OCTOBER 2015 Kenya Toward a National Crop and Livestock Insurance Program BACKGROUND REPORT Kenya Toward a National Crop and Livestock Insurance Program BACKGROUND REPORT © 2015 International Bank for Reconstruction and Development / International Development Association or /// The World Bank /// 1818 H Street NW Washington DC 20433 Telephone: 202-473-1000 Internet: www.worldbank.org This work is a product of the staff of The World Bank with external contributions. The findings, interpretations, and conclusions expressed in this work do not necessarily reflect the views of The World Bank, its Board of Executive Directors, or the governments they represent. The World Bank 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 The World Bank concerning the legal status of any territory or the endorsement or acceptance of such boundaries. Rights and Permissions ** ** The material in this work is subject to copyright. Because The World Bank encourages dissemination of its knowledge, this work may be reproduced, in whole or in part, for noncommercial purposes as long as full attribution to this work is given. Any queries on rights and licenses, including subsidiary rights, should be addressed to the Office of the Publisher, The World Bank, 1818 H Street NW, Washington, DC 20433, USA; fax: 202-522-2422; e-mail: pubrights@worldbank.org. Photo Credits ** ** Photos specifically credited are done so under Creative Commons Licenses. The licenses used are indicated through icons showcased next to each image. b 2.0 Attribution License (http://creativecommons.org/licenses/by/2.0/legalcode) bd Attribution No-Derivatives License (http://creativecommons.org/licenses/by-nd/2.0/legalcode) ba Attribution Share Alike License (http://creativecommons.org/licenses/by-sa/2.0/legalcode) If not indicated otherwise, photos used in this publication have been sourced from the following locations with full rights: World Bank Flickr Website United Nations Flickr Website All non-Creative Commons images in this publication require permission for reuse. K E N YA : A G R I C U LT U R E I N S U R A N C E S O L U T I O N S A P P R A I S A L i Table of Contents 05 Acknowledgments 06 Acronyms and Abbreviations 07 Introduction 09 Institutional Frameworks 09 Rationale for Public-Private Partnerships in Agricultural Insurance 11 Public And Private Sector Functions 23 Institutional Framework 25 Institutions 27 Livestock Insurance for Pastoralists Located in ASALs in Northern Kenya 27 Context 29 Proposals For Large-Scale Livestock Insurance For Pastoralists Located In ASALs In Northern Kenya 40 Fiscal Costing Assumptions And Scenarios 43 Welfare Impacts Of Index-Based Livestock Insurance In HSNP Countries 48 Crop Insurance 48 Context 50 Description of Potential Agricultural Insurance Programs for Crops 54 Fiscal Costing Assumptions and Scenarios 58 Welfare Impacts of Area Yield Insurance for Maize and Wheat in Kenya 65 Conclusion SECTION i ii K E N YA : A G R I C U LT U R E I N S U R A N C E S O L U T I O N S A P P R A I S A L Table of Contents ANNEXES 67 Annex A. Possible Options for Coinsurance Pools in Kenya 70 Annex B.1. Index Based Livestock Insurance (IBLI) Program 71 Annex B.2. Assumptions and Parameters for Fiscal Costing Scenarios for Livestock 75 Annex B.3. Summary of Modeling and Simulations of Welfare Analysis for Livestock 82 Annex C.1. Assumptions and Parameters for Fiscal Costing Scenarios for Crops 85 Annex C.2. Summary of Modeling and Simulations of Welfare Analysis for Crops 94 Bibliography 96 Endnotes SECTION i K E N YA : A G R I C U LT U R E I N S U R A N C E S O L U T I O N S A P P R A I S A L iii Figures 11 Figure 1 —  Toward an integrated private risk management and insurance framework for different segments of Kenya’s crop and livestock producers 12 Figure 2 —  Potential public sector roles for GOK in support of agricultural insurance development in Kenya 31 Figure 3 —  How government could support financial protection for different segments of the population: Example of pastoralists in the four current HSNP counties (Mandera, Marsabit, Turkana, and Wajir) 35 Figure 4 —  Illustrative calculated pure loss cost rates for 12-month NDVI asset protection cover at District and Division level 46 Figure 5 —  Potential short-term impacts of livestock insurance on income available for consumption 47 Figure 6 —  Potential impacts of livestock insurance on herd accumulation 47 Figure 7 —  Potential impacts of livestock insurance on probability of falling into poverty trap 51 Figure 8 —  Types of Agricultural Insurance Products 51 Figure 9 —  Coverage Level and Insurance Payouts in AYII 55 Figure 10 —  Estimated AYII Risk Premium Rates for Maize at District Level 55 Figure 11 —  Estimated AYII Pure Premium Rates for Wheat at District Level 62 Figure 12 —  Potential impacts of AYII on net income available for consumption 70 Figure 13 —  Translating NDVI data into estimated livestock mortality and IBLI payouts 70 Figure 14 —  IBLI seasonal sales periods, contract cover period and contract payout dates SECTION i iv K E N YA : A G R I C U LT U R E I N S U R A N C E S O L U T I O N S A P P R A I S A L Tables 13 Table 1—  Agricultural insurance data collected by GoK 34 Table 2—  Comparison of Uruguayan pasture NDVI cover and proposed Kenyan NDVI cover 40 Table 3—  Proposed livestock safety net and insurance program for Kenya’s four HNSP counties 41 Table 4—  Fiscal costing projections for macro-level asset protection coverage 42 Table 5—  Fiscal costing projections for top-up and nontargeted pastoralists options 56 Table 6—  Variation of Premium Rates According to Different Coverage Levels 63 Table 9—  Fiscal cost per household of achieving different policy goals in different insurance scenarios 66 Table 10—  Illustrative fiscal costing for agricultural insurance programs, 2016 and 2019 73 Table 12—  Fiscal costing projections for macro-level asset protection coverage 74 Table 13—  Fiscal costing projections for top-up and nontarget pastoralists options 81 Table 14—  Summary statistics of pastoral households in four HSNP counties 83 Table 15—  Yield, Area, and Premium Rate Data for Maize 85 Table 16—  Yield, Area, and Premium Rate Data for Wheat 86 Table 17—  Estimation of potential cost of additional data collection activities for AYII 90 Table 18—  Summary statistics of maize and wheat growing households 92 Table 19—  Summary of key impact indicators by contract variations SECTION i K E N YA : A G R I C U LT U R E I N S U R A N C E S O L U T I O N S A P P R A I S A L 05 Acknowledgments This technical report of the Agriculture Insurance The report benefited greatly from the data Solutions Appraisal was led and prepared by Daniel and information provided by the Ministry of Clarke (Disaster Risk Financing and Insurance Agriculture, Livestock and Fisheries; and special Program, Finance and Markets Global Practice, World acknowledgements are extended to Kenneth Bank Group) in collaboration with the Ministry of Ayuko (Director, State Department of Agriculture, Agriculture, Livestock and Fisheries, Kenya, and Ministry of Agriculture, Livestock and Fisheries) and with contributions from the following: Barry Maher Vincent Ngari (Deputy Director, State Department (Disaster Risk Financing and Insurance Program, of Livestock, Ministry of Agriculture, Livestock Finance and Markets Global Practice, World Bank and Fisheries). We also gratefully acknowledge the Group); Felix Lung (Disaster Risk Financing and support, inputs, and feedback from Kenya’s National Insurance Program, Finance and Markets Global Treasury, the National Drought Management Practice, World Bank Group); Sarah Coll-Black Authority, Kimetrica, the UK Department for (Labor and Social Protection Global Practice, International Development, and the Tegemeo World Bank Group); Richard Carpenter, Sommarat Institute of Agricultural Policy and Development. Chantarat, James Sinah, Andrea Stoppa, and Charles Stutley (Consultants, World Bank Group); Andrew We gratefully acknowledge funding support from the Mude (International Livestock Research Institute); Ministry of Foreign Affairs of the Netherlands and and Michael Mbaka (Financial Sector Deepening U.S. Agency for International Development (USAID) Kenya). Overall guidance was provided by Olivier through the World Bank’s Agricultural Insurance Mahul (Program Manager, Disaster Risk Financing Development Program. The Agricultural Insurance and Insurance Program, Finance and Markets Global Development Program is part of the World Bank– Practice, World Bank Group) and Smita Wagh Global Facility for Disaster Reduction and Recovery (Finance and Markets Global Practice, World Bank (GFDRR) Disaster Risk Financing and Insurance Group). This nonlending technical assistance (NLTA) Program. Contributions of the International report has been prepared as part of an NLTA to the Livestock Research Institute (ILRI) are supported by Ministry of Agriculture, Livestock and Fisheries to the UK Department for International Development, support the ministry in deciding whether to establish Australian AID, and the European Union, which fund a large-scale public-private partnership in agricultural ILRI’s Index Based Livestock Insurance Agenda. insurance. This report can be read in conjunction with the accompanying World Bank policy note, “Kenya: Toward a National Crop and Livestock Insurance Program” (2015), and “Kenya: Agricultural SECTION Sector Risk Assessment Risk Prioritization” (2015), a report by the World Bank Agricultural Risk Management Team. A 06 K E N YA : A G R I C U LT U R E I N S U R A N C E S O L U T I O N S A P P R A I S A L Acronyms and Abbreviations AIDP Agriculture Insurance Development Program AIM Agricultural Insurance Manager ALRMP Arid Lands Resources Management Project ARC Africa Risk Capacity ASAL arid and semi-arid land AYII area yield index insurance CCE crop-cutting experiment eMODIS enhanced Moderate Resolution Imaging Spectroradiometer FSD Financial Sector Deepening GFDRR Global Facility for Disaster Risk Reduction and Recovery GIS geographic information system GoK Government of Kenya GPS Global Positioning System HSNP Hunger Safety Net Program IBLI Index Based Livestock Insurance ILRI International Livestock Research Institute IRA Insurance Regulatory Authority MALF Ministry of Agriculture, Livestock and Fisheries MPCI multi-peril crop insurance NAIP National Agricultural Insurance Policy NDV Normalized Difference Vegetation Index NPCI named peril crop insurance PPP public-private partnership SACCO savings and credit cooperative SDA State Department of Agriculture SDL State Department of Livestock TLU Tropical Livestock Unit TSU technical support unit SECTION USAID U.S. Agency for International Development WII weather index insurance A Currency: Kenyan shilling (K Sh) K E N YA : A G R I C U LT U R E I N S U R A N C E S O L U T I O N S A P P R A I S A L 07 Introduction This Agriculture Insurance Solutions /// insurance, agriculture credit will remain insufficient Appraisal proposes the technical concept of to fully meet the needs of farmers and herders. a potential public-private partnership (PPP) Moreover, international experience suggests in agricultural insurance in Kenya, covering /// that agricultural insurance programs will both crop and livestock. It lays out the rationale not scale up unless based on a balanced /// for the proposal, offers an insurance PPP solution partnership between the public and private for the identified challenges, and makes a suggestion sectors. In recent years, numerous private sector for the required underlying institutional framework. /// agricultural insurance pilots have been implemented This technical report is meant to guide further policy in Kenya with support from donor partners for making and technical development processes and index-based crop insurance. However, to date most to form the basis for further discussion among all of these programs have failed to reach significant involved stakeholders. scale. Overall, experience from other countries In Kenya agriculture is risky, and that risk /// suggests that both the government and the private has large human and economic costs. /// sector must play a role in developing the agriculture Agriculture is key to the Kenyan economy, generating insurance market.2 approximately 30 percent of annual gross domestic Recognizing the importance of its product and approximately 50 percent of revenue /// involvement, the government of Kenya from exports. It is also an important source of (GoK) is collaborating with the World Bank employment: over 61 percent of the population has to investigate how agriculture insurance can jobs in agriculture.1 But agriculture in Kenya is a risky help transform the agriculture sector and activity, often unirrigated and highly vulnerable to the promote food security, economic growth, impacts of climate change. and shared prosperity in the medium to Despite the recognized need for a /// long term. The collaboration, which is undertaken /// commercially oriented, internationally with the Agriculture Insurance Development competitive, and modern agricultural sector, Program (part of the World Bank Group’s Disaster rural lending in Kenya is low. Agricultural lending /// Risk Financing and Insurance Program), aims to accounted for only 4.3 percent of total lending in understand how insurance could form part of a Kenya in 2012 (Central Bank of Kenya 2012). A strategy to derisk agriculture value chains and more large-scale agricultural insurance program would generally function within a broader risk management support resilient, viable expansion of agriculture framework. The GoK has identified the agriculture credit to farmers by removing agriculture risk from sector as a key area of focus under the Kenya SECTION the balance sheet of rural banks and cooperatives, Vision 2030 plan to promote Kenya’s transition thereby making them more robust to agricultural to a middle-income country; and agricultural shocks. Without adequate coverage of agricultural insurance is a stated priority of government, as A 08 K E N YA : A G R I C U LT U R E I N S U R A N C E S O L U T I O N S A P P R A I S A L reflected in the Second Medium Term Plan (MDP 2013). The government is exploring initiatives to Box 1—  Key International Lessons further derisk the agriculture sector value chain for the Design and Implementation in order to facilitate better access to markets and unlock access to credit, which would in turn allow of Agricultural Insurance farmers to purchase higher-yielding technology • Agricultural insurance programs are challenging to (seeds, fertilizers, plant protection chemicals, etc.) develop and successfully sustain. and increase their incomes. These initiatives aim • Carefully designed and well-implemented agricultural simultaneously to ensure food security in Kenya and insurance programs can support a range of government transform the agriculture sector. policy objectives, such as increased access to credit, improved agricultural productivity, reduced vulnerability, This technical report investigates the /// and social protection. institutional policy and design issues associated with agriculture insurance PPP • Agricultural insurance should be considered by structures, as well as their fiscal cost and government alongside other potential agricultural risk management and social protection interventions, welfare benefit. Taking into account Kenya’s /// since other interventions may offer higher benefit- current agriculture insurance policy and government cost ratios or be a precondition for successful institutions, our analysis sought to identify sound agricultural insurance. policies and institutional structures that would • Agricultural insurance programs are more effective and unlock the innovative potential of the private sector efficient when underwritten by the private insurance in agriculture insurance. For both crop and livestock, sector and actively supported by government under we analyzed the current market to understand what carefully designed PPPs. high-quality products could be developed in the • Financial support to agricultural insurance programs can short, medium, and long term to meet Kenya’s needs. provide a faster, more cost-effective way of supporting agricultural producers’ recovery from shocks than ad hoc post-disaster relief. • Cost sharing between government, donors, and farmers may differ for different segments of the population depending on policy priorities. Sources: Mahul and Stutley 2010; World Bank–GFDRR Disaster Risk Financing and Insurance Program 2014. SECTION A K E N YA : A G R I C U LT U R E I N S U R A N C E S O L U T I O N S A P P R A I S A L 09 Institutional Frameworks index insurance for smallholder farmers on a retail Rationale for Public- basis at scale—has partly to do with the absence of Private Partnerships in certain fundamental building blocks required by an agricultural insurance market: (i) improved access to Agricultural Insurance inputs, husbandry, and irrigation; (ii) reliable access The agricultural insurance market in Kenya /// to weather data; and (iii) a supportive regulatory has failed to reach scale. With the exception of /// framework (FSD 2013). some small-scale pilots and niche retail activity, the But international experience suggests that private sector is currently not providing agricultural /// the absence of these building blocks is not crop and livestock insurance (GoK 2014a). This the only reason why agricultural insurance chapter considers possible causes for the agricultural markets fail and pilots do not scale up.3 Most insurance market’s failure in Kenya, the rationale /// of the following reasons are applicable to Kenya: for and benefits of a public-private partnership (PPP) for developing agricultural insurance, and • Lack of agriculture data. As discussed /// /// the appropriate functions of the public and private below, there is very little reliable agricultural sectors within a PPP. The final section suggests a data available in Kenya. This is a serious vision for a PPP and makes recommendations for constraint on the development of agricultural next steps. insurance products. • Lack of capacity, especially for catastrophe Weather index insurance (WII) has been /// /// risk. Insurers do not have the capacity to cover considered a potential solution for fostering a /// catastrophe risk associated with drought, flood, viable agricultural insurance market, but the and other typical agricultural risks. Although approach has achieved mixed results. Although /// international reinsurance is available, it is a number of small-scale agricultural insurance pilots expensive, particularly where there is a lack of have been commenced, only one, the UAP Syngenta data. program, has so far scaled up. This program, currently in its fifth full year of implementation, • High distribution costs. Given that farms /// /// now insures more than 87,000 farmers annually tend to be small and spread over wide areas, in Kenya (during long rains 2014) where crop WII agricultural insurance typically carries very high SECTION and crop-credit provision are automatically linked. distribution costs. These are exacerbated by the The reason why WII has not scaled up in Kenya— lack of established branch or agent networks in that is, why it has been difficult to establish viable rural areas. 01 10 K E N YA : A G R I C U LT U R E I N S U R A N C E S O L U T I O N S A P P R A I S A L Photo Credit: Daniel Clarke • High loss assessment costs. In relation to /// /// Furthermore, flaws in the design of post- /// traditional indemnity insurance, the costs of disaster relief mechanisms often result in assessing losses are usually extremely high. This is the crowding out of insurance. If famers expect /// especially true for small insured farm units, where post-disaster relief from government, development the premium volume generated is usually very agencies, or nongovernmental organizations, they have little incentive to purchase insurance. low and insufficient to cover the costs of the loss assessment. International experience suggests that /// sustainable, scaled-up agricultural • High development costs. Although index /// /// insurance programs are based on a strong insurance lowers the transaction cost, public-private partnership—one involving it carries extremely high development engagement, innovation, and action from and other start-up costs. These start-up both partners. The failure of the agricultural /// costs cannot Expensive premiums. Small insurance market in Kenya provides a clear + ./// /// farmers are unwilling, and may be unable, to pay justification for intervention by the government for commercially priced agricultural crop and of Kenya (GoK), but that intervention is further livestock insurance. justified by the severe challenges that public sector– only and private sector–only approaches face. These • Poor understanding of insurance. Farmers’ range from inefficient delivery, distribution, and /// /// poor understanding of agricultural insurance claims settlement in the case of the former, and reduces demand and may lead to purchase of underinvestment in necessary data in the case of the inappropriate products. latter. A strong partnership between the public and • Lack of an enabling legal and regulatory /// the private sectors will allow Kenya to build on each framework. As discussed below, Kenya’s /// sector’s comparative advantages. Insurance Act does not support index insurance, 4 and a regulatory framework for microinsurance is SECTION still being developed. 01 K E N YA : A G R I C U LT U R E I N S U R A N C E S O L U T I O N S A P P R A I S A L 11 microinsurance or retail product, and most of these Public And Private programs are not achieving scale. Sector Functions Traditional indemnity-based multi-peril crop /// Overview insurance (MPCI) is not well suited to the risk transfer needs of subsistence farmers Few functions belong exclusively to either and pastoralists. It is therefore necessary to /// the public sector or the private sector; rather, /// identify other risk transfer solutions to meet these most are shared. For example, both the public and needs. In the short to medium term, potential /// private sectors have separate functions in relation to linkages between existing social safety net programs data, marketing and outreach, and risk financing. The and applications of macro-level index insurance shared nature of the functions both strengthens the programs could be explored as part of an integrated arguments for a PPP framework and influences the risk management framework (figure 1). institutional framework’s design. The Kenyan agricultural insurance market The PPP framework for agricultural insurance /// /// and wider risk management will need to is subject to market inefficiencies that the support public and private sector institutions GoK can help to overcome through a number in identifying, developing, and distributing of mechanisms. These include (i) collecting /// the appropriate risk transfer solution to reliable agricultural insurance data, (ii) conducting each segment of the farming population. /// appropriate outreach to potential policyholders, (iii) Currently, most traditional indemnity-based crop providing or supporting the risk financing of the and livestock insurance in Kenya is targeted at small catastrophic layer of reinsurance, (iv) supporting to medium-size commercial farmers and dairy cattle the design of appropriate insurance products, and producers. On the other hand, index insurance is (v) establishing and implementing an enabling legal being promoted by the donors as a small-scale farmer and regulatory environment. Figure 1 —  Toward an integrated private risk management and insurance framework for different segments of Kenya’s crop and livestock producers • Medium/large farm units Commercial Crop and • Commercial dairy & beef herds MPCI Livestock Producers • Mechanized production (top 5% of Kenyan • High access to credit Named farmers) • High levels of input use Peril • Produce for sale Index • Smallholder farmers Insurance Semicommercial Crop • Smallholder livestock producers and Livestock Producers • Some assets (middle 20% of Kenyan • Some access to credit farmers) • Part consumption/part sale SECTION Subsistence Farmers/ • Very small/ no land 01 Pastoralists • Very few assets (bottom 75% of Kenyan • Subsistence farming Social Safety-Net Programs: Macro-Level Crop and Livestock Index Insurance farmers) • Nomadic Pastoralists 12 K E N YA : A G R I C U LT U R E I N S U R A N C E S O L U T I O N S A P P R A I S A L The full participation of the private sector /// Insurers expect to recover product development is critical for the successful implementation costs over time through the premium paid. In of an agriculture insurance program. The /// the case of agricultural insurance, however, the following are considered to be principally private high costs and the limited financial capacity of sector functions: (i) product design and rating, policyholders make this recovery unrealistic. (ii) risk acceptance and underwriting, (iii) decisions Therefore, although product development and about risk retention and reinsurance strategies, (iv) technical support are private sector functions, the supplementary data collection, (v) the marketing of support of government and development institutions crop and livestock insurance products, and (vi) the (such as the International Livestock Research distribution of these products. As indicated above, Institute [ILRI], World Bank, UK Department for many functions are shared by the private and public International Development, and U.S. Agency for sectors. The public sector plays a role in both risk International Development [USAID]) is likely to be financing and data collection; and although the necessary, at least in the short to medium term. private sector is responsible for product design and rating, the government will have a strong interest Care will need to be taken to mitigate the risk /// in the price of the product—and therefore in the of crowding out private sector innovation or product’s rating—where it provides a subsidy. of subsidizing tasks that the private sector is able to undertake. Once products have been Product development and ongoing technical /// /// support are costly. Given the actuarial and other /// developed and demonstrated to be actuarially sound, specialist expertise required to design and price insurers should be able to support their continued new actuarially sound and sustainable agricultural development; and once agricultural insurance insurance products, and to support their ongoing has reached scale, the premiums should be able development, the costs are likely to impose a to support the costs of developing new products significant entry barrier to commercial insurers. without public sector support. Figure 2 —  Potential public sector roles for GOK in support of agricultural insurance development in Kenya Data Outreach Risk Financing Collect Link to social safety nets Public Sector Audit Link to Credit Reinsurance Manage Premium Subsidies Promote Coinsurance Pool Finance Awareness Building Financial Support Support Product Enabling Environment Design & Development SECTION Product Development & Institutional Framework Pricing (Short Run) Legal Framework 01 Technical Support for Consumer Protection Insurers (Long Run) K E N YA : A G R I C U LT U R E I N S U R A N C E S O L U T I O N S A P P R A I S A L 13 Functions Of The Public Sector on an ongoing basis and informs risk management strategies and systems, and it offers other benefits Data Collection, Auditing, And Financing as well: (i) reliable data can be used for Ministry of Agriculture, Livestock and Fisheries (MALF) policy Effective insurance solutions require good- decisions (about subsidies for fertilizer, water, /// quality data; without such data, sustainable seed, or irrigation); (ii) they can improve farm- insurance markets are unlikely to develop. level understanding of risks to empower farmers /// For insurance purposes, data must be sufficient and to undertake better risk management techniques adequate to enable the design and rating of products; (crowding in of good mitigation); and (iii) they they must be relevant, so that products offer reliable protection; they must be reliable enough can put a price on risk by (for example) informing to be accepted by international reinsurers, whether farmers if they should stop growing a crop in a through audit or otherwise; they must be timely, so given location. that claims can be paid quickly; and they must be The GoK intends to play an important role in cost-effective. /// collecting agricultural insurance data, both for livestock and for crop insurance. Given The different categories of risk and the /// /// that the collection and management of most data for different insurance schemes in Kenya require different types of, and investments in, data. /// agriculture insurance is expensive and nonrivalrous,5 For example, crop insurance and livestock insurance the function is usually more efficiently undertaken require different types of data available from different through a monopoly. For example, it does not make sources (such as ground-based data or remote- economic sense for every insurer to set up its own sensing data, including satellite data, on agricultural weather stations in the same area to capture the production or weather variables). Investment in same data. Thus the public sector has a natural reliable data allows monitoring of risk dynamics role to play. In Kenya, as in many countries, the Table 1—  Agricultural insurance data collected by GoK Data Type Public Institution in Charge of Collection Kenya Meteorological Department under the Meteorological data Ministry of Environment, Water and Natural Resources Kenya National Bureau of Statistics and Ministry Time series crop production and yield data of Agriculture, Livestock and Fisheries (MALF) Crop and livestock damage data MALF Arid Lands Resource Management Program Further livestock statistics (ALRMP) and USAID’s Pastoral Risk Management Project SECTION Source: Government of Kenya 2014a. 01 14 K E N YA : A G R I C U LT U R E I N S U R A N C E S O L U T I O N S A P P R A I S A L collection of agricultural insurance data is largely Given the current lack of high-quality /// coordinated by government agencies. This is true agricultural data, Kenya requires a strong for all agricultural insurance data apart from claims audit function to ensure data quality and data, which insurers collect themselves (see below). access to international reinsurance markets. /// There may be other sources of data, but the main Agricultural shocks are covariate in nature, so it is important to off-load some of this risk outside the responsibility for collection lies with public sector country through international reinsurance markets. institutions (see table 1). But reinsurers have high standards for the data The GoK intends to investigate the benefits of /// they use to develop and price insurance products, and they charge significantly higher premiums if outsourcing some parts of the data collection they have concerns about how the data are audited. to private providers. This approach has been tried A transparent process for auditing insurance /// in India, for example, where crop-cutting experiments data will ensure the quality of the data and in (CCEs) that support area yield indexes for insurance turn allow local insurers to leverage international are outsourced by several state governments to reinsurance markets. private sector agents; it is too soon to judge whether this approach will be successful, however (World Some of the concerns about data quality are /// being addressed through a series of data Bank 2011b). Outsourcing does not make the activity collection guidelines developed by the GoK. a private sector function but rather an outsourced /// These guidelines are currently under review and public sector function. This is an important include the Kenya Agricultural Data Collection and distinction, as ownership of the function suggests Management Guideline, a complementary training control; where public functions are outsourced, manual, and a list of standards and guidelines for greater checks and balances will need to be built into food and agricultural data collection. However, much the structure to protect the public sector interest. work remains to be done, including (i) implementing the guidelines, (ii) providing for integrated databases Most publicly collected agricultural insurance /// of agricultural insurance data, and (iii) introducing data are perhaps not of insurable quality. Data /// clear protocols regarding access to and charges levied are often incomplete, missing, or unavailable. The for use of agricultural insurance data. MALF report (GoK 2014a)6 suggests various reasons The discussion above suggests that for this: /// considerable investment is required in the 1. Data collection coverage is low. The Kenya collection, management, and audit of data. /// Meteorological Department operates 92 synoptic To avoid wasted investment, it would be prudent to agrometeorological automated weather stations undertake a preliminary analysis of the data available across Kenya, but they are mainly located in the from public and private sector sources in Kenya. This major towns in the central and southern regions; analysis could be used to: total coverage of the country would require more 1. Produce a data gap analysis. than 1,250 automatic weather stations. 2. Determine how to fill the data gap with 2. There has not been a farm-level census since 1999. agricultural insurance products whose design requires minimum investment in the 3. MALF field extension officers are underfunded. data infrastructure. SECTION 4. MALF data on crop production seem to be 3. Explore the extent to which data can be sourced unavailable or not systematically maintained. externally as a substitute for local data (e.g., using 01 satellite or remote-sensing data). K E N YA : A G R I C U LT U R E I N S U R A N C E S O L U T I O N S A P P R A I S A L 15 Outreach Photo Credit: Daniel Clarke The GoK intends to provide general outreach /// campaigns, it should ensure that associated products support in relation to agricultural insurance are developed and offered in tandem. with the objective of expanding market The GoK will consider various ways awareness. Achieving scale is fundamental to the /// to support outreach for agricultural /// sustainability of agricultural insurance programs insurance products: because it helps to spread the cost of providing /// insurance among numerous policyholders. 1. Linking to rural lending. Rural banks and /// /// However, low levels of financial literacy in the target microfinance institutions have the potential to market and poor understanding of the potential reach a large number of rural farmers in Kenya. benefits of insurance often prevent programs from Linking agriculture insurance to rural credit reaching scale. Although the marketing of specific may help to promote very broad outreach while at the same time deepening access to financial insurance products is a function of the private services (i.e., both credit and insurance). The sector, government can play a more general role imposition of a legal obligation to purchase aimed at building financial literacy among potential insurance on taking agricultural credit can lead policyholders and at helping them understand the to poor incentives. However, banks may impose types and potential benefits of agricultural insurance. the requirement as part of the package they offer The government should exercise caution /// farmers, and this requirement can be supported in this role, however. Experience has shown /// by government. that government consumer education and 2. Conducting financial literacy campaigns. /// /// marketing campaigns may be unsuccessful and even Unless potential policyholders have a basic counterproductive if the insurance products are level of financial literacy, it will be difficult for not available (for example, because insurers do not insurers to sell agricultural insurance products. have the necessary distribution channels in place) With a greater degree of financial understanding, SECTION or if insurers are not trusted (for example, because farmers can better weigh the risks and benefits of slow claims payment and low claims ratios). As of insurance products. It is expected that county the government develops any financial awareness governments will play an essential role in this. 01 16 K E N YA : A G R I C U LT U R E I N S U R A N C E S O L U T I O N S A P P R A I S A L 3. Raising awareness of insurance. Beyond /// /// Risk Financing basic financial literacy, potential policyholders Given the high costs for development, need to understand the types and benefits of /// distribution, claims assessment, and risk agricultural insurance if they are to purchase it. financing, agricultural insurance is unlikely Raising awareness of insurance should be regarded to succeed without some public sector as a shared role. Government may be better able subsidy. Development costs are an upfront charge; /// than the private sector to utilize the media, but distribution costs can be mitigated through the campaigns are unlikely to be effective unless the development and use of alternative distribution private sector also plays a role, specifically by channels; and claims assessment costs can be providing effective training to insurance agents and mitigated through product design. The cost of by developing clear product documentation. financing the risk, however, is ongoing and must be met on a year-to-year basis. An insurance product 4. Linking to the Hunger Safety Net Program /// cannot be sustainable unless the risk financing costs (HSNP). Linking livestock insurance to the HSNP /// are fully met. It is therefore perhaps inevitable that should help to increase outreach by targeting the public sector will have to provide support for risk based on poverty data that was collected as part financing. of the HSNP. This linkage will also serve to lower Insurance offers governments an efficient transaction costs by enabling more efficient /// mechanism for providing financial support collection of premiums and distribution of claims. to vulnerable farmers and pastoralists in the event of crop failure or significant livestock Although the Insurance Regulatory Authority (IRA) is losses. Well-designed insurance products are an already engaged to a limited extent in raising public /// efficient method of transferring extreme agricultural awareness of insurance, the GoK and the county risk. But agricultural insurance is unlikely to be governments also have roles to play. The GoK’s role purchased by vulnerable farmers and pastoralists. in financial literacy and market awareness campaigns Thus governments may decide that purchasing is to develop a strategy; the role of the country insurance on their behalf is more efficient than governments is to lead implementation through the relying on other support mechanisms such as post- devolution process. disaster relief. Photo Credit: Daniel Clarke SECTION 01 K E N YA : A G R I C U LT U R E I N S U R A N C E S O L U T I O N S A P P R A I S A L 17 Governments often provide support for the /// This form of risk layering offers a number /// financing of risk through direct premium of advantages: /// subsidies with the objective of incentivizing 1. By covering the catastrophic layer of risk, the insurers to enter the market and increasing government reduces the premium paid by agricultural the take-up of insurance product. However, this producers, as the premium does not include the /// approach has potential drawbacks, and there may be price of the catastrophic risk. better ways for government to reduce the premium cost to farmers than direct premium subsidies, such 2. If the government decides to withdraw the as through risk financing. subsidy, the nonsubsidized commercial layer can still continue to be sold on a sustainable basis, Providing public stop-loss reinsurance would /// as the noncatastrophic risk is fully priced. build on international experience that has The commercial layer will cover all but the demonstrated the efficiencies gained by catastrophic risk. splitting the risk into layers. For example, there 3. Significant efficiencies are obtained through the effects /// are three layers of risk under the Mongolian livestock of risk pooling at the national level, in both the insurance scheme, which has now reached national commercial layer and the catastrophic layer. scale: 4. Government can optimize the cost of capital by 1. The first layer of risk (up to 6 percent livestock managing the amount and type of reinsurance, mortality), which covers the more frequent or other types of risk transfer instruments, that it low-impact events, is borne by the insured purchases. Selectively transferring a portion of livestock herders. the catastrophic risk to the international market, 2. The second layer of risk (between 6 percent and and retaining the balance of the risk, lowers the 30 percent livestock mortality) is covered by government’s total costs; an indirect premium commercial insurers, through a pool, for which the subsidy costs significantly less than direct policyholders pay a fully priced rate.7 This is the premium subsidy. noncatastrophic layer of risk. This risk financing approach could be /// 3. The third layer of risk (over 30 percent livestock considered for crop insurance in the Kenyan mortality) is covered by the government under context. Given the limited availability of data and /// a stop loss agreement entered into with the need to develop affordable products for farmers, commercial insurers. This is the catastrophic layer the government’s assumption of a role in risk of risk. The government does not charge for the financing could have significant benefits for a crop stop loss agreement. insurance program. In order to achieve the most efficient pricing for the risk, the government intends The commercial insurers reinsure part of their to consider in the medium term a risk layering liability under the commercial layer to the approach— similar to that used in Mongolia—under international reinsurance market, while the which it provides support for the higher layers government reinsures a portion of its risk under the of risk. catastrophic layer to the international reinsurance SECTION market (Mahul and Skees 2007). 01 18 K E N YA : A G R I C U LT U R E I N S U R A N C E S O L U T I O N S A P P R A I S A L For livestock insurance, where the objective /// sound and sustainable agricultural insurance is to reduce the vulnerability of pastoralists, products and to support their ongoing development, direct premium subsidy may be necessary in the costs are likely to impose a significant entry the short term. In the initial years, any livestock /// barrier to commercial insurers. Insurers expect insurance product should be based on high-quality to recover product development costs through satellite data; these products would not be subject the premium, over time. However, in the case to large increases in the premiums for poor data of agricultural insurance, the high costs and the quality. In addition, as the primary objective of the 8 limited financial capacity of policyholders make this livestock insurance program is to reduce vulnerability unrealistic. Thus although product development and of households in Kenya’s arid and semi-arid lands technical support are private sector functions, the (ASALs), the beneficiaries will be low-income support of government, together with development households that would not be able to afford to pay institutions (such as ILRI and the World Bank), is for the insurance. Thus the provision of premium likely to be necessary, at least in the short to medium subsidies by the GoK could be considered a viable term. option; the GoK intends to make sure that subsidies Care will need to be taken to mitigate are clearly targeted and options considered for their /// the risks of crowding out private sector gradual withdrawal over time. In particular, the innovation or subsidizing tasks that the GoK will carefully consider the challenges linked to private sector is able, and would otherwise maintaining the long-term financial stability of the be willing, to undertake. Insurers should be able insurance scheme and will consider devising a clear /// to support the continued development of products exit strategy or long-term financing before embarking once they have been found to actuarially sound, and on premium subsidies. premiums should be able to support the costs of The provision of agricultural insurance /// developing new products once agricultural insurance through a coinsurance arrangement is has reached scale. recommended later in this chapter. Although Setting And Implementing An Enabling Legal /// the establishment of nonstatutory coinsurance pools And Regulatory Environment is a private sector function, the initial push for this effort may need to come from the public sector. (See Efforts to establish an enabling environment /// below for more details.) for insurance should take a number of general considerations into account. Traditional /// Support for the Design and Ongoing indemnity-based agricultural insurance should be Development of Insurance Products regulated like any other line of insurance, although special regulatory provisions may be required in As stated above, there may be need for relation to catastrophe risk. Recognizing that the /// public sector support in product development current Insurance Act and Regulations do not enable and on-going technical support in the Kenya to comply with international standards, the short to medium term, with the support of IRA has led the process to develop a new Insurance Government together with development Bill and Insurance Regulations that would enable institutions (such as ILRI and the World Bank). /// substantial compliance with international standards. SECTION Product development and ongoing technical /// For index insurance, an appropriate legal /// support are costly. Given the actuarial and other framework needs to be established. Given that 01 /// /// expertise required to design and price new actuarially index insurance pays against an agreed-upon index K E N YA : A G R I C U LT U R E I N S U R A N C E S O L U T I O N S A P P R A I S A L 19 rather than on the basis of actual losses, there is some issue consumer protection regulations that cover question about whether index risk transfer products Poor-value products can be properly classified as insurance at all. As the • Lack of disclosure current Insurance Act does not recognize index- based insurance, the introduction of index insurance • Unfair contract terms products carries both legal and regulatory risk. Under • Delays in insurance payments the proposed new Insurance Bill, index risk transfer products can be classified as insurance, subject to A number of countries have specific /// certain general criteria. The bill also provides for agricultural insurance legislation. This/// supporting regulations concerning index insurance to legislation is not usually intended to cover regulatory be issued by the IRA. The enactment of the bill and and supervisory issues, but rather to make statutory issuance of regulations would significantly reduce the provision for a specific institutional framework legal and regulatory risks associated with developing (such as a statutory coinsurance pool or statutory new index insurance products. The government reinsurance arrangements) and to govern the intends to expedite the legislative process. provision of subsidy. In relation to subsidy, the legislation may obligate government to provide a The primary responsibility for the /// certain level of subsidy, to take the subsidy outside implementation of the legal and regulatory the usual budgetary process, and/or to establish a framework for insurance lies with the IRA. /// framework or arrangements that govern the use of Once the new Insurance Act has been enacted, the the subsidy and ensure it is not mischannelled or IRA will need to issue appropriate regulations. We used inefficiently. This framework could include recommend that the IRA consider including at least a body to make decisions relating to the subsidy, the following in relation to index insurance: audit processes, etc. Whether such legislation is 1. Detailed criteria for determining whether an index required in Kenya will depend on the institutional product can be classified as insurance framework that is eventually adopted and the level and types of subsidy that are to be provided for in 2. Rules allowing for composite (i.e., index and the long term. It is therefore too soon to make any traditional) products and dual-trigger products recommendations. 3. General requirements in relation to indexes aimed at reducing basis risk Driving The Process For Change 4. Restrictions on persons to whom index insurance Considerable work is required to build the /// may be sold (aimed at ensuring an appropriate necessary foundations for agricultural insurable interest) insurance, to design and market appropriate products, and to establish an appropriate 5. Key requirements for issues to be included in the institutional framework. As discussed, these policy document /// tasks will require an effective PPP. Without the 6. Specific provisioning requirements active involvement of both the public and private 7. Consumer protection requirements sectors, it will not be possible to develop a mature, scaled-up agricultural insurance market in Kenya. Consumer protection is relevant to both /// However, it is unlikely that the process will even traditional and index insurance. Consumer /// commence unless the GoK takes the initiative, protection concerns are often exacerbated in rural encouraging insurers to engage and to collaborate, SECTION settings, where farmers lack financial literacy and a for example, through a coinsurance pool. The GoK full understanding of both the product’s details and intends therefore to mobilize and allocate adequate its broader implications. We recommend that the IRA financial and human resources to lead this process. 01 20 K E N YA : A G R I C U LT U R E I N S U R A N C E S O L U T I O N S A P P R A I S A L Functions Of The Private Sector may be an acceptable substitute or proxy for data that are not available in Kenya (such as crop or Data weather data). For example, Normalized Difference Vegetation Index (NDVI) data are available from Together with the public sector, private sector the U.S. National Oceanic and Atmospheric /// insurers play a primary role in the collection Administration. Where data in the public domain of some product-specific data. Although the or available from commercial providers will enable /// collection, management, and audit of aggregate product design, it may be more efficient to use weather and agricultural data are primarily public these data than to establish systems for collecting, sector functions, commercial insurers have functions managing, and storing data in Kenya, even if the related to collecting and storing product-specific public sector contributes toward the cost. The data, such as data relating to sales, distribution, feasibility of using such data should form part of the and claims. Moreover, international reinsurance “data gap analysis” recommended above. companies require a party other than the government to be involved in either collecting or auditing data to Outreach ensure independence and transparency. This leaves a key role for the private sector. Outreach and product marketing are /// primarily private sector functions. As The private sector could also cover some or /// indicated, the public sector may have a role to play /// all of the cost of collecting and managing in raising financial literacy and general awareness agriculture data. For example, an access fee could of agricultural insurance, but outreach should be /// be levied on all parties that wish to use the data. This regarded as part of distribution, which is clearly a approach has been adopted for Motor Third-Party private sector function. Insurers sell insurance, and Liability in Turkey: the government is responsible even if public sector agencies are used as part of the for the collection and management of data, while distribution process, distribution remains a private all insurance companies that wish to use the data to sector function. Furthermore, the private sector develop and price insurance products must pay an may be better able than the public to (i) employ (equal) access fee. What is important here is that innovative distribution channels;9 (ii) leverage the data are equally available to all users on the same the significant outreach infrastructure in place; terms, an arrangement that encourages competition. and (iii) respond quickly to shifts in the market. As the design and rating of agricultural /// Most importantly, however, competition among insurance products are also private sector private insurers can increase speed, scale, and the functions, private sector insurers should effectiveness of outreach. play a role in advising the GoK on their data needs. That is, they should specify (i) the data they /// Design And Development Of Agricultural require; (ii) the form in which the data are required; Insurance And Related Tasks and (iii) the quality of the data. Insurers are responsible for the design /// The private sector can play a key role in /// and development of agricultural insurance developing and providing commercially products, although they may receive SECTION available data. Data that are publicly available /// financial and other public sector support at no charge or from commercial providers, such in the short to medium term. Such support 01 /// as remote-sensing data (including satellite data), may be necessary in the early years when the costs K E N YA : A G R I C U LT U R E I N S U R A N C E S O L U T I O N S A P P R A I S A L 21 would be unsupportable through the premiums.10 However, design and development remain private sector functions. Insurers are required by the regulatory regime in Kenya and elsewhere to take full responsibility for the insurance products that they sell, including the actuarial pricing of those products. Specialized professional and technical skills and experience are required to design, develop, and price all insurance products, including agricultural insurance products. Where insurers do not have the resources in house, they are permitted to outsource them, but insurers remain fully responsible for all outsourced services, including those provided by or through the public sector. The institutional framework must be designed with this in mind. Claims adjustment and settlement are also /// private sector functions. The comparative /// advantage of private insurers here is founded on their (i) existing outreach channels; (ii) knowledge of the clients (as they are responsible for distribution); and (iii) greater ability to innovate. A good example of private sector innovation is offered by India, Photo Credit: Daniel Clarke where cell phone technology is used to video record, geotag, and upload the results of CCEs to a database, important that agricultural insurance underwriters allowing insurance companies to access the data in and loss adjusters receive the appropriate specialist real time (World Bank 2011b). This mechanism has training. To ensure the long-term sustainability of improved the quality of the CCE procedure—by the approach, and given the expertise of private enabling insurance companies to witness the CCE insurers, this function should be taken on by the being carried out, the video recording acts as an audit private sector. However, this is another area in which mechanism—and it has also made the CCE procedure public financial and other support could be provided more timely, which greatly speeds up the process of in the early years, particularly in relation to new and payouts for area yield index insurance (AYII). A final technical areas, such as index insurance. reason for the private sector’s comparative advantage is that it is better suited to respond to the potential Risk Financing complexity of claims adjustment processes. Underwriting agricultural insurance products /// Private insurers must properly train their /// and financing the risk is a core private sector insurance and distribution staff. Given the function. The insurance business involves the SECTION /// /// highly technical nature of insurance production, it acceptance of insurance risk and the financing of is important that insurance staff have the required that risk. Although the public sector may have some skills to carry out their tasks. It is particularly risk financing functions, the function primarily 01 22 K E N YA : A G R I C U LT U R E I N S U R A N C E S O L U T I O N S A P P R A I S A L Photo Credit: Daniel Clarke belongs to private sector insurers. Insurers are Insurers may reinsure their insurance risk /// required by the legal and regulatory framework, and with national, regional, or international the IRA, to take responsibility for the management reinsurers as a substitute for holding and financing of their insurance risk. capital to support that risk. The negotiation /// and conclusion of reinsurance contracts is part of Through pooling and diversifying their insurers’ risk management process. Thus even where /// insurance risk, insurers are able to reduce the the public sector offers risk financing support, for price of the risk, which should result in lower example in relation to catastrophe risk, insurers premiums to policyholders. By underwriting must decide whether that support is adequate to /// through a coinsurance pool, as elaborated later in the enable them to underwrite the products. SECTION chapter, private insurers can significantly lower cost to policyholders. 01 K E N YA : A G R I C U LT U R E I N S U R A N C E S O L U T I O N S A P P R A I S A L 23 • The institutional frameworks already operating in Institutional Framework other countries and the experiences and lessons First Steps learned in those countries • The applicability of international experience in the Establishment Of A Coordinating Body Kenyan context Although significant work on the institutional /// • The legal and regulatory implications, including framework has already been undertaken, whether specific legislation or regulations will be further work is required and important required policy decisions must be made before the framework can be finalized. One issue, for /// The work of the task force will contribute toward example, involves how much financial and other formulation of the National Agricultural Insurance support the GoK is prepared to provide to agricultural Policy (NAIP), as recommended below. insurance in the short, medium, and long term. The considered views of stakeholders will need to National Agricultural Insurance Policy be sought, including various GoK departments and (NAIP) agencies, county governments, the IRA, and insurers. The MALF report recommended that the GoK The design of a firm and final institutional framework /// expedite the policy process for formulating at this stage would therefore be premature. and finalizing the NAIP, which would serve International experience demonstrates /// as a guiding framework for developing the that agricultural insurance is more likely to Kenyan agricultural insurance market (GoK succeed under a PPP that is formalized within 2014a). The NAIP should address the following: /// a well-designed institutional framework. /// • The GoK’s objectives for agricultural insurance, International experience has also demonstrated that including social objectives, such as preferential the establishment of the institutional framework is promotion and support programs for agricultural a necessary precondition for the design of specific insurance for small and marginal farmers agricultural insurance products. One of the reasons • Definition of the functions, roles, and obligations why many donor-funded pilots fail to scale up is the of each party to the PPP lack of institutions to follow through once the donors or development agencies have left. It is important, • Establishment of the institutions most suitable therefore, to give priority to the institutional for delivering the functions the GoK wishes framework, even ahead of product design. implemented The government intends to establish /// The GoK intends to make formulating the NAIP a a Program Steering Committee with priority. Once finalized, the NAIP will provide the representation from the GoK and the private blueprint for the institutional framework. sector that will examine options for an institutional framework and specifically The formulation of the NAIP will be /// consider the following: /// considered a process rather than a discrete task that can be completed in the immediate • The appropriate functions of the public and future. The work undertaken by the task force private sectors /// should therefore feed into the development of the SECTION • The options for an institutional framework, NAIP, which should be regarded initially as a work in building on those presented in the MALF report progress. As the work moves forward, the NAIP will (GoK 2014a) be adjusted accordingly. 01 24 K E N YA : A G R I C U LT U R E I N S U R A N C E S O L U T I O N S A P P R A I S A L Interim Framework could be given to establishing the iHub as soon as possible within an existing Considerable work is required on product institution, such as MALF. The iHub’s function /// development. This is likely to include an analysis /// /// and work undertaken could then be transferred to of the data required and available, the costs of designing agricultural insurance products, the market another institution when the PPP is fully established demand (including the willingness and ability of (or remain with MALF). potential policyholders to pay for the insurance), and the appetite of private sector insurers in Kenya and Coinsurance And Coinsurance Pools national, regional, and international reinsurers to As it is unlikely that a fully competitive participate in agricultural insurance. /// insurance market will be viable in Kenya, the Existing institutions could be used on /// task force should consider establishing a an interim basis to commence the work. coinsurance pool. A competitive Kenyan market /// The functions could then be absorbed into the /// is hampered by the high costs of designing and institutional framework, once finalized. For example, distributing agricultural insurance to small farmers. the MALF report recommended the formation of a national agricultural insurance Web-based data and Hence some form of cooperation between insurers information iHub; this would link end-users, including is needed. Establishing a coinsurance pool would agricultural risk managers, insurers, and MALF staff, meet this need and also enable the pooling of risk, with the main institutions involved in agriculture and which should result in lower insurance premiums. agricultural risk management and with their databases The concept of coinsurance is further elaborated on (GoK 2014a). An iHub will be needed, whatever the in Annex 1. institutional framework eventually will be, and work on it could start immediately. The MALF report There are many ways to structure a /// specifically suggested that coinsurance pool, each with different “the starting point for the iHUB project would be features, advantages, and disadvantages. to define exactly what minimum (priority) key The core principles are detailed in box 2. /// data is required for agricultural insurance purposes and to then check with . . . organisations what data and information they currently hold in their own databases, and the software formats of this data and Box 2—  Core principles for a time-series available and missing data. This would coInsurance pool result in the production of a data and statistics 1. Insurers share the costs of certain core activities, such catalogue covering the data held by each organisation as product design and pricing. (GoK 2014a, 199).” 2. Certain administrative costs are shared, such as Defining the priority key data for agricultural /// claims administration. insurance could be used to undertake the 3. Other activities may be shared, depending on the pool gap analysis recommended above. The detailed /// design, including handling of distribution costs. SECTION proposal is set out in GoK (2014a). 4. There is at least some risk pooling. This may include Given that the data work is a foundation block 01 presenting a pooled portfolio of insurance to reinsurers, /// for future product development, consideration enabling a lower reinsurance cost. Risk pooling should reduce the cost of risk, which would then lower the cost of the premiums. K E N YA : A G R I C U LT U R E I N S U R A N C E S O L U T I O N S A P P R A I S A L 25 Institutions Photo Credit: Daniel Clarke Given that public sector functions are Whether or not a separate entity is /// established, certain core functions will need /// spread between different GoK ministries, to be undertaken. These include /// departments, and bodies, it is important (i) coordinating the implementation of the PPP to ensure that public policy on agricultural from a policy perspective; insurance is effectively coordinated. The (ii) conducting original risk assessment and risk /// MALF report recommended that the GoK consider mapping studies on behalf of MALF; establishing the Agricultural Risk Management (iii) coordinating the implementation of the NAIP Agency to coordinate public policy and support the with the private sector insurers; individual private sector companies that sign up for (iv) assisting private sector insurers in product the PPP (GoK 2014a). A separate entity of this kind marketing and education programs for farmers, including the allocation of subsidies; would help to ensure that the GoK could effectively carry out its functions in the PPP. The task force (v) providing data and statistics and general assistance related to agricultural insurance will consider this and other possible options for products; ensuring that the policy agenda is driven forward (vi) conducting program research and development; and that the PPP is implemented. If a separate entity and is established, costs should be kept to a minimum, SECTION (vii) coordinating donor technical assistance meaning the entity will be small with a core staff programs for agricultural risk management and of specialists. insurance in Kenya. 01 26 K E N YA : A G R I C U LT U R E I N S U R A N C E S O L U T I O N S A P P R A I S A L The institutional framework will need to /// cover monitoring, supervising, accounting, and auditing for any public sector subsidy provided as well as advising the GoK on the size of the subsidy. The National Treasury will /// have a key interest in this function, which could be housed within it or within the coordinating unit. The National Treasury clearly has strong experience in public financial management, but it would be necessary to ensure that the Ministry of Finance staff also has, or has access to, the technical capacity to undertake this function. Given the high costs of technical tasks related /// to agriculture insurance, the GoK should consider establishing a technical support unit (TSU) to house technical expertise centrally. /// As already discussed, technical functions belong to the private sector. Given their costs, however, there is significant advantage in having insurers coordinate and centralize these functions, and the GoK may choose to support them. TSUs are typically present in countries where some degree of competition exists among private insurance providers or distributers. A TSU can have a wide range of responsibilities, such as (i) data analysis; (ii) insurance demand assessments; (iii) product design and rating, including basis risk analysis; (iv) design of operating systems and procedures; (v) training for stakeholders; (vi) awareness campaigns; (vii) analysis of any public subsidies; and (viii) the development of catastrophe risk models and other risk assessment tools. However, the establishment of a TSU is not /// the only option. For example, if the private insurers /// went the route of a fully incorporated, capitalized, and staffed pool insurance company, the TSU would not be required, since its functions could be performed by the managing underwriting unit of the pool. SECTION 01 K E N YA : A G R I C U LT U R E I N S U R A N C E S O L U T I O N S A P P R A I S A L 27 Livestock Insurance for Pastoralists Located in ASALs in Northern Kenya severe droughts. Because they rely on livestock Context for their livelihoods, high livestock deaths can Kenya’s Experience With have devastating effects, rendering many of these Livestock Insurance households among the most vulnerable in Kenya. The economic analysis presented in section 2.4 Kenya has a lengthy history of livestock /// shows that without any form of livestock insurance accident and mortality insurance for the protection, the poorest households (<5 Tropical commercial livestock dairy sector, but until Livestock Units, or TLUs13) and vulnerable poor recently the insurance market did not offer (5 - 10 TLUs) are very likely to lose all their livestock, any cover to meet the risk transfer needs of and therefore their livelihoods, in severe drought the many resource-poor pastoralists located events. in the arid and semi-arid lands (ASALs) of northern Kenya. Following the devastating drought /// To help pastoralists manage drought risk and /// losses in the livestock sector between 2008 and protect their animals, insurance solutions 2011—an estimated 9 percent of the national cattle were developed by the International herd was lost, with total livestock losses valued at Livestock Research Institute (ILRI), together K Sh 699 billion (GoK 2012)—the government has with its technical partners at Cornell signaled its major commitment under the Second University and University of California–Davis. /// Medium Term Plan (2013–2017) to provide funding The logistical challenges of working in the ASAL regions suggested that an index-based insurance for a national livestock insurance scheme (MDP product would be appropriate. Developing the 2013). Insurance Based Index Product (IBLI) involved two Drought is the most pervasive hazard, natural /// years of comprehensive research aimed at designing, or otherwise, encountered by pastoralist developing, and implementing market-mediated households in the ASAL regions; it can lead to index-based insurance products that livestock widespread death of livestock and severely keepers—particularly in the drought-prone ASALs— SECTION deplete livestock assets for the affected could purchase to protect themselves from drought- households. Many pastoralist households in related asset losses. The IBLI product is based on 02 /// the ASALs are now regularly hit by increasingly a satellite Normalized Difference Vegetative Index 28 K E N YA : A G R I C U LT U R E I N S U R A N C E S O L U T I O N S A P P R A I S A L (NDVI)14 cumulative-season drought index, which State Department of Livestock is combined with a predicted livestock mortality Interest in Large-Scale Drought index to insure pastoralists against drought-related Insurance For Pastoralists in deaths to their livestock (cattle, camels, sheep, and ASAL Regions goats). It provides full-value animal cover to enable the insured pastoralists to restock their herd after the As part of its plans to promote and /// drought event. strengthen livestock insurance provision in the ASALs, the government of Kenya (GoK) The commercial sale of IBLI was launched in /// has proposed creating a national livestock Marsabit, northern Kenya, in January 2010 as insurance scheme under the Second Medium a voluntary retail insurance product and was Term Plan. An indicative budget of K Sh 2,000 /// marketed to individual pastoralists. The IBLI /// million–2,500 million over the fiscal years 2013/14 demand assessment studies identified affordability to 2017/18 was identified to support the national as a constraint to uptake, and since launch in 2010 livestock insurance scheme (MDP 2013). donor partners have financed premium subsidies in the order of 40 percent of the full premium With this objective in mind, the GoK, through /// costs. In 2010, UAP Insurance Company was the the State Department of Livestock (SDL) underwriter, while Equity Insurance Agency was the within the Ministry of Agriculture, Livestock insurance agent. Swiss Re provided reinsurance for and Fisheries (MALF), approached the World the product. The IBLI program has gone through Bank’s Disaster Risk Financing and Insurance various adjustments since it was launched, and Program in 2014 to ask for technical support APA Insurance Company became the underwriter in developing a public private partnership for Marsabit and Isiolo Counties in August 2012 (PPP) in livestock insurance to support and August 2013 respectively. In Wajir County, a pastoralists. The MALF-World Bank team /// Sharia-compliant version of IBLI is currently being partnered closely with ILRI and the Financial Sector implemented by Takaful Insurance Company with Deepening (FSD) Kenya program to benefit from the support from Mercy Corps. considerable practical experience those institutions have in Kenya. While the current program has driven /// innovation in IBLI product development, pricing, and distribution, several challenges remain to be met to achieve large-scale uptake. In 2010, when IBLI was first launched, /// pastoralists bought nearly 2,000 policies to insure about 6,000 TLUs (ILRI 2013). Since then, however, the program has struggled to achieve scale and sustainability in spite of making payouts to insured pastoralists in response to droughts in 2011 and again in 2012. SECTION 02 K E N YA : A G R I C U LT U R E I N S U R A N C E S O L U T I O N S A P P R A I S A L 29 support any transition in the medium term toward Proposals For Large-Scale increasing contributions from targeted recipients Livestock Insurance For and the pastoralist population at large. Pastoralists Located In Under the plan, SDL would be assisted in /// ASALs In Northern Kenya designing and implementing a macro-level pasture drought index insurance program; Livestock Insurance Options starting in 2015, SDL would purchase the insurance on behalf of approximately 71,000 Under the proposed plan, the SDL would vulnerable pastoralists located in the four /// collaborate with the National Drought Hunger Safety Net Program (HSNP) counties Management Agency (NDMA) to develop a of Mandera, Marsabit, Turkana, and Wajir. large-scale index-based livestock insurance /// The SDL would insure itself at the national level, program to cover pasture drought risk. In /// with insurance payouts triggered by a satellite-based order to quickly build a critical mass of covered households, a macro-level product will form the index at a local/pastoralist level. The SDL would then foundation of a sustainable livestock insurance channel payouts to pre-identified pastoralists on the market. Grafted onto this will be the concrete triggering of the index, or underwriting insurer(s) provision of “top-up” cover for targeted beneficiaries would provide payments to these pastoralists who wish to expand their coverage, and a voluntary directly. An element of cost sharing between central SECTION purchase cover for all nontargeted pastoralists. This 15 and county governments will be explored, with the structure is designed to gauge the level of untapped plan initially proposing that the entire cost of the demand for voluntary livestock insurance and to compulsory coverage be paid by SDL. 02 30 K E N YA : A G R I C U LT U R E I N S U R A N C E S O L U T I O N S A P P R A I S A L Start-up implementation of the SDL macro- /// government subsidies for the top-up and voluntary level pasture drought index insurance purchase options have tentatively been set at 50 program is recommended for the four HSNP percent and 25 percent, respectively, but these counties because infrastructure systems may change; final percentages will be guided by and procedures are already in place there. the program experience with partially subsidized This infrastructure allows for (i) verifying voluntary contracts provided at the outset. identification and registration of eligible pastoralists who will be the beneficiaries Integrating Existing Social Protection of this insurance; and (ii) delivering timely and Insurance Programs for insurance payouts in kind or in cash to Pastoralists in the Target Counties the individual beneficiaries. In the HSNP /// Linkages Between Programs in the ASAL Region counties, the NDMA Secretariat has partnered with implementing nongovernmental organizations to The overall framework for the insurance /// register 375,000 households and their dependents program should be mindful of various and to classify them into four main income/ insurance products already being distributed wealth status categories or poverty bands. The and social protection measures currently HSNP program is currently targeting the poorest being developed in the four HSNP counties. /// 100,000 households under its regular program of These include bimonthly cash payments and plans to scale up this program in times of extreme drought. The goal is to • The ILRI-developed IBLI contracts currently complement the HSNP program by implementing being offered by APA Insurance and Takaful the SDL insurance program with approximately Insurance Africa on a voluntary basis 71,000 vulnerable pastoralists who are just above • The HSNP protection the poverty criteria for inclusion in the HSNP cash transfer program (see below for further discussion). • The new SDL-led IBLI initiative for macro-level as In addition, the HSNP payment system might be used well as top-up and voluntary coverage as a way to distribute the payouts (complementary to To avoid overlap between the three /// the use of mobile payment options). programs, the HSNP classification of To supplement the macro-level pasture /// households according to wealth/poverty drought index product, the insurance status should be used to target each program offers a top-up option for eligible insurance program to different poverty pastoralists, plus voluntary policies that groups. The poorest 100,000 households would /// would be sold to all pastoralists on an continue to be covered by the HSNP under the individual basis. The macro-level product as /// regular bimonthly cash transfer program, which is described above would be supplemented by top-up 100 percent financed by the GoK and donors using insurance policies to be purchased by pastoralists a variety of funding mechanisms, including Africa on a voluntary basis; the costs of any such top-up Risk Capacity (ARC) index insurance payouts (see cover would be shared between the GoK and the below for further discussion). Adopting the HSNP pastoralists. This product would also be offered— poverty ranking, the macro-level pasture drought SECTION and partially subsidized—for voluntary, individual index insurance program—funded entirely by the purchase by all pastoralists, independent of whether government—would apply to registered vulnerable 02 they are covered by the macro-level product. The pastoralists immediately above the HSNP’s target K E N YA : A G R I C U LT U R E I N S U R A N C E S O L U T I O N S A P P R A I S A L 31 beneficiaries. Finally, relatively wealthier (though still IBLI pilot over the past five years, (ii) offer choice low-income) pastoralist households could be covered to individual livestock producers not targeted by by the top-up and voluntary SDL cover as well as SDL for its program, (iii) enable an assessment ILRI’s IBLI product being marketed by APA Insurance to be made over the next two to three years of and Takaful Insurance of Africa. This layering the voluntary demand for livestock insurance by approach is illustrated in figure 3. individual pastoralists in the ASAL regions, and— based on the assessment findings—(iv) allow SDL The SDL macro-level product for the targeted /// to decide whether to introduce its own top-up and vulnerable category of pastoralists should be voluntary individual index-based livestock insurance linked with the ILRI-developed IBLI product products and programs. currently being offered on a voluntary basis by APA Insurance and Takaful Insurance Possible Linkages with ARC of Africa. This approach will ensure that all In 2014, Kenya and four other African /// /// pastoralists (not just those covered by the SDL countries—Mauritania, Mozambique, Niger, program) can purchase livestock insurance. It will and Senegal—joined a new African drought also allow the SDL and the GoK to do the following: index insurance facility under the ARC (i) learn from the major technical design expertise initiative (ARC 2014).16 ARC is an initiative of the /// and implementation experience gained under the Commission of the African Union’s Department of Figure 3 —  How government could support financial protection for different segments of the population: Example of pastoralists in the four current HSNP counties (Mandera, Marsabit, Turkana, and Wajir) Income Level Livestock Safety Net Cost Share and Insurance Program Middle-income 100% of premium and above Unsubsidized livestock insurance covered by farmers (>1 USD/day) Premium cost Low-income Partially subsidized livestock insurance sharing at (<1 USD/day) 50%/25% Macro-level insurance program for Ultra-poor Premium 100% approximately 71,000 ultra-poor pastoralists (<0.5 USD/day) subsidized by GoK above HSNP poverty levels Hunger Safety Net Program (HSNP), providing Hardcore poor Premium 100% scalable cash transfers to 100,000 hardcore (<0.3 USD/day) subsidized by GoK poor households SECTION Note: Classification is based on distribution of livestock holding size for Marsabit County, which may not be similar in other HSNP counties 02 32 K E N YA : A G R I C U LT U R E I N S U R A N C E S O L U T I O N S A P P R A I S A L Rural Economy and Agriculture and the World Food other circumstances would be too poor to Program, It aims to create a pan-African-owned pool afford insurance premiums. The final decision /// index insurance fund to underwrite catastrophic on eligibility will involve technical input from the weather events, initially to cover drought; in the future, coverage could be expanded to include other SDL and other ministries. However, eligibility for weather risks such as flood.17 The program is insured the public subsidy should be limited to vulnerable by ARC Insurance Company Limited (ARC Ltd.), households that reside in the four pilot counties, domiciled in Bermuda, and is reinsured by specialist where vulnerable households are defined as a international reinsurers of this class of business. specific, not-yet determined number of households In 2014, Kenya purchased protection from the /// ranked just above the eligibility cut-off point for ARC and determined that one of the primary benefits under the HSNP. purposes of the ARC program in Kenya would be to support the scalability mechanism of For the macro-level index insurance program, /// HSNP. Coverage has been purchased for both the the GoK should purchase 100 percent of /// long rains/long dry season and the short rains season (maximum payout US$30 million each). The primary cover for 5 TLUs per eligible household. /// use of the cover will be to lessen the fiscal burden to The number of households covered will depend GoK of meeting the cost of scaling up the HSNP. on the resources the SDL has available to support SDL could consider exploring possible links /// the scheme. For illustrative purposes the team has between the SDL macro-level insurance chosen annual to analyze three scenarios: budgets coverage and the ARC program in Kenya. /// of K Sh 100 million, K Sh 200 million, and K Sh 300 Given the clear complementarities between million to analyze. these programs, the synergies available should be leveraged. For top-up coverage for eligible households, /// Eligibility and Subsidy the amount of subsidy should be determined by the SDL. An initial suggestion is that the GoK /// The macro-level SDL product is intended to provide a subsidy for 50 percent of the actuarially /// provide insurance to pastoralists who, at this stage, would not be able to afford commercial calculated commercial premium, with the other 50 premiums. The GoK therefore intends to provide /// percent paid by pastoralists. This subsidy would be a public subsidy for the product. Given the subsidy, subject to a cap of 5 TLUs per pastoralist. eligibility criteria will be required to ensure that targeting is in line with the government’s objective For voluntarily purchased individual /// of supporting the most vulnerable pastoralist coverage, the GoK will need to decide households. As eligibility and subsidy both affect whether it will also be subsidized and what product design, decisions are required on these issues before the product design can be finalized. cap per pastoralist will apply. Over time, the /// GoK may plan to reduce the size of public subsidies. In principle, the macro-level program under SECTION A 25 percent premium subsidy has been tentatively /// which SDL would finance 100 percent of the insurance premiums is targeted at poor and applied, subject to change going forward (see section 02 vulnerable pastoralist households who under 3.3). K E N YA : A G R I C U LT U R E I N S U R A N C E S O L U T I O N S A P P R A I S A L 33 Design Options for SDL Macro-Level Similar approaches have been implemented /// Livestock Index Insurance Product in the following countries: /// • Spain, United States, and Canada. In these For implementation in calendar year 2015, a /// /// /// countries, the NDVI pasture drought index macro-level livestock insurance product for insurance programs operate as voluntary micro- the SDL is proposed: its central objective is level individual livestock producer programs. The to effect timely cash payouts to vulnerable cover period is defined as the normal pasture/ pastoralists at the onset of drought in order grazing growing season (which usually coincides to keep breeding stock alive. Under this scheme, with the spring and summer rainy seasons of /// the GoK would be the party entering the insurance maximum pasture and biomass production), and contract. If it received a payout, it would in turn make the basis of the sum insured is usually calculated payments to pastoralists as identified above. according to the nutritional requirements of the livestock/costs of purchasing supplementary For this macro-level cover, satellite data livestock feeds in the event of loss of pasture /// (NDVI) would be used to create a pasture and grazing due to drought. Regular payouts are drought index. This would enable the made during the cover period for each month development of an index that measures (or time period as defined) that the NDVI policy the onset of drought-related pasture and is triggered. All three programs attract heavy grazing degradation and that triggers early government premium subsidy support. payouts (so that pastoralists can purchase • Mexico. The federal and state governments animal feeds to keep their core breeding /// /// purchase macro-level NDVI/pasture drought animals alive). The advantage of this approach in cover used in the events of catastrophic losses /// comparison to the existing ILRI product (interim in pasture and grazing to finance payouts to period) is that payouts could be triggered earlier in the many small, vulnerable livestock producers the season, i.e., during the onset of drought—before (owning <50 livestock units) who are eligible for reduced pasture/grazing creates a disaster. Under state-funded natural disaster assistance under this approach, pastoralists would not be forced into the CADENA program. One hundred percent of untimely sales of livestock at very reduced prices, the premium cover is borne by federal and state and their animals would be saved from starvation, governments together (80:20 ratio). Since the disease, and ultimately death. Pastoralists would use program was introduced in 2006, it has been the funds provided to preserve livestock (through massively scaled up such that in 2011, a total of buying fodder, migrating, culling, etc.) rather than almost 60 million hectares of grazing lands were having to replace it. This approach enables a faster insured in 21 states, and nearly 4 million head of mobilization and increased effectiveness of the livestock were protected (World Bank 2013a). emergency response.18 Among other important • Uruguay and Argentina. In 2011–2013, the /// /// welfare gains it offers, such an approach could World Bank assisted the governments of Uruguay help facilitate income smoothing and reduce asset SECTION and Argentina in designing NDVI/pasture drought depletion; this protection of assets would increase macro-level products protecting livestock and household resiliency to future shocks. issuing early payouts (World Bank 2012, 2013b). 02 34 K E N YA : A G R I C U LT U R E I N S U R A N C E S O L U T I O N S A P P R A I S A L Table 2—  Comparison of Uruguayan pasture NDVI cover and proposed Kenyan NDVI cover Product Approach taken in Uruguay Considerations for Kenya feature NDVI/pasture index; 5km x 5km (2,500 ha). Required index and data will be based on the existing NDVI database created and maintained Index by ILRI: NDVI/pasture index; 250m x 250m (eMODIS). Pasture growing season: 7 months (September- Still being explored. Northern Kenya has two March). rainy seasons: long rains (March-June) and short Cover period rains (October-December); ideally both would be covered. Police Section (equivalent to a municipality). Still being explored. Preliminary discussions with Homogeneous NDVI signature and individual ILRI indicate that index products could be more Insured unit livestock herd data are registered at this effective at a scale smaller than the division level. administrative level for foot-and-mouth disease control purposes. Beef cattle (breeding cows and heifers only). All households in the HSNP poverty census above Insured interest Program has been designed to cover all registered the cut-off point for the regular program. beef cattle herds in Uruguay. Based on nutritional requirements of insured Still being explored. For example, more input cattle during the insurance cover period, is being collected from livestock experts on assuming animals are fed on supplementary feed nutritional requirements, etc. Sum insured rations that can be purchased locally. Monthly payout frequency, because once pasture Different payout frequencies being explored degradation is visible on NDVI, the insured cattle (monthly, every 3 months, seasonal basis); to be Payout are already suffering from starvation. determined based on pastoralists’ needs. parameters Sum paid out along gradual trigger with entry and Gradual and binary trigger options being exit point. explored, i.e., paying out either in full or not at all, for simplicity. Source: World Bank 2013b SECTION Note: eMODIS = enhanced Moderate Resolution Imaging Spectroradiometer. 02 K E N YA : A G R I C U LT U R E I N S U R A N C E S O L U T I O N S A P P R A I S A L 35 Figure 4 —  Illustrative calculated pure loss cost rates for 12-month NDVI asset protection cover at District and Division level SECTION Source: World Bank 02 36 K E N YA : A G R I C U LT U R E I N S U R A N C E S O L U T I O N S A P P R A I S A L The SDL macro-level NDVI product is /// light on how early drought index insurance currently under development, and a payouts could be used to support the protection prototype contract has been designed that of breeding herds (e.g., by buying fodder and will be presented to key stakeholders— animal feed supplements, migrating/transporting including SDL and livestock producer animals to other grazing lands, controlled associations in the ASAL regions—for review, destocking of animals, etc.). refinement, and finalization. The product • The most appropriate definition of the /// draws on the lessons and experience of the Mexican, /// insured unit. In determining whether the Argentine, and Uruguayan livestock NDVI programs, /// insured unit should be the department, division, while taking the local Kenyan context into account. or a smaller area with homogeneous grazing/ A comparison of the Uruguayan macro-level NDVI rangeland conditions affected in a similar way by cover and the proposed macro-level cover for SDL in drought, it will be important to take into account Kenya is presented in table 2. the seasonal migration patterns of the pastoralists An example of the outputs of the macro- /// as they move their nonbreeding herds to their level prototype NDVI pasture drought index dry season grazing lands, which may be outside insurance cover is given in figure 4, which the defined insured geographic unit where they shows the calculated pure loss costs (average normally reside. Setting the insured unit at a very expected payouts) for an annual policy small localized level may invalidate the operation for the four HSNP counties at district and of such an NDVI cover. divisional levels. The calculated pure premium 19 • Ways to integrate any macro-index /// /// rates are presented for the four HSNP counties insurance program with NDMA and SDL (and their divisions)—Turkana, Marsabit, Wajir, and drought-response plans for the livestock Mandera—and in general terms reflect the increasing sector. Possible approaches include controlled drought risk exposure in natural rangelands from /// destocking programs, livestock watering, pasture west to east as measured by eMODIS NDVI by month and grazing conservation measures, government for the 13-year period 2001 to 2013. Decisions that emergency livestock feed programs (if these will need to be taken with local stakeholders in due currently exist), and veterinary support programs course include the size of insured unit (county or during times of drought. A key point is that under division) and whether to market top-up or voluntary a macro-level index insurance program aimed at cover using differential premium rates in each keeping animals alive, it will be important to avoid insured unit. suggesting to pastoralists that they do not need To determine whether the proposed macro- /// to destock their herd in times of acute drought, as level NDVI pasture drought degradation they will receive insurance payouts. cover, with its emphasis on early payouts to • Presence of local public or private forage /// keep breeding animals alive, is appropriate to markets in times of drought in the ASALs. /// the pastoralist production systems of Kenya’s These markets are essential, and could be (i) a ASALs, further research will be required into GoK-SDL fodder supply program in place at the the following key areas: onset of drought; and/or (ii) a program under /// SECTION • Use of early payouts to protect breeding /// which private traders are incentivized to truck herds. Focus groups discussions with pastoralists fodder from Kenya’s surplus regions to drought- 02 /// in the target HSNP counties are planned to shed stricken regions. K E N YA : A G R I C U LT U R E I N S U R A N C E S O L U T I O N S A P P R A I S A L 37 Registration and Distribution understand how the insurance product operates Channels (how the triggers are based on satellite data, what the trigger points for payout are, etc.). Households SDL Macro-Level NDVI Program must understand their coverage in order to promote the voluntary purchase market. Better understanding Registering pastoralist households eligible for will also encourage voluntary take-up by spreading /// the GoK subsidy could be done automatically, awareness of the product. through the existing HSNP database; this approach would be cost-effective and save time, but it could also present some complications: (i) Failure to explain the product /// to beneficiaries could lead to poor awareness of the benefits being provided and of claims procedures; (ii) if the program is not well known in the region, both political visibility and broader awareness of the product would be reduced; (iii) without insurance awareness creation, potential financial inclusion gains would be lost; (iv) confirming beneficiaries’ inclusion in the program would be hard; and (v) beneficiaries might never understand the benefit being provided to them by the GoK. It is preferable, therefore, to handle initial /// registration in person and not automatically. One option would be to enroll pastoralists into the insurance program when Equity Bank opens bank accounts with them and they receive their bank cards. All pastoralists /// registered under the HSNP poverty census are scheduled to receive a bank account and a bank card by end of the first quarter 2015. When pastoralists receive their bank cards, they could also get an explanation of the program, including benefits, payments procedures, issuance of cards and pins, confirmation of persons and identification details, awareness/education, and consumer protection issues. Although more time-consuming and costly, /// this method will support the development SECTION of a sustainable market. It has the key benefit /// of ensuring that eligible households understand the insurance coverage they are being given and, further, 02 38 K E N YA : A G R I C U LT U R E I N S U R A N C E S O L U T I O N S A P P R A I S A L Top-up and Voluntary Purchase Products Legal and Regulatory Issues and the Role of Insurers There will be significant distribution /// challenges beyond the subsidized cover, Both legal issues and regulatory issues /// particularly as insurance will not be linked (beyond consumer protection) must be to credit. The challenges would emanate from /// considered during product design. The role of /// potentially high operational costs associated with insurers and insurance intermediaries should also the sales and service process, which to date have be considered. Ultimately, however, the regulatory been significant in northern Kenya. The distribution framework is a matter for the Insurance Regulatory should primarily be the responsibility of the private Authority (IRA), which should be kept informed as sector. The top-up could potentially happen at the product is designed. the time of registration, when beneficiaries of the government-supported program could opt for The IRA regulates and supervises the /// additional coverage. This approach could utilize the insurance sector under the current Insurance proposed distribution channels, or the underwriter Act. However, if the new principles-based Insurance /// could devise a cost-effective model for distribution Act under consideration is enacted, it will enable the of this additional cover. Voluntary purchase could IRA to develop new regulations that will foster an be done through developed or parallel network enabling environment for livestock insurance. and infrastructure. Contract Design (Macro-level Policy) Ultimately, the registration and distribution For the proposed SDL macro-level NDVI /// process will be led by the private sector, /// pasture drought index insurance policy, though GoK support may be needed in the insurance contract will be purchased the short term. A private sector–led initiative by GoK-SDL, not by individual pastoralists. /// would demonstrate viable long-term business /// The policy would be set up as a macro-level policy, opportunities and hence be more sustainable. In purchased by the GoK for the benefit of eligible the long run, it would also provide an opportunity pastoralists. Payment would be made either to the for expanding more financial opportunities to the GoK (which would then pay covered pastoralists) or target communities, and the overall execution would directly to the pastoralists. A point to note here is cost less than if the initiative was government led. that a beneficiary does not have an automatic right The private sector–led process would undertake to enforce a master policy against the insurer; this fresh registration of beneficiaries and develop and arrangement makes the program much easier to manage payment infrastructure under contractual implement. Enforcement rights are dependent on arrangement with the government. The tendering the terms of the policy. process would be used to select the provider(s) or consortium to offer registration and/or distribution Contract Design (Voluntary processes. Purchase Product) The voluntary purchase product will be /// available to all pastoralists but will be SECTION offered as a voluntary top-up to those benefiting from the macro-level product. For 02 /// these pastoralists, the top-up should operate as an K E N YA : A G R I C U LT U R E I N S U R A N C E S O L U T I O N S A P P R A I S A L 39 extension of the macro-level policy—that is, as an Role of Private Sector Insurers extension of the same “master policy.” This would be by far the most straightforward approach. An effective PPP requires the engagement /// and willing participation of private sector For all other pastoralists, the freely purchased /// insurers at an early stage. /// insurance policy would be more difficult to fit within the constraints of a master policy. /// The macro-level policy would need to be /// However, issuing a series of individual policies would purchased from a local insurer or insurers. /// add to the transaction cost. Under this scenario, one Under the Kenyan Insurance Act, insurance must option would be to allow pastoralists to enforce the be purchased through the local market unless there policy against the insurer, though this option has is insufficient capacity. If reinsurance is purchased, possible cost implications, as the insurer might have it could be placed into the international market to deal with pastoralists seeking payment on the basis after any compulsory cessions to national/regional of losses even though the index had not triggered. reinsurers. The inability of the GoK to legally This possibility would need to be factored into the purchase directly from the international market premium as an additional risk (which would be very would almost certainly add some transaction cost. difficult to cost). The product design includes a single master /// policy plus a series of individual policies. Form of Contract /// Both the master policy and the individual policies Even though eligible pastoralists will not /// would need to be purchased from a local insurer or contribute to the premium payable under insurers. the macro-level policy issued to SDL, the insurance contract/policy should be kept as straightforward as possible. There are a number /// of reasons for this: • The contract/policy design will most likely become a precedent. • Eligible pastoralists will be paying for the voluntary top-up. • A similar contract form should be used for the master policy and for the individual policies. Other issues to consider will be the potential /// use of electronic policy acceptance. This is not /// an issue directly covered under the current legal and regulatory framework, although there is precedent for it in relation to other products. SECTION 02 40 K E N YA : A G R I C U LT U R E I N S U R A N C E S O L U T I O N S A P P R A I S A L Table 3—  Proposed livestock safety net and insurance program for Kenya’s four HNSP counties Number of pastoralists Government’s Form of financial expected to be covered Income level of contribution to cost protection against across four counties over beneficiary of premium or welfare disasters next five years (out of payments (%) 470,000 total) Unsubsidized livestock Middle-income 0 0 insurance (US$1/day or more) Partially subsidized livestock Low-income 15,000 (by 2019) 50, 25a insurance (below US$1/day) Ultra-poor Macro-level insurance (below national rural 71,000 100a program poverty line of US$0.5/day) Hardcore poor HSNP scalable cash transfers (below national food 100,000 100b poverty line of US$0.3/day) Note: The four HSNP counties are Mandera, Marsabit, Turkana, and Wajir a. Contribution is from State Department of Livestock, based on annual assumed budget of K Sh 300 million per year; the two figures refer to subsidies for the top-up and voluntary purchase options respectively, b. Contribution is from National Drought Management Authority. Fiscal Costing Assumptions The optional top-up coverage for pastoralists /// enrolled in the program will also be available And Scenarios from the outset of the program. In addition, pastoralists who are not part of the initial The following assumptions underlie the /// target group will also have the option illustrative estimation of the potential fiscal costs of the programs described above: /// to purchase the NDVI-based insurance coverage on a voluntary basis. The first layers /// The macro-level NDVI-based index insurance /// of both the top-up option and the coverage for product for livestock asset protection will nontargeted pastoralists will be partially subsidized. target the poor “vulnerabale pastoralists” who are above the poorest 100,000 Table 3 shows the relationship of the SDL macro- beneficiaries of the HSNP program. The /// level automatic NDVI livestock insurance and the program will be implemented in the four counties voluntary top-up cover at year 3 with the HSNP cash SECTION transfer program. The figures used are indicative covered by HSNP (Mandera, Marsabit, Turkana, and Wajir) and is expected to cover free of charge and subject to change as the insurance program 02 approximately 71,000 pastoralists. is concretized. K E N YA : A G R I C U LT U R E I N S U R A N C E S O L U T I O N S A P P R A I S A L 41 Table 4—  Fiscal costing projections for macro-level asset protection coverage SCENARIO KSh 100 million SCENARIO KSh 200 million SCENARIO KSh 300 million (US$ 1.2 million) (US$ 2.3 million) (US$ 3.5 million) Case A Case B Case A Case B Case A Case B Budget available for Macro-level asset 100,000,000 200,000,000 300,000,000 protection coverage -K Sh No. of pasotralists eligible for livestock 11,592 33,636 24,204 69,636 36,815 105,636 asset protection coverage Estimated Financial Costs of the Macro-Level to approximately 12,000 to 106,000 Livestock NDVI Program pastoralists. Specifically, a budget of K Sh 100 /// million would cover 12,000 to 34,000 pastoralists, Following specific indications from SDL, we a budget of K Sh 200 million would cover 24,000 to /// analyze here three public support scenarios 70,000 pastoralists, and a budget of K Sh 300 million for the macro-level asset protection scheme: would cover 37,000 to 106,000 pastoralists. budgets of K Sh 100 million, K Sh 200 million, and K Sh 300 million. The projected fiscal costs /// To cover a reference target group of 37,000 /// of the programs are summarized in table 4 and table to 106,000 (mid-point of 71,000) pastoralists 5. For each budget scenario, two extreme cases are under the macro-level insurance option, presented. Case A is structured by selecting, within under which SDL would be the insured and a reasonable range of variation, the more costly responsible for payment of premium, fiscal extremes of the key parameters (i.e., higher values resources of K Sh 300,000 (US$ 3.5 million) per TLU insured, a higher number of TLUs per policy, or above would be required.20 /// and a higher insurance premium estimate). This Estimated Fiscal Costs of the Top-Up and defines a lower bound for the number of potential Voluntary Livestock Insurance Programs pastoralists to be covered with the reference budget available. On the other extreme, Case B takes into A partially subsidized top-up option for /// account the less expensive options, thus identifying covering an additional 5 TLUs will be offered, the higher bound for the number of potential with a budget requirement ranging from pastoralists to be covered. roughly K Sh 2 million in 2015 to K Sh 16 million in 2019. This scenario is estimated /// Depending on the policy choices made assuming a progression over five years from 1,000 SECTION /// and the parameters selected, budgetary to 10,000 pastoralists who voluntarily purchase the support ranging from K Sh 100 million to coverage, together with tentative public support of K Sh 300 million would provide coverage 50 percent of the premium cost. 02 42 K E N YA : A G R I C U LT U R E I N S U R A N C E S O L U T I O N S A P P R A I S A L The other option, for all nontargeted /// As next steps, SDL, MALF, and other /// pastoralists, allows purchase of the NDVI key stakeholders will need to consider asset protection coverage with tentative funding arrangements to cover the public support of 25 percent for up to 10 TLUs costs of premiums as well as design and insured; this generates an additional budget implementation for the macro-level NDVI requirement of K Sh 5 million to K Sh 16 insurance program, and for the top-up and million between 2015 and 2019. This projection /// nontargeted pastoralist programs that carry assumes that in the five-year interval, 1,000 to 5,000 premium subsidies. One option would be to use /// nontargeted pastoralists purchase the coverage. the proposed National Livestock Insurance Fund to finance the premiums and other program costs, The two additional insurance schemes to /// including those for registration of beneficiaries, be implemented at year 3 of the program education and training programs, program design, could in aggregate increase the budget and implementation and auditing. requirements by roughly K Sh 6 million in 2015 and K Sh 31 million in 2019. /// In summary, for the combination of (i) the /// macro-level asset protection coverage, (ii) the top-up option, and (iii) the expansion to nontargeted pastoralists, the estimated fiscal costs are estimated at around K Sh 306 million at program inception in 2015 and at around K Sh 331 million by 2019. /// Table 5—  Fiscal costing projections for top-up and nontargeted pastoralists options 2015 2016 2017 2018 2019 Cost of public support for top-up option 1.6 5.1 8.6 12.1 15.6 (million K Sh) Cost of public support for nontargeted pastoralists 4.7 8.6 11.7 14.1 15.6 (million K Sh) Total cost for GoK 6.3 13.7 20.3 26.2 31.2 (million K Sh) Total cost for GoK 0.1 0.2 0.2 0.3 0.4 SECTION (million US$ at 85 K Sh/US$) 02 K E N YA : A G R I C U LT U R E I N S U R A N C E S O L U T I O N S A P P R A I S A L 43 1999–2013, average livestock mortality rates were Welfare Impacts Of Index- 9–18 percent per year. As recurrent droughts reduce Based Livestock Insurance households’ average herd size, they lead to increasing In HSNP Countries poverty and food insecurity in the region. The four HSNP counties are among the /// To cope with droughts, households in this /// poorest counties in Kenya, with the majority region rely heavily on food aid, risk sharing of the population depending heavily on within communities, and other emergency livestock both for income and food. As /// response and welfare programs, but they are estimated by the HSNP household survey, the ultra- largely uninsured. Under the HSNP, one of the key /// poverty rate in 2012 was 46.8 percent, with average government welfare programs, households receive consumption expenditure per capita per month of cash transfers of approximately K Sh 3,500 every 1,746 K Sh.21 The share of livestock income in total two months. HSNP impact evaluation results find household economic income ranges from 25 percent that the cash transfer has reduced poverty and that to 80 percent, and the share is larger for the poorer it was used as a safety net during the 2011 drought quintiles. Livestock production is the key source in the region. Other coping mechanisms appear of livelihood in this region. Alternative productive livelihood appears very limited, and is accessible very limited—for example, households are credit only to the wealthiest. Alternative livelihoods for constrained and have limited access to financial the poor majority are petty trading, casual labor, and services—leaving households uninsured against small cropping. catastrophic herd losses from droughts. Livestock holdings provide a good proxy for /// We have conducted a detailed economic /// welfare in this region, with the poor owning analysis of the likely impact of livestock small herds but relying more heavily on insurance on four categories of pastoralist: livestock than those who are better off. From poorest, ultra-poor, poor, and nonpoor. We /// /// the longitudinal monthly household survey by Arid develop a dynamic model to explore the potential Land Resource Management Project (ALRMP) in varying welfare impacts on a representative pastoral these counties, the average herd size owned by household in each of the four wealth groups: households was 10.4 TLUs during 1999–2013. On the poorest with small herd, the ultra-poor with average, households in the two poorest quintiles own less than 5 TLUs, those in quintile 3, 4, and 5 own vulnerable herd, the poor with medium herd, and respectively 5–10 TLUs, 11–20 TLUs, and more than the nonpoor with large herd (see annex B.3). In 20 TLUs. Livestock production systems could vary the economic model, a household earns income across and within the four counties, with relatively each season from milk production and livestock larger mobile pastoralists (i.e., those with larger herd off-take, which it uses for consumption and to size) in the relatively low and arid lands of northern accumulate herd for the next season. Normally, Marsabit, Mandera, and Wajir. the own herd grows each season, but it sometimes Livestock production in this region is prone decreases in size due to livestock mortality arising SECTION /// to droughts that can cause catastrophic herd from various idiosyncratic factors (e.g., disease, losses. Extreme droughts have occurred four times sickness, accidents, as well as the covariate 02 /// over the past 10 years. According to ALRMP data for catastrophic droughts). 44 K E N YA : A G R I C U LT U R E I N S U R A N C E S O L U T I O N S A P P R A I S A L Catastrophic herd losses from droughts /// could have immediate welfare effects by reducing livestock income available for consumption. While severe droughts can immediately push the better-off poor and nonpoor into poverty, they can likewise push the ultra-poor and the poorest, whose livestock income is extremely low, into destitution. The black lines in figure 5 depict these /// impacts. Specifically, a 1-in-4-year drought could push livestock income of the poorest to the level of destitution (K Sh 13/day/person). A 1-in-8-year drought might also push income of the relatively better-off households below the food poverty line, and could bring the nonpoor households into poverty. We find that free provision of asset /// protection livestock insurance could reduce vulnerability but would not likely provide an immediate exit from poverty. The long-dash /// red lines in figure 5 depict these patterns. This result is in contrast to the existing HSNP cash transfer program, which transfers approximately K Sh 3,500 to the poorest eligible households every two months. The analysis considers the likely impact of /// As the green lines show, direct cash transfers could asset protection livestock insurance. For the /// potentially produce immediate poverty reduction asset protection product, we assumed design features effects for some groups, e.g., the ultra-poor, who similar to those of the IBLI asset replacement have been boosted up above the food poverty line in some good years. contract designed by ILRI, except that payouts were made early as opposed to at the end of the season. By itself, direct cash transfer could still leave /// The analysis assumes that monthly insurance payouts poor beneficiaries vulnerable to falling into could allow effective early interventions that enable poverty in extreme years. Complementing the insured pastoralist to keep insured livestock cash transfer with the free provision of livestock insurance might provide a more alive. Sensitivity analysis was also performed with sustainable exit from poverty. Especially for varying assumptions, and the key results did not vary /// the ultra-poor, asset protection livestock insurance significantly from our main assumption. Our analysis coverage could immediately protect cash transfer considers both short-term and long-term impacts beneficiaries from falling into poverty in a 1-in-6 year SECTION of this livestock insurance, along with the proposed extreme drought; see the long-dash and short-dash forms of government support, on the four distinct green lines in figure 5. This is the intention of the 02 subsets of the population. current plan to scale up from the HSNP. K E N YA : A G R I C U LT U R E I N S U R A N C E S O L U T I O N S A P P R A I S A L 45 The biggest impacts of livestock insurance /// impacts of droughts vary across different herd are expected to be realized in the longer groups, so will the potential impacts of livestock term, as the insurance helps pastoralists insurance and related government support. build up the herds over time and keep them Livestock asset protection insurance that at or above the viable size needed to stay /// is designed to keep the core breeding stock out of the poverty trap. Existing academic /// alive during severe droughts could have research (e.g., Lybbert et al. 2004; Barrett et al. large long-term impacts, most notably by 2006; Chantarat et al. 2014) finds that a critical protecting poor households with vulnerable herd size of about 10–15 TLUs (see annex B.3) is herds from a poverty trap. As shown in figure 6, necessary to sustain a viable herd accumulation in /// this region. Given limited productive nonlivestock free asset protection insurance and top-up coverage livelihood options, and given the need for seasonal might provide enough cash for effective early migration as adaptation to climate variability, intervention and allow households to save and grow pastoral households in this region consume a good their viable herd, which otherwise could collapse. portion out of their own herd each season (e.g., The overall impacts could be large when insurance through direct slaughtering or off-taking for cash). coverage is offered with cash transfers, which could This necessary consumption tends to slow down and also relax the required consumption out of the own disrupt natural herd growth, especially for very small herd. Based on our simulation exercise with large herds. Households with small herd sizes (below the numbers of replicated years, figure 7 further depicts critical threshold) thus tend to deplete their herds the expected probability of falling into the poverty over time. Furthermore, as poor households tend to trap (losing viable herd) five years after being hit be credit constrained, they are unable to restock their by drought–induced livestock losses at different herds up to the economically viable and sustainable magnitudes. It appears that free insurance and 50 levels. With a small, collapsing herd size and low percent subsidized top-up coverage could reduce consumption, these households easily fall into the the probability of falling into the poverty trap by poverty trap that researchers have found in this up to 60 percent. And if these schemes were to be vulnerable pastoral region. combined with cash transfer, altogether they could reduce the probability by up to 80 percent. This is The existence of a viable herd threshold in contrast to cash transfer alone, which could offer /// size implies that catastrophic herd loss from temporary poverty reduction while still leaving drought could be irreversible, especially when beneficiaries vulnerable to falling back into poverty droughts have led to livestock losses below in extreme drought years. the viable level. Figure 6 provides an example of /// common herd accumulation over time for different For better-off households with medium and /// herd groups. It shows that especially for poor larger herds, livestock insurance could help households with a vulnerable herd size around the to increase herds over time by stabilizing viable threshold, a big 1-in-6-year herd loss could herd accumulation. With a typical insurance /// push herd size down to a level that will not recover policy, the insured might need to sacrifice his average without a restocking intervention. For small herds, income to pay for protection that reduced variability extreme droughts could speed up herd collapse but did not increase productivity; but livestock SECTION and move households toward destitution. For large insurance could have a productivity improvement herds, droughts could disrupt and slow down herd effect through stabilizing herd accumulation. accumulation over time. Overall, just as the potential Figure 6 suggests that commercial asset protection 02 46 K E N YA : A G R I C U LT U R E I N S U R A N C E S O L U T I O N S A P P R A I S A L insurance could be attractive to better-off households, Overall, these varying insurance impacts can /// given that it could be less costly for the insurance be used to appropriately target public support contract to disburse early payouts to keep livestock for livestock insurance. Support mechanisms /// alive than to replace lost livestock, and given the targeted to ensure an effective safety net among possible multiplier effects from protecting the critical the vulnerable group could be very cost-effective in reducing poverty in the long run. As we see, poverty breeding herd through herd accumulation. increases (and herds decline) in this region over time Livestock insurance might have the smallest /// due to recurring droughts. A safety net intervention long-term welfare effects on the poorest that can keep the vulnerable households from joining (who own small and nonviable herds), since the ranks of the poor would allow the government by itself livestock insurance is unlikely to help to concentrate its limited resources on bringing the existing poor out of poverty. For the poorest, a them to reach a viable herd size. Figure 6 and /// combination of cash transfer and effective insurance figure 7 suggest that combining direct cash transfers could work to reduce vulnerability (and immediate with free livestock insurance might help stabilize poverty). But if the long-term goal is to move the the herds of the poorest households and slow down poorest households out of poverty through pastoral their herd collapse in the short run (e.g., cash transfer production, complementing livestock insurance with could potentially relieve necessary consumption out interventions that promote restocking toward a viable of owned herd), but the scheme might not alter the herd could be critical. For the larger herd groups, probability of falling into the poverty trap for this promotion uptakes of the (potentially cost-effective) small-herd group. commercial livestock insurance could be effective.22 Figure 5 —  Potential short-term impacts of livestock insurance on income available for consumption SECTION 02 K E N YA : A G R I C U LT U R E I N S U R A N C E S O L U T I O N S A P P R A I S A L 47 Figure 6 —  Potential impacts of livestock insurance on herd accumulation Figure 7 —  Potential impacts of livestock insurance on probability of falling into poverty trap SECTION 02 48 K E N YA : A G R I C U LT U R E I N S U R A N C E S O L U T I O N S A P P R A I S A L Crop Insurance more than 30 years until issues in performance Context and unsustainable financial losses led to its closure Kenya’s Experience in Crop Insurance between 1977 and 1978. As well documented in the recent Ministry of /// Interest in agricultural crop and livestock /// Agriculture, Livestock and Fisheries (MALF) insurance reemerged in the mid-2000s. Two /// report (GoK 2014a), Kenya has a long tradition main routes were explored: (i) the development of developing agricultural policy programs for of a Kenyan market crop insurance capability to risk management purposes. /// 23 underwrite traditional indemnity-based multi-peril crop insurance (MPCI) for medium- and large-scale In Kenya, government support to agricultural /// commercial farmers, and (ii) the introduction of insurance dates back to 1942, with the index-based insurance as a potential retail product formation of the Guaranteed Minimum Return to market to small and marginal crop and livestock scheme. The objectives of this scheme were /// producers (in situations where operating traditional twofold: (i) to encourage food production to meet indemnity-based crop and livestock insurance Kenya’s basic food needs by providing seasonal crop programs would be prohibitively expensive). credit to farmers growing strategic food crops (such as wheat and maize) through a system of guaranteed The reemergence of interest in agricultural /// prices for output; and (ii) to provide these farmers insurance in Kenya began in 2006, when with crop insurance in order to compensate them four local private insurance companies for droughts, pests, diseases, and other natural perils came together to form a crop and livestock (Sinah 2012; Kerer 2013). The system operated for insurance consortium or pool agreement SECTION 03 K E N YA : A G R I C U LT U R E I N S U R A N C E S O L U T I O N S A P P R A I S A L 49 identified as Agricultural Insurance Manager to willing private sector players. An underlying (AIM). The role of the AIM consortium was to design, /// assumption is that agricultural insurance is nested rate, and implement traditional indemnity-based within the broader context of integrated risk crop and livestock insurance covers including MPCI. management systems (such as improved extension The pool operated from 2008 to 2010, when it was for better asset protection, seed development, etc.). disbanded. Since 2010, several of the companies have continued to underwrite their own separate crop and The GoK proposes the development of a /// livestock portfolios. dedicated public-private partnership (PPP) in agricultural insurance and plans to invest Kenya has subsequently witnessed an /// resources in supporting it financially. From increased interest in developing a crop /// weather index insurance (WII) product. This /// a program design point of view and for illustrative effort is being led for the most part by the Syngenta purposes, the approach analyzed below is that of area Foundation for Sustainable Agriculture and the yield index insurance (AYII) for maize and wheat Financial Sector Deepening (FSD) Kenya program crops. AYII would be electively retailed through through a public sector and donor-sponsored credit institutions via their lending operations for initiative.24 agricultural inputs. Government’s interest in a Once the PPP framework for crop insurance /// new generation of agricultural has been implemented, appropriate solutions insurance tools for other agricultural sectors could be In order to reduce risk and promote growth /// also developed. The State Department of in the agricultural sector, the government of Agriculture (SDA) intends to extend future Kenya (GoK) is now placing new emphasis on analyses to horticulture, coffee, and tea. /// the development of insurance solutions for Maize and wheat have been selected as the sectors agriculture. The GoK intends to foster a generation /// to start with, given their relevance in terms of of innovative and widespread insurance products food security (maize in particular), their major by addressing the conditions that so far have been contribution to agricultural value added, and the hampering their development. availability of readily implementable insurance GoK’s key assumption is that a well- /// solutions. structured agricultural insurance program, As discussed in detail below, AYII seems the with participation by both public and private /// appropriate tool for reaching the operational players, could unlock access to production credit and stimulate investment in productive scale that would allow the GoK to meet its inputs. It is now clear that, to be successful, an /// policy objectives in the grains sector. The case /// insurance scheme needs to reach a scale large enough of India, which has the largest insurance program in to operate effectively, both in terms of the risk the world per number of farmers insured (34 million transfer objectives and of the insurance industry’s farmers/20 percent of farmer households), offers an commercial interests. International experience shows inspiring example of a PPP in agricultural insurance SECTION that this scale is rarely achieved without the active developed in an emerging country.25 participation of government in building appropriate institutions and in providing financial support 03 50 K E N YA : A G R I C U LT U R E I N S U R A N C E S O L U T I O N S A P P R A I S A L Description of Potential The more traditional NPCI products (e.g., /// those covering hail or frost) and MPCI Agricultural Insurance products (covering all risks combined) are already offered by several insurance Programs for Crops companies in Kenya, but given the prevalent operating conditions in Kenyan agriculture, Rationale for Selecting AYII in an they may not be suitable for large-scale Agricultural Insurance PPP Framework application. Such products are probably better /// suited for medium- and large-scale commercial A PPP in agricultural insurance could make agriculture than for small-scale subsistence farming, /// use of any of four types of insurance products and they require a strong network of loss adjusters. for crops: multi-peril crop insurance, weather index In addition, the moral hazard and adverse selection /// insurance, area yield index insurance, and named challenges they pose are difficult to manage. peril crop insurance (NPCI) (see figure 8). WII is an interesting innovation that has In a mature agricultural insurance program /// been extensively piloted in Kenya and is /// these four different contract typologies would now starting to be retailed in niche markets. /// not necessarily be alternative solutions, While not affected by moral hazard and adverse but could be complementary in offering a selection, WII covers essentially weather perils wide range of risk management tools to (mainly drought). Product design requires significant select from. However, in the case of a nascent PPP /// customized research and development activities that, system, the GoK will need to concentrate efforts and together with its significant exposure to basis risk, resources on the approach that best suits its policy limits the adoption of WII on a widespread scale. objectives, leaving other approaches to develop as GoK’s motivations for investing in a PPP the system gains momentum. Figure 1 and figure 8 /// based on an AYII scheme seem to be summarize the conditions in which the use of the supported by the welfare impact analysis different products is more appropriate. presented in section 3.4. /// The GoK has chosen to promote the AYII Operating Modalities /// development of an AYII program for maize and wheat production. By definition, AYII is /// The key feature of AYII is that it does not /// based on an indexed approach, where the underlying indemnify crop yield losses at the individual index is crop yield of a defined area called an insured field or grower level.26 Rather, an AYII product unit. In AYII, the actual yield of the insured crop makes indemnity payments to growers in the insurance unit is compared to the threshold according to yield loss or shortfall against an yield. If the former is lower than the latter, all insured average area yield (the index) in a defined farmers in the insurance unit are eligible for the same geographical area. An area yield index policy /// rate of indemnity payout. establishes an “insured yield” that is expressed as a percentage (termed the “coverage level”—see figure AYII provides wide peril coverage if designed /// 9) of the historical average yield for selected crops in appropriately: it is not affected by adverse the defined geographical area that forms the insured selection and moral hazard, and it has a unit. Farmers whose fields are located within the SECTION standardized design that can lead to rapid insured unit may purchase optional coverage levels, scalability. The main drawback of AYII is or insurers may offer only one coverage option in the 03 “basis risk” (explained below). /// insured unit. K E N YA : A G R I C U LT U R E I N S U R A N C E S O L U T I O N S A P P R A I S A L 51 Figure 8 —  Types of Agricultural Insurance Products What are the various types of insurance products for agriculture? Transaction Moral hazard and Claims settle- What is it? costs adverse selection Basis risk ment time • Multi-peril crop insurance (MPCI) is a traditional indemnity insurance Multi-peril crop product against all perils Farm High High Low Medium insurance • Payouts are determined through a farm-level loss assessment process. • Area yield index insurance is based on average losses at the regional Area-yield index level, rather than farm level. Village insurance Medium Low Medium Medium • It is often based on crop cutting experiments. • Weather index insurance is based on weather parameters (such Weather index as rainfall, temperature, or soil Village moisture) correlated with farm- insurance level yields or revenue outcomes Low Low High Low * Basis risk with index insurance arises when indices are imperfectly correlated with farmers’ losses. Some farmers with losses may not receive payouts while some farmed without losses may receive payouts. Source: Authors 2014 Figure 9 —  Coverage Level and Insurance Payouts in AYII 2000 1800 Yield (kg per hectare) Reference average yield 1600 1400 Coverage Level at 80% of average yield 1200 1000 Yield shortfall to be compensated by insruance 800 payout 600 400 200 0 Crop season with no payment Crop season with payout Sources: Authors 2014 SECTION 03 52 K E N YA : A G R I C U LT U R E I N S U R A N C E S O L U T I O N S A P P R A I S A L The actual average yield for the insured /// conducted on an individual farmer and field-by-field crop is established by a statistical sample of basis, but rather according to a pre-agreed random field measurement (usually involving crop sampling of crop yields on plots within the insurance cuttings) in the insured unit, and an indemnity unit. is paid by the amount that the actual average yield falls short of the insured yield coverage The main drawback of AYII is basis risk, or /// level purchased by each grower. /// the potential difference between the insured area yield outcome and the actual yields The key advantages of the area yield achieved by individual insured farmers within /// approach are that moral hazard and adverse selection are minimized, and the costs the insured area. Basis risk arises where an /// of administering such a policy are much individual grower suffers severe crop yield losses due reduced. The policy responds to yield loss at /// to a localized peril (e.g., hail, or flooding by a nearby a defined area level, and not at the level of the river) that does not have a large impact on the area- individual farmer, so if the insured unit is large level average yield. In such cases the farmer who enough, no farmer can influence the yield indemnity has incurred damage does not receive an indemnity. payments—minimizing adverse selection and Basis risk may also arise where individual farmer SECTION moral hazard. Administration costs are also greatly crop production and yields are highly heterogeneous reduced because there is no need for pre-inspections (different) within the same department (that is, 03 on individual farms, and loss assessment is not where an area-based approach invalid). K E N YA : A G R I C U LT U R E I N S U R A N C E S O L U T I O N S A P P R A I S A L 53 Program Requirements Because data requirements for AYII are /// very specific, they may go beyond what a Developing a functional and effective AYII /// traditional system of agricultural statistics program for maize and wheat in Kenya requires. Specific public support will likely be requires two key steps: (i) define homogeneous /// needed to allow for additional CCE activity. /// producing zones (the insured units) with high levels of correlation between farmers of the same unit; and For an AYII program, the value of defining /// (ii) generate an accountable, reliable, and statistically appropriate insurance units and of accurate system of measuring actual average area developing a suitable data collection system yields in the defined insured unit, and define the is paramount. We therefore suggest that a basis for triggering payouts where actual yields fall specific multi-stakeholder study on how best short of the insured yield(s). to organize the data collection system be carried out. In addition to selected SDA officials, /// As for any agricultural insurance program, /// the team for such a study should also include staff historical data for structuring and rating from the Kenya Bureau of Statistics, agricultural AYII are fundamental. Ideally, for each of the research institutions, the insurance industry, defined insured units, yield data for the past and any other interested party. The study would 15 years or more would be available. If such need to cover items such as (i) risk profile–based data are not available, logistical and financial identification of insurance units; (ii) statistical support to the insurance industry will be sampling methodology for identification of plots critical. The GoK will likely need to help the private /// for CCEs; (iii) number of CCEs per insurance sector overcome challenges related to data in the unit; (iv) procedures, roles, and responsibilities inception phase. As the program develops, the data for carrying out CCEs, with a potential view to will be collected and compiled, thus generating the outsourcing the activities for which government basis for a well-established and actuarially sound personnel may be overtasked; (v) training and insurance program. accreditation for government and/or private sector personnel to ensure consistency with international The data for AYII are usually collected reinsurer data collection standards; and (vi) reliable /// through crop-cutting experiments (CCEs), in auditing procedures to make sure that the national which samples of crops are harvested, dried, and international insurance community can have and weighed, and yield values are inferred. confidence in the quality of the data collected.27 /// This work is usually carried out by government extension officers but could be outsourced to private Finally, those designing and implementing /// entities if a great many CCEs are to be carried out. an AYII program for Kenyan agriculture In this scenario, extension officers could play a key must take into account lessons learned role in auditing the data collection activities. It is from similar programs in other countries, worth noting that SDA is currently in the process and should also take advantage of the of improving its data collection system in order latest developments in technology. Elements /// to harmonize it with international and regional like real-time data transfer through mobile phone standards. To this end, a dedicated set of guidelines connections, digital video recording, remote-sensing was published in January 2014 (GoK 2014b). Efforts performance indicators, GIS (geographic information by the GoK to update and improve the data collection system) mapping, GPS (Global Positioning System) SECTION system (partly in response to the devolution process georeferencing, etc. will increase the possibilities of started in 2012) will certainly help in developing an assessing production losses in an efficient, effective, AYII program. and transparent way. 03 54 K E N YA : A G R I C U LT U R E I N S U R A N C E S O L U T I O N S A P P R A I S A L above.31 Figure 10 and figure 11 illustrate the spatial Fiscal Costing Assumptions distribution of the estimated district-level premium and Scenarios risk rates for AYII policies for maize and wheat, respectively. The objective of this section is to provide /// indicative references on the potential fiscal It is very important to note that the pure /// cost of developing an AYII scheme for maize premium rates presented in this report are and wheat producers. In order to develop such /// purely indicative and that the basic analysis projections, it is necessary to estimate the potential carried out in this context aims only to cost of insurance policies and to define the key highlight the diverse risk exposure of the assumptions for potential policy choices and for the different areas of Kenya. The responsibility expected uptake of the proposed insurance products. to perform appropriate actuarial analyses for underwriting purposes lies with the The cost of insurance is made up of several insurance industry. In addition, we emphasize /// key components, such as the cost of risk (in /// that the maize data used in the analysis are technical terms, the “pure risk premium”) composed of annual yield values that do not account and the charges required to cover data for production performance in the individual long collection, reinsurance fees, administration and short rainfall seasons. This makes estimating costs, tax, profits, and any other cost of doing yield variability in the individual seasons more business. Such charges are often estimated difficult and increases the uncertainty in estimating as a multiple of the cost of risk and, for the the pure risk rate. purpose of this analysis, we assume that they will double the pure risk premium. An /// Despite the limitations in the production data /// approximated way of estimating the insurance available, it is still possible to identify rough premium is indeed to start from the pure risk operational estimates of the fiscal costs premium and scale it up by a comprehensive loading of an AYII program.32 However, for potential /// factor defined as “premium multiple.” With the implementation activities, specific care should be premium multiple set at 2, a pure risk premium rate taken in developing seasonal-based contracts. of (for example) 6 percent in a particular area means An essential assumption underlying this that the final commercial premium rate in that area— /// fiscal costing exercise is that the GoK will the rate at which at which the policy will be sold—is provide direct financial support to the AYII 12 percent.28 scheme.33 The first means for channeling A preliminary assessment of the pure /// public support will be to finance the cost of risk premiums for both maize and wheat risk. The analysis assumes that the GoK will was carried out on the basis of historical cover a 50 percent share of such costs. Risk /// production records provided by SDA at financing support can be structured in many ways, district level and assuming an 80 percent and for the purpose of this analysis we assume that coverage level. As indicated in more detail in /// 29 it would come as a dedicated “risk financing fund” annex C.1, the data were carefully analyzed, revised, covering part of reinsurance costs, or as premium SECTION and detrended.30 Pure premium rates were then subsidies. determined on the basis of the historical payout 03 performance at the coverage level mentioned K E N YA : A G R I C U LT U R E I N S U R A N C E S O L U T I O N S A P P R A I S A L 55 Public support for AYII will also entail /// amount of subsidized insurance could be introduced. providing resources to complement the data Differentiations could be also made between maize collection activities needed for operating the and wheat production activities, given that the latter insurance scheme. The current fiscal scenarios /// are traditionally carried out in larger and more assume that the GoK will cover the cost of the sustainable production units. activities needed to complement the estimation An important dimension to be defined for process carried out by the public extension service, /// determining the value insured per district including costs for equipment, labor, management, is the expected take-up rate of insurance and auditing.34 However, more complex arrangements products (identified, in insurance terms, as can be envisioned in which the private stakeholders the “degree of penetration”). As a tentative also play relevant roles in supporting the data reference, the fiscal scenarios have been collection process. developed by starting at 3 percent of The current costing exercise does not /// cultivated area at the beginning of the distinguish between commercial and program in 2016, and reaching 15 percent for subsistence farming, although different maize and 25 percent for wheat in 2023.35 The /// supporting schemes could be envisioned for penetration rate is clearly difficult to predict, as it is the two farming typologies. For example, in the /// a function of many variables, some under the control areas where agricultural production is carried out of the program and some not. These projections are by smallholders at subsistence level, the program based on the assumption that AYII will be retailed in could be operated in a more socially oriented connection with agricultural input credit operations fashion by having higher levels of support (for that are currently accessed by less than 5 percent example, near to full premium subsidy support). of farmers (see section 4.4). The availability of At the same time, in more commercially oriented AYII should allow financial institutions to expand production environments, specific limits to the their agricultural lending operations, generating Figure 10 —  Estimated AYII Risk Figure 11 —  Estimated AYII Pure Premium Premium Rates for Maize at District Level Rates for Wheat at District Level SECTION 03 56 K E N YA : A G R I C U LT U R E I N S U R A N C E S O L U T I O N S A P P R A I S A L a mutually reinforcing process that could lead to the program has reached significant scale a progressive increment in the take-up of both (see table 7 for maize and table 8 for wheat, insurance and credit. respectively). The bulk of the estimated fiscal /// support—nearly 90 percent of resources provided by Under the assumptions presented in the the GoK—would be directed to maize production. /// analysis, and excluding expenses related to other support activities, the direct fiscal costs to be borne by the GoK for supporting the development of a national AYII program for maize and wheat would total approximately K Sh 140 million (US$ 1.6 million) at the start of the program, and K Sh 740 million (US$ 9 million) per year in 2021 assuming that Box 3—  Coverage Levels and Premium Rates Table 6 provides an example of how the premium rates of hypothetical AYII contracts vary according to different coverage levels. In the /// /// simulations presented below, for a coverage level of 80 percent, the premium rate would be below the adopted 15 percent cap only for the district of Uasin Gishu. In order to meet the 15 percent threshold, it would be necessary to reduce the coverage level to 70 percent for Kajiado, and to 50 percent for Machakos. These simple examples show the clear tradeoff between cost and coverage of AYII policies. This tradeoff is driven by the underlying risk; and where risk proves excessive, insurance may not represent an economically viable proposition Table 6 —  Variation of Premium Rates According to Different Coverage Levels Coverage Level District 80% 70% 60% 50% Machakos 31% 26% 20% 14% Kajiado 17% 11% 7% 4% U/Gishu 4% 3% 3% 2% The case of Machakos suggests how combining seasonal production data in one annual observation may distort the perception of the risk profile in the area. The extremely high premium rate estimated for Machakos is due to the inclusion of the March-May rainfall season, significantly drier than the October-December one. Seasonal production data would provide different insurance premium rates for the two seasons, leading to different risk management recommendations. SECTION 03 K E N YA : A G R I C U LT U R E I N S U R A N C E S O L U T I O N S A P P R A I S A L 57 Table 7—  Fiscal costing projections for AYII for maize, 2016–2023 2016 2017 2018 2019 2020 2021 2022 2023 Insurance Penetration (as a % of cultivated 3.0% 4.7% 6.4% 8.1% 9.9% 11.6% 13.3% 15.0% area) Penetration (hectares) 61,517 96,670 131,822 166,975 202,128 237,280 272,433 307,586 Premium volume (million KSh) 253 398 543 687 832 977 1,122 1,266 Projected public support as a share of 50% 50% 50% 50% 50% 50% 50% 50% premium volume (%) Cost of premium subsidy for GOK (million 127 199 272 344 416 489 561 633 KSh) Additional costs for data collection / yield 0.2 0.6 1.2 1.7 2.3 2.9 3.5 3.8 estimation (million KSh) Number of farmers covered (per season) 25,632 40,279 54,926 69,573 84,220 98,867 113,514 128,161 Total cost for GoK (million KSh) 127 200 273 345 418 491 564 637 Total cost for GoK (million USD at 85 KSh/ 1.5 2.3 3.2 4.1 4.9 5.8 6.6 7.5 USD) Table 8—  Fiscal costing projections for AYII for wheat, 2016–2023 2016 2017 2018 2019 2020 2021 2022 2023 Insurance Penetration (as a % of cultivated 3.0% 6.1% 9.3% 12.4% 15.6% 18.7% 21.9% 25.0% area) Penetration (hectares) 3,819 7,820 11,821 15,822 19,823 23,824 27,825 31,827 Premium volume (million KSh) 24 49 74 98 123 148 173 198 Projected public support as a share of 50% 50% 50% 50% 50% 50% 50% 50% premium volume (%) Cost of premium subsidy for GOK (million 12 25 37 49 62 74 87 99 KSh) Additional costs for data collection / yield 0.01 0.04 0.07 0.11 0.14 0.18 0.21 0.24 estimation (million KSh) Number of farmers covered 1,273 2,607 3,940 5,274 6,608 7,941 9,275 10,609 Total cost for GoK (million KSh) 12 25 37 49 62 74 87 99 Total cost for GoK (million USD at 85 KSh/ 0.1 0.3 0.4 0.6 0.7 0.9 1.0 1.2 USD) SECTION 03 58 K E N YA : A G R I C U LT U R E I N S U R A N C E S O L U T I O N S A P P R A I S A L Welfare Impacts of Area 2004), Kenya Integrated Household Budget Survey data (2005)36, KARI (2009)37, and Yield Insurance for Maize Tegemeo Institute (2010) —show that 38 23–30 percent of maize and wheat farmers and Wheat in Kenya report using high-yielding technology and Maize and wheat productions are crucial /// hybrid seeds. And while almost 50 percent /// for the livelihoods and food security of of farmers reported using some kind of smallholder and medium-scale farmers in input credit, less than 5 percent of these Kenya. Maize-growing areas span the country and /// farmers reported obtaining credit from can be classified into three production zones with formal financial institutions. Other sources of /// distinct production systems and socioeconomic production credit include cooperatives, savings and conditions. The low-potential zone occupies low- credit cooperatives (SACCOs), local traders, input yielding and high-risk production in Eastern and suppliers, and other informal financial institutions. Central Provinces, where the majority of farmers are Statistically, input loans have been relatively small, poor smallholders (median farm size is 1.5 hectares) just enough to afford minimum input costs, and who use subsistent production technology (we have been offered at interest rates of 8–19 percent sometimes refer to this zone as the subsistent maize per year. Among other things, limited access to zone). The medium-potential zone occupies the agricultural credit has thus served as a key supply- relatively higher-yield, lower-risk regions of Nyanza side constraint to productive agricultural investment. and Western Provinces, where farmers are slightly Maize and wheat productions are better off but still smallholders. The high-potential /// significantly exposed to extreme production zone occupies the high-yielding production regions risk. We used detrended district-level yield of Rift Valley Province, with relatively larger-scale data for the 30-year period 1983–2012 farmers (with 2.5 hectares of land on average). (obtained from the MALF report [GoK 2014a]) Maize production is one of the main livelihood to determine how often maize production bases and is mainly used for home consumption, falls below 80 percent of the district average. especially in the low- and medium-potential zones. The analysis showed that these shortfalls This is in contrast to the high-potential zone, where are 1-in-3-year events in the low-potential production is relatively more commercialized. Wheat zone and 1-in-4- to 1-in-5-year events in the production is concentrated in smaller regions of the other two maize zones and wheat region. Eastern and Rift Valley Provinces, is relatively more /// The low-potential zone thus appears to have a larger commercialized, and is practiced by relatively larger- exposure to production risk than the other zones, scale farmers (with 3.0 hectares of land on average). with significant drops in production—below 50 (See annex C.2, table 18 for summary statistics of percent of the district average—occurring once in maize- and wheat-growing households.) five years. While Kenyan farmers have established Low investment in productive inputs and /// various informal risk-sharing mechanisms that limited access to production credit have allow unaffected farmers to help affected farmers impeded efforts to improve productivity in reduce consumption shortfalls from shocks, these SECTION both maize and wheat productions in Kenya. /// mechanisms tend to be ineffective against extreme Data from a variety of sources—Tegemeo production shocks, which generally affect whole 03 Institute’s household survey data (2000, communities. K E N YA : A G R I C U LT U R E I N S U R A N C E S O L U T I O N S A P P R A I S A L 59 Uninsured production risk could have /// The model is then calibrated using a combination of significant welfare impacts on Kenyan maize 30 years of district production data from MALF and and wheat farmers, not only by increasing detailed household survey data from the Tegemeo their vulnerability but also by reinforcing Institute (2000, 2004)39 covering key maize-growing both supply- and demand-side constraints areas of the country. Overall, maize yields vary on smallholders’ adoption of productive significantly across the three production zones, with inputs. Extreme production risk directly affects /// the highest CV40 of 0.49 in the subsistent maize zone, welfare by reducing the income/food available following by 0.34 in the high-potential zone, 0.35 in for consumption, especially among the poor the wheat region, and 0.29 in the medium-potential smallholders whose livelihoods rely extensively zone. Input costs vary from 50 percent to 75 percent on these crop productions. Exposure to extreme of the expected crop revenues. For both crop production risk could further reduce investment productions, we thus assume that a farmer needs to incentives, especially among risk-averse poor farmers. Empirically, according to the household take an input loan at a median rate of 60 percent of survey, farmers have always cited uninsured risks as expected revenue and at a cost of 17 percent interest one of the key reasons for their underinvestment in per year. production. At the same time, as agricultural loan Net income available for consumption and portfolios would also be exposed to large default /// expected loan repayment rates vary greatly risk following extreme production shortfalls, lenders with frequency and severity of shocks in all tend to limit the supply of agricultural credit or offer credit at relatively high interest rates. Overall, zones. The black lines in figure 12 reflect annual /// through the direct effect on vulnerability and the maize and wheat income after netting out input indirect effect on productivity, exposure to uninsured loan repayment and thus the net income that risk could increase the probability that Kenyan would be available for household consumption. farmers will fall into poverty. Our simulations considered both price and yield variability, and thus variations in net incomes reflect Our empirical analysis reviews the variations of both. As expected, the net incomes /// significance of, and variations in, exposures (realized in 1-in-2-year frequency) are very low— and welfare impacts of covariate production lower than the national food poverty line (at K Sh risk on representative farmers in the key 988 per capita per month) in the subsistent and maize and wheat production zones. We /// develop a simple economic model to explore the medium-potential maize zones, and slightly above potential welfare impacts on a representative farmer the food poverty line in the high-potential maize in each of the three distinct maize zones and overall zone and in the wheat region.41 wheat production region (see annex C.2). In each Net incomes available for consumption could production zone, we assume that a representative /// drop to or below zero at a frequency of once farmer owns a farm of median size, produces with the in three years in the subsistent zone and at zone-specific production system, and realizes zone- specific crop yields and variability. A representative a frequency of once in four years in others. /// farmer is credit constrained and so needs to take In these cases, the farmer has no income left for SECTION an input loan at the beginning of the cropping year consumption and/or is unable to repay the full loan. to purchase required minimum inputs. The loan is repaid using crop income obtained after the harvest. 03 60 K E N YA : A G R I C U LT U R E I N S U R A N C E S O L U T I O N S A P P R A I S A L and that is possible within a 15 percent maximum commercial premium rate. Given differences in yield variations, insurance coverage varies across zones, with 50 percent, 85 percent, and 80 percent in low-, medium-, and high-potential maize zones respectively, and 75 percent in the wheat region. In figure 12, net income available for consumption is then plotted in red. As expected, AYII would reduce net income in good years, as a farmer would need to pay for the insurance premium, which is loaded at a multiple of 2. But the AYII payout could stabilize net income in bad years. Especially for households that rely /// extensively on crop production as their main consumption, AYII that reduces variability in crop production will also reduce households’ vulnerability to food insecurity. But when /// an extreme shock—one with at least a 1-in-4-year frequency—occurs, this high-coverage commercial AYII generally does not guarantee enough income for full repayment of input loans in any of the zones. A 1-in-10-year production risk could further There could several reasons for this: (i) commercial /// force farmers in all production zones to AYII is quite expensive; (ii) there is still basis risk accumulate debt of up to 80 percent of associated with AYII because it provides protection their expected income each year. In reality, /// only with respect to the district-level yield, not with however, a farmer might not use all crop income to respect to the individual yield; and (iii) there are pay back a loan. To make this scenario more realistic, other background risks due to uninsured variations we computed the expected loan repayment rates assuming that a farmer will try to pay back the loan in prices. after meeting the necessary subsistent consumption AYII could potentially increase the ability /// at 30 percent of the food poverty line. As shown in of farmers to pay back input loans in annex C.2, table 19, the expected loan repayment the bad years and so increase expected rates in all zones could be reduced by extreme loan repayment rates of rural lenders’ shocks.42 loan portfolio (see annex C.2, table 19). /// /// Area yield index insurance could potentially /// Commercial AYII could stabilize loan stabilize consumption in the years when repayment rates in bad years and so increase extreme shocks affect the entire community. /// expected loan repayment by as much as SECTION We first explore the potential of high-coverage 10 percent when the farmer faces 1-in-10- AYII that pays out based on a district-level yield year production risk relative to the case 03 index at a coverage level specific to each zone, without AYII. /// K E N YA : A G R I C U LT U R E I N S U R A N C E S O L U T I O N S A P P R A I S A L 61 Public support that reduces the commercial /// (relative to the average cost) and the expected premium rate could significantly improve the yield improvement (relative to the average yield) welfare impacts of AYII on maize and wheat from a detailed maize production study by the households. We show that a 50 percent Kenya Agricultural Research Institute (KARI) in reduction in commercial premium (which 2009 and from a wheat production’s gross margin allows the farmer to pay a fair premium rate), study of DASS (see annex C.2). Farmers who can could potentially allow AYII to stabilize net afford to invest 126 percent and 138 percent more income available for consumption above in productive inputs could improve maize yields by zero even with the extreme 1-in-10-year risk as much as 196 percent in the high-potential maize in all but the risky subsistent zone. Thus, for /// zone and 182 percent in the medium-potential zone, farmers who always use all crop income to repay respectively. Similar but less significant evidence their loan, fair AYII could ensure full loan repayment is also found for wheat farmers, for whom 133 even in extremely bad years. Even if farmers meet percent more investment in productive input could their subsistent consumption level before repaying enhance yield by up to 139 percent. The productivity the loan, fair AYII could increase expected loan gain from increasing productive investment, repayment by as much as 20 percent in extreme years however, could be limited in the subsistent maize relative to the case without AYII. production zone by the zone’s low production potential and scarce rainfall. Thus the extra cost of If insurance could further unlock access to /// expensive input appears to outweigh the additional agricultural credit and enhance farmers’ yield improvement. investment incentives, even commercial AYII could potentially crowd in sustainable So while commercial AYII might be too /// increases in productivity in line with a key expensive as a stand-alone insurance in this recommendation of the Kenya Vision 2030. /// setting, if it could unlock access to credit, Various studies have documented positive effects it could potentially crowd in significant of derisking agricultural production on productive improvement in income and reduce the investment and credit demand—e.g., Cai et al. (2009) probability of falling into poverty for farmers in China; Galarza and Carter (2011) in Peru. Existing in all zones except the low-potential zone. agricultural programs in Kenya have also successfully As the green lines in figure 12 show, the allowed banks to expand lending to farmers using crowding in effect of even the commercial insurance as a prerequisite for loans and/or by AYII could improve expected net income bundling insurance with credit directly. available for consumption by more than double in the medium- and high-potential Since the above analysis shows that AYII /// zones and by about 65 percent in the wheat can remove some production risk from region. The significant productivity gain from rural lending institutions and thus increase expanded credit with commercialized AYII expected loan repayment rates, we explore could further result in 67 percent and 30 the potential impacts of this effect by percent reductions in the probability of falling allowing insured farmers to access larger into poverty for farmers in the high and loans for investment in expensive but more medium zones, respectively. This crowding- SECTION /// productive inputs (hybrid seeds, fertilizer, in effect could be smaller for the better-off wheat equipment). We used the crop- and zone- farmers, who already use relatively more expensive 03 /// specific evidence of expensive input cost markups inputs and achieve relatively higher productivity. 62 K E N YA : A G R I C U LT U R E I N S U R A N C E S O L U T I O N S A P P R A I S A L The crowding in effect of AYII might be limited for double the production in high- and medium- farmers in the low-potential zone, however. potential zones, and almost double the By subsidizing AYII and using AYII to crowd /// production in the wheat region. The program in productive input loans, the government could potentially reduce farmers’ chance of could further ensure sustainable and falling into poverty by 78 percent, 39 percent, significant increases in productivity and thus and 29 percent in the high-potential zone, in agricultural GDP, and in this way contribute medium-potential zone, and wheat region, toward achievement of the Kenya Vision respectively (see annex C.2, table 19). These 2030. In turn, this approach could move /// many small- and medium-scale farmers in poverty reduction effects come about as the AYII some production regions out of poverty. and credit enhance farmer’s productivity, and as AYII Subsidized AYII with extended productive acts as a safety net to protect yield shortfalls in bad input loans could potentially more than years. Figure 12 —  Potential impacts of AYII on net income available for consumption SECTION Note: The net income available for consumption depicted in the figure reflects crop income after any loan repayment. It does not account for the potential that households might use some part of this for saving before consumption. Thus this should be 03 viewed as an upper bound of income that will be available for consumption. K E N YA : A G R I C U LT U R E I N S U R A N C E S O L U T I O N S A P P R A I S A L 63 Table 9—  Fiscal cost per household of achieving different policy goals in different insurance scenarios Public cost (KSh) per unit Policy Crop zone Free provision Subsidize AYII objectives Subsidize AYII Cash transfer of AYII + unlock credit Low potential maize no effect no effect no effect 1,394 1% reduction Medium potential maize 27,453 no effect 118 1,169 in poverty High potential maize 4,425 no effect 420 1,832 Wheat 5,408 no effect 1,722 2,355 Low potential maize 761 3,774 no effect 802 1% Medium potential maize 937 1,328 255 1,206 reduction in High potential maize 3,540 4,963 739 2,917 vulnerability* Wheat 4,804 5,235 1,679 4,175 Low potential maize 2.01 no effect no effect 1.00 1 KSh increase Medium potential maize 2.02 no effect 0.10 1.00 in expected High potential maize 2.01 no effect 0.08 1.00 income Wheat 2.00 no effect 0.19 1.00 * Measured by probability of falling below zero net income available for consumption Overall, the welfare impacts of AYII vary /// improvement through increased productive input across production zones with different use is low. degrees of risk exposures, and AYII might The welfare impacts of AYII could also vary not be suitable as an intervention to improve /// slightly across different insurable indexes smallholders’ productivity in the subsistent and coverage levels. Changing from a district- maize production region. Given this region’s /// level yield index to a division-level index, for which /// low expected yield and large exposure to production correlations of individual yields with the measured risk, AYII with large coverage could be too expensive area yield are potentially larger, could achieve a to be useful for these farmers. But the coverage larger reduction in net income variability (see level (currently at 50 percent) affordable within a annex C.2, table 19). The performance of AYII in 15 percent commercial premium could also be too reducing income variability also declines as one low to effectively insure net income and expected moves from the high coverage, with 15 percent loan repayment against extreme shocks. Even with maximum premium rate, to the lower coverage level 50 percent premium reduction through public affordable within 10 percent commercial premium. supports, extreme production shock could still cause This analysis assumes that there could be effective serious shortfalls of consumption and expected loan insurance demand even at the high commercial rate. SECTION repayment. AYII is unlikely to unlock credit access and so improve productivity for smallholders in this Government’s support to development of 03 /// subsistent zone, where the potential for productivity an AYII program could be a cost-effective 64 K E N YA : A G R I C U LT U R E I N S U R A N C E S O L U T I O N S A P P R A I S A L strategy for meeting various policy would cost as little as K Sh 0.08 per household per objectives. To identify the most cost-effective types /// year for the same scheme to improve productivity of support that can achieve different policy objectives and so increase household income by K Sh 1. And targeted to different subsets of maize and wheat this is clearly cheaper than direct one-to-one cash farmers, we compute the cost per household (in transfer. The combination of government subsidies Kenyan shillings) per year of four types of support— for AYII and crowding in of input credit would not explained below—to achieve three outcomes (see be an effective policy tool for smallholders in the table 9): (i) 1 percent reduction in the poverty rate subsistent maize zone, however. relative to the baseline without the program; (ii) 1 For a cost-effective tool to reduce percent reduction in the vulnerability rate (measured /// vulnerability of smallholders in the subsistent by the probability of net income falling below zero); maize zone, the government could freely and (iii) an increase of K Sh 1 in net income available provide AYII coverage as a social protection for consumption when targeted to each of the program. This could lead to a 1 percent reduction maize and wheat areas . The four types of support /// in vulnerability (i.e., the probability of household’s include (i) free provision of AYII, (ii) 50 percent net income falling to zero) and would cost the subsidization of AYII, (iii) 50 percent subsidization government about K Sh 761 per household per year. of AYII and facilitation of access to input credit, This is cheaper for government than providing only and (iv) direct cash transfer program, the cost of a 50 percent subsidy for AYII (which could cost which we compare to the costs of the first three K Sh 3,774 per household per year) and the direct interventions. For the high-coverage AYII program cash transfer (which could cost about K Sh 802 per (at 15 percent maximum premium rate), the free household per year). If the policy goal is to reduce provision of AYII could cost government from K Sh poverty, however, government’s support through 2,642 per household per year (in the subsistent maize AYII would not be the appropriate policy tool zone) to K Sh 34,448 per household per year (in the relative to the direct cash transfer program. wheat region). The cost is thus reduced by half when the government subsidizes only 50 percent of AYII’s premium cost. The types of public support that reduce /// AYII’s commercial premium and unlock the agricultural credit market could be the cost-effective tools that allow government to reduce poverty and vulnerability and to improve productivity among the median farmers (smallholders) in the medium- and high-potential maize zones and the wheat region. In the medium-potential zone, it would /// cost as little as K Sh 118 per household per year for the government to reduce the poverty rate by 1 percent through subsidizing AYII and crowding SECTION in input credit access. This compares to K Sh 1,169 per household per year if the government tried 03 to achieve the same goal through cash transfer. It K E N YA : A G R I C U LT U R E I N S U R A N C E S O L U T I O N S A P P R A I S A L 65 Conclusion This report provides a detailed technical /// For crop insurance, investments in data and /// analysis of how the government of Kenya linkage to credit are key. Investments in data /// (GoK) could develop agriculture insurance will be required to develop high-quality products public-private partnerships (PPPs) to support that provide meaningful coverage to farmers, reduce basis risk, and ensure that payments are made when rural livelihoods and help to raise Kenya to necessary. The welfare analysis has highlighted middle-income status, as described in Kenya the importance of linking insurance to credit as a Vision 2030. This report, written with the guidance way of empowering rural farmers to make capital /// of the GoK, analyzes possible PPP structures and investments on their farms, raising household explores the options for developing crop and income, and increasing the size of farms. The livestock insurance programs in the short, medium, main costs to the GoK include developing the data and long term. market infrastructure and some form of support for financing the cost of risk. This report is meant to guide the GoK in /// key policy decisions based on the potential For livestock insurance, linking to the /// fiscal cost and potential welfare benefits of Hunger Safety Net Program (HSNP) in the developing an agriculture insurance PPP. /// four northern counties is the key initial step. /// Both the fiscal costing analysis, which estimates Building on the scalable component of the HSNP, the the resources required to develop the PPP, and the analysis provides details of the costs and benefits of a GoK-funded livestock insurance scheme that will welfare analysis, which looks at how agriculture reduce the vulnerability of low-income families, in insurance could benefit farmers, are provided toward addition to laying the foundation for a politically that end. Leading the way for the African continent, sustainable livestock insurance market. Initially, these decisions will help to establish an appropriate it is envisaged that a macro-level product will be policy framework for an effective agriculture developed and given to vulnerable households. Over insurance PPP in Kenya. time, top-up coverage will be made available to the covered households, in addition to other households On the institutional side, the report discusses in the target counties. /// key next steps toward developing a National Agricultural Insurance Policy (NAIP), in Fiscally, the approach suggested in this /// addition to the potential institutions to report would entail certain costs to the GoK be established in support of the PPP’s through its involvement in the PPPs. These public aspect. are outlined in table 10. SECTION /// /// 04 66 K E N YA : A G R I C U LT U R E I N S U R A N C E S O L U T I O N S A P P R A I S A L Table 10—  Illustrative fiscal costing for agricultural insurance programs, 2016 and 2019 Annual fiscal cost Estimated number of Average cost per producer Program Description (2016) (K Sh millions) producers covered per year (K Sh) Maize: area yield index insurance 127 25,632 5,000 Wheat: area yield index insurance 12 1,273 9,500 Pastoralists: satellite-based livestock protection insurance 300 71,000 4,200 (fully subsidized) Pastoralists: satellite-based livestock protection insurance 14 5,250 2,600 (partially subsidized) TOTAL 453 103,155 Annual fiscal cost Estimated number of Average cost per producer Program Description (2019) (K Sh millions) producers covered per year (K Sh) Maize: area yield index insurance 345 69,573 5,000 Wheat: area yield index insurance 49 5,274 9,200 Pastoralists: satellite-based livestock protection insurance 300 71,000 4,200 (fully subsidized) Pastoralists: satellite-based livestock protection insurance 31 15,000 2,100 (partially subsidized) TOTAL 725 160,847 SECTION 04 K E N YA : A G R I C U LT U R E I N S U R A N C E S O L U T I O N S A P P R A I S A L 67 Annex A. Possible Options If a coinsurance pool is established as an /// insurer, the pool company underwrites the risks for Coinsurance Pools /// directly in its own right. A pool company that underwrites risks must, of course, be licensed in Kenya to write insurance business and must be fully As discussed in chapter 1, it is unlikely that insurers capitalized as an insurer. will be able to compete within a fully competitive Other coinsurance pools, whether or not market for agricultural insurance. The purpose of this /// /// established solely by contract or as a special section is not to make detailed recommendations for (noninsurer) company, usually share the following a pool structure, but to demonstrate the variety of features: pool structures that could be considered. 1. Each insurer accepts a pre-agreed share in all the Nonstatutory Coinsurance Pools risks that are covered by the pool agreement. Insurance pools can be statutory (i.e., established 2. All premiums are paid into the pool, less an by specific legislation) or nonstatutory (i.e., not amount to cover expenses. established by specific legislation). 3. The pool manager or administrator assesses and Different structures are commonly used to establish settles claims. nonstatutory insurance pools: 4. f there is an underwriting gain, the surplus 1. A coinsurance pool may be established by the (beyond any reserve retained in the pool) is paid participating insurers as an insurer in its own to each insurer in accordance with its agreed right, so that it is the pool itself that issues the share. insurance contracts and assumes the risk on behalf 5. If there is an underwriting loss, the insurers of the insurers. In this case, either the pool would contribute to the loss in accordance with their sell its own insurance contracts or the insurers agreed share. would sell insurance contracts as intermediaries (i.e., agents) on the pool company’s behalf, the risk being underwritten by the pool company. 2. The insurance contracts may be written by the insurer pool members, on an individual basis, but with the risk ceded to the pool. In this case, the pool may be either (i) a special pool company established by the insurers; or (ii) an arrangement between the insurers whose terms are set out in a pool agreement. 3. The insurance contracts may be written by a lead insurer on behalf of the other insurers that are members of the pool. Again, under this scenario, the pool may be a special company established SECTION by the insurers or an arrangement between the insurers set out in a pool agreement. 05 Photo Credit: Neil Palmer (CIAT) 68 K E N YA : A G R I C U LT U R E I N S U R A N C E S O L U T I O N S A P P R A I S A L Box 4—  Benefits and limitations of coinsurance pool arrangements Coinsurance pools offer these benefits: They achieve economies of scale through operating as a single unit with shared (pooled) administration and operating functions. These lead /// /// to costs savings from (i) reduced staffing requirements (fixed costs); (ii) shared costs of product research and development, and actuarial services including rating; and (iii) reduced costs of underwriting and claims control and loss adjustment. There are cost advantages to companies when they purchase common account (pooled) reinsurance protection rather than trying to place /// their own reinsurance program. The advantages arise from (i) a stronger negotiating position with reinsurers; (ii) larger and more balanced /// portfolio and better spread of risk; (iii) reduced costs of reinsurance due to pooled risk exposure; and (iv) reduced transaction costs (reinsurance brokerage, etc.). There is no competition on rates in a soft market, and pools can maintain technically set rates. Most pools operate as the sole insurance /// /// provided or monopoly (as in Austria, Senegal, Spain, and Turkey, for example), and there is therefore no competition on pricing. Pools are able to maintain underwriting and loss adjustment standards. Under a pool monopoly arrangement, the pool manager can ensure /// /// that common and high standards are maintained in the underwriting of crop and livestock insurance and in the adjusting of claims. Where companies are competing against each other for standard crop insurance business, there is often a problem of varying loss adjustment standards between companies. Within a PPP, governments can more easily coordinate support to a pool than to individual insurers. Governments seeking to coordinate /// /// national agricultural insurance policy and planning and specific support functions (e.g., provision of premium subsidies, research and development, education and training) can work more easily with a pool than with individual insurers, each of which may have very different priorities for agricultural insurance. Coinsurance pools have these limitations: When a pool acts as the sole agricultural insurer, lack of competition in the market may result. This could (i) limit the range of products and /// /// services offered by the monopoly pool underwriter; (ii) restrict the range of perils insured; (iii) restrict the regions where agricultural insurance is offered and/or the type of farmer insured; and (iv) lead to a lack of competitiveness in premium rates charged by the pool. Source: Mahul and Stutley 2010. If a pool is established solely through a contractual It is important to appreciate that where the insurers arrangement, the “pool” is not a legal person and write their own insurance contracts and cede the does not have the power to contract. The pool could risk to the pool, each participating insurer typically not, therefore, write insurance contracts. accepts a pre-agreed share of all the risks ceded to the pool, not just the risks that the insurer has If the insurers enter into their own individual written. insurance contracts, the insurance business is conducted under their individual licenses. The Management of a coinsurance pool, where /// /// capital of the participating insurers supports the the pool is incorporated as a (noninsurance) SECTION risk. The position may be rather more complicated company, involves the pool company acting as the if the insurance contracts are underwritten by a lead pool manager or administrator. Where a special 05 insurer on behalf of the other insurers. pool company is not incorporated, the pool K E N YA : A G R I C U LT U R E I N S U R A N C E S O L U T I O N S A P P R A I S A L 69 may be managed by a lead insurer; by a technical Benefits of an Agricultural Insurance Pool management unit contracted or employed by, or on All coinsurance pools offer benefits but also have behalf of, the participating insurers; or by a third limitations. These are summarized in box 4. party such as a broker, another nonparticipating insurer, or a reinsurer. The participating insurers International Precedents typically share the management costs in accordance with their proportionate risk share. If a Program Steering Committee is established to address the institutional framework for agricultural Statutory Coinsurance Pools insurance, it could consider a number of precedents: (i) the Turkish Agricultural Insurance Pool Statutory insurance pools are often, but not (TARSIM); (ii) the Spanish Agricultural Insurance /// /// necessarily, corporate bodies. Usually, statutory Pool (AGROSEGURO); and (iii) the proposed coinsurance pools are part of a national or regional Mongolian Index-Based Livestock Reinsurance program and are established as part of a public- Company (which will have features of a pool and a private partnership (PPP). Relevant legislation reinsurance company). typically provides for the governance of the pool and sets out the pool’s functions. The legislation may also The Turkish and Spanish pools are considered in cover other matters, such as the provision of some more depth in the MALF report (GoK 2014a). form of subsidy. Because they are established by legislation, statutory pools take many forms and may be structured very differently to a typical voluntary pool. The legislation may establish a coinsurance pool, but not as a corporate body. For example, the pool may be established as a contractual arrangement between participating insurers. In this case, although the legislation would set out the functions of the pool, those functions would not usually include acting as an insurer, since the pool is not a legal person. Of course, the legislation may establish a corporate body to act as manager of the pool, but not to write insurance contracts. The legislation establishing the pool would usually provide the pool with exclusive rights in relation to the business underwritten by the pool. This is necessary to prevent nonpool insurers undermining the pool by offering similar insurance products at a lower, nonsustainable, price. SECTION Statutory coinsurance pools sometimes operate as hybrids, with some limited reinsurance functions. 05 70 K E N YA : A G R I C U LT U R E I N S U R A N C E S O L U T I O N S A P P R A I S A L Annex B.1. Index Based Livestock Insurance (IBLI) Program Figure 13 —  Translating NDVI data into estimated livestock mortality and IBLI payouts Response Insurance Predicted NDVI data Function Contract Indenmity purchased (measurement mortality (payout (estimation payouts (pricing precision options) precision index options) option) options) Source: ILRI 2014. Figure 14 —  IBLI seasonal sales periods, contract cover period and contract payout dates 1 year contract coverage LRLD season coverage SRSD season coverage Jan Feb Mar Apr Jun Jul Aug Sept Oct Nov Dec Jan Feb Sale period for Period of NDVI observations for constructing LRLD mortality LRLD index Sale period for Period of NDVI observations for constructing SRSD SRSD mortality index Predicted LRLD mortality is announced. Indemnity payment made if IBLI is triggered. SECTION Predicted SRSD mortality is announced. Indemnity payment made if IBLI is triggered. 06 Source: ILRI 2013 Note: LRLD = long rain /long dry season; SRSD = short rain / short dry season. K E N YA : A G R I C U LT U R E I N S U R A N C E S O L U T I O N S A P P R A I S A L 71 Table 11—  IBLI livestock insurance results, 2009-2012 (US$) No. No. Tropical Total sum Premium Paid Avg. no. Avg. sum Avg. premium Avg. premium Sales Year contracts Livestock Units insured (TSI) by Herders insured TLUs insured TLUs rate per herder period sold insured (TLUs) (US$) (US$) per herder per herder (%) (US$) Jan/Feb 2010 1,974 5,965 1,118,437 46,602 3 187.5 4.2 23.6 Jan/Feb 2011 595 1,229 230,437 9,033 2.1 187.5 3.9 15.2 Aug/Sept 2011 509 836 158,750 6,122 1.6 187.5 3.9 12 Aug/Sept 2012 219 413 77,437 3,150 1.9 187.5 4.1 144 Total 3,297 8,443 1,583,061 64,907 2.6 187.5 4.1 16.3 Source: ILRI 2013 Note: Exxchange rate: 1US$ = 1 KShs 0.80. pastoralists to be covered on the basis of the given Annex B.2. Assumptions budget. and Parameters for The parameters presented in table 12 are based on Fiscal Costing Scenarios the following assumptions and considerations: for Livestock • Sum insured per TLU: Each TLU is valued at /// ////// K Sh 7,000 in Case A, and at K Sh 5,000 in Case B. 1. Fiscal costing for the macro-level insurance Values lower than K Sh 5,000 are not considered /// coverage for asset protection /// meaningful. The fiscal costing scenarios for the macro-level • Number of TLUs insured per vulnerable /// Normalized Difference Vegetation Index (NDVI)– pastoralist: The number of eligible TLUs has /// based insurance coverage for livestock asset been set at seven in Case A and at five in Case B. protection have been developed in order to calculate Five TLUs are considered the level below which the number of pastoralists covered on the basis of the insurance coverage would not provide useful budget references provided by the State Department support to pastoralists’ livelihoods. of Livestock (SDL): K Sh 100 million, K Sh 200 • Sum insured per pastoralist: The reference million, and K Sh 300 million. /// /// sum insured per pastoralist is obtained by For each budget scenario, two extreme cases are multiplying the number of TLUs to be covered presented (see table 12). Case A is structured by by the selected value of 1 TLU. The parameters selecting, within a reasonable range of variation, selected in table 12 determine a range of sums the more costly extremes of the key parameters insured between K Sh 49,000 (Case A) and K (i.e., higher values per tropical livestock unit [TLU] Sh 25,000 (Case B). The difference between the two extremes is significant because it highlights insured, a higher number of TLUs per policy, and how the policy choices that go into selecting a higher insurance premium estimate). This will the relevant parameters influence the support define a lower bound for the number of pastoralists provided to pastoralists. SECTION to be covered for the reference budget figure. Case B takes into account the less expensive options, • Premium rate: As the NDVI asset protection 06 /// /// thus identifying the higher bound of the number of product is still in the design phase, actual 72 K E N YA : A G R I C U LT U R E I N S U R A N C E S O L U T I O N S A P P R A I S A L estimates for the potential premium rates of • Cost of education and training: The cost /// /// the program are not available. Hence, while the for education and training has been set at K Sh necessary elaborations are being carried out, the 5 million for Case A and at K Sh 4 million for current average premium rates of the Index Based Case B. Livestock Insurance (IBLI) products are used. The • Cost of payout distribution: Given that IBLI scheme allows pastoralists to select between /// /// the enrolled pastoralists will all be equipped two trigger options. The average premium for with bank accounts, costs for distributing the the products with the lower trigger (hence the payouts should be minimal (i.e., the cost of the version that provides payouts more frequently) bank transfer operation), and could be included is 16.06 percent, while the average premium for in the costs budgeted by insurance companies. the higher trigger option is 9.24 percent. Hence, However, raising awareness about program approximating such figures, the premium rate for payouts is important, so a lump sum of K Sh 2.0 Case A has been set at 15 percent and the premium million for Case A and of K Sh 1.5 million for Case rate for Case B at 10 percent. B will likely be used for dedicated information • Premium per pastoralist: The premium per /// /// campaigns on payout distribution. pastoralist is obtained by applying the selected • Cost of contract design and data premium rate to the sum insured per pastoralist. /// processing: Costs of K Sh 1 million have been The premium amount for Case A is set at K Sh /// assumed for handling NDVI data processing and 7,350 and for Case B at K Sh 2,500. Again, the monitoring the contract. spread between the two figures is quite significant, and this has relevant implications for the cost of • Cost of auditing: An auditing cost of 1 percent /// /// the program. of the value of the program has been assumed. • Cost of registration and enrollment per /// 2. Fiscal costing for the top-up and nontarget /// pastoralist: The cost for registration and /// purchases /// enrollment has been set at K Sh 500 for Case A As mentioned above, optional top-up coverage for and at K Sh 250 for Case B. The target pastoralists pastoralists enrolled in the program will be made belong to the Hunger Safety Net Program available in year 3 of program implementation. framework, and they could therefore be registered In addition, pastoralists who were not part of the automatically without generating any specific initial support program will also have the option cost. However, the technical analysis presented in to purchase the NDVI-based insurance coverage as chapter 2 suggests that registration for households a nontarget group of pastoralists. The first layers eligible for GoK subsidy should not be automatic of both the top-up option and the nontarget group but rather done in person. In-person registration coverage will be partially subsidized. helps to create a sustainable market for livestock insurance by ensuring that pastoralists understand The suggestion to make the top-up and nontarget the details of how the scheme operates. In group coverage available at year 3 of program addition, these activities could help spread implementation is motivated by the significant awareness of the insurance product, which could challenges to be faced when moving beyond a SECTION both promote the purchase of the top-up option fully subsidized coverage scenario. In addition, and encourage nontarget pastoralists to take up because the NDVI-based asset protection scheme 06 the insurance product. is still in the design phase, its performance will K E N YA : A G R I C U LT U R E I N S U R A N C E S O L U T I O N S A P P R A I S A L 73 Table 12—  Fiscal costing projections for macro-level asset protection coverage Scenario: KS h 100 million Scenario: K Sh 200 million Scenario: K Sh 300 million (US$ 1.2 million) (US$ 2.3 million) (US$ 3.5 million) Case A Case B Case A Case B Case A Case B Budget available for macro-level asset protection 100,000 100,000 200,000 200,000 300,000 300,000 coverage (K Sh) Sum Insured per Tropical Livestock Unit (TLU) (K Sh) 7,000 5,000 7,000 5,000 7,000 5,000 No. of TLUs insured per vulnerable pastoralist 7 5 7 5 7 5 Sum Insured per pastoralist (K Sh) 49,000 25,000 49,000 25,000 49,000 25,000 Premium Rate {as a share of sum insured) 15.00% 10.00% 15.00% 10.00% 15.00% 10.00% Premium per pastorarist (K Sh) 7,350 2,500 7,350 2,500 7,350 2,500 Cost of registration and enrollment per pastoralist 500 250 500 250 500 250 (K Sh) Cost of education and training {lump sum) (K Sh) 5,000,000 4,000,000 5,000,000 4,000,000 5,000,000 4,000,000 Cost of payout distribution (lump sum) (K Sh) 2,000,000 1,500,000 2,000,000 1,500,000 2,000,000 1,500,000 Cost of contract design and data processing (lump 1,000,000 1,000,000 1,000,000 1,000,000 1,000,000 1,000,000 sum) (K Sh( Cost of auditing (lump sum) (K Sh) 1,000,000 1,000,000 2,000,000 2,000,000 3,000,000 3,000,000 No of pastoralists eligble for livestock asset 11,592 33,636 24,204 69,636 36,815 105,636 protection coverage need to be carefully accessed before it is launched rate of 12.5 percent (average of rates assumed for on a semicommercial basis. Thus it remains to be the macro asset protection). determined whether for these additional options the • Number of vulnerable pastoralists to GoK may support the asset protection structure, the /// purchase top-up option: The assumed IBLI product, or both. However, for the purposes /// of this analysis, given that the values of selected take-up progression for the top-up option starts parameters have been defined on the basis of the IBLI with 1,000 policies in year 1 and reaches 10,000 experience, the simulations would apply in any case. policies after five years of implementation. The parameters presented in table 13 are based on the • Premium volume: The premium volume is /// /// following assumptions and considerations: obtained by multiplying the premium cost by the number of pastoralists purchasing the coverage. Top-Up Option ////// ////// • Projected public premium support: It is /// /// • Reference premium cost per pastoralist: assumed that the government of Kenya (GoK) will /// /// The reference premium cost per pastoralist has cover 50 percent of the cost of the coverage. been obtained by assuming the average standard conditions developed for the macro-level Expansion to Nontarget Pastoralists SECTION coverage. Hence, the K Sh 3,125 value derives /// /// from a sum insured per TLU of K Sh 5,000, five • Sum insured per TLU: In analogy with the top- 06 /// /// additional TLUs to be covered, and a premium up option, the value of a TLU is set at K Sh 5,000. 74 K E N YA : A G R I C U LT U R E I N S U R A N C E S O L U T I O N S A P P R A I S A L Table 13—  Fiscal costing projections for top-up and nontarget pastoralists options 2015 2016 2017 2018 2019 Top-Up Option Reference premium cost per pastoralist (K Sh) 3,125 3,125 3,125 3,125 3,125 No. of vulnerable pastoralists to purchase top-up option 3,000 3,250 5,500 7,750 10,000 Premium volume (K Sh) 3,125,000 10,156,250 17,187,500 24,218,750 31,250,000 Projected public premium support (%) 50% 50% 50% 50% 50% Cost of public support for top-up option (K Sh) 1,562,500 5,078,125 8,593,750 12,109,375 15,625,000 Expansion to non-target pastoralists Sum insured per Technical Livestock Unit (TLU) (K Sh) 5,000 5,000 5,000 5,000 5,000 Maximum no. of eligible TLUs per pastoralist (K Sh) 10 10 10 10 10 Values of additional sum insured per pastoralist (K Sh) 50,000 50,000 50,000 50,000 50,000 Premium Rate (as a share of sum insured) 12.5% 12.5% 12.5% 12.5% 12.5% Premium per pastoralist (K Sh) 6,250 6,250 6,250 6,250 6,250 No of non-target pastoralists to purchase coverage 1,000 2,000 3,000 4,000 5,000 Premium volume 6,250,000 12,500,000 18,750,000 25,000,000 31,250,000 Projected public premium support (%) 25% 25% 25% 25% 25% Projected public premium support (K Sh) 1,562,500 3,125,000 4,687,500 6,250,000 7,812,500 Costs for implementation as a share of premium support 200% 175% 150% 125% 100% (%) Costs for implementation (K Sh) 3,125,000 5,468,750 7,031,250 7,812,500 7,812,500 Cost of public support for non-target pastoralists (K Sh) 4,687,500 8,593,750 11,718,750 14,062,500 15,625,000 Total cost for GoK (million K Sh) 6.3 13.7 20.3 26.2 31.3 Total cost for GoK (million USD at 85 K Sh/USD) 0.1 0.2 0.2 0.3 0.4 • Maximum number of eligible TLUs per /// pastoralist is obtained by multiplying the number pastoralist: Pastoralists not belonging to the /// of TLUs to be covered by the selected value original target group will be able to purchase per TLU. supported coverage for a maximum of 10 TLUs. SECTION • Premium rate: Same as for top-up option. /// /// • Value of additional sum insured per • Premium per pastoralist: Same as for top-up 06 /// /// /// pastoralist: The reference sum insured per option. /// K E N YA : A G R I C U LT U R E I N S U R A N C E S O L U T I O N S A P P R A I S A L 75 • Number of nontarget pastoralists to /// Annex B.3. Summary of purchase coverage: The assumed take-up Modeling and Simulations /// progression for the nontarget group purchases starts with 1,000 policies and reaches 5,000 of Welfare Analysis policies after five years of implementation. • Premium volume: Same as for top-up option. /// /// for Livestock • Projected public premium support: It is /// /// A. A Dynamic Economic Model /// /// assumed that the GoK will cover 25 percent of the 1. Household consumption and /// cost of the coverage. livestock accumulation: /// • Projected public premium support: The Consider a dynamic model of a representative /// /// projected public premium support is obtained pastoral household, whose livelihood relies primarily by applying the share of premium that will be on livestock production. At the end of each season supported by the GoK to the estimated premium t=LRLD,SRSD where LRLD refers to long rain-long volume. dry season (March-September) and SRSD refers to • Costs for implementation as a share of /// short rain-short-dry season (October-February), premium support: Implementation costs /// this household earns and consumes from total refer to extension, marketing, capacity building, income from milk production m(Ht ) out of their own training, and infrastructure deployment. They are livestock Ht of which they can sell the milk at the estimated by referring to the IBLI experience and on-going market price ptmm. The income available for to the parameters that International Livestock consumption each period is thus ptmm(Ht ). Research Institute (ILRI) researchers have If milk production income is not enough for developed for future projections. consumption, household can also consume out of The ratio of implementation costs to premium their own herd by off-taking (sale or slaughter) some support cost is 5:2 in the short term, and nearly 1:1 of their livestock at the ongoing market price pth. in the medium term. These are the references that Household can also use left over milk production have been adopted for estimating these costs. income to invest more in its herd by buying livestock at the ongoing market price. • Costs for implementation: The actual costs /// /// for implementation are obtained by applying Household makes intertemporal decisions by the assumed percentage share to the projected choosing optimal consumption and herd investment premium support figures. each period to maximize their expected lifetime utility function, of which they draw welfare gain • Cost of public support for nontarget from consumption as well as livestock.44 Let /// pastoralists: The cost is the sum of the represent the rate at which household discounts /// projected premium support and the costs for future. Let ot represent the net livestock off-take (the implementation. number of herd sold and slaughtered netting out herd purchased) at the end of each season, we write household’s intertemporal decision as ∞ SECTION maxc E∑ βtu(ct, Ht ) subject to ct = ptmm(Ht )+ pthot t (t=0) H(t+1) = (1+bt+1-mt+1)(Ht -ot ) 06 76 K E N YA : A G R I C U LT U R E I N S U R A N C E S O L U T I O N S A P P R A I S A L At the end of each season, household herd—netting herd (e.g., selling off herd for consumption or out net herd off- take—would be accumulated toward slaughtering herd), then it will try to accumulate the next season herd. Herd can grow at natural livestock to maximize its herd size. biological birth rate bt+1 and is also subjected to mortality shock in that period mt+1 . The optimal herd off-take each season can thus be written as Droughts that could lead to catastrophic livestock mortality could thus affect household herd, which And the optimal herd accumulation dynamic is thus could have immediate effect on reducing current milk production income and longer-term effect on disrupting herd accumulation in the following Those with small herd will meet by off-taking out of periods. own herd at the rate faster than the net herd growth. 2. Poverty trap and economically viable herd /// Their herd thus tends to decline—instead of grow— in arid and semi-arid land (ASAL) region /// over time. The herd accumulation dynamic above could thus imply the existence of an economically With limited productive nonlivestock livelihood viable herd H* necessary to sustain seasonal herd options and the need for seasonal migration as growth each period: adaptation to climate variability, pastoral households in the ASAL region consume a good portion out of their own herd each season (e.g., through direct slaughtering or off-taking for cash). This necessary Furthermore, as poor households tend to be to credit consumption out of own herd each season tends to slow down and disrupt natural herd growth, constrained, they have difficulty restocking their especially for very small herds. Existing academic herds up to the economically viable and sustainable research (e.g., Lybbert et al. 2004; Barrett et al. 2006; levels. So while we should expect household with Chantarat et al 2014) has thus identified the existence Ht ≥H* to grow herd over time, those with Ht 1. Livestock prices are also uncertain. During droughts that could cause large livestock mortality, animals tend to be weak, and together with lower demand in the local market this could cause livestock price to drop. We thus describe a joint relationship among SECTION ptm , m t, m(NDVIt ) in a joint multivariate normal distribution with a correlation matrix capturing meaningful correlations of these three series. 06 78 K E N YA : A G R I C U LT U R E I N S U R A N C E S O L U T I O N S A P P R A I S A L While milk prices could also be uncertain, we observe other fixed administrative costs. Total premium per relatively stable prices across different seasons in insured TLU is each area. We thus assume that they are deterministic ρi=xE(πp) where i= r , p. at their mean level. 4. NDVI index based livestock insurance /// /// As household needs to increase herd off-take to pay for insurance premium when the cost is beyond the Using objectively measured NDVI data to trigger milk production income, we can write the optimal insurance payout, NDVI-based livestock insurance herd accumulation dynamics with asset protection for the Hunger Safety Net Program (HSNP) counties insurance insuring Hp unit of herd as could be of two forms: (i) An asset replacement insurance. This /// /// form aims to compensate insured household where (πp)Hp reflects the amount of insured herd that for livestock losses by making payout at the household could save using asset protection’s early end of each season if m(NDVIt ) is above a indemnity payout. Thus (πp) reflects the effectiveness predetermined strike level m*. Thus the seasonal of early intervention, made possible through early indemnity payout per insured TLU is indemnity payout πp in keeping the insured herd that π = max[m(NDVIt )-m*, 0] × p r survived from drought-related mortality. where p is a replacement cost per TLU. The If (πp)=m(NDVIt ), early intervention would be product was already designed by ILRI and has very effective in keeping all the insured herd been on sale in two of the four HSNP counties. that survived from drought-induced mortality. (ii) An asset protection insurance. This form ////// /// If (πp)=max[m(NDVIt )-m*,0], asset protection aims to provide timely cash to allow insured contract would thus make equivalent payout to the household to engage in actions (e.g., purchase comparable asset replacement contract. And if this forage supplement or water, or migrate to better effective early intervention can be achieved with forage/water sources) to save its livestock from comparable payout frequency and intensity, asset the slow-onset drought. It makes payout as protection insurance would be cheaper and so more early as possible at the end of every month in cost effective than the asset replacement counterpart. the coverage season when monthly NDVI falls Basis risk: Note that both forms of livestock insurance below a predetermined strike level NDVI*. The are written NDVI not actual mortality rate. Basis seasonal payout is thus the sum of the monthly risk – when indemnity payment deviates from or could payout not allow household to save their individual herd losses – would exist. The value to farmers will thus depend on the how closely individual herd mortality tracks that of where exit is the minimum level of NDVI that m(NDVIt ) especially for the case of asset replacement will allow insured household to receive 100% and so insurance will be valuable to pastoral household as payout each month and c represents the cost to r(mt, m(NDVIt ) ) -> 1. keep animal alive each month. 5. Public supports /// /// Actuarial fair premium per insured TLU for these SECTION contracts is equal to the expected indemnity payout. We assume that public support could result in Insurance company will however add some premium s% reduction in insurance premium rate (the free 06 multiple x > 1 to the commercial premium to cover provision of macro-level asset protection will have K E N YA : A G R I C U LT U R E I N S U R A N C E S O L U T I O N S A P P R A I S A L 79 s=100%) and will cover up to a prespecified herd size. averaged livestock prices were obtained from Total public cost per household i is thus 2005-2012 household survey data collected by Arid Land Resource Management Project S=sρHi (ALRMP) in all four counties. B. Calibrating Economic Model with /// 2. Milk price (K Sh/liter): ,ptm = 49 Actual Data /// Mean inflation adjusted milk price obtained from 1999-2012 ALRMP household survey data in all /// Livestock production /// four counties. 1. Sublocation and division seasonal livestock 3. Milk production (liter/season): m(Ht )= % milking mortality (%) and livestock price animal × averaged milk produced /TLU/ day × 180 × f(mt , m(NDVIt ),pth)~N(μm , μm(NDVI ) , σm , σm(NDVI ) ,σ p h ,r ) TLU =0.28 × 1 × 180 ×TLU t t t t t • Sublocation mortality mt: μ=0.11, σ=0.15 Parameters obtained from ILRI’s index-based livestock insurance impact evaluation household • NDVI-predicted division averaged livestock survey in Marsabit, 2009-2012. mortality m(NDVIt ): μ=0.11, σ=0.13 4. Natural herd growth rate (% per season): bt=0.2 • Division averaged TLU price (K Sh): μ=19,843, Obtained from ILRI’s index-based livestock σ=5,981 insurance impact evaluation household survey in • Common correlation matrices for the Marsabit, 2009-2012. three variables 5. Herd distribution (TLU): Ht obtained from HSNP Livestock impact evaluation household survey 2009-2012 in Sublocation Division Price all the four counties. Sublocation 1 Division 0.5 1 NDVI index based livestock insurance /// /// Livestock Price -0.4 -0.2 1 6. Index: Our analysis considered the impact of asset replacement that triggers monthly payout Our analysis was done on a representative based on monthly NDVI. Since the actual design pastoral household at the sublocation level with of monthly trigger is still in progress, we assume the assumption that perfect risk sharing exist at that this asset protection contract triggers payout this level. Sublocation average yields were thus based on ILRI’s predicted livestock mortality used to represent mortality of our representative index household. 7. Coverage level: When predicted livestock mortality index is above 15% similar to Long-term mean and standard deviation of the ILRI’s product NDVI-predicted division averaged livestock mortality were obtained from ILRI’s constructed 8. Sum insured (K Sh/TLU/season): c=4500 mortality indexes, which IRLI used to underwrite Drawing on discussion with some officials at the its asset replacement contract. The series were Ministry of Agriculture, Livestock and Fisheries constructed for each and every division in the and based on the recent droughts experience SECTION four HSNP counties from 1982-2013. Long-term in Wajir, Taita and Laikipia, we estimate that it mean and standard deviation of the subdistrict- would cost 25 K Shs per day to keep 1 TLU alive averaged seasonal mortality rates and division during drought. 06 80 K E N YA : A G R I C U LT U R E I N S U R A N C E S O L U T I O N S A P P R A I S A L 9. Pure premium rate = 9% per year means, standard deviations and correlation matrices obtained above. 10. Premium multiple (% of fair rate): x=200% This is a common rule of thumb in the industry 3. For each simulated year in each replicate, we estimated key outcome variables for four levels 11. Effectiveness of asset protection in reducing of starting herd sizes: 5 TLUs, 10 TLUs, 20 TLUs, livestock mortality: πp=max[m(NDVIt )-15%,0] and 40 TLUs. We assume that monthly insurance payouts could allow for effective early interventions, which 4. Finally, we calibrated our economic model using would enable the insured pastoralist to perfectly empirical data and estimated 100 replicates of avert all the predicted drought-related mortality 100-year series of key outcome variables of the beyond 15% of insured livestock. representative households in the scenarios with and without insurance and across government 12. Minimum subsistent consumption: c is supports. assumed at 30% of annual food poverty line of a representative farming household with 4.7 adult equivalent members (according to the HSNP household survey data) calculated at national food poverty line of rural regions at K Shs 988 per month per adult equivalent. 13. Government supports represented as premium reduction (%): s=100%, 50%, 25% C. Simulations /// /// We took the following steps to simulate key outcome indicators: 1. In order to describe the joint distributions of the seasonal sublocation-averaged livestock mortality rates, NDVI predicted division averaged livestock mortality rates and division average TLU prices, we first computed their long-term means, standard deviations and correlation matrices of the deviation of mortality rates from their location-specific long-term means. These statistics were calculated using variations over the 16 seasons from 2005 to 2012, when ALRMP and ILRI’s index overlap. 2. We then simulated, 100 replicates of 100 years series of these three levels of area yields assuming that their joint distribution follows 3-variable SECTION truncated multivariate normal distribution with 06 K E N YA : A G R I C U LT U R E I N S U R A N C E S O L U T I O N S A P P R A I S A L 81 Table 14—  Summary statistics of pastoral households in four HSNP counties Socioeconomics * Mean SD Mandera Marsabit Turkana Wajir Household member (adult equivalent) 4.7 1.8 4.7 4.5 4.5 5.275697 Monthly consumption expenditure/adult eq. 1746 789 2133 1363 1346 2202 Poverty headcount (2005 National poverty line) 47% 18% 72% 73% 20% % with seasonal food shortage 60% 33% 45% 78% 71% % receiving food aid 71% 69% 91% 51% 72% Main Source of Income Livestock production (rearing, sale of livestock/ 47% 41% 54% 41% 53% product) Casual Labor 17% 29% 18% 3% 19% Employment/Salary 2% 4% 3% 0% 1% Business and trade 6% 3% 4% 7% 8% Petty trade 18% 12% 5% 42% 13% Remittances and gifts 8% 9% 14% 4% 5% Statistics by income quartile Q1 Q2 Q3 Q4 % households who own livestock 89% 78% 89% 96% 93% % engage in livestock production 51% 61% 59% 54% 31% % share of livestock in total economic income 68% 63% 67% 74% 68% Mean number of livestock owned by household (TLU) 10.5 14.9 5.1 9.0 11.4 14.8 Livestock production * * Mean SD Mandera Marsabit Turkana Wajir Herd size 10.4 14.9 11.5 11.7 9.0 11.6 Herd composition % Cattle 23% 19% 33% 36% 14% 42% % Camel 18% 0% 44% 36% 30% 43% % Smallstock 15% 0% 44% 28% 56% 15% % Milking animal 28% 18% Livestock mortality and price statistics Sub-location TLU mortality (%) 0.15 0.18 0.18 0.09 0.14 0.12 S.D. 0.20 0.13 0.14 0.17 Division NDVI-predicted TLU mortality (%) 0.15 0.11 0.20 0.10 0.18 0.11 S.D. 0.09 0.14 0.08 0.11 Division averaged TLU price (KSh/TLU) 19,844 5,981 18,075 20,412 19129 24,448 Division averaged milk price (KSh/liter) 49 17 51 56 48 45 SECTION * From Hunger Safety Net Programme (HSNP) Impact Evaluation 2009-2012 panel household survey in 4 counties. * *From Arid Land Resource Management Project (ALRMP) Monthly Drought Monitoring Survey 1999-2012 in 4 counties 06 Milk and livestock prices are inflation adjusted using 2013 as base year 82 K E N YA : A G R I C U LT U R E I N S U R A N C E S O L U T I O N S A P P R A I S A L Both the maize and wheat series presented data Annex C.1. Assumptions reporting issues (e.g., confusions between metric and Parameters for Fiscal tons, kilograms, and bags), so data were revised and corrected when compiling mistakes were evident. Costing Scenarios for Crops The assumptions adopted in the analysis are the The fiscal analysis presented in section 3.3 is based following: on annual production data for maize and wheat at the district level provided by the State Department • The reference figures for cultivated area are of Agriculture. The administrative classification equivalent to the average of the latest five years refers to the 73 “pre-2012” districts (see table 15 for available. For yields, reference is made to the the list of districts). In such a data set the number average yield recorded in the period 2008–2012 of observations per district is quite heterogeneous, for maize, and in the period 2011–2012 for wheat and there are many gaps. For maize, complete (data for the 2008–2010 wheat campaigns were series ranging from 1983 to 2012 are available for mostly not available). only 50 percent of the districts; in the remaining 50 • Yield data have been detrended with respect to a percent of districts the series are shorter (as few trend reference composed of an average of linear, as six observations in some instances). However, exponential, and moving average trends. for 15 districts—accounting for over 50 percent • The price at which maize and wheat production of the maize cultivated area—the time series are have been valued is K Sh 34/kg for maize and K Sh acceptably long and start at the latest in the mid- 46/kg for wheat. 1990s. One significant limitation of the maize data set is that it is composed of annual yield values, • The coverage level was set at 80 percent. which make accounting for yield variability in the • Progressively increasing insurance take-up has biannual production areas impossible. While this been projected, with a rate of 3 percent assumed limitation would be more problematic in a potential at the beginning of the program in 2016, and rates implementation phase, from a fiscal analysis of 15 percent for maize and 25 percent for wheat perspective the data can still provide the basis for reached by 2023. initial rough operational estimates of the fiscal costs. • The number of farmers involved in the program The data for wheat also shows many gaps; has been estimated by dividing the projected unfortunately, data are missing for 2008 and 2009, cultivated area by the median farm size, which were critical years for wheat production respectively 1.5 ha for maize and 3.0 ha for wheat. (2009 in particular). The presence of these gaps in The reason for selecting the median, and not the recent and sensitive years has a significant impact average, is linked to the possible introduction of on the quality of the simulations. The wheat data caps in the number of hectares per farm insured set is smaller than the maize data set, as 95 percent under the supported program. In addition, for of cultivated area is concentrated in five districts maize, the biannual production pattern in 75 only (Meru Central, Laikipia, Narok, Nakuru, and percent of cultivated area has been accounted Uasin Gishu). Given that wheat production in other for by dividing the estimated number of farmers SECTION districts is sparse and of low quality, the analysis has by 1.6, also considering that in the biannual focused only on the five main production districts production regions cultivated area may be lower 07 (see table 16). in the less favorable season. K E N YA : A G R I C U LT U R E I N S U R A N C E S O L U T I O N S A P P R A I S A L 83 Table 15—  Yield, Area, and Premium Rate Data for Maize Yield (kg/ha) Cultivated Area (ha) Pure Risk Capped Insurance Province District Avg ‘08-’12 Avg ‘08-’12 Premium Rate x District Premium Rate x District 1 Central Thika N.A. N.A. N.A. N.A. 2 Central Kiambu East 1,442 4,778 6.6% 13.2% 3 Central Kiambu West N.A. N.A. N.A. N.A. 4 Central Kirinyaga 1,108 20,671 8.4% 15.0% 5 Central Murang’a North 697 19,508 10.8% 15.0% 6 Central Murang’a South 1,443 30,085 8.6% 15.0% 7 Central Nyandarua North 2,038 8,892 6.3% 12.6% 8 Central Nyandarua South 2,152 2,171 1.8% 3.5% 9 Central Nyeri South 741 13,332 7.5% 14.9% 10 Central Nyeri North N.A. N.A. N.A. N.A. 11 Coast Taita Taveta 880 16,599 12.8% 15.0% 12 Coast Kwale 1,265 45,120 5.5% 10.9% 13 Coast T/River 1,484 8,893 6.7% 13.5% 14 Coast Mombasa 777 1,264 8.5% 15.0% 15 Coast Lamu 1,890 18,065 6.1% 12.2% 16 Coast Malindi 970 15,853 4.8% 9.5% 17 Coast Kilifi 885 52,811 4.9% 9.9% 18 Eastern Embu 1,390 19,722 6.4% 12.8% 19 Eastern Isiolo 543 736 1.8% 3.5% 20 Eastern Kitui 700 43,463 14.9% 15.0% 21 Eastern Machakos 713 140,485 15.6% 15.0% 22 Eastern Makueni 640 95,984 12.0% 15.0% 23 Eastern Marsabit N.A. N.A. N.A. N.A. 24 Eastern Mbeere 709 26,326 3.1% 6.2% 25 Eastern Meru central 1,739 37,104 10.0% 15.0% 26 Eastern Meru North 1,519 59,291 4.7% 9.3% 27 Eastern Meru South 1,405 15,262 7.8% 15.0% 28 Eastern Moyale 299 456 1.8% 3.5% 29 Eastern Mwingi 514 37,160 15.1% 15.0% 30 Eastern Tharaka 1,135 12,033 10.7% 15.0% 31 North Eastern Ijara 128 145 1.8% 3.5% 32 North Eastern Garrisa 580 356 1.8% 3.5% 33 North Eastern Wajir 222 785 1.8% 3.5% 34 North Eastern Mandera 329 1,385 18.9% 15.0% 35 Nairobi Nairobi N.A. N.A. N.A. N.A. 84 K E N YA : A G R I C U LT U R E I N S U R A N C E S O L U T I O N S A P P R A I S A L Table 15—  Yield, Area, and Premium Rate Data for Maize (continued) Yield (kg/ha) Cultivated Area (ha) Pure Risk Capped Insurance Province District Avg ‘08-’12 Avg ‘08-’12 Premium Rate x District Premium Rate x District 36 Nyanza Bondo 1,134 19,921 13.5% 15.0% 37 Nyanza Gucha 2,240 17,894 4.4% 8.7% 38 Nyanza H/Bay 1,429 41,438 1.8% 3.5% 39 Nyanza Kisii 2,258 37,678 1.8% 3.5% 40 Nyanza Kisumu 1,394 18,140 1.9% 3.8% 41 Nyanza Kuria 2,277 13,533 2.9% 5.9% 42 Nyanza Migori 1,515 56,209 1.8% 3.5% 43 Nyanza Nyamira 2,019 61,632 1.8% 3.5% 44 Nyanza Nyando 1,502 9,243 1.8% 3.5% 45 Nyanza Rachuoyo 1,432 15,220 1.8% 3.5% 46 Nyanza Siaya 1,316 35,740 3.0% 6.0% 47 Nyanza Suba 1,263 7,569 2.2% 4.3% 48 Rift Valley Baringo 1,848 18,593 7.4% 14.7% 49 Rift Valley Bomet 1,938 34,234 8.4% 15.0% 50 Rift Valley Bureti 2,288 16,164 1.8% 3.5% 51 Rift Valley Kajiado 1,819 16,173 8.4% 15.0% 52 Rift Valley Keiyo Marakwet 1,696 39,364 9.4% 15.0% 53 Rift Valley Kericho 2,730 28,775 1.8% 3.5% 54 Rift Valley Koibatek 1,748 10,021 4.0% 8.1% 55 Rift Valley Laikipia 2,068 31,902 4.3% 8.6% 56 Rift Valley Marakwet 2,829 17,592 2.8% 5.5% 57 Rift Valley Nakuru 2,183 71,375 9.9% 15.0% 58 Rift Valley Nandi 2,768 77,603 1.8% 3.5% 59 Rift Valley Narok 1,849 38,884 12.3% 15.0% 60 Rift Valley Samburu 1,649 795 1.8% 3.5% 61 Rift Valley T/Mara 3,070 60,325 4.8% 9.5% 62 Rift Valley T/Nzoia 3,829 101,272 5.3% 10.6% 63 Rift Valley Turkana 1,319 1,631 3.7% 7.3% 64 Rift Valley U/Gishu 3,588 86,650 1.8% 3.5% 65 Rift Valley W/Pokot 2,068 23,416 4.7% 9.4% 66 Western Bungoma 2,529 79,872 1.8% 3.5% 67 Western Busia 1,219 30,416 1.8% 3.5% 68 Western Butere 1,899 4,366 1.8% 3.5% 69 Western Kakamenga 2,026 50,140 2.3% 4.5% 70 Western Lugari 2,921 19,359 1.8% 3.5% 71 Western Mt. Elgon 2,969 15,718 2.6% 5.2% 72 Western Teso 1,255 7,930 1.8% 3.5% 73 Western Vihiga 1,217 33,513 3.1% 6.3% K E N YA : A G R I C U LT U R E I N S U R A N C E S O L U T I O N S A P P R A I S A L 85 Table 16 —  Yield, Area, and Premium Rate Data for Wheat Yield (kg/ha) Avg last Cultivated Area (ha) Pure Premium Capped Insurance Province District available 5 years Avg ‘11-’12 Rate x District Premium Rate x District Eastern Meru central 2,264 16,078 12% 15% Rift Valley Laikipia 2,178 5,468 3% 5% Rift Valley Nakuru 2,850 26,111 6% 11% Rift Valley Narok 2,681 49,982 10% 15% Rift Valley U/Gishu 2,703 29,668 2% 3% As mentioned in section 3.3, the fiscal scenarios by the extra CCEs should be 100 percent. In order to include the provision of public support for additional account for equipment, management, and auditing data collection activities that should complement costs, an approximated overhead of 50 percent has the government of Kenya procedures for estimating been added to the cost of carrying out the CCEs. production and area data operated at county level. Although many different arrangements can be envisioned, including the possibility of outsourcing Annex C.2. Summary of some functions to the private sector, the present analysis assumes that the additional crop-cutting Modeling and Simulations of experiments (CCEs) required would be carried out Welfare Analysis for Crops by the public extension service.46 Hence, in terms of costing, the government of Kenya will cover expenses A. A Simple Economic Model for equipment, labor, management, and auditing. 1. Crop production /// /// The calculations that lead to the estimation of the supplementary data collection costs are presented Consider a one period model in key crop regions with many farmers. Each period, farmers’ crop production in table 17. For simplicity, reference is made to an yields yi kilograms per hectare of land and can enjoy area of 10km x 10km (10,000 hectares), for which a total income of yi pi K Shs per hectare of cultivated hypothetical number of 10 additional CCEs would be land, where yi is the crop price per kilogram. foreseen. It is estimated that a team of two people can carry out four CCEs per day, and that a man- At the beginning of each season, farmers are credit day salary for such an activity could be set at K Sh constrained and so needs to take out loan L K Shs to 2,500. The cost of the supplementary CCE activity purchase inputs (e.g., seeds and fertilizer). Farmers is obviously a function of the area to be covered. In then pay back the loan at the end of the harvest with the beginning the CCEs can be carried out in the crop income. areas where the AYII programs are piloted; but if the 2. Risk programs are to expand significantly, all the areas /// /// should be surveyed and yield databases developed Both crop price and yield are uncertain. Crop price pi SECTION for them. This is why, despite the fact that the is assumed to follow a uniform distribution, projected penetration of AYII in 2022 is 15 percent U(pi, ph). Crop production also faces various kinds of for maize and 25 percent for wheat, the area covered risk including both farm-specific risk (e.g., disease 07 86 K E N YA : A G R I C U LT U R E I N S U R A N C E S O L U T I O N S A P P R A I S A L or illness of farm labor) and covariate risk (e.g., ry ,y ̅ -> 1. On the other hand, when farm-specific i droughts and floods that tend to affect all farmers in shocks dominate the covariate ones, ry ,y ̅ will deviate i the area). With the presence of common covariate largely from one. shocks, we should thus expect individual crop yields 3. Area yield index insurance (AYII) to track average yields in their area to some extent. /// /// The contract is designed to protect farmers from In order to understand this empirical relationship, we covariate shocks that could affect all farmers in describe joint distribution of individual yield yi and _ the area and that are not effectively managed by the average yield across all farmers in the area y with existing ‘mutual risk sharing mechanisms’ within the a bivariate normal distribution as community. Specifically, AYII compensates insured _ f(yi ,y)~N(μy , μy ̅ , σy , σ y ̅ , ry ,y ̅ ) farmer at an expected crop price p per kilogram i i i _ when area averaged yield y falls below a prespecified where μy ,μy represent ̅ long-term average levels of coverage level y*. Indemnity payout per insured i individual and area-averaged crop yields, σy , σ y hectare can thus be written as i ̅ describe long-term standard deviations of the two _ yield series and ry ,y ̅ represents correlations of the π=max[0,y*- y ]×p i two series observed in the empirical data. In the areas where the coverage level is set as some percentage of with large exposure to common covariate shocks, we the expected area yield, i.e. y* = coverage × μy_ . should expect individual yields to move together with the area-averaged yield and thus Actuarial fair premium per insured hectare for this contract is equal to the expected indemnity payout. Insurance company will however add some premium Table 17—  Estimation of potential cost of multiple x > 1 to the commercial premium to cover additional data collection activities for AYII other fixed, administrative costs. Total premium per insured hectare can be written as No of hectares in an area of 10km x 10km 10,000 Number of CCs per 10km x 10km area 10 ρ=xE(π) . Number of people on a CC team 2 With AYII offering protection of income shortfall Number of CCs carried out in a day by a CC team 4 from area yield variability, farmer’s insured crop Number of man-days needed to cover each 10,000 ha 5 income per hectare can thus be yi pi+ π-ρ . Man-day cost in Ksh 2500 Basis risk: Note that insurance is written on area Cost of labor in KSh for each 10,000 ha 12500 yield, not individual yield. While this resolves Maize: Reference total cultivated area 127,306 Wheat: Reference total cultivated area 127,306 2016 2017 2018 2019 2020 2021 2022 2023 Share of area covered by additional CCs 5% 15% 30% 45% 60% 75% 90% 100% Overhead for equipment, management, auditing, 50% 50% 50% 50% 50% 50% 50% 50% etc. Maize: Additional costs of yield data collection 0.192 0.577 1.153 1.730 2.307 2.884 3.460 3.845 (million KSh) Wheat: Additional costs of yield data collection 0.012 0.036 0.072 0.107 0.143 0.179 0.215 0.239 (million KSh) K E N YA : A G R I C U LT U R E I N S U R A N C E S O L U T I O N S A P P R A I S A L 87 asymmetric information and reduces transaction reduce poverty rate by 1% (based on national food cost, it also could limit the value of insurance to poverty line, 2005) relative to the baseline without individual farmers because of basis risk, which occurs the program in these production zones. Direct cash when indemnity payment deviates from individual transfer program to the poor was further used as losses. The value to farmers will thus depend on counterfactual program for cost-benefit analysis. The the how closely individual yields track that of area K Sh cost per farming household per year that can average. AYII will be valuable to farmer as ry ,y -> 1. reduce poverty rate by 1% was computed as (poverty i gap x poverty line)/poverty rate. 4. Loan repayment /// /// 6. Values of AYII /// /// Input credit is obtained at the interest rate r._ If • Value to farmers AYII reduces vulnerability by farmers always pay back their loans using crop /// /// providing buffer against sharp drop of net crop income as much as possible, then net income income available for consumption in the event of available for consumption for farmer i _who severe shocks cultivates a median farm size Ai hectares of maize will be • Value to lenders: Based on our assumption that /// /// farmers will try to pay back loan after meeting Ci=(yi pi + π-ρ-(1+r)L)×Ai required consumption, AYII thus will increase Loan default is however possible and can be partial loan repayment rate on average. To make this or total. While full repayment is an option, we more assumption more realistic, lenders can make realistically assume that farmer will try to payback insurance a prerequisite for obtaining loan and/or their loan as much as they can after meeting their link insurance with loan directly. With increasing subsistent consumption c (set at 30% of food poverty loan repayment, lenders could eventually be line).47 willing to extend more credit to farmers. • Potential crowding in value of AYII through Farmer ‘s loan repayment will be /// credit market: In the medium term, insurance /// LRi=max [(1+r)L,yi pi+ π-ρ-c ]×Ai could enhance agricultural productivity by promoting smallholder farmers’ adoption of 5. Public supports productive inputs (e.g., new technology, hybrid /// /// We assume that public support could result in s% seeds). This could be true when AYII relaxes reduction in insurance premium rate and will cover demand-side constraint (i.e., enhancing farmer’s the whole cultivated farm of representative farmer. investment incentives and credit demand when Total public cost per farmer i is thus agricultural production is derisked) as well as supply-side constraint (i.e., allowing lenders to S=sρAi increase credit supply to farmers). Farmer i’s net income available for consumption when AYII Cost-benefit analysis of public support to agricultural unlocks access to credit allowing him to afford insurance program assumed that in the very first more expensive but productive input with yield years, development of insurance program would be markup αy>1 per hectare and higher cost (and possible only with public support and that one of the larger loan size) relative to the current required SECTION program’s key policy objectives was to reduce poverty level with mark up of αL>1: among smallholder farmers. We then computed K Sh cost per farming household per year that could Cih=(αy (yi pi + π-ρ)-αL (1+r)L)×Ai 07 88 K E N YA : A G R I C U LT U R E I N S U R A N C E S O L U T I O N S A P P R A I S A L B. Calibrating Economic Model with Country-level average maize producer prices Actual Data (1991-2011) obtained from FAOSTAT and 2012- 2013 from Regional Agricultural Trade Intelligence Crop production Network (RATIN). Prices were inflation adjusted /// /// 1. Sublocation, division and district yield (kg/ha): with 2013 as base year. _ f(yi,y f(yi ,y)~N(μy , μy ̅ , σy , σ y ̅ , ry ,y ̅ ) i i i 3. Working capital loan (% of expected revenue): with μsubl = μdiv = μdist = μ and σsubl = σdiv = σdist = σ _ Li=60%μy_ P • Low-potential maize zone: μ=703, σ=347 From gross margin studies (KARI 2009, etc.), • Medium-potential maize zone: μ=1426, σ=414 total input costs range from 50% to 75% of • Low potential maize zone: μ=2892, σ=991 average crop revenue. This figure was also similar to total working capital loan reported in Tegemeo • Wheat region: μ=2505, σ=881 household survey. The median level is 60%. • Common correlation matrixes for all zones 4. Yield and cost markup rates with respect to Sublocation Division District high-cost, more productive input invested (% Sublocation 1 of expected yield and cost): αy , αL vary across Division 0.85 1 crops and production zones. They were estimated District 0.75 -0.81 1 from the ratio of yields and costs of high versus low input crop productions of three to five Our analysis was done on a representative farmer representative small-scaled farmers with less at the sublocation level with the assumption than 4 hectares of land in some key growing that perfect income risk sharing exists at the provinces in each zone. Maize data were derived sublocation level. Sublocation average yields from KARI (2009)’s Assessment of Costs of Maize were thus used to represent yields of our Production, Marketing and Processing in Kenya: representative farmer. A Maize Grain-Maize Meal Value Chain Analysis. Mean and standard deviations obtained from Wheat data were obtained from DASS’s (2010) detrended annual district yields are from the gross margin analysis. Ministry of Agriculture, Livestock and Fisheries from 1983 to 2013.48 Correlation metric is from Crop zone Studied county Markup (% of average) 2-year Tegemeo panel household survey in 2000 Production Yield cost and 2004. Survey covers 15-60 representative Low- households in representative sub-locations, potential Machakos, Eastern 233% 308% maize locations, divisions and districts in all production Med- Kirinyaga, zones. Sample size varies by relative populations. potential Central 182% 138% maize Districts selected for analysis in all zones are those Laguri, Western 200% 123% High- with large maize and wheat-growing areas with potential maize Narok, Rift 193% 129% available data in both Tegemeo’s household survey Valley and district-level yield data.49 Mean 196% 126% Narok, Rift 139% 132% SECTION 2. Producer price (K Sh/kg): pi~U(pl , ph ) Wheat Valley Nakuru, Rift 139% 133% • Maize: pl =22.7, pl =45.5 Valley 07 • Wheat: pl =34.9, pl =58.1 Mean 139% 133% K E N YA : A G R I C U LT U R E I N S U R A N C E S O L U T I O N S A P P R A I S A L 89 C. Simulations /// /// 5. Farm size (hectare): We took the following steps to simulate key outcome Ailow = Aimedium = 1.5 , Aihigh = 2.5 , Aiwheat = 3 indicators and zone-specific longitudinal series of representative and area yields and prices from their This was obtained from Tegemeo joint distribution: household survey. 1. Using the two-year Tegemeo household data, we Area yield index insurance (AYII) /// /// constructed two years of annual area-average _ 6. Indexes: y, we constructed both division and yields at sublocation, division and district level by district average yields, as the goal was also to averaging individual yields across households in evaluate the AYII with these two different indexes. each area in each year. 7. Premium multiple (% of fair rate): x =200%. 2. In order to describe the zone-specific joint This is a common rule of thumb in the industry. distributions of the three levels of yields, we computed zone-specific means, standard 8. Coverage level: deviations and correlation matrices of sub- location, division and district yields. These High coverage Low coverage statistics were calculated using variations over Zone (15% maximum (10% maximum rate) rate) the two years and across respective area yields Coverage Fair Coverage Fair within each zone. 50 premium premium 3. The relatively short temporal coverage Low- potential maize 50% 7.30% 30% 4.50% of household data could have resulted in underestimation of temporal variations of these Medium- potential maize 85% 6.60% 75% 4.20% series. We thus complemented the data with High- longitudinal detrended district-level yield data potential 80% 7.40% 65% 4.20% maize and computed zone-specific moments. While Wheat 75% 6.60% 65% 4.50% means of these three levels of area yields were comparable within each zone, standard deviations were a lot smaller in the two-year data. Means and 9. Interest rate on working capital loan (% per year): standard deviations of these sublocation, division, r=17% and district yield series in each zone were then Weighted average commercial bank lending assumed to be similar to that estimated from the rate as of April 2014 obtained from Financial 1983–2013 district yield series. Sector Deepening. 4. For each production zone, we simulated 100 _ 10. Minimum subsistent consumption: c is replicates of 100-year series of these three assumed at 30% of annual food poverty line of a levels of area yields assuming that their joint representative farming household with five adult distribution followed 3-variable multivariate equivalent members (statistics from Tegemeo normal distribution, with zone-specific means survey) calculated at national food poverty line and standard deviations obtained from 1983–2013 of rural regions at K Shs 988 per month per district yield data and correlation matrixes adult equivalent. obtained from the variations within the two-year SECTION household data. 11. Public supports represented as premium reduction (%): s=50% 07 90 K E N YA : A G R I C U LT U R E I N S U R A N C E S O L U T I O N S A P P R A I S A L Table 18—  Summary statistics of maize and wheat growing households Mean Median SD Min Max Household member (adult equivalent) 5.2 5.1 1.9 0.6 13.2 Poverty headcount (national rural poverty line (2005)) 41.5% Yield and price statistics* Maize in low potential zone (kg/ha) 703 671 347 18 2,085 Maize in medium potential zone (kg/ha) 1,426 1,342 414 159 3,165 Maize in high potential zone (kg/ha) 2,892 2,790 991 397 5,325 Wheat (kg/ha) 2,505 2,617 881 26 4,581 Aggregated maize price (Ksh/kg) 34 34 7 23 45 Aggregated wheat price (Ksh/kg) 46 45 8 35 58 Maize producing households Cultivated land size (ha) 2.5 1.5 4.5 0.1 110.0 Low potential zone 2.2 1.5 2.9 0.1 21.0 Medium potential zone 1.7 1.5 1.6 0.1 11.5 High potential zone 2.9 2.5 6.0 0.1 110.0 % households who own land 52% 67% 48% 0% 100% % with two croping seasons a year 42% 39% 38% 0% 100% % use purchased hybrid seed 23% 19% 33% 0% 100% % households with maize sale 18% 0% 27% 0% 100% Low potential zone 8% 0% 18% 0% 100% Medium potential zone 7% 0% 17% 0% 100% High potential zone 28% 20% 31% 0% 100% % maize income from total econ income 29% 23% 20% 3% 100% Low potential zone 70% 70% 10% 50% 100% Medium potential zone 28% 31% 20% 3% 100% High potential zone 28% 23% 20% 4% 100% 5. For each simulated year in each replicate, we 6. Finally, we calibrated our economic model using also randomly drew one price realization from empirical data and estimated 100 replicates of a uniform distribution specified with 10-year 100-year series of key outcome variables for minimum and maximum national aggregate, the representative farmer in each zone in the inflation-adjusted price observed empirically from scenarios with and without AYII and across 1991 to 2013. contract variations. SECTION 07 K E N YA : A G R I C U LT U R E I N S U R A N C E S O L U T I O N S A P P R A I S A L 91 (continued) Wheat producing households Cultivated land size (ha) 7.6 3.0 23.6 0.0 240.0 % households who own land 60% 100% 48% 0% 100% % with two croping seasons a year 0% 0% 0% 0% 0% % use purchased hybrid seed 26% % households with wheat sale 78% 90% 30% 0% 100% % wheat income from total econ income 29% 23% 20% 0% 86% Credit access* % households with input credit 46% 39% 14% 0% 100% Purpose of credit Fertilizer 81% 79% 21% 0% 100% Seed 9% 9% 41% 0% 100% Other equipments 10% 3% 43% 0% 100% Credit source AFC 1% 1% 6% 0% 14% Commecial banks 1% 1% 3% 0% 8% Cooperatives/Saccos 25% 39% 21% 0% 100% Local trader/companies 10% 10% 14% 0% 100% NGOs/MFIs 1% 1% 5% 0% 11% Money lenders 2% 1% 32% 0% 100% Friend/relatives, ROSCAs, etc. 6% 6% 26% 0% 100% Lending rates Commercial banks (5-yr statistics) 16% 15% 2% 14% 20% * From 30 years detrended district-level yield data from 1983-2012 obtained from the Ministry of Agriculture Inflation adjusted aggregated producer pricesFAOSTAT and Regional Agricultural Trade Intellegence Network ** Household data from 2000, 2004 household survey of Tegemeo Agricultural Monitoring and Policy Analysis Project National rural poverty line is 1,562 KSh/capita/month *** Data from Kenya Integrated Household Expenditure Survey 2005 ****Monthly FSD data on commercial bank’s weighted average lending rates from 2005-2014 SECTION 07 92 K E N YA : A G R I C U LT U R E I N S U R A N C E S O L U T I O N S A P P R A I S A L Table 19—  Summary of key impact indicators by contract variations District-level yield index High coverage Low coverage Commercial AYII 50% subsidized Commercial AYII 50% subsidized Commercial 50% subsidized Commercial 50% Impact indicators No insurance w/ increased w/ increased w/ increased w/ increased AYII AYII AYII subsidized AYII investment investment investment investment Maize - Low potential zone Coverage = 50%, fair rate = 7.3% Coverage = 30%, fair rate = 4.5% Net income available for 10,721 9,408 10,721 10,234 10,721 consumption per year (Ksh). - - - - (18,940) (17,713) (17,713) (18,383) (18,383) Std.Dev in parenthesis Probability of falling into 100% 100% 100% - - 100% 100% - - poverty* Loan repayment (%) after min. 59% 56% 59% - - 58% 59% - - consumption Maize - Medium potential zone Coverage = 85%, fair rate = 6.6% Coverage = 75%, fair rate = 4.2% Net income available for 21,723 17,647 21,723 54,342 61,753 19,447 21,723 57,615 61,753 consumption per year (Ksh). (25,840) (23,250) (23,250) (41,500) (41,500) (24,075) (24,075) (43,210) (43210) Std.Dev in parenthesis Probability of falling into 90% 95% 90% 60% 55% 90% 90% 60% 55% poverty Loan repayment (%) after min. 84% 82% 86% 95% 97% 83% 85% 96% 97% consumption Maize - High potential zone Coverage = 80%, fair rate = 7.4% Coverage = 65%, fair rate = 4.2% Net income available for 73,582 58,965 73,582 237,515 266,227 66,847 73,582 252,999 266,227 consumption per year (Ksh). (99,641) (85,534) (85,534) (168,014) (168,014) (92,757) (92,757) (180,753) (180,753) Std.Dev in parenthesis Probability of falling into 45% 50% 45% 15% 10% 50% 50% 15% 10% poverty Loan repayment (%) after min. 91% 90% 93% 98% 99% 91% 92% 98% 98% consumption Wheat Coverage = 75%, fair rate = 6.6% Coverage = 65%, fair rate = 4.5% Net income available for 104,466 87,274 104,466 169,047 194,750 94,179 104,466 179,372 194,750 consumption per year (Ksh). (136,714) (117,954) (117,954) (176,345) (176,345) (125,404) (125,404) (185,782) (185,782) Std.Dev in parenthesis Probability of falling into 35% 40% 35% 30% 25% 40% 35% 30% 25% poverty (%) Loan repayment (%) after min. 92% 92% 94% 95% 96% 92% 94% 95% 96% consumption K E N YA : A G R I C U LT U R E I N S U R A N C E S O L U T I O N S A P P R A I S A L 93 (continued) Division-level yield index (with reduced basis risk) High coverage Low coverage Commercial AYII 50% subsidized Commercial AYII 50% subsidized Commercial 50% subsidized Commercial 50% subsidized Impact indicators No insurance w/ increased w/ increased w/ increased w/ increased AYII AYII AYII AYII investment investment investment investment Maize - Low potential zone Coverage = 50%, fair rate = 7.3% Coverage = 30%, fair rate = 4.5% Net income available for 10,721 9,402 10,717 10,227 10,717 consumption per year (Ksh). - - - - (18,940) (16,483) (16,483) (17,289) (17,289) Std.Dev in parenthesis Probability of falling into 100% 100% 100% - - 100% 100% - - poverty* Loan repayment (%) after min. 59% 55% 58% - - 57% 58% - - consumption Maize - Medium potential zone Coverage = 85%, fair rate = 6.6% Coverage = 75%, fair rate = 4.2% Net income available for 21,723 17,652 21,733 54,346 61,766 19,450 21,733 57,615 61,766 consumption per year (Ksh). (25,840) (21,765) (21,765) (39,530) (39,530) (23,075) (23,075) (41,151) (41,151) Std.Dev in parenthesis Probability of falling into 90% 95% 90% 60% 55% 95% 90% 60% 55% poverty Loan repayment (%) after min. 84% 82% 86% 95% 97% 83% 85% 96% 97% consumption Maize - High potential zone Coverage = 80%, fair rate = 7.4% Coverage = 65%, fair rate = 4.2% Net income available for 73,582 58,952 73,579 237,492 266,224 66,837 73,579 252,980 266,224 consumption per year (Ksh). (99,641) (84,314) (84,314) (154,653) (154,653) (90,487) (90,487) (178,876) (178,876) Std.Dev in parenthesis Probability of falling into 45% 50% 45% 10% 5% 50% 50% 15% 10% poverty Loan repayment (%) after min. 91% 91% 94% 98% 99% 92% 93% 99% 99% consumption Wheat Coverage = 75%, fair rate = 6.6% Coverage = 65%, fair rate = 4.5% Net income available for 104,466 87,271 104,462 169,044 194,745 94,185 104,462 179,380 194,745 consumption per year (Ksh). (136,714) (111,404) (111,404) (173,571) (173,571) (123,509) (123,509) (183,904) (183,904) Std.Dev in parenthesis Probability of falling into 35% 40% 35% 30% 25% 40% 35% 30% 25% poverty (%) Loan repayment (%) after min. 92% 93% 95% 96% 97% 93% 94% 96% 96% consumption * At national food poverty line (2005) at KSh 988 per month per adult equivalent. For a representative household of 5 equivalent adults, food poverty line is calculated at 988*12*5 = Ksh 59,280 per year. 94 K E N YA : A G R I C U LT U R E I N S U R A N C E S O L U T I O N S A P P R A I S A L Bibliography Barrett, C. B., M. Carter, M. Ikegami. 2012. “Poverty Traps and Social ———. 2014a. __Kenya: Situation Analysis for a National Agricul- Protection.” Working paper, Cornell University, University of tural Insurance Policy (NAIP)__. Nairobi: Government of Kenya, California–Davis, and International Livestock Research Institute. Ministry of Agriculture, Livestock and Fisheries. Barrett, C. B., P. P. Marenya, J. McPeak, B. Minten, F. Murithi, W. ———. 2014b. __Kenya Agricultural Data Collection and Man- Oluoch-Kosura, F. Place, J. C. Randrianarisoa, J. Rasambainari- agement Guideline__. Nairobi: Government of Kenya, State vo, and J. Wangila. 2006. “Welfare Dynamics in Rural Kenya and Department of Agriculture. Madagascar.” __Journal of Development Studies__ 42 (2): 248–77. ILRI (International Livestock Research Institute). 2013. __Index Cai, H., Y. Chen, H. Fang, and L. Zhou. 2009. “Microinsurance, Based Livestock Insurance (IBLI) in Northern Kenya: Project Trust and Economic Development: Evidence from a Randomized Document__. January 2013 revision. Nairobi: ILRI. Natural Field Experiment.” NBER Working Papers 15396, National Bureau of Economic Research, Cambridge, MA. KARI (Kenya Agricultural Research Institute). 2009. “Assessment of Costs of Maize Production, Marketing and Processing in Kenya: Central Bank of Kenya. 2012. __Annual Report__. Nairobi: Central A Maize Grain-Maize Meal Value Chain Analysis.” KARI, Nairobi. Bank of Kenya. Kerer, J. 2013. __Background Paper on the Situation of Agricultural Chantarat, S., A. G. Mude, C. B. Barrett, and C. G. Turvey. 2014. Insurance with Reference to International Best Practices__. “Welfare Impacts of Index Insurance in the Presence of a Poverty Nairobi: Adaptation to Climate Change and Insurance (ACCI), Trap.” Working paper, Cornell University and Australian National German Agency for International Cooperation (GIZ), and Minis- University. try of Agriculture, Livestock and Fisheries (MALF). Clarke, D., and R. Vargas-Hill. 2013. “Cost-Benefit Analysis of the Lybbert, T., and C. B. Barrett, S. Desta, and D. L. Coppock. 2004. African Risk Capacity Facility.” IFPRI Discussion Paper 01292, “Stochastic Wealth Dynamics and Risk Management among a International Food Policy Research Institute, Washington, DC. Poor Population.” __Economic Journal__ 114, (498): 750–77. FSD (Financial Sector Deepening). 2013. __Review of FSD’s In- Mahul, O., and J. Skees. 2007. “Managing Agricultural Risk at the dex-Based Weather Insurance Initiatives__. Nairobi: FSD. Country Level: The Case of Index-Based Livestock Insurance in Mongolia.” Policy Research Working Paper 4325, World Bank, Galarza, F., and M. Carter. 2011. “Risk Preferences and Demand for Washington, DC. Insurance in Peru: A Field Experiment.” Working Paper 11-08, Departamento de Economía, Universidad del Pacífico, Lima; Mahul, O., and C. Stutley. 2010. __Government Support to Agricul- revised January 2011. tural Insurance__. Washington, DC: World Bank. GoK (Government of Kenya) 2006. Kenya Integrated Household MDP (Ministry of Devolution and Planning). 2013 Second Medium Budget Survey 2006/06. Ministry of Planning and National Devel- Term Plan, 2013-17. Government of Kenya, Ministry of Devolu- opment. Nairobi. tion and Planning, Nairobi.) SECTION ———. 2012. __Kenya Post Disaster Needs Assessment, Drought Smith, A., H. Smit, and D. Chamberlain. 2011. __Beyond Sales: New 08 2008–2011__. Nairobi: Government of Kenya. Frontiers in Microinsurance Distribution: Lessons for the Next Wave of Microinsurance Distribution Innovation__. Geneva: K E N YA : A G R I C U LT U R E I N S U R A N C E S O L U T I O N S A P P R A I S A L 95 International Labour Organization. Producers in South West Buenos Aires Province: Feasibility Study, Final Report__. World Bank, Washington, DC. Sinah, J. 2012. __Index-Based Weather Insurance: International and Kenyan Experiences__. Nairobi: Adaptation to Climate Change ———. 2013a. __Mexico: Agriculture Insurance Market Review__. and Insurance (ACCI), German Agency for International Cooper- World Bank, Washington, DC. ation (GIZ), and Ministry of Agriculture, Livestock and Fisheries (MALF). ———. 2013b. __Uruguay: NDVI Pasture Index-based Insurance for Livestock Producers in Uruguay—Feasibility Study, Final Woodard, J. 2012. “A Spatial Econometric Analysis of Loss Experi- Report__. World Bank, Washington, DC. ence in the U.S. Crop Insurance Program.” __Journal of Risk and Insurance__ 79 (1): 261–85. ———. 2014. __Financial Protection against Natural Disasters__. Global Facility for Disaster Reduction and Recovery (GFDRR). World Bank 2011a. “Burkina Faso—Risk Management in the Cotton World Bank, Washington, DC. Sector—Index Insurance Feasibility Study—Draft Report.” Agri- cultural Risk Management Team, World Bank, Washington, DC. ———. 2011b. __Enhancing Crop Insurance in India__. Washington, DC: World Bank. ———. 2011c. “India Crop Insurance Non-Lending Technical Assis- tance: Summary of Policy Suggestions.” World Bank, Washington, DC. ———. 2012. __NDVI Pasture Index-based Insurance for Livestock SECTION 96 K E N YA : A G R I C U LT U R E I N S U R A N C E S O L U T I O N S A P P R A I S A L Endnotes 1 World Development Indicators, World Bank, Washington, DC 14 NDVI is a very good indicator of pasture growth, grazing quality, (accessed November 3, 2014), http://data.worldbank.org/data-catalog/ and the impact of drought on pasture degradation over time. world-development-indicators. 15 The main differences between a micro-level individual pastoralist 2 See Mahul and Stutley (2010) for a comprehensive review of pasture/grazing-drought NDVI index insurance program and a government support to agricultural insurance. macro-level program are these: (i) under the micro-level program, individual pastoralists purchase their own policy and are the 3 See in particular Kerer (2013). insured for their declared number of animals (TLUs), while under the macro-level program the insured is government (or another 4 Support for index insurance is included in a proposed bill for a new appointed entity), which purchases a single policy on behalf of Insurance Act. a defined target audience of pastoralist households (termed the beneficiaries); (ii) under a macro-level policy, premium payments 5 Nonrivalrous goods are those that may be consumed by many at the are usually fully covered by the insured (government), and the same time at no additional cost (e.g., national defense or a piece of beneficiaries do not contribute at all toward the costs of insurance scientific knowledge). premiums; and (iii) under the macro-level policy, the beneficiaries have no legal rights to make any claim against the policy, as they are 6 This report, issued by the GoK, was based on data supplied by not deemed to be insured. MALF and was developed with technical assistance from the German Agency for International Cooperation (Deutsche Gesellschaft für 16 See ARC, “First Risk Pool,” http://www.africanriskcapacity.org/ Internationale Zusammenarbeit, or GIZ). countries/risk-pool-1. 7 The fully rated price includes the full price of the risk and an 17 This initiative is being supported by the Rockefeller Foundation, administrative loading to cover the ongoing costs of the insurers, the UK Department for International Development, the Global although not the development costs. Facility for Disaster Reduction and Recovery, and the International Fund for Agricultural Development. 8 Reinsurance companies add potentially large “data uncertainty” increases to insurance premiums if they have concerns about the 18 According to Clarke and Hill (2013), compared with an emergency data quality, thus significantly increasing the cost for farmers. assistance baseline in which cash or food is provided seven to nine months after harvest, an early payout when combined with 9 A good example is the Kilimo Salama scheme in Kenya, which is improved contingency planning will lead to substantial speed, cost, supported by the Global Index Insurance Facility and which uses and targeting gains. Speed benefits could be as large as a nine- mobile phones as a point-of-sale device Other distribution channels month improvement. include cash-based retailers, utility companies, or third-party bill 19 Under its old constitution, Kenya comprised eight provinces, payment providers. For a discussion of innovative distribution each headed by a provincial commissioner. The provinces (mikoa channels, see Smith, Smit, and Chamberlain 2011. in Swahili) were subdivided into districts (wilaya). There were 69 10 See the discussion above. districts at the 1999 census. Districts were then subdivided into 497 divisions (taarafa). The divisions were further subdivided into 2,427 11 Possible options for coinsurance pools are set out in annex A. locations (kata) and 6,612 sublocations (kata ndogo). Under the Constitution of 2010, the districts became counties (there are now 12 Countries that have TSUs include Brazil, Chile, France, Ghana, 46) and the divisions became subcounties (there are 290). SECTION Italy, Mexico, Poland, the Russian Federation and Spain. 20 A full description of the assumptions and the parameters adopted 08 13 An adult cow or 10 goats are equal to 1 TLU; a camel is equal to in the fiscal costing scenarios for the livestock insurance options is 1.43 TLUs. presented in annex B.2. K E N YA : A G R I C U LT U R E I N S U R A N C E S O L U T I O N S A P P R A I S A L 97 21 The ultra-poverty rate is based on a national rural poverty line level is de facto reduced to lower levels (see box 3 for a more equivalent to US$0.5/day. detailed discussion). 22 These results are parallel to findings in Chantarat et al. (2014) and 33 See section 3.4 for more detailed argumentation on the need for Barrett et al. (2012). public support. 23 For further discussion of Kenya’s experience with agricultural 34 A detailed description of how these costs have been estimated is insurance, see GoK (2014a), which is the source of the information presented in annex C.1. presented in this section. 35 A higher take-up rate has been assumed for wheat since farming 24 GoK (2014a) details the different schemes and product approaches units are generally larger than for maize, and the value chain is and thoroughly explains why these faltered or were inappropriate. generally more integrated with the financial environment. Some of the key challenges were the lack of entrepreneurship of the insurance companies involved, and potentially the costs of 36 Data underlying GoK 2006 the products. 37 http://www.kalro.org/ 25 For more details on the National Agriculture Insurance Scheme of India and its modifications see, World Bank (2011b). 38 http://www.tegemeo.org/ 26 The material presented in this section has been adapted from 39 http://www.tegemeo.org/ World Bank (2011a). 40 The CV, or coefficient of variation, is the standard deviation 27 Along these lines, some of the suggested topics to be covered divided by the mean and expressed as a ratio or percentage variation in the revision analysis of India’s National Agriculture Insurance around mean. Scheme were the following: (i) establishment of a standardized national manual on CCEs; (ii) systematic training and certification 41 Since maize and wheat households would potentially earn of loss adjusters; (iii) commission of randomized, independent, high- income from other sources of livelihood, poverty measures based quality CCE audits to be conducted alongside the standard CCEs; on household crop income relative to either the national food (iv) standardized statistical approach to handle outlier yields in the poverty line (K Sh 988 per capita per month, according to the Kenya calculation of the area yield; and (v) implementation of an auditing Integrated Household Budget Survey [KIHBS 2005]) or the national system, such as video recording, satellite imagery, and/or additional rural poverty line (K Sh 1,562 per capita per month, according to CCEs on plots adjacent to the official CCE plots (World Bank 2011c). KIHBS [2005]), would reflect only the upper bound of poverty incidence in the region. Since maize income constitutes the majority 28 In insurance transactions it is customary to refer to “premium of economic income of those households in the low-potential areas, rates” where the cost of the policies is expressed as a share of the poverty measures for this group could well reflect their actual value insured. poverty incidence. 29 The coverage level determines the cases in which a payout is 42 We note that our model assumes away the potential that farmers triggered; e.g., any time the recorded yield level in a specific area falls below 80 percent of the reference average yield, a payout is issued. can save in a good year and draw on their saving to consume and See figure 9 for a graphical representation of the role of the coverage pay back loans in a bad year. Our results for the expected loan level. repayment rate should thus be interpreted as the lower bound of the potential rate. 30 Annex C.1 also presents the district breakdown adopted in the analysis. 43 This annex draws in part on GoK 2014a. 31 In technical terms this process is defined as a historical 44 This reflects the reality of the pastoral households in the regions, burn analysis. where livestock also provides intrinsic value beyond just serving as store of wealth. 32 The selected coverage level (80 percent) generates premium rates that for some districts would be excessive and not sustainable. 45 See Woodard et al. 2012 for detail. Hence in order to generate more realistic projections, commercial premium rates were capped at a maximum of 15 percent. In a 46 This leads to a conservative cost estimate, in particular if SECTION potential implementation phase, it will be important to assess the compared to situations in which CCEs would have to be outsourced 08 tradeoff between the cost of the policies and their actual coverage to a private entity and extension officers would mainly have an capacity. In districts where the capping is binding, the coverage auditing function. 98 K E N YA : A G R I C U LT U R E I N S U R A N C E S O L U T I O N S A P P R A I S A L 47 This should capture the important feature from reality that farmers will prefer to satisfy their basic needs before relaying any loan. 48 Longitudinal district yield data were detrended assuming an average combination of linear, exponential, and moving average trend (Stutley’s method). We also used estimated trends in the longitudinal data to detrend the two-year yield data. 49 Districts considered in low-potential maize zone include Kitui, Machakos, and Makueni in Eastern Province and Muranga, Kirinyaga, and Nyeri in Central Province. Districts in medium-potential maize zone include Kisumu, Siaya, and Nyamira in Nyanza Province; Vihiga and Busia in Western Province; and Meru in Eastern Province. Districts considered in high-potential maize zone are Nakuru, Trans Mara, Trans Nzoia, and Uasin Gishu in Rift Valley Province and Bungoma and Kakamenga in Western Province. 50 Because our temporal coverage was limited and could results in underestimation of actual temporal variations, we decided to exploit spatial variations of the area yield within each zone as well as with the assumption that variations in area yields within each homogenous zones could represent variations of yield realizations over time in that zone. SECTION 08 The World Bank /// /// 1818 H Street NW Washington DC 20433 Telephone: 202-473-1000 Internet: www.worldbank.org