephone: 202-473-1000 Internet: www.worldbank.org/agriculture