Evaluation of socio-economic feasibility of pilot micro-insurance schemes for cotton producers in Mozambique

Transcription

1 Provision of Funds from the Food and Agriculture Organization of the United Nations Development of risk management strategies for cotton producers in Mozambique (CFC/CFC/27/FT/FA) Evaluation of socio-economic feasibility of pilot micro-insurance schemes for cotton producers in Mozambique Final report FAO Consultant ETH Zürich Grégoire Tombez and Raushan Bokusheva Zürich, December 2011 Institute for Environmental Decisions / IED Agricultural Economics - Agri-food & Agrienvironmental Economics Group 1

2 Executive summary In Mozambique, the agricultural sector contributes up to 23 percent of the gross domestic product (GDP) and up to 20 percent of the total exports. During the 2009/10 season, the main cash crops were cotton, sesame and tobacco. Due to cotton s economic relevance but also in view of the quality of the data available in Mozambique, cotton seems to be the best choice for developing a pilot insurance scheme. In view of the availability of yields and price data at the district level, the eligible provinces are Nampula and Cabo Delgado. The study focuses on the districts Montepuez, Namuno, Ribaue and Meconta due to their proximity to a weather station. In 2010, the World Bank analyzed the vulnerability of the different actors in the cotton value chain based on the expected losses and the capacity to manage risks. Exchange rate risks, international cotton price volatility and adverse weather events are risks causing higher vulnerability of the different actors along the cotton value chain. In Mozambique at the national level, the annual average yield of cottonseed fluctuates between 300 and 600 kg/ha. The average yield fluctuates widely from year to year. The goals of the first phase of the project were i) to collect data on yields, prices and adverse weather events; ii) to identify and characterize the climatic and market risks and determine their effect on income of actors in the cotton sector; and iii) to describe the prevailing relevant conditions for risk management in Mozambique and the possible management options for the risks identified. The objectives of the second phase were i) to define the most appropriate insurance schemes for cotton and to design a pilot instrument; ii) to quantify the performance of the insurance; iii) to estimate the potential for embedding the micro-insurance in a credit scheme; iv) to develop proposals for delivery mechanisms; and finally, v) to propose a road map to resolve the existing gaps and implement the most appropriate risk management options for Mozambique. There are many different risk management instruments farmers can use to protect themselves from adverse weather events as well as from market risks. The instruments can be divided into two categories: i) on-farm measures and ii) strategies for sharing risk with others. In Mozambique, limited access to credit, especially in rural areas, the absence of technologies reducing farmers dependency on weather conditions and low market integration limit farmers ability to apply on-farm measures. In this context, risk-sharing instruments can be attractive. This study focuses particularly on these risk-sharing instruments: agricultural insurance to manage production risk and contracts and hedging products to manage price risk. Agricultural insurance can be divided into two categories: traditional indemnity-based insurance and indexbased insurance. Traditional indemnity-based insurance is characterized by high administrative 2

3 and transaction costs. The contracts are subject to moral hazard and adverse selection problems. These problems are worsened by the institutional and legal framework in Mozambique. The production system characterized by small-scale farm units increases further administrative and transaction costs and the potential risk of moral hazard. In the case of index insurance, indemnities are based on an index (rainfall shortfall, high temperature, area yield etc.). A high correlation between the index and the yield shortfall is a precondition for developing successful index insurance. The main problem of index insurance is the potential mismatch between the index and the yield shortfall, the so-called basis risk. However, index insurance overcomes the problem of moral hazards: the policy s behavior cannot affect whatever the index is measuring. If the contracts are well developed, index insurance also overcomes the problem of adverse selection because the sellers and buyers share the same information regarding the likelihood of a weather event triggering potential payment. In the Mozambican context, index insurance schemes are more appropriate than traditional indemnity-based insurance. For each study, district meteorological, hydrological and agricultural indicators were computed to establish the link between weather and yields: i) yield shortfall is not necessarily caused by low cumulative rainfall; ii) yield variability is also caused by other uninsurable parameters; iii) rainfall distribution plays a decisive role; and iv) coverage against low and excessive rainfall is needed. The best results were obtained with an agricultural indicator, the sum of the days of crop drought stress (DCDS) in the entire crop season. In this model, in addition to rainfall (RF), the potential evapotranspiration (PET), the soil available water (SAW), the actual evapotranspiration (AET) and the growing cycle of cotton are considered. There are two critical periods in the cotton growing cycle. The first is shortly before and directly after the sowing period. Farmers need a minimum quantity of water at the beginning of the rainy season to start plowing their plots, as the seeds need water to start germinating. The second critical period is during flowering; it is the crucial yield formation period. In Montepuez, the index explains 71.2 percent of the yield variability. The thresholds for insurance payouts can be fixed at 62 DCDS and 85 DCDS. In these conditions, payouts are triggered i) if the DCDS do not reach the minimum level of 62 (each missing day of crop drought stress triggers a payout of 38 kg cottonseed) or ii) if the DCDS exceed the maximum level of 85 (each supplementary day of crop drought stress triggers a payout of 16 kg cottonseed). The maximum payout is fixed at 545 kg/ha and corresponds to the DCDS value of 48 and 120 days. For such a scheme, the fair premium would be 85 kg cottonseed/insured ha (15.6 percent of the average value of production per hectare). The proposed scheme allows the variance of income to be reduced by 3

4 45 percent. In the report, other possible insurance schemes (other threshold levels, drought coverage, area-based yield index) are described. During the last 15 years, the farmgate prices of first-quality cottonseeds fluctuated between US$0.15 and US$0.35/kg (except for 2011, US$0.50/kg). There is a high correlation between the Cotton A Index price and the local farmgate price (correlation coefficient, 0.93). The current price setting mechanism transfers the price volatility from the international markets to the national market. By fixing the minimum price before the beginning of the growing season instead of the end, price risk can be shifted from vulnerable cotton growers to less vulnerable ginneries (meso level). There are different price risk management tools, and each is described in detail in the report. Option contracts seem to be the most appropriate strategy at the meso level for shifting the risk to international markets. Based on the findings of this study and international experiences, the most appropriate scheme for smallholder cotton growers is index insurance bundled with a loan or an input package. Concerning the index, area-based yield and the number of days of crop drought stress perform similarly. To shift price risk from vulnerable farmers to less vulnerable ginneries, price setting should take place before the start of the growing season. Option contracts are effective tools for shifting the risk from ginneries to international markets. Furthermore, these contracts can be used similarly to hedge against exchange rate fluctuations. However, limiting factors still have to be resolved before launching a pilot project: The availability of farm data is a prerequisite for developing an insurance scheme trusted by the insurance and reinsurance sector. Reliable historical data on yields at the household level located close to existing weather stations must be collected. New weather stations installed in areas with high farm density and improvements in collecting farm yield data in these areas are highly recommended. Farmers willingness to pay for an insurance product is influenced by their educational background. Non-governmental organizations (NGOs), the government and ginners can play a key role in educating farmers on the potential benefits of risk management strategies. Before making a link with a loan and/or the supply of improved seeds, development constraints in these markets should be addressed. 4

5 Study objectives The overall objective of the FAO project Development of risk management strategies for cotton producers in Mozambique (CFC/CFC/27/FT/FA) is to assess the prefeasibility of crop insurance in Mozambique covering at least the risk of drought and if possible other perils. The objectives have been established in consultation with the National Cotton Institute of Mozambique (IAM) (Terms of reference in annex 11). This project covers various tasks: to collect data on yields, prices and adverse weather events; to identify and characterize the climatic and market risks and determine their effect on the income of actors in the cotton sector; to describe the prevailing relevant conditions for risk management in Mozambique and the possible management options for the risks identified; to define the most appropriate insurance schemes for cotton and to design a pilot instrument; to quantify the performance of the insurance; to estimate the potential for embedding micro-insurance in a credit scheme; to develop proposals for delivery mechanisms; to propose a road map to resolve the existing gaps and implement the most appropriate risk management options for Mozambique. 5

6 Index Executive summary 2 Study objectives 5 Index 6 Index of figures 7 Index of tables 7 Glossary 8 List of abbreviations Introduction and background information Agricultural sector in Mozambique Subject of the study Cotton sector Cotton plant growing cycle Risks in the cotton sector Vulnerability Characterization of production risks Characterization of market risks Natural hedge Risk management instruments Recent experience with agricultural insurance Recent experience with commodity price risk management Toward an integrated risk management strategy Quantitative analysis Production risk management: weather index Production risk management: area-based yield index Price risk management: option contracts Delivery channels Conclusions Recommendations Development of weather index insurance: next steps 74 References 79 Annexes A 6

7 Index of figures Figure 1: Distribution of Mozambican individual farms as a function of size. (data from Censo Agro Pecuário, 2009/2010) 11 Figure 2: National yields of major crops (except cotton) over time. Established from INE and FAO dataset. 12 Figure 3: Imported quantities and values of agricultural products in Mozambique (average ). Source: FAO STAT 13 Figure 4: Exported quantities of agricultural products in Mozambique (average ) [in t]. Source: FAO STAT 13 Figure 5: Cottonseed production at the provincial and national levels ( ). Source: IAM 14 Figure 6: Structure of the cotton industry in Mozambique. Source: Global Development Solutions, LLC 15 Figure 7: Main agroclimatic regions and agricultural research stations. Source: Instituto Nacional de Investigação Agronómica. 17 Figure 8: Yields at the national level ( ). Source: IAM. 20 Figure 9: Detrended cotton yields* [kg/ha] at different aggregation levels (district, province, country) ( ) 22 Figure 10: Average rainfall pattern in Ribaue and Montepuez with the driest and wettest seasons in the past 30 years. 23 Figure 11: Average rainfall patterns in Ribaue and Montepuez with the two harvest years with the best and worst yields in the past 16 years. 24 Figure 12: Exchange rate [MZN/US$]. Source: OANDA and Banco de Moçambique. 25 Figure 13: Cottonseed farmgate and Cotton A Index price ( ) expressed in kg cottonseed (left axis) and in kg cotton lint (right axis). Source: calculated from data from IAM (farmgate price), World Bank (Cotton A Index price) and OANDA (exchange rates). 26 Figure 14: Cotton A Index with moving average (20 weeks), standard deviation (20 weeks) and Bollinger bands. Source: InteractiveData Figure 15: Overview on risk management instruments 29 Figure 16: Scheme to determine the need for insurance, based on price variability, yield variability and price/yield correlation. Source: Meuwissen et al., Figure 17: Detrended yield vs. cumulated rainfall (Jan-Apr) in the Montepuez and Ribaue districts. 63 Figure 18: Detrended yield vs. days of drought stress (5 months) in the Montepuez and Namuno districts. 63 Figure 19: Detrended yield vs. days of crop drought stress (during the whole season) in the Montepuez and Namuno districts. 64 Figure 20: Detrended yield vs. DCDS in the Montepuez district with statistical dependence. 64 Figure 21: Payout structure of the proposed weather index based insurance scheme. 65 Figure 22: Payout structure of the proposed drought index based insurance scheme. 66 Figure 23: Yield vs. weather index yield estimate. 66 Figure 24: Income with vs. without weather index insurance. 67 Figure 25: Yield in Namuno (district) vs. yield in Cabo Delgado (province). 68 Figure 26: Payout structure of the proposed area-based yield insurance scheme (Namuno district). 69 Figure 27: Yield vs. area-based yield estimate. 69 Figure 28: Income with vs. without area-based yield insurance. 70 Figure 29: Transactions and main relationships within the cotton supply chain in Mozambique. 72 Index of tables Table 1: Cotton plant growing cycle Table 2: Preliminary identification of risks in the Mozambican cotton sector Table 3: Vulnerability to risky events based on expected loss and capacity to manage risk Table 4: Coefficients of variation and correlation coefficients of yields at different aggregation levels 21 Table 5: Correlation coefficients between national yields of the main crops in Mozambique Table 6: Indemnity and Index-based crop insurance products Table 7: Overview of commodity price risk management instruments Table 8: Road map for resolving the existing gaps and implementing index insurance in the Mozambican cotton sector

8 Glossary Adverse Selection Basis Risk Censo Agro Pecuário Cotton A Index Far East FOB Detrended yields People who buy insurance often have a better idea of the risks they face than do the sellers of insurance. People facing large risk are more likely to buy insurance than people exposed to a smaller risk. The insurance will automatically select people with a large risk and exclude people with smaller risk. This is called adverse selection. Index insurances may not be an appropriate tool in all circumstances. While an index should be highly correlated with individual losses, there will always be some variance between index and individual losses. This potential mismatch is known as basis risk. The Censo Agro Pecuário is a national agricultural survey conducted in 2000 and 2010 that includes, among others, data on households demographics, agriculture production and marketing, land and other assets, livestock holdings, agricultural technology use, food security indicators, nonfarm income sources and adult mortality and morbidity. Since 2004, the A Index represents offering prices based on CFR Far Eastern main ports terms. Cotton A Index is compiled from the cheapest five offering prices to mills of a basket of growths (currently 19), on the assumption that these are likely to be the cottons most frequently traded on the day in question. The mechanism of averaging the cheapest quotations has been proved to be a reliable method of calculating the Index over its forty-year existence. The trend in yields hinders the combination of yields observed over different periods of time: the cotton yields observed in the 1980s are not directly comparable to those observed today. To address the problem of structural change in yields levels observed over time, a variety of methods for detrending yield data have been adopted. Least squares regression techniques are generally used to estimate the linear or non-linear trend. Deviations from this trend are then added to the yield estimate to produce series of yields that can be compared over time, so called detrended yields. Ginning outturn ratio The ginning outturn ratio is the conversion ratio from seed to lint cotton. Moral Hazard People respond to incentives, and incentives are determined by costs and benefits. Because insurance reduces the costs of an adverse event for the 8

9 insured, he will reduce his efforts to mitigate this risk without insurance. This undesirable change in behavior is called moral hazard. New Mozambican Metical (MZN) Side-selling The New Mozambican Metical is the national currency of Mozambique Side-selling is the action to sell goods to one buyer that have been previously contracted to a different buyer to get better conditions as previously signed. 9

10 List of abbreviations AAM AET CBOT CC CCPs CME CNSP DCDS DDS DFID EDI FAO GDP GOT ha IAM INE MADER MCE MFI MPCI MZN NGO NISCO NCDE NMA NYBOT OTC PET RF SAW UNCTAD USD WFP WR WRSI ZCE The Mozambique Ginners Association Actual Evapotranspiration Chicago Board of Trade Correlation Coefficient Counter-Cyclical Payments Chicago Mercantile Exchange National Commission for Price and Wages Sum of Days of Crop Drought Stress Sum of Days of Drought Stress UK Department for International Development Ethiopian Drought Index Food and Agriculture Organization of the United Nations Gross Domestic Product Ginning Outturn hectare National Cotton Institute of Mozambique Instituto Nacional de Estatistica Ministry of Agriculture and Rural Development Multi Commodity Exchange Micro Finance Institutions Multiple Perils Crop Insurance New Mozambican Meticals Non-Governmental Organization Nyala Insurance Company National Commodity and Derivatives Exchange National Meteorological Agency New York Board of Trade Over-the-Counter Potential Evapotranspiration Rainfall Soil Available Water United Nations Conference on Trade and Development United States Dollars World Food Programme Seasonal crop water requirement Water Requirement Satisfaction Index Zhengzhou Commodity Exchange 10

11 1. Introduction and background information 1.1. Agricultural sector in Mozambique The agricultural sector contributes up to 23 percent of the Mozambican GDP and up to 20 percent of the total exports. The total area of the country is square kilometers including hectares (ha) of arable land. The farms cultivate ha, i.e. about 16 percent of the arable land (Censo Agro Pecuário, /10). Although available land is abundant, the amount of land under effective use is limited because of the lack of draught power and limited access to water. Ninety-nine percent of farms are smallholder farms (<10 ha) with an average size of 1.45 ha (Figure 1). Nampula, Zambézia, Tete and Figure 1: Distribution of Mozambican individual farms as a function of size. (data from Censo Agro Cabo Delgado are the provinces with Pecuário, 2009/2010) the highest number of farms with percent, percent, 9.83 percent and 8.88 percent, respectively, of the total number of Mozambican farms. During the 2009/10 season, the main food crops cultivated were maize ( ha), cassava ( ), sorghum ( ha) and sweet potatoes ( ha). The main cash crops were cotton ( ha), sesame ( ha) and tobacco ( ha) (Census, ). However, the amount of cultivated area can vary widely from year to year. In addition, yields fluctuate widely from year to year. The trend in yield is positive for most crops except for sweet potatoes (Figure 2). The year 1992 had particularly low yields for all crops due to a severe drought. 1 The Censo Agro Pecuário is a national agricultural survey conducted in 2000 and 2010 that includes, among others, data on household demographics, agriculture production and marketing, land and other assets, livestock holdings, agricultural technology use, food security indicators, non-farm income sources and adult mortality and morbidity. 11

12 Food crop: Maize Food crop: Cassava Yield [kg/ha] Yield [kg/ha] Harvest Year Harvest Year Food crop: Sorghum Food crop: Sweet.Potato Yield [kg/ha] Yield [kg/ha] Harvest Year Harvest Year Cash crop: Sesame Cash crop: Tabacco Yield [kg/ha] Yield [kg/ha] Harvest Year Harvest Year Figure 2: National yields of major crops (except cotton) over time. Established from INE and FAO dataset. In Mozambique, it is not possible to own land privately (Bachke, 2009), only to lease it for up to 50 years with a guaranteed second period of another 50 years. Small-scale farmers access to land is governed through a mix of customary laws, rights transfer from generation to generation and renting of the land. In general, small-scale farmers do not have access to improved farming technologies and tools. The agricultural production system is characterized by basic on-farm practices and low use of agricultural inputs. Fertilizers and chemicals are mainly applied in tobacco and cotton production since they are provided by local processing companies. On average ( ), rice, wheat and maize represented the three main imported agricultural products in terms of quantity in Mozambique (Figure 3). In terms of value, palm oil replaces maize in the ranking. Concerning exports, in terms of quantity, raw 12

13 sugar, maize and cashew nuts were the three main agricultural export products from 2004 to 2008 (Figure 4). In terms of value, tobacco holds first place ahead of raw sugar and cotton lint. Figure 3: Imported quantities and values of agricultural products in Mozambique (average ). Source: FAO STAT Figure 4: Exported quantities of agricultural products in Mozambique (average ) [in t]. Source: FAO STAT 1.2. Subject of the study Due to cotton s economic relevance but also in view of the quality of the data available in Mozambique, cotton seems to be the best choice for developing a pilot insurance scheme. Furthermore, risk management is a priority in the IAM s action plan. By extension and in a second phase, other cash crops can be included in the analysis to estimate the potential benefit of an insurance scheme for other sectors. In view of the availability of data on yields and prices at the district level, the eligible provinces are Nampula and Cabo Delgado. Furthermore, both provinces have a high potential for cotton production, increasing the benefit of the project. In these provinces, most weather stations are situated in coastal areas. However, three weather stations are situated inland. The weather station located in Montepuez delivers data for the districts Montepuez and Namuno, the weather station located in Ribaue covers the district Ribaue and the weather station situated in Nampula covers the district of Meconta. Henceforth, the study focuses on these four districts Cotton sector In years past, the cotton sector suffered due to civil war, internal politics, a fluctuating exchange rate, irregular weather and other secondary factors. After the civil war, the sector began showing some promising signs of recovery but in the last two years (2008/09 and 2009/10) again showed some worrying signs of weakness. In 2009/10, farmers (2.2 percent of all farmers) cultivated cotton in Mozambique. Cotton covered ha, or about 13

14 2 percent of the entire cultivated area. Production amounted to tons. In comparison, production in amounted to tons, i.e. three times more than in In , the number of smallholders producing cotton was estimated at About 70 percent of the cotton is Figure 5: Cottonseed production at the provincial and produced in Nampula and Cabo national levels ( ). Source: IAM Delgado (Figure 5). Considering the strategic role of the cotton sector in Mozambique, the government established the Cotton Institute of Mozambique (IAM) in 1991.The IAM activities are supervised by the Ministry of Agriculture and Rural Development (MADER). IAM is tasked with coordinating the production of seed cotton, performing marketing functions on behalf of the cotton sector and ensuring proper processing of raw cotton. IAM is also tasked with a wide variety of functions, including: Statistical monitoring and analysis Market supervision Cotton lint classification Conflict resolution Marketing and promotion of cotton production throughout the country Advising MAP on awarding new concessions. IAM s activities are funded through a % export levy on cotton produced by smallholder farmers (family sector), and through the state budget (World Bank, 2005). The Mozambique Ginners Association (AAM) was launched in 1998, to represent cotton processing companies. AAM s mandate is to promote the coordination between members, develop a dialogue between government and civil society, and to undertake initiatives to help develop the cotton sector. To help ensure coordination with the cotton farming sector, AAM is represented in the board of the IAM (Figure 6) (World Bank, 2005). 14

15 Figure 6: Structure of the cotton industry in Mozambique. Source: Global Development Solutions, LLC The cotton sector is not fully liberalized. The ginning companies have administrative concessions in which they have the right to purchase all cotton and the responsibility to provide support to any small cotton producing farmer. Thy concessionary system has been successful in re-launching the cotton sector after independence and in the absence of a working credit and input market. However, this has created local monopoly and monopsony. This loss of balance in market power, forces the government to fix a minimum price for cotton at the beginning of each growing season. The National Commission for Price and Wages (CNSP) define this minimum price in negotiation between MADER, IAM, cotton companies and smallholder representatives. The process of price determination begins with two parallel proposals. First, the cotton companies through the AAM advance a price proposal based on a residual method which takes into account all costs (ginning costs, seed cotton marketing costs, financial costs and the costs of input provision). At the same time, IAM draws up its own proposal, and the two negotiate. The agreed price is typically accepted by CNSP and officially announced shortly before the beginning of the harvest. Until 1999, a part of the lint cotton production was dedicated to the textile industry. After transition to a market oriented economy, the textile industry faced a shortage of capital and weak local demand. The low productivity could not attract sufficient investment to enable competitive spinning, weaving and finishing. Hence, most of firms were not able to meet the 15

16 high quality standards and reduce costs of production and thus disappeared. Since 2000, the whole production of lint cotton is exported to international markets Cotton plant growing cycle The growing cycle and the local rainfall pattern provide important information for analyzing the correlation between yield losses and weather events. There are two critical periods in the cotton growing cycle. The first is just before and soon after the sowing period; at the beginning of the rainy season, farmers need a minimum quantity of water to start plowing their plots for the seeds to start germinating. The second is the flowering period; it is the crucial yield formation period. In Mozambique, if adverse weather events occur during these periods, the effects are particularly devastating. Table 1 illustrates the typical cotton plant growing cycle in Mozambique. Table 1: Cotton plant growing cycle Source: adapted from IAM

17 Mozambique can be divided into ten different agroclimatic zones (R1 10) (Figure 7). Nampula and Cabo Delgado are composed of regions R7, R8 and R9. R7 covers a large part of Cabo Delgado and Nampula. In this region, there is an important human and agro-ecological potential for cotton production. These provinces are located from 200 to m above sea level. The texture of the soils varies from sandy to clay, consistent with the topography (Instituto Nacional de Investigação Agronómica, 2011). Information about the other regions can be found in annex 1. Figure 7: Main agroclimatic regions and agricultural research stations. Source: Instituto Nacional de Investigação Agronómica. 17

18 2. Risks in the cotton sector Table 2 presents the risks causing the most disruptive losses in the Mozambican cotton sector. These risks are not independent of each other; rather, many are interlinked. For example, farmers credit default can be a direct consequence of a pest outbreak or of an adverse weather event. On the other hand, ginners credit default and crop substitution are direct consequences of international lint cotton price volatility. Hence, two categories of risks can be built: the primary, exogenous risks (cause) and the secondary, endogenous risks, resulting from others (effect). Table 2: Preliminary identification of risks in the Mozambican cotton sector Source: adapted from World Bank

19 2.1. Vulnerability In 2010, the World Bank analyzed the vulnerability of the different actors in the cotton value chain based on the expected losses and the capacity to manage different sources of risk (Table 3). Exchange rate risks, international cotton price volatility, crop substitution and adverse weather events cause higher vulnerability (T1, dark grey) of the different actors along the cotton value chain. These risks require priority when developing solutions. The World Bank s analysis is qualitative and provides the basis for developing quantitative statistical models. Table 3: Vulnerability to risky events based on expected loss and capacity to manage risk 2 Source: World Bank The resulting matrix demonstrates five sets of vulnerabilities to the identified risks in terms of their priority, from the highest vulnerability containing the risks in the boxes with the darkest shade marked T1 (tier 1) upper-left corner toward the risks ranked with lowest vulnerability shown in the boxes with the clear shades toward the right bottom of the table marked T5 (tier 5). There are in between three additional intermediate vulnerability levels lighter shaded. 19

20 2.2. Characterization of production risks In Mozambique at the national level, the annual average cottonseed yield fluctuates between 300 and 600 kg/ha and can fluctuate widely from year to year (temporal fluctuations). Since 1981, there has been a slightly upward trend in yields (Figure 8). Yield [kg/ha] Mozambique C. DELEGADO NAMPULA Linear (Mozambique) y = x R² = Harvest Year Figure 8: Yields at the national level ( ). Source: IAM. The detrended yields at the district, province and country What are detrended yields? levels can be observed in Figure The trend in yields hinders the combination of yields 9. As yields have been detrended, observed over different periods: the cotton yields the graphs do not show effective observed in the 1980s are not directly comparable to those observed today. To address the problem of yields but show the variation in structural change in yield levels observed over time, yield over time. Table 4 shows the many methods for detrending yield data have been coefficient variation of yield over adopted. Least squares regression techniques are time for each district and the generally used to estimate the linear or non-linear correlation coefficient between trend. Deviations from this trend are then added to yields at different aggregation the yield estimate to produce a series of yields that levels. Various observations can can be compared over time, the so-called detrended be made: First, the coefficients of yields. variation are rather high at the district level; yields fluctuate widely. Second, the variation in yields is the highest at the lowest level of aggregation, i.e. the district level. The coefficient of variation can be expected to be even higher at the household level. Third, in Cabo Delgado, there is a good correlation 20

21 between yields at the district level and yields at the provincial level (correlation coefficient, ) and at the national level (correlation coefficient, ). This homogeneity emphasizes the potential for embedding an area-based yield insurance scheme in developing risk management strategies in What is a correlation coefficient? A correlation coefficient measures the extent to which as one variable increases the other variable tends to increase. The correlation is +1 in the case of a perfect Cabo Delgado. By contrast, the positive linear relationship, 1 in the case of a perfect correlation between district and negative linear relationship and some value between provincial yields is negligible in 1 and +1 in all other cases. Nampula (correlation coefficient, ). Table 4: Coefficients of variation and correlation coefficients of yields at different aggregation levels Province District Coefficient of variation Correlation coefficient (with the yield at the province level) Correlation coefficient (with the yield at the national level) C.Delgado Balama Chiúre Montepuez Namuno C. Delgado NAMPULA Meconta Mecuburi Moma Monapo Mogovolas Eráti Lalaua Ribaue Nampula MOZAMBIQUE Mozambique Source: elaborations on IAM dataset. 21

22 Country: Mozambique Detrended yield [kg/ha] Province: CaboDelgado (=CD) Harvest Year Province: Nampula (=N) Detrended yield [kg/ha] Detrended yield [kg/ha] Harvest Year Harvest Year District (CD): Montepuez District (N): Ribaue Detrended yield [kg/ha] Detrended yield [kg/ha] Harvest Year Harvest Year District (CD): Namuno District (N): Meconta Detrended yield [kg/ha] Detrended yield [kg/ha] Harvest Year Harvest Year Figure 9: Detrended cotton yields* [kg/ha] at different aggregation levels (district, province, country) ( ) 3 3 Note: To estimate the trend model, the trend for each district has been used. If no trend was statistically significant, the yield has not been detrended. Due to the high similarity in provincial and national trends, the national trend has been used to detrend national and provincial yields. 22

23 As seen above (section 2.1), in addition to market risks (exchange rate and price volatility), weather is the main production risk. Because of a low capacity to cope with weather, cotton growers are very vulnerable. Weather data from the Montepuez What is a rainfall pattern? Rainfall patterns are important tools for comparison in and Ribaue weather stations have rainfall over years. One can reasonably assume that been used to describe rainfall the rainfall in a particular month or season will be patterns because the weather relatively the same in the same month or season in the stations are located in cottonproducing areas where yield data following year. Hence, deviation from the average rainfall pattern can be observed. are available. The rainy season begins earlier in Montepuez (November) than in Ribaue (December) (Figure 10). In addition, the year with the most extreme weather conditions in the past 30 years was a dry season in Ribaue and a wet season in Montepuez Ribaue Montepuez Average Dryest Year Wettest Year Rainfal [mm] Rainfall [mm] sep oct nov dec jan feb mar apr may 0 sep oct nov dec jan feb mar apr may Figure 10: Average rainfall pattern in Ribaue and Montepuez with the driest and wettest seasons in the past 30 years. In Ribaue, the wettest season of the last 30 years and the harvest year with the lowest yield are the same (2008, 250 kg/ha) (Figures 10 and 11). The rainfall pattern of the year with the second lowest yield (2004, 300 kg/ha) is very similar to the average pattern; the only difference is that February and March were slightly drier than the average (Figure 11). Other factors may have played a role in limiting the yield during this year. In addition, rainfall in September or October generally has a negative impact on yield, which may be due to a false start in the growing season. In Montepuez, the year with the lowest yield (2001, 254 kg/ha) was not characterized by a low cumulated rainfall. Here, the point in time when it rained seems to have played a decisive role; rainfall was excessive in November and December and insufficient in March, an important period in crop and yield development (flowering). In Montepuez, the second worst yield was obtained in 2000 (342 kg/ha). However, the rainfall 23

24 pattern was very similar to an average year. This reduction in yield may be the result of other parameters variability Rainfall [mm] Ribaue Rainfall [mm] Montepuez Average 1st lowest yield 2nd lowest yield 2nd highest yield 1st highest yield sep oct nov dec jan feb mar apr may 0 sep oct nov dec jan feb mar apr may Figure 11: Average rainfall patterns in Ribaue and Montepuez with the two harvest years with the best and worst yields in the past 16 years. At the national level, cotton yield is not strongly correlated with the yield of other crops (Table 5). Hence, crop portfolio diversification is an effective on-farm strategy for coping with production risks. From the insurer s perspective, proposing insurance schemes covering different crops will not raise the insurer s risk exposition. Maize and sorghum are the only pair of crops for which the correlation coefficient is higher than 0.5. If the maize yield is high, the sorghum yield is generally high, and vice versa if the maize yield is low. Table 5: Correlation coefficients between national yields of the main crops in Mozambique Cotton Maize Sorghum Sesame Cassava Sweet Potato Tobacco Cotton Maize Sorghum Sesame Cassava SweetPotato Tabacco 1.00 Source: elaborations on INE dataset. 24

25 2.3. Characterization of market risks Exchange rate Exchange rate variation (Figure 12) has a large impact on the variation in local prices. For further analysis, local prices are converted into US dollars. As farmers sell cotton directly after harvesting, the exchange rate at harvest was used to convert New Mozambican meticals (MZN) into US dollars (US$). There is a negative trend in the exchange rate MZN/ US$. Figure 12: Exchange rate [MZN/US$]. Source: OANDA and Banco de Moçambique. Cotton price During the past 15 years, the farmgate prices of first-class cottonseed fluctuated between US$0.15 and US$0.35/kg. On average, the farmgate price for second-class quality is 20 percent lower than for first class (Figure 13). To compare cottonseed prices with lint cotton international prices, the ginning outturn (GOT) ratio must be taken into account. The ginning outturn ratio is the conversion ratio from seed to lint cotton. In Mozambique, the average ginning outturn ratio is 34 percent (World Bank, 2005). With a GOT ratio of 34 percent, the farmgate price (first class) falls between 35 percent and 55 percent of the price for the Cotton A Index at harvest. There is a high correlation between the Cotton A Index price and the local farmgate price (correlation coefficient, 0.93). The actual price-setting mechanism transfers the variation from the international markets to the local market. 25

26 Figure 13: Cottonseed farmgate and Cotton A Index price ( ) expressed in kg cottonseed (left axis) and in kg cotton lint (right axis). Source: calculated from data from IAM (farmgate price), World Bank (Cotton A Index price) and OANDA (exchange rates). The international market price reflects the interaction between demand and supply: The price is set to equate the quantity being supplied with that being demanded. Agricultural commodity price fluctuations are the result of the coexistence of four circumstances: i) exogenous events (weather or others) that can dramatically change the supply; ii) inelastic demand that amplifies the effects of supply changes on price; iii) government intervention disrupting the trade in cotton; and iv) market actors imperfect information about available supplies and strength of demand. Habitually, the Cotton A Index fluctuates between US$0.65 and US$1.80/kg cotton lint. In the beginning of 2011, the volatility increased widely while the price peaked at US$5.30/kg cotton lint. In addition to the price, the volatility also fluctuated widely over time (Figure 14). In Mozambique, the minimum price between cotton farmers and ginning companies is fixed by the Mozambican Government in April just before the seed-cotton commercialization campaign. Henceforth, cotton farmers bear price risk during the whole season from the sowing period to the moment when the minimum price is fixed. Ginning companies bear price risk from the moment when the minimum price is fixed to the moment when they sell cotton lint to international trading/processing companies. 26

27 Figure 14: Cotton A Index with moving average 4 (20 weeks), standard deviation 5 (20 weeks) and Bollinger bands 6. Source: InteractiveData Natural hedge Whether prices and yields are negatively correlated is important. If they are, low yields are generally compensated with high farmgate prices. This phenomenon is called natural hedge. It occurs when production shortages induce price increases. In Mozambique, the correlation coefficient between cotton aggregated yields at the national level and the farmgate price is 0.08, meaning that the correlation is very weak. There is no natural hedge. This arises from the fact that the farmgate price is determined according to the international cotton price and that Mozambican production has no influence on international cotton price development. 4 The Moving Average shows the mean instrument price value for a certain period. When one calculates the moving average, one averages out the instrument price for this time. As the price changes, its moving average either increases or decreases. 5 Standard Deviation (StdDev) measures the market volatility. This indicator characterizes the scale of price changes relating to the moving average. Thus, if the indicator value is large, the market is volatile, and the bar prices are rather dispersed relating to the moving average. If the indicator value is not large, it means that the market volatility is low and the bar prices are rather close to the moving average. 6 Bollinger Bands are plotted a certain number of standard deviations away from the moving average. When the markets become more volatile, the bands widen, and during less volatile periods, the bands contract. 27

28 3. Risk management instruments In general, many different risk management instruments can be used by farmers to protect themselves from adverse weather events as well as from market risks. The instruments can be divided into two categories: i) on-farm measures (including crop diversification, other income-generating activities, access to market information, use of stress resistant seed varieties, pesticides, irrigation, off-farm employment) and ii) risk-sharing strategies (insurance, hedging products and contract agreements, e.g. forward, futures, option or swap contracts) (Figure 15). In the Mozambican cotton sector, the most widespread on-farm measure is probably crop diversification; most cotton farmers also grow food crops such as maize, cassava or sorghum. Price condition is one of the most important factors influencing production area. Unfortunately, as reference, the farmers have only the price they get during the last crop season. The second measure is to replant the cotton if the rainy season starts later than usual. This measure induces additional labor costs but allows farmers to cope with the lack of rainfall at the beginning of the growing season. Pesticides are generally not used optimally. Food crops are usually allocated pesticides first, and cash crops are often neglected. Because of poor market access, improved seed varieties are not widely available to farmers. Irrigation is very limited because of poor credit access and high investment costs. Fertilizers are spread only marginally. If farmers are able to reach the desired degree of risk protection with on-farm measures, there may be no need for risk-sharing strategies. In Mozambique, poor access to rural credit and the low level of technology adoption due to the low market integration limit farmers opportunity to apply on-farm measures and increase the farmers dependence on weather variations. Risk-sharing instruments are adapted to cover risk that cannot be managed efficiently with on-farm measures. Because on-farm measures are not available for all farmers in Mozambique, identifying this part of risk is not possible. In this context, it is more attractive to enhance on-farm measure availability first through developing market and credit access before introducing risk-sharing strategies. Once these measures are widely available, risk-sharing strategies can be introduced. This study focuses particularly on these two types of risk-sharing instruments: - agricultural insurance (to manage production risks) and - contracts and hedging products (to manage price risk). 28

29 Figure 15: Overview on risk management instruments 3.1. Recent experience with agricultural insurance Agricultural insurance can be divided into two categories: traditional indemnity-based insurance and index-based insurance (Table 6). Traditional indemnity-based insurance Description Indemnity-based insurance provides indemnities when crop yield falls below a specified level due to a variety of natural causes. The different experiences with indemnity-based insurance show that risk must have three characteristics to be insurable. First, the likelihood must be 29

30 readily quantifiable. Second, the damage the risk causes must be easy to attribute and value, and third, neither the occurrence of the event nor the damage it causes should be affected by the insured s behavior (Hazell, 1992). Insurable risks include some production risks (e.g. losses due to catastrophic weather events such as drought, flood and hail), health risks and some asset risks. Indemnity-based insurance products fall into two categories: i) damagedbased indemnity policies and named-peril crop insurance and ii) yield-based indemnity products, which include multiple perils crop insurance (MPCI) yield shortfall cover, and crop revenue insurance, which combines protection against the loss of physical crop yield and loss of market price (World Bank, 2008). Is indemnity-based insurance applicable in Mozambique? There are several challenges with this type of contract. First, losses must be recorded and compared with individual What is moral hazard? People respond to incentives, and incentives production histories. This leads to high are determined by costs and benefits. administrative and transaction costs. Two Because insurance reduces the costs of an adverse event for the insured, he or she will additional problems are moral hazard and reduce his or her efforts to mitigate this risk. adverse selection (see boxes), which This undesirable change in behavior is called come into play with such contracts. moral hazard. These high administrative and transactions costs coupled with the control of moral hazard and adverse selection problems make these schemes very costly (Hazell, 1992). Furthermore, these problems are worsened by the What is adverse selection? People who buy insurance often have a better idea of the risks they face than do the sellers of insurance. People facing a large risk are institutional and legal framework in more likely to buy insurance than people exposed to a smaller risk. The insurance will Mozambique. The production system automatically select people with a large risk characterized by small-scale farmers and exclude people with a smaller risk. This is implies high administrative and called adverse selection. transaction costs and the potential risk of moral hazard. These components make indemnity-based insurance products very unattractive in Mozambique. From then on, it is advisable to take into consideration alternative insurance schemes that can undermine these problems. 30

31 Table 6: Indemnity and Index-based crop insurance products Type of insurance Traditional Crop Insurance Damage-based indemnity insurance (named-peril crop insurance) Yield-based crop insurance (MPCI) Crop revenue insurance Greenhouse insurance Index-based Crop Insurance Weather index insurance and normalized difference vegetation index/satellite insurance Area-yield index insurance Source: World Bank 2008 Description Insurance in which the claim is calculated by measuring the amount of damage in the field soon after damage occurs. This figure, less a deductible expressed as a percentage, is applied to the preagreed sum insured, which may be based on production costs or expected crop revenue. Where damage cannot be measured accurately immediately after the loss, the assessment may be deferred until later in the crop season. Damage-based indemnity insurance is best known for hail but is also used for other named-peril insurance products, including frost, excessive rainfall, and wind. Insurance in which an insured yield (for example, tons/hectare) is established as a percentage of the historical average yield of the insured farmer. The insured yield is typically percent of the average yield on the farm. If the realized yield is less than the insured yield, an indemnity is paid equal to the difference between the actual yield and the insured yield, multiplied by a preagreed value of sum insured per unit of yield. Yield-based crop insurance typically protects against multiple perils (many different causes of yield loss), because it is generally difficult to determine the exact cause of the loss. Insurance that combines yield based crop insurance (MPCI) with protection against market price reduction at the time of sale of the crop. As of 2009, this product was marketed on a commercial basis only in the United States for grains and oilseeds with future contracts quoted on the Chicago Board of Trade. Insurance that combines coverage to greenhouse structures and equipment and conventional crop insurance (usually restricted to named perils) for the greenhouse crop. Insurance in which the indemnity is based on realizations of a specific weather parameter measured over a prespecified period of time at a particular weather station. The insurance can be structured to protect against index realizations that are either so high or so low that they are expected to cause crop losses. An indemnity is paid whenever the realized value of the index exceeds or falls short of a prespecified threshold. The indemnity is calculated based on a preagreed sum insured per unit of the index (for example, dollars/millimeter of rainfall). In the case of satellite insurance, indexes are constructed using time-series remote sensing imagery (for example, applications of false color infrared waveband to pasture index insurance, where the payout is based on a normalized difference vegetation index, which relates moisture deficit to pasture degradation). Research is being conducted on applications of synthetic aperture radar to crop flood insurance. Insurance in which the indemnity is based on the realized (harvested) average yield of an area such as a county or district. The insured yield is established as a percentage of the average yield for the area (typically percent of the area average yield). An indemnity is paid if the realized average yield for the area is less than the insured yield, regardless of the actual yield on a policyholder s farm. This type of index insurance requires historical area yield data on which the normal average yield and insured yield can be established. 31

32 Weather index based insurance Description In the case of weather index based insurance, indemnities are based on an index (rainfall shortfall, high temperature etc.). A high correlation between the weather index and the yield shortfall is a precondition for developing successful weather index based insurance. Index insurance overcomes the problem of moral hazard: The policy s behavior cannot affect whatever the index is measuring. If the contracts are well developed, index insurance also overcomes the problem of adverse selection because the sellers and buyers share the same information regarding the likelihood of a weather event that will trigger payment. This advantage also facilitates access to the reinsurance market. Finally, there are no loss adjustment costs. The amount of loss can be calculated using the coverage value and the index level, and the payment can be paid directly and rapidly to the policyholder (Skees et al., 2007). Index insurance may not be an appropriate tool in all circumstances. Although an index should be highly correlated with individual losses, there will always be some variance between the index and individual losses. This potential mismatch is known as basis risk. Basis risk occurs when a policyholder experiences a loss but does not receive a payment because the index level is not met or conversely when an insured receives a payment but does not experience a loss. There are different types of basis risks (Skees et al., 2007): Spatial basis risk is the difference in outcomes between the places where the index is measured and where a loss event occurs. For example, it may rain at the weather station, but the rain may not extend to a farmer a few kilometers away. Temporal basis risk refers to the timing of the loss event considered. The lack of rainfall may have different impacts depending on the stage of crop development. A part of this basis risk can be addressed through developing contracts that consider the crop growing cycle and different stages of the growing season. Loss-specific basis risk occurs when the losses are poorly related to the index. Basis risk is a limiting factor in developing index insurance. Index insurance may not be appropriate in regions with different agroclimatic zones. Carefully considered contract design and rigorous stakeholder capacity-building can mitigate the incidence of basis risk and help avoid undue expectations about the benefits and advantages of index insurance (Barrett et al., 2007). 32

33 As seen above, weather index based insurance offers many advantages in low-income countries. Several projects are ongoing around the world, and we can take advantage of the lessons emerging from these projects (annex 1). These projects are at different development stages. This report will focus on three different countries. Malawi was selected because of many similarities with the Mozambique case (region, context etc.). India was selected because of the stage of the project (scaling-up) and Ethiopia because of the diversity of the pilot projects conducted (micro- and macro-level weather index insurance). Malawi: micro-level drought index insurance bundled with a loan for groundnut farmers The drought index insurance product developed in Malawi is bundled with a loan and an input package containing improved groundnut seed. A part of the yield is used to repay the loan. The farmer can obtain a loan to purchase improved seed varieties only if he subscribes to drought index insurance. Weather index insurance protects the loan. If the rainfall level falls under a predetermined level, the lender receives an indemnity, and the farmer does not have to pay back the entire loan. The Malawian pilot project tests the capacity and willingness of banks and insurance companies to provide index based weather insurance, and the impacts on credit supply and costs. The 2005/06 groundnut pilot year faced some problems related to input quality, product communication and loan repayment. The pilot project also demonstrated some difficulty in educating farmers about weather index based insurance, as there was a general lack of knowledge of or experience with insurance, and many farmers had never accessed credit from a formal financial source before. However, the stakeholders who participated in the program believed that these problems could be overcome with improvements in program design and that the results were promising enough to continue the program in 2006/07 (Syroka, 2009). Difficulties The banks involved in the pilot project therefore faced a dilemma: although the weather insurance product provided protection against adverse weather risk, the product did not strengthen the contractual relationships within the groundnut supply chain. Side-selling was a serious problem with no clear solution. As a result, the banks agreed that the weather insurance tool would have more utility in a commodity sector with stronger supply chains, such as tobacco, where contract farming arrangements were common, and paprika, tea, 33

34 coffee and cotton, where they are developing. Weather index based insurance is only one tool for mitigating the risks in agricultural finance and supply chain relationships. This insurance focuses on only one or two aspects of production risk. The groundnut pilot project revealed that problems related to production, marketing and sales could still undermine credit repayment, and therefore the value of the insurance policy. What is side-selling? Side-selling is the action of selling goods to one buyer that have been previously contracted to a different buyer to get better conditions than previously agreed to. The main advantage of the index-based approach is that the payout is not based on the condition of the crop, but rather on the indisputable rainfall record available in real-time so that claims can be automatically triggered to farmers when adverse rainfall events occur. To date, the index based insurance contracts piloted in Malawi cover the value of the input loan, not the crop. If there is drought, the insurance payout repays part of the costs of the loan. Insofar as loan default risks are reduced, credit costs should decline, and banks should be willing to extend larger quantities of credit to more farmers. Expanding the weather insurance project in Malawi will also provide an opportunity for the local insurance industry to grow a larger, more diversified portfolio of risk. Opportunities Conclusion and recommendations According to the experience in Malawi and before designing weather index based insurance, the following prerequisites must be met: Malawi, with 22 government-supported full meteorological weather stations, 21 subsidiary agrometeorological stations and more than 400 rainfall stations has one of the strongest infrastructures in the region (Malawi Meteorological Services, 2011) (1 full meteorological station/5 400 km 2 ). This weather station infrastructure and planned investment in new rain gauges may make scaling up the project at the national level feasible. According to Syroka (2009), this level of infrastructure was the minimum, and replicating Malawian-type insurance in countries with less developed Weather data 34

35 weather station networks is not recommended. In most African countries, crop-yield and production data are nearly nonexistent. Thus, product designers must rely on plant growth simulation models. Even though this method may appear appealing, the assumption that all farmers use the same technology and farm the same types of soil must be made. This is questionable in view of the diversity of soil types and farming systems in most African countries. Good production and weather data sets allow this diversity to be taken into account and to limit basis risk. The pilot project highlighted that many insurance companies in Malawi were waiting to learn from several years of pilot testing before committing more fully and expanding the program. The complex method of product development (crop yield modeling) may reduce the capacity of the local insurance industry to develop products further independently. The pilot project also highlighted that interested local insurance companies should be involved from the beginning of the project. The best prospects for integrating index based insurance into wider agricultural risk strategies occur within relatively well developed supply chains. The revelation that some farmers were side-selling and defaulting on their loans is an important lesson from this pilot project. Consideration of possible moral hazard problems and the advantages and disadvantages of links between actors in the value chain should be important for any future index insurance product. Production data Capacity building and commitment of local partners Supply chain Important questions remain about how to address repayment risks in supply chains where contract farming arrangements are not common, the viability of targeting small farmers with this product is unclear and whether weather index based insurance can contribute to either reducing the cost of credit or expanding access to credit in the agricultural sector. Malawi: macro-level drought index insurance (maize) A weather risk management tool has been developed to help the Malawian Government manage the financial impact of drought-related national maize production shortfalls. The instrument is an index-based weather derivative contract designed to transfer the financial risk of severe and catastrophic national drought that adversely impacts the government s budget to the international risk markets. This innovative risk management instrument was pioneered in by the Government of Malawi, with the assistance of the World Bank, and was the first for a sovereign entity in Africa. Several piloting seasons will be 35

36 necessary to understand the scope and limitations of such contracts, and their role in the government s strategy, contingency planning and operational drought response framework (Syroka et Nucifora, 2010). India: micro-level drought and floods index insurance for groundnut farmers In 2003, rainfall-index insurance was introduced in India to address high credit default rates and increase lending opportunities in rural areas. The policies were sold directly to groundnut farmers and were bundled with a loan (Skees et al., 2007). A key variable in the introduction of index insurance in India is likely the fact that the regulatory framework binds insurance companies to collect 5 percent of their total premium in the rural sector. The insurance companies strategy was clearly to reach this level with the help of index insurance. They might pay less attention to the sustainability of the proposed insurance schemes. Demand for index insurance has been sensitive to price and to endorsement from a trusted third party. However, without loan coupling, uptake has remained low, even when the price of insurance was less than its expected value. These results are consistent with the view that in addition to price and liquidity, trust and financial literacy significantly influence uptake (Giné et al., 2010). Giné et al. identified many factors that may discourage participation such as household credit constraints, limited understanding of the product, limited trust in insurance providers or high transaction costs that raise the price of the contracts. Farmers perceived a number of problems with the product 7 : Difficulties Rainfall measurements were taken too far from their village and did not represent the rainfall the farmers had (spatial basis risk). Farmers would prefer a simple linear relationship between the rainfall and the claim amount. They were unable to appreciate the trigger points and different slab rates. Farmers would like to receive phase wise payouts subject to maximum limits; that is, the farmers would prefer to have two or three consecutive contracts with separate payouts that would trigger cash payments at the exact time of need. 7 Based on a meeting with farmer customers of the rainfall insurance pilot in village Pamireddi Pally, Mandal:Atmakoor, District: Mahaboobnagar on January 29, 2004, documented by D Sattaiah AVP, BASIX. 36

37 Frequent interactions between the insurance company representative and the farmers were needed to clarify doubts and questions about the product. Despite these difficulties, farmers expressed strong interest in buying weather insurance bundled with a loan. They preferred an improved perhectare product that allows them to scale up the purchase according to individual exposures. Since there were sufficient data and a good correlation between the index and losses and insurers were willing to offer the product and farmers willing to buy it, a scaling-up of the project took place. The insurance industry could diversify the product portfolio to address farmers needs. Opportunities Conclusion and recommendations The Indian project with a pilot project and scaling-up stage highlights the following findings: In India, the density of meteorological station is relatively good. The problems reside in the data update. Data at most stations are recorded manually, and it frequently takes days for insurers to get data from public weather stations. This delays the timely settlement of farmers payouts and may discourage participation. The product must be simple enough for farmers to understand and must pay out for events the policyholder cares about. In addition, farmers should be educated about the benefits of such products. In this Indian project, it was not clear who should bear the cost of educating potential clients and how detailed the message should be. In a region where liquidity is a major constraint, lenders could offer loans to pay for the premiums. This solution protects individual farmers only partly against adverse events but may increase credit availability in rural areas. It is easier to sell products in regions where a positive past insurance payout has occurred. At the beginning, insurance providers could introduce the insurance policy without collecting premiums until the first payout occurs. This allows a better understanding of the underlying mechanism by the policyholder and trust in the insurance product. As insurers will not be willing to enter a market with such conditions, a strategy of financing the Weather data Marketing channels Liquidity constraint Trust building 37

38 first payout should be developed. For the lenders, the potential benefit of index insurance was clear. It is an attractive way to mitigate credit default risk. Therefore, the focus of research should be on the demand side. Client education Ethiopia: macro- and micro-level weather index insurance pilot projects In 2006, the World Food Programme (WFP) bought weather index insurance against extreme drought for regions in Ethiopia during the agricultural season. The WFP used payments to provide food aid and sustain vulnerable farmers (macro-level insurance). The Ethiopian Drought Index (EDI) was developed using meteorological data provided by the National Meteorological Agency (NMA) coupled with water balance models. The index had an 80 percent correlation with the number of food aid recipients between 1994 and 2004 and demonstrated that the index could be a good indicator of human needs when a drought occurs. The policy was not renewed in 2007 due to the lack of donor support, but other pilot projects followed this initiative. In 2009, a micro-level product was developed by Nyala Insurance Company (NISCO), with guidance from the WFP, to insure smallholders against severe drought events. The product was crop specific (pea beans) and based on a simplified version of the WRSI (Water Requirement Satisfaction Index) (Hazell et al., 2010). The index measured the balance between water demand and supply during the growing season. The WRSI is computed as the ratio between actual evapotranspiration (AET) and the seasonal crop water requirement (WR). This simplified model assumes a zero soil water holding capacity. During the three main cultivation phases (initial, mid- and final), the number of decades with deficit rainfall was counted. There is a water deficit within a ten-day period if required rainfall 10d >actual rainfall 10d. At the end of each phase, the payout is computed according to the value of the corresponding rainfall deficit. Contracts were structured from a micro-perspective. They were sold to the farmers through a farmer cooperative to secure the farmers participation (trusted delivery channel). When drought occurred in 2009, the growth of pea beans was impeded, and payouts were triggered for the 137 farmers participating in the pilot project. There is no government regulation regarding index insurance in Ethiopia. Without appropriate regulation, the future scalability and sustainability of index insurance is threatened. This may lead to poor understanding of the product or reduced product appreciation. In Ethiopia, the government provides lending guarantees to agriculture that has a significant impact on Difficulties 38

39 the market for and the development of a weather index based insurance product. Indeed, the government promises to assume a borrower s debt obligation if that borrower defaults. This guarantee currently affects the incentives for banks to become interested in market-based risk management approaches (World Bank, 2006). The macro-level pilot project showed that it is feasible to use market mechanisms to finance drought risk in Ethiopia and that it is possible to develop objective, timely and accurate indicators for triggering drought assistance. Thus, ex-ante resources can give the Ethiopian Government and donors the incentive to put contingency plans in place, allowing earlier response to shocks and growing interest in funding (WFP, 2007). The Ethiopian Government considered the pilot project a good learning experience in providing insurance in rural areas and believes that index insurance has the potential to increase production by sharing risk and assisting farmers in adopting new technologies. As a result of the pilot projects, awareness among farmers and financial institutions of the role of index insurance increased. This has had a number of positive effects, including opening up smallholders access to loans (as farmers are using credit to finance premium payments) and connecting agricultural insurance in Ethiopia to international financial markets (Hazell et al., 2010). The pea bean pilot project also involved ongoing communication with farmer cooperatives. They reach farmers directly and communicate clearly and effectively, without misunderstanding. Opportunities Conclusion and recommendations From these pilot projects, some advice can be drawn: In general, sufficient data for developing large-scale weather insurance contracts are lacking. This is due to a relatively thin geographical distribution of stations and to missing data. At the micro level, the pilot project failed to identify any organizations that could be used to reach clients effectively. Financial institutions were the prime candidates, but the government s lending guarantee minimized incentives for financial service providers to participate in market-based risk management strategies; the government s lending guarantee and the index- Weather data Marketing channels 39

40 based insurance competed with each other. A strong local insurance partner is essential for the scalability and sustainability of an index insurance product. Experiences in pilot projects have led the WFP to conclude that sustainability and scalability will not be achieved unless the product development is locally owned and managed. In all pilot projects, the local company NISCO played a key role. Nevertheless, further capacity building was needed to give the insurance companies and banks the ability to offer appropriate and diversified products and sustain market growth. At the pilot project scale, an insurer company was capable of assuming the whole risk. On a larger scale, a reinsurance scheme must be developed to shift part of the risk to international markets. According to NISCO, the biggest obstacle to achieving scale is the low level of awareness of insurance and education among smallholders. During the dry season, producers are interested only in price, and during the wet season, they are busy taking care of their crops. One option is to bring farmers together and pay them a per diem to learn about index-based insurance. Capacity building and commitment of local partners Risk capacity Client education Is weather index insurance applicable in Mozambique? The index insurance pilot projects highlighted some preconditions and recommendations for successfully developing and implementing an index insurance product. We can now use the outcome of these projects (Malawi, India and Ethiopia) to analyze the situation in Mozambique: In Mozambique, weather station density is very poor (approximately 1 station/ km 2 ). It is possible to develop index insurance only for the region where a weather station is available and limit the potential of scaling up index insurance. The current weather station network in Mozambique (33 synoptic stations, 13 agromet stations, 48 climatological stations and 7 AWS) suffers from missing data, principally during the civil war ( ). Furthermore, most weather stations are near cities, and very few are in rural areas. Initial insurance pilot projects can be established even without historical weather data or real-time weather data services. Market Weather data 40

41 players in international markets will not trust the insurance scheme if they cannot be assured of good data for risk pricing. In general, production data are available only at an aggregated provincial level. Exceptions are Nampula and Cabo Delgado where production and yield data are available for the main producing districts. The lack of yield and production data at the household level does not allow an analysis of the basis risk problem. In Mozambique, smallholders are hard to reach. Farmer associations are very weak if they are not supported by an NGO or the government. One alternative could be to use input providers or output processors as marketing channel aggregators. In the Mozambican cotton sector, cotton ginning companies play both roles. Therefore, the companies could act as an intermediary to supply insurance products. The pilot project in Malawi gives an example of combining an insurance product with the input market. This possibility must be taken into account carefully since there is a lack of improved seed market in Mozambique and the input market is biased; ginning companies distribute seed free and have the right to buy all the cotton produced in their own concession. A package with seed and index insurance will compete with the free seed actually distributed by the ginning companies. The second alternative is to use microfinance institutions and to link the insurance product with a loan. The main problem is that, actually, microfinance institutions are not widely established, and the costs of reaching individual farmers are very high. The third alternative is to establish a compulsory system in which all producers are protected by the index insurance bought by the government and in which no marketing channels are necessary. International experiences show that this alternative may be adapted to manage catastrophic events. A weather index based risk management instrument was pioneered in by the Government of Malawi, with the World Bank s assistance. Several piloting seasons will be necessary to understand the scope and limitations of such contracts. Local insurance partners are interested in developing and marketing index insurance products. The majority of national insurance companies have no experience in developing these products. Only Hollard Moçambique has previous experience. The lack of available meteorological and yield data Production data Marketing channels Capacity building and commitment of local 41

42 has led the company to stop its research in this field of activity. However, the company remains interested in developing this product. To strengthen the contractual relationship within the supply chain, farmers associations should act as an intermediary among input suppliers, farmers and processors. Farmers associations can help reduce side-selling and the credit default rate. The beginning of the planting season is a crucial time in terms of liquidity for many smallholders. Without credit access, it would be very difficult for a farmer to pay the premium. Here again, we can take advantage of the experience in Malawi and in India where the product is bundled with a loan. Bundling increases considerably the producers willingness to pay a premium. Bundling also increases the bank s willingness to lend to the rural sector as loans are partly protected. National insurance companies expressed the need to develop a reinsurance strategy. Drought risk is systemic and can affect large regions, and companies may not have the capacity to face this risk. An advantage of index insurance is reduced information asymmetry and increased international reinsurance company interest. Smallholders have little experience with agricultural insurance. Educating the farmers is necessary. This consequent effort should not be undermined, and pointing out the actors responsible for this task is essential. partners Welldeveloped supply chain Liquidity constraint Risk capacity Client education In view of the similarities, the pilot project in Malawi is a very good example to illustrate the challenges of developing functioning weather index insurance for cotton producers. The major problem in Malawi is the weakness in the contract relationship and can be solved with the help of strong and good established farmers associations. 42

43 Area-based yield index insurance Description Area-based yield insurance contracts provide the insured farmer with an indemnity each time the yield at a given aggregation level falls below its critical level (i.e. strike), regardless of the realized yield on individual farms. Typically, it is assumed that the individual farm s yield will have only a small impact on the area-based yield, and therefore, area-based yield crop insurance contracts do not provide incentives for moral hazards or adverse selection. Areabased yield insurance provides farmers whose individual yields are strongly correlated with area yields with considerable protection against yield and, therefore, income variation (Smith et al., 1994). Thus, area-based yield insurance provides effective risk management only in areas where yield risks are largely systemic (Skees et al., 1997). As with any index, basis risk is an important factor affecting the risk management effectiveness of area-based yield insurance. The higher (lower) the positive correlation between individual and aggregated yield, the lower (higher) the basis risk. Mali: area-based yield insurance bundled with a loan In Mali, all cotton producers are organized in cooperatives and sell their product to a national unique buyer. Cooperatives are compulsory intermediaries to obtain a loan. The scheme developed in Mali is targeted toward cooperatives of producers and is linked to the cooperatives credit contracts. If a cooperative opts for insurance, the area in which cotton is grown by members of the cooperative is insured. The premium is pre-financed like a regular input by the bank that allocates the loan. The average yield level in the district triggers indemnity payments. Insurance payments serve first to pay back cooperative loans (Guirkinger, 2011). The main difficulty of this project is to minimize basis risk. Basis risk arises from the fact that the cooperative s yield only imperfectly correlates with the district level. There are two types of unfortunate situations: Difficulties - the cooperative yield is above the district strike-point, but the insurance pays out (the so-called false positive); - the cooperative yield is below the district strike-point, but no payment occurs (the so-called false negative). In this project, basis risk with an area-based yield was lower than with the weather- or satellite-based index. That is why area-based yield 43

44 insurance was chosen for this project (Guirkinger, 2011). An innovation was introduced to minimize basis risk. Reducing the geographical area used to compute the index reduces basis risk but increases the scope of moral hazards. A double-trigger structure provides a solution to this dilemma by enabling basis risk to be reduced while remaining immune to perverse incentives. The idea is to keep the district trigger and to add a coop trigger so that insurance pays out only if two conditions are fulfilled: The district yield is below the district strikepoint, and the coop yield is below the coop strike-point (Guirkinger, 2011). Opportunities In the Malian cotton sector, cooperatives are well organized and can be used as channels to target smallholders with insurance products. Historical production data are available at the household level, and the contract performance can be estimated. Conclusion and recommendations Is area-based yield insurance applicable in Mozambique? The first constraint in developing area-based insurance is the correlation between individual farmers yield and aggregated yield (e.g. at the district level). If this correlation is too low, basis risk reduces contract performance dramatically. As historical production data at the household level are not available in Mozambique, estimating a contract s performance is impossible. The alternative is to develop area-based insurance at the meso or macro level using district and provincial production data. Second, cooperatives are less developed in Mozambique than in Mali where they act as intermediaries for input purchasing and cotton marketing. This marketing channel cannot be used in the same way as in Mali. 44

45 3.2. Recent experience with commodity price risk management Recently, many trading instruments (Table 7) have been developed to manage price volatility. They can be divided into two categories: i) instruments for managing intra-annual price risk (during one growing season) and ii) instruments for managing inter-annual price risk (between different growing seasons). In this section, the mechanisms of these instruments and their ability to reduce price volatility in the Mozambican context are discussed. Table 7: Overview of commodity price risk management instruments Instruments Designed to manage intra-annual price risk: 2.1 Forward contracts 2.2 Futures contracts 2.3 Option contracts Designed to manage inter-annual price risk: 2.4 Swaps 2.5 Commodity bonds and loans 2.6 Fonds de lissage Forward contracts Forward contracts are agreements between a buyer and a seller to purchase (long position) and to sell (short position), respectively, a specified amount of a commodity at a pre-set price on a fixed date. Payment occurs at maturity when physical delivery occurs. At maturity, if the actual price (spot price) is higher than the price agreed in the forward contract, the buyer makes a profit, and the seller suffers a corresponding loss. However, if the spot price is lower than in the forward contract, the buyer loses, and the seller profits. Forward contracts have two important features. First, no cash transfer occurs when the contract is signed. The seller is constrained to deliver the commodity at maturity, but the buyer pays no money up front. Second, the unique guarantee that a forward contract will be honored is the parties reputation (UNCTAD, 1998). A major advantage of a forward contract is that it eliminates the risk of price changes for the buyer and the seller. In an agricultural context, a forward contract allows the producer to know the price before choosing the level of input intensity. Moreover, forward contracts guarantee delivery of the Description and characteristics Advantages 45

46 commodities to the buyer and a physical market to the seller for the commodities produced. Forward contracts also have two main disadvantages. First, the possibility of profiting from favorable spot market developments is lost. Second, there is a major counterparty risk, if the counterpart does not honor its commitments profiting from favorable spot market development (UNCTAD, 1998). Forward contracts are relatively old price risk management instruments. Presently, most forward trade is OTC (over-the-counter), with transactions made directly or through brokers. Forward contracts are widely used for all commodities and in all regions. For instance, a large part of the world s cotton is traded through three- to 12-month forward contracts (UNCTAD, 1998). Forward contracts must be signed between two players along the supply chain. In the Mozambican cotton sector, there are two possibilities: i) between cotton farmers and ginning companies or ii) between ginning companies and international cotton trading/processing companies. At the first transaction step, a ginning company should establish a contract with cotton-producing smallholders. Due to the large number of smallholders, this makes the operation very costly. Additionally, due to the concessionary system, ginning companies already have a guarantee that farmers will deliver their production at harvest. In addition, the minimum price is fixed shortly before the harvest. Consequently, in this context, forward contracts with farmers are unattractive for ginning companies. On the contrary, ginning companies can hedge price risk during the processing period. At harvest, they can negotiate forward contracts with international trading/processing companies to sell cotton lint at a predetermined price and on a specified date. Unfortunately, with this solution, farmers remain exposed to price risk during the growing season. A solution for protecting farmers would be to shift the minimum price definition from the end to the beginning of the growing season. Thus, price risk will be borne by ginning companies during the whole season from the sowing period to the moment when they sell the cotton lint on international markets. At sowing time, the ginning companies can subscribe to forward Disadvantages Use and experiences Cotton forward contracts in Mozambican context Solution proposed 46

47 contracts at a fixed price on the date when the companies sell the cotton lint. The weakness of this coverage scheme is that ginning companies bear production risk. Indeed, they must estimate the cottonseed and lint production at the beginning of the season to establish forward contracts. If the production is lower than estimated, the ginning companies must buy the cotton lint at the spot price to honor their forward contracts and vice versa. However, if weather conditions explain the variability in yield and production, weather index products can be used to address this issue. A second problem is that the minimum price for cottonseed is fixed in New Mozambican meticals (MZN) and the price for cotton lint is fixed in US dollars (US$). With this scheme, ginners bear also currency risk. This problem can be solved with the assistance of option or futures contracts. For example, we can take a simplified situation where production can be estimated and the currency exchange MZN/US$ is stable. In August 2010, Cotton Index A was traded at US$3 300/tonne. The IAM fixed the national minimum price for cottonseed at MZN /tonne (US$520) for the 2010/11 growing season. Simultaneously, cotton companies estimated their production quantities and negotiated forward contracts with international processing/trading companies at an agreed price of US$3 300/tonne with delivery at the end of the ginning period, e.g. in October 2011 or another fixed date. This strategy allows ginneries to know the exact price before making decisions about production amount and input use, regardless of spot market developments. Example Futures contracts The mechanism of a futures contract is similar to that of a forward contract. A futures contract also an agreement to purchase (long position) or sell (short position) a predetermined quantity of a commodity at a preset price, with the transaction expected to occur at a future date. Nevertheless, there are some differences between futures and forward contracts (UNCTAD, 1998): Description and characteristics 1. Futures trade happens on organized exchanges through exchange institutions whereas most forward contracts are traded OTC. 2. Futures contracts have standardized contract terms (quantity, delivery term etc.) while forward contracts can be tailor-made to 47

48 match specific hedging needs. 3. Futures contracts require initial cash transfers for margin payments and may require daily settlements to adjust margins to adverse price movements. Forward contracts require cash payments only at maturity. 4. Futures contracts imply very little counterparty risk because the exchange institution guarantees the fulfillment of the contractual obligations. Forward contracts may involve a high degree of counterparty risk. 5. Futures transactions result normally not in actual delivery of the underlying commodity; physical delivery is possible but not expected. Forward contracts include the expectation of physical delivery. The goal of the exchange institution the so-called clearinghouse is to minimize the risk of default by either party. Thus, the exchange requires both parties to put up an initial amount of cash, the so-called margin. Since the futures price changes daily following spot price fluctuations, the difference between the prior agreed prices and the daily futures prices is also settled daily. The exchange institution draws money out of one party s margin account and puts it into the other s so that each party has the appropriate daily loss or profit. On the delivery date, the amount exchanged is not the price specified in the contract but the spot value; any gain or loss has been previously settled (Redhead, 1996). Futures cumulate the advantages of forward contracts (elimination of price risk and guarantee of delivery of the commodity needed and guarantee of a market for the commodity produced [spot market]). In addition, there is no need to negotiate contract specifications since the contracts are standardized. Profits and losses are paid in real time; counterparty risk is minimal. Moreover, the initial position can be easily reversed by buying a short or long position as the futures trader does not need to possess the underlying commodity. Finally, physical delivery is not necessarily implied; futures traders are not necessarily supply chain players. Advantages 48

49 On the other hand, all eventual losses must be paid with the margin and in real time; working capital is frozen in the margin. Furthermore, similarly to forward contracts, the possibility of profiting from favorable spot market developments is lost. In the end, the prices of the hedged product and the futures contract may diverge, and thus, in particular cases, futures contracts may not provide a perfect coverage against spot market price fluctuations. The two main locations where arbitrage occurs for agricultural commodity futures markets are in Chicago. The Chicago Board of Trade (CBOT) is where agricultural commodities corn, soybean, soybean oil, soybean meal and wheat futures are traded. The Chicago Mercantile Exchange (CME) is where the agricultural commodities lean hogs, live cattle, stocker cattle and feeder cattle are traded. Cotton futures are mainly traded at the New York Board of Trade (NYBOT). In 2004, two Indian exchanges (National Commodity and Derivatives Exchange (NCDE) and Multi Commodity Exchange [MCE]) and a Chinese exchange (Zhengzhou Commodity Exchange [ZCE]) began to trade futures on cotton. Futures contracts can also be used as forward contracts to hedge against international price volatility by ginning companies to obtain a guaranteed cotton lint selling price. The only difference is that they must engage capital to assume margins. Hedging on the futures market can lower the ginning companies exposure to counterparty risk compared with forward contracts. Because being part of the supply chain is not necessary to trade futures contracts and the government (the IAM) is active in national price fixing and applies export tax, the IAM can play a key role in price risk management strategies. For example, to protect farmers from price volatility, the government can set the minimum price before the sowing period starts. Simultaneously, the government can guarantee a selling cotton lint price to the ginning companies for the end of the ginning period. At the same time, the IAM could buy short positions on the futures market. At the end of the ginning period, when ginning companies sell their product on international markets, and if the price rises against the beginning of the growing season, the government suffers a loss on its futures contract. The government can impose an export tax, so that the Disadvantages Use and experiences Cotton futures contracts in Mozambican context Solution proposed 49

50 final price for the ginners corresponds to the guaranteed price. With this tax, the government can reimburse the loss on the futures contract. Inversely if the price falls during the growing season, the government can use its profits to subsidize cotton lint exports so that the ginners get the guaranteed price. The advantages of this solution are that all farmers can benefit from the earlier price set and that the actual institutional framework is suitable for the application of this compulsory system. For example, we can take the same simplified situation as above and the hypothesis that the price of futures contracts is perfectly correlated with the spot price. In August 2010, the Cotton A Index is traded at US$3 300/tonne. The government sets the national minimum price for cottonseed at MZN /tonne (US$520) for the 2010/11 growing season. Simultaneously, the IAM guarantees the ginners a selling price of US$3 300/tonne cotton lint in October 2011 and estimates national production at tonnes. The IAM takes a short position on the cotton futures market for US$ and pays a margin of US$ (10 percent of the futures contract value). At the end of the ginning period, there are three possible scenarios: Example 1. The price is lower than US$3 300/tonne: e.g. US$2 000/tonne ( 39 percent). In October 2011, ginning companies sell their production at the international spot price of US$ The government realizes a profit of US$ on its short position (+39 percent). This money can be used to subsidize ginning companies with US$1 300/tonne cotton lint. In total, the ginning companies get the guaranteed price of US$3 300/tonne cotton lint. A second possibility is that the government buys the cotton lint from the ginning companies at a price of US$3 300 and resells it on the international market at US$ The loss made in this operation will compensate the profit from a short position on the futures market. 2. The price is still US$3 300/tonne. In October 2011, ginning companies sell their product at the international price of US$ The IAM gets its margin back and realizes neither profits nor losses on its short positions. 3. The price is higher than US$3 300/tonne: e.g. US$4 000/tonne (+21 percent). 50

51 In October 2011, ginning companies sell their product at the international spot price of US$ The IAM loses its margin and suffers a supplementary loss of US$ on its short positions (total loss of US$ [ 21 percent]). The IAM must pay this loss in real time. The Exchange will call for a supplementary margin when the initial margin is not sufficient to cover the losses. The IAM must pay this loss and deduct a tax of US$700/tonne cotton lint. In total, the ginners get the guaranteed price of US$3 300/tonne cotton lint. As we can see, this strategy needs much working capital, but the government can assume the price risk and shift it to international markets. In the United States, countercyclical payments (CCPs) provide benefits to producers with eligible historical production whenever the effective price for a commodity is less than the target price. Thus, when market prices fall, the payments increase. The main problem in this strategy is that production should be known at the beginning of the growing season. Without this information, it is not possible to hedge efficiently against price risk. In Mozambique, national production can fluctuate widely from year to year, limiting the efficiency of this strategy. Option contracts Option contracts are another type of hedging product. In an option contract, the counterpart has the right but not the obligation to purchase or sell a specified amount of a commodity. For example, if one buys an option contract to have the right to sell (the put option) a specified amount of a commodity on a future date at a predetermined price (the strike price) and the spot price rises, one has the right to sell the commodity on the spot market, profiting from the higher price than on the contract. Inversely, if a trader buys an option contract to have the right to purchase (the call option) a specified amount of a commodity at a predetermined price, he or she can profit from the lower price if the spot price falls. To achieve this right, a premium must be paid. The premium is determined by the difference between the current spot price and the strike (the predetermined price), the length of time the price protection is needed and the volatility of the market. As opposed to the margin of a futures contract, this premium is always lost. Description and characteristics 51

52 The main advantage of an option contract is that the hedger can always take advantage of positive spot market developments. Additionally, capital needs are known at the beginning and will not change during the season; the costs of protection are known up front. Similarly to futures contracts, option contracts are available in standardized form on exchanges and tailor-made for over-the counter transactions. Tailor made vs. standardized Tailor-made instruments are developed for a special need (specific expiry date, quantities, strike price). Standardized instruments have a standard expiry date, quantity and strike price). Over-the-counter vs. exchanges Over-the-counter trading occurs directly between two parties. It is different from exchange trading, which occurs through specific frameworks facilities for trading (i.e. exchanges). Advantages In times of volatile prices, premiums can be expensive. The option contract value is correlated not only with the price change but also with the volatility of the underlying commodity price. The financial institutions that sell option contracts are highly exposed to risks as these institutions need to pay margin calls. Most of the option trade on agricultural commodity exchanges is the result of speculative activity. One example where options are directly traded by farmers to hedge against price fluctuations is Mexico. The Mexican Government decided to set up an intermediary organization to allow farmers and processors to buy agricultural options. In developing countries, however, participation in option markets is fairly limited. In September 2005, the Malawian Government concluded an option contract giving the government the right, but not the obligation, to buy additional maize at a price fixed at the time the contract was signed. The contract allowed the government to purchase a maximum of tons of maize at a cost of approximately US$ The UK Department for International Development (DFID) provided the financing to pay the option premium up front, and the World Bank provided technical support. This was an over-the-counter contract on maize from South Africa. Delivery Disadvantages Use and experiences 52

53 costs are included in the contract. This contract contributed to reduced uncertainty over transport costs. Reducing uncertainty over transport costs reduces the basis risk that exchange prices and local prices are not correlated. This strategy had the advantage of showing the private sector how the Malawian Government would proceed when a food shortage and a period of high maize prices occur. This reduced the negative impact of food aid distribution on local and regional trade. Option contracts can also be used in the strategy described above substituting forward or futures contracts. Option contracts reduce the counterparty risk and the uncertainty about working capital needed for futures contracts. As in the Malawian example, the Mozambican Government could buy put option contracts and secure cotton lint export prices for a predetermined quantity. If the cotton lint spot price falls during the season, the option can give to the ginning companies the opportunity to sell a part of their production at the strike price. The government, using its rights to sell at the strike price, resells the product directly on the international market. Premium financing can come from international donors or directly from the cotton private sector by slightly increasing the cotton lint export tax (compulsory coverage) or by providing the opportunity to ginners to participate in the premium financing if they want to profit from the secured price. The national production is estimated at between and tons of cotton lint. In August 2010, the Cotton A Index traded at US$3 300/tonne. The government decided to buy put option contracts for tons at a predetermined price of US$2 500/tonne. The total premium cost was US$ At the end of the ginning period (October 2011), the national cotton lint production was tonnes. Concerning the price evolution, there are two possible scenarios: Cotton option contracts in Mozambican context Solution proposed Example 1. At the end of the ginning period, the price is higher than US$2 500 (e.g. US$3 000). It is more profitable to sell cotton lint on the spot market than to use the right to sell at US$ The coverage is not used, and the premium is lost. 53

54 2. At the end of the ginning period, the price is lower than US$2 500 (e.g. US$2 000). The government has the right to sell tons at US$ The government buys and resells two-thirds of the production of all ginning companies at a price of US$ The rest of the production is sold on the spot market. In this way, ginners had an average price of US$2 330, and they made a profit of US$ compared with the situation without an option contract. The premium had to be paid up front; thus, the total benefit from this situation is US$ Swaps The advantages of swaps are similar to forward contracts. The difference lies in the mechanism. Although a forward contract is a direct agreement between the seller and the buyer, a financial institution plays the role of intermediary in the case of swaps. In addition, swaps are purely financial instruments. Sellers and buyers continue to deal on the spot market; they agree only to exchange a stream of cash flow at fixed intervals. In swaps, the seller (or producer) agrees with the financial institution on a fixed price, a price index and a quantity. When the producer operates on the market, the financial institution pays or collects the difference between the fixed price and the price index. If the price index is lower than the fixed price, the financial institution pays the price difference to the producer. If the price index is higher than the fixed price, the producer pays the difference to the financial institution. The financial institution purchases a similar and inverse contract with a buyer (or consumer). Usually, swaps are middle- or longterm arrangements. The first advantage of swaps is to lock in long-term prices and thus secure investments by securing future cash flows. The difference with futures is the absence of a margin call as a counterpart the financial institution is known. Similar to forward contracts, because of the lack of guarantee, there is a high counterpart risk. Moreover, positions are difficult to revert when an arrangement is made. However, in the case of an over-the-counter contract, design and set-up costs are high, and it is difficult to assess the fair price Description and characteristics Advantages Disadvantages 54

55 for the deal. Actually, the commodity swap market is very small compared to the interest-rate and currency swap market. Moreover, commodity swaps are largely oil swaps. For agricultural goods, liquid and well-established futures markets limit the need for swaps. Swaps can be used in the same way as forward contracts between two actors of the supply chain. The Mozambican framework makes establishing such contracts between farmers and ginners very difficult as a contract has to be established between each producer and the financial institution. However, swaps can be subscribed between ginners and international cotton trading/processing companies to lock in the price. Swaps, similar to futures and options, can also be used to hedge against currency risk. Use and experiences Swaps in Mozambican context Commodity bonds and loans Commodity loans and bonds tie the repayment of a loan or bond to commodity prices using the mechanisms described above. For example, commodity-linked loans tie the interest and/or principal payment to the price of a commodity. The loan is reimbursed with the value equivalents (using a reference price) of fixed amounts of a commodity. Regarding commodity bonds, they can be divided into two categories: i) forward-type bonds and ii) option-type bonds. In forward-type bonds, the principal and/or the interest is linked to a commodity price or price index. In option-type bonds, the holder of the bond has the right to buy or sell a commodity at a predetermined price, in addition to his or her conventional bond. In risk-hedging strategies, forward-type commodity bonds are often issued by producers. Option-type bonds are mainly used to reduce the cost of financing. Commodity bonds and loans can act as a vehicle to obtain access to capital at more accessible terms. These tools hedge price risks and simultaneously raise investment financing. Because there is no initial transfer by the issuer, working capital is not necessary to subscribe a commodity-linked loan. Thus, repayment is directly linked to the borrower s risk, i.e. the commodity price. This loan reduces the lender s risk exposure and can be used as a long-term risk management instrument. Description and characteristics Advantages 55

56 Due to set-up costs underlying the issuance, a high volume is required to reach sustainability of the product. Thus, a fairly sophisticated distribution network is needed to market this tool. Most commodity loans and bonds issued so far have been linked to precious metals and fuels, but in the agricultural sector, some are issued for coffee and cocoa (Priovolos, 1991; Rutten and Youssef, 2007). In the mid- 1980s, a Zambian cotton grower expanded its operations with a loan for which the interest rate was linked to the international prices of cotton (Glen, 1999). Commodity loans and bonds cannot overcome the problem of the lack of a legal framework to guarantee loan repayments and the high costs of reaching numerous individual farmers as limiting factors for developing rural credit in Mozambique. However, these tools can reduce the risk exposure of the lender, overcome the problem of the lack of collateral and enhance credit access in rural areas. Microcredit institutions can use commodity-linked loans to reduce risk exposure. A variable interest rate can provide a natural hedge if the rate is linked to price developments. A bank can offer loans to farmers or farmer associations and tie the interest rate and/or repayment conditions to a cotton price index. If the price rises, cotton farmers have to pay a higher interest rate and inversely a lower interest rate if the price falls. The following example will attempt to illustrate the product. A cotton farmer association wants to purchase inputs for its members. The bank doubts that the association can reimburse the bank if the price of cotton falls. The price of cotton is volatile; thus, there is a real risk that the association s income will not be sufficient to repay the loan. It would be logical for the bank to protect its loan with the help of risk management instruments. The subscription of a forward contract is excluded because there is a high risk that the farmers will side-sell their production if the price rises. The bank decides to include an options contract in the agreement that would protect the association if the price of cotton falls and would allow the farmers to take advantage of a possible price increase. An options contract is costly, but without it, the association would not obtain the loan, and its members Disadvantages Use and experiences Commodity loans and bonds in Mozambique Solution proposed Example 56

57 would not benefit from a minimum price. Fonds de Lissage / Stabilization fund The Fonds de Lissage is an instrument developed to help manage cotton price volatility in Burkina Faso. Some fluctuation around the trend price is tolerated, and an intervention takes place only if the price falls below the bottom price or rises above the ceiling price. The trend price is calculated with the aid of the effective price of the two last growing seasons and the projected price of the three coming seasons. The Cotton A Index Far East FOB can be used. The projected price for the next season can be calculated from the Cotton A Index Forward and, for the following season, from the projected prices published by the World Bank. The trend price must be corrected with the currency exchange rate. The upper and lower bounds (ceiling and bottom prices) are directly calculated from the trend price. For example, the upper bound would be 110 percent of the trend price and the lower bound 95 percent. Trend, ceiling and bottom prices are calculated and published before the beginning of the growing season. From this point, the bottom price can be seen as a guaranteed price for the cotton growers. In compensation, producers and ginners commit themselves to commit a part of their profit if the price rises above the ceiling price. The Fonds de Lissage provides security to farmers and ginners by setting the bottom price before the growing season starts. Thus, this instrument can serve as an intermediate step to shift the risk from the producer and ginners to the international markets. The Fonds de Lissage is not an instrument for reducing price risk. A strategy for shifting the risk from the Fonds to international markets using the instruments described above should be developed. Since many variables must be taken into account (production volume, exchange rate etc.), it is not possible to affirm that the Fonds is definitively sustainable. Additionally, this mechanism leads habitually to overproduction and, applied on a large scale, destabilizes the market. This is why this mechanism has been abolished in many developed countries. However, on a small scale, this mechanism is completely viable. The Agence Française pour le Développement (AFD) introduced this tool in 2008 in Burkina Faso. The AFD provided technical assistance and Description and characteristics Advantages Disadvantages Use and 57

58 allocated a loan of in 2007 to launch the Fonds and to create a reserve. Although it seems to show positive results and all the actors along the supply chain are involved in the Fonds, it is too early to deduce some lessons. Thus, the association managing the Fonds is actually introducing a market-based mechanism to shift the risk to international markets. The Fonds de Lissage could be implemented in Mozambique. Five conditions must be met: i) the price-setting mechanism must be indexed on the price trend on international markets; ii) the determination of the upper and lower bounds, indemnities to producers and ginners should be based on a verifiable criterion and calculated with the help of a predetermined mathematic formula; iii) this mechanism should be compatible with other price risk management instruments to guarantee its sustainability; iv) the management of the Fonds should be delegated to a third party (bank) that supervises the different actors in honoring the agreements; and v) the whole supply chain should be involved and the mechanism should be described in detail in a rule book. experiences Fonds de Lissage in Mozambique 3.3. Toward an integrated risk management strategy The instruments described above can be used alone or in combination. Bringing all discussed aspects together, this chapter aims to select the most appropriate instruments and describe how they can be combined with each offer according to the Mozambican context. The first selection has to be made between production risk and the price risk management instruments. The three most important criteria are the price variability, the yield variability farmers face and the correlation between prices and yields (Figure 16) (Meuwissen et al. 1999). In the Mozambican cotton sector, farmers face high yield variability as well as high price variability. Moreover, the cotton price at the farmgate and national yield do not show any remarkable correlation (r = 0.08). 58

59 Figure 16: Scheme to determine the need for insurance, based on price variability, yield variability and price/yield correlation. Source: Meuwissen et al., Production risk management Traditional crop insurance is not considered in this report. Many countries, such as Malawi, have experimented with traditional crop insurance without success. In general, traditional crop insurance is not viable multi-peril crop insurance, which covers many production risks and adjusts payouts on individual farm losses, is plagued by moral hazards, adverse selection due to asymmetric information, and high monitoring, administrative and transaction costs. Many studies have underlined the limitations of traditional crop insurance: i) It depends on subsidies; ii) it tends to distort incentives; iii) it can be inequitable as large farmers tend to absorb most of the subsidies; and iv) it requires high levels of expertise in loss adjustment (Hess and Syroka, 2004). Therefore, traditional crop insurance is not considered in this report. To manage production risk, the two remaining instruments are selected and analyzed. Insurance schemes are discussed in the next section. Price risk management Due to the large number of smallholders, the use of price risk management tools by farmers and ginners will engender high and prohibitive administrative and transaction costs. Thus, these instruments are being considered for ginners and international trading/processing companies. Option contracts and Fonds de Lissage are being considered in this report. 59

60 Option contracts appear particularly suitable for managing intra-annual price variability because they give the possibility of benefiting from positive price development and because the premium liquidity needs are known up front. The Fonds de Lissage appears particularly indicated to manage inter-annual price variability as it can offer a minimum price to farmers at the beginning of the season; this mechanism tolerates some of the price volatility and engenders any counterpart risks such as swap contracts or commodity loans and bonds. 60

61 4. Quantitative analysis 4.1. Production risk management: weather index Methodology of instruments design According to the US National Drought Mitigation Center, drought can be analyzed from four different perspectives: meteorological, hydrological, agricultural and socioeconomic. These perspectives are defined as follows: - Meteorological drought is usually defined in terms of deviations of precipitation from normal levels and the duration of dry periods in a region. - Hydrological drought takes surface and subsurface water into account. This type of drought deals with stream flow, reservoir levels and groundwater. Hydrological drought is induced by an extended period experiencing a lack of precipitation. Since lack of precipitation requires longer periods to affect these water supplies, hydrological droughts usually lag behind meteorological droughts. - Agricultural drought refers to a situation where the soil moisture is no longer sufficient to meet the needs of crops due to the lack of precipitation or other adverse events. The different growing phases and their specific water requirements are taken into account. - Socioeconomic drought associates the supply and demand of some economic goods with elements of meteorological, hydrological and agricultural drought. Socioeconomic drought refers to a situation that occurs when the water shortage begins to affect people and their quality of life. The three first types of drought can be modeled with the help of precipitation data. For each study district, meteorological, hydrological and agricultural indicators were computed to establish the link between weather and yields. The deviation of precipitation from the long-term average was computed for each decade and each month of the year. These meteorological indicators were then used to explain What are the potential evapotranspiration (PET) and the actual evapotranspiration (AET)? The PET is the amount of evaporation that would occur if a sufficient water source were available. The AET is the sum of evaporation and plant transpiration at a definite time. If the AET is considered the net result of atmospheric demand for moisture from a surface and the ability of the surface to supply moisture, then the PET is a measure of the demand side. 61

62 yield variability. Second, hydrological indicators have been computed in a Soil Available Water Model. In this model, in addition to rainfall (RF), the potential evapotranspiration (PET), the soil available water (SAW) and the actual evapotranspiration (AET) are considered. In agreement with Mather et al. (2008), the PET was assumed to be constant at 5 mm/day. Thus, the PET for a decade is 5 mm/day * 10 days = 50 mm. 5, Where d is the length of the considered period. Soil available water for the period t is given by: Where the SAW value is not allowed to exceed 200 mm. Actual evapotranspiration is computed as follows: For AET t <=PET = 5 mm/d The drought stress (DS) is then determined for each day in the following way: 1 Third, the specific water requirement and physiology of cotton during the different growing phases have been considered. This water requirement (WR) is computed as follows: where 0<=Kc t =>1 The parameter Kc t was obtained based on expert interviews, agrometeorological studies and models such as the FAO water requirement satisfaction index. The crop drought stress (CDS), an agricultural drought indicator, was then obtained for each day: 1 In the Soil Available Model, the start of the growing season is determined when the SAW reaches a critical threshold (60 mm). In addition to meteorological indicators, the sum of days of drought stress (DDS) and days of crop drought stress (DCDS) in a given growing phase are used to explain yield variability. In the following phase, the statistical significance of these relationships is explored. 62

63 Weather-based indices A regression analysis was conducted by employing cumulated rainfall in several different periods as an explanatory variable for yield variability. However, it was not possible in any study districts to find a meteorological indicator that would explain a significant part of yield variation. For example, cumulated rainfall during the flowering period (January-April) has no significant influence on yield levels (Figure 17). Montepuez Namuno Detrended yield [kg/ha] Detrended yield [kg/ha] Accumulated Rainfall (jan-apr) Accumulated Rainfall (jan-apr) Figure 17: Detrended yield vs. cumulated rainfall (Jan-Apr) in the Montepuez and Ribaue districts. Further, hydrological indicators were formed obtaining the total of days of drought stress during the different periods of the growing season. The dependence of yield on this sum, during the different periods, was analyzed (Figure 18). A quite high dependence between the number of days of drought stress during the five first months of the growing season in Montepuez and in Namuno was found (both are located in Cabo Delgado). Montepuez Namuno Detrended yield Detrended yield Days of drought stress (5 months) Days of drought stress (5 months) Figure 18: Detrended yield vs. days of drought stress (5 months) in the Montepuez and Namuno districts. The low and high number of days of drought has an impact on yield. Before the statistical results are interpreted, the model has to be refined. By introducing the water requirement for 63

64 cotton, the agricultural indicators can be computed (Figure 19). Agricultural indicators allow weighting drought stress in a function of the different growing phases of the crop. Montepuez Namuno Detrended yield Detrended yield Days of drought stress (5 months) Days of drought stress (5 months) Figure 19: Detrended yield vs. days of crop drought stress (during the whole season) in the Montepuez and Namuno districts. Considering specific water requirements allows two trends to be observed (Figure 20). First, if the number of DCDS is lower than 70, yield is positively correlated with the number of DCDS (regression coefficient estimate +37.9). Inversely, if the number of DCDS is higher than 70, yield begins to be negatively correlated with the number of DCDS (coefficient estimate 15.8).Optimal yields were realized in the season with DCDS. The model accounts for 71.2 percent of the variability in yields and can be used to estimate yield losses. Montepuez Detrended yield Days of crop drought stress Figure 20: Detrended yield vs. DCDS in the Montepuez district with statistical dependence. Insurance product Since a good estimate for losses could be established for the Montepuez district, a pilot insurance scheme can be developed for this region. In the last 20 years ( ), the detrended yield averaged 545 kg/ha. If we assume that the scheme should pay out in years when the yield falls under the long-term average, the thresholds can be fixed at 62 DCDS 64

65 and 85 DCDS. If the DCDS do not reach the minimum level of 62, payouts are triggered. Payouts depend on the number of missing DCDS to reach the minimum level; each missing day counted of drought stress triggers a payout of 38 kg of cottonseed. Inversely, if the number of DCDS exceeds the maximum level of 85, each supplementary day counted of drought stress triggers a payout of 16 kg cottonseed (Figure 21). The maximum payout of 545 kg/ha occurs if the sum of DCDS does not reach 48 days or overcomes 120 days. Based on historical data, the fair premium of such a scheme would amount to 85 kg cottonseed/insured ha (15.6 percent of average production value). With such a premium, the loss ratio would have been 1 in the past 17 years. In a fair premium, the risk premium and administrative and transaction costs are not taken into account. Most payouts would have been triggered when rainfall was excessive. Payout [in kg cotton seed/ha] Days of crop drought stress Figure 21: Payout structure of the proposed weather index based insurance scheme. Since this scheme provides protection against excessive and limited rainfall, developing a scheme for a specific peril only could be useful (Figure 22). That way, an index-based drought insurance scheme can be proposed. Payouts are triggered according to the same rule as above but only if the sum of the DCDS is above the strike level. Based on historical data, the fair premium of such a scheme would amount to 9 kg cottonseed/insured ha (1.7 percent of the average production value), and payouts would have been triggered twice in the past 17 years. 65

66 Payout [in kg cotton seed/ha] Days of crop drought stress Figure 22: Payout structure of the proposed drought index based insurance scheme. Viability and value of the weather insurance scheme Due to the lack of historical yield data at the household level, assessing the basis risk and the potential benefit of the insurance scheme at this level is not possible. Hence, district data were used to assess the basis risk and the reduction in income volatility. The DCDS index explains 71 percent of the total amount of yield variability; the remaining variability constitutes the basis risk (Figure 23).The goal of a yield insurance product is to reduce the income variability of the insured. Thanks to the proposed scheme, it is possible to reduce the variance of income by 45 percent and the coefficient of variation by 26 percent (Figure 24). To carry out these calculations, it was assumed that the price at the farmgate was constant at US$0.195/kg for cottonseed ( average) Index Yield Estimate Historical yield 700 Yield [kg/ha] Harvest Year Figure 23: Yield vs. weather index yield estimate. 66

67 Income without insurance Income with insurance Income [USD/ha] Harvest Year Figure 24: Income with vs. without weather index insurance. Note: empty points in the yield estimates time series correspond to the years for which the weather data were not available. 67

68 4.2. Production risk management: area-based yield index Design methodology As seen previously, in Cabo Delgado, yields at the district level are strongly correlated with the yield at the provincial level. Therefore, provincial yield variability can be used to explain yield variability in different districts. Area-based yield index Regression analysis was conducted to bring to light an eventual relationship between yields at different aggregation levels. Once that yield is detrended, yields in Montepuez are only poorly related to those of the province. On the contrary, yields recorded in Namuno are strongly linked to those of the province (Figure 25). Provincial yield variability can explain more than 80 percent of the yield variability at the district level. Namuno Yield Namuno [kg/ha] Yield Cabo Delgado [kg/ha] Figure 25: Yield in Namuno (district) vs. yield in Cabo Delgado (province). Insurance product If we assume that the scheme should trigger payouts in the years when the Namuno district yield falls under its long-term average (474 kg/ha), the corresponding threshold for the province yield can be fixed at 509 kg/ha. That way, if the provincial yield falls under this threshold, it would trigger payouts. For each kilogram under this level, the scheme would pay out 0.87 kg cottonseed/ha. For example, if the provincial yield is 400 kg/ha (109 kg below the threshold level), 94.8 kg cottonseed are paid out per ha (109*0.87) for the Namuno district (Figure 26). For this scheme, the fair premium would amount to 62 kg cottonseed/ha (13.1 percent of the average production value). 68

69 Payout [in kg cotton seed/ha] Yield Cabo Delgado [kg/ha] Figure 26: Payout structure of the proposed area-based yield insurance scheme (Namuno district). Viability and value Here again, the lack of data at the farm level impedes the basis risk analysis in the case where the analysis is applied to smallholder farmers. Indeed, if yields at the farm level are strongly correlated with aggregated yields at the district or provincial level, the area yield turns out to be a powerful instrument in risk management. Because of the high correlation between yields at different levels of aggregation, the scheme described above is very efficient (Figure 27). It allows the variance of income to be reduced by 71 percent and the coefficient of variation by 46 percent in the district yield for Namuno (Figure 28). Nevertheless, four districts produce more than 90 percent of the cottonseed in Cabo Delgado. Hence, it is not surprising that district yields are strongly correlated with provincial yields Index Yield Estimate Historical Yield Yield [kg/ha] Harvest Year Figure 27: Yield vs. area-based yield estimate. 69

70 Income without insurance Income with insurance Income [USD/ha] Harvest Year Figure 28: Income with vs. without area-based yield insurance Price risk management: option contracts Since in the current situation the minimum price is set at the end of the growing season, farmers bear cottonseed price risk from the point when they sow to the moment the price is set, i.e. they have to make their decisions under price uncertainty. Once the minimum price is fixed, the risk shifts to ginneries, which bear it until they sell the cotton lint they produced on international markets. Because of small size and poor access to financial services, farmers have no tools to cope with price risk. In contrast, ginneries have access to financial instruments, and some already apply hedging strategies. Due to the ability to hedge against price risk, ginneries are less vulnerable to price volatility than farmers. The proposed solution is to shift price risk from farmers to ginneries. It would be enough to fix the minimum price before the sowing period starts instead of the end. Ginneries would bear price risk during the whole season but could buy put option contracts with a strike value equivalent to the cotton lint price on international markets observed at the beginning of the season. Thanks to options and premium payments, ginneries have the right but not the obligation, during the whole season, to sell cotton lint at the strike price. Cotton option contracts are traded in the United States (NYCBOT), India (NCDE and MCE) and China (ZCE), and everybody is free to trade them. Because of the possibility of taking profit from positive price development and the low counterpart risk, option contracts seem to be the more adapted tool to hedge against price risk. However, the Mozambican context put some constraints on the use of option contracts. 70

71 Example In this example, the exchange rate is considered constant. In October 2009, the IAM fixed the minimum price at 8.1 MZN/kg for the 2010 harvest year. At this time, cotton lint was traded at US$1.47/kg in international markets. A ginning company will sell its production output only one year after the price setting and bears the risk of price volatility until this date. The company management decides to buy a put option, with a strike price of US$1.47/kg and valid for more than one year. The company pays the premium of US$0.15/kg. With this put option contract, the company can sell cotton lint at the minimum price of US$1.47/kg, no matter what the price in international markets at the moment of selling. If the price rises, the company generates an additional profit. Constraints The first constraint is linked to production uncertainty. Indeed, before purchasing an option at the beginning of the growing season, production has to be estimated. If the estimation is bad, there is a risk of hedging for cotton that will not be produced. As yield and production area fluctuate widely in Mozambique, a precise estimation of production is not possible. The second restriction is linked to access. Before trading option contracts, the trader must have a perfect understanding of the underlying mechanism. It should not be a problem for ginneries that belong to an international group and that have fairly good knowledge of the world market as well as much experience in fiber marketing (Plexus, Olam, CNA and Dunavant). In contrast, smaller ginneries need capacity building before entering option markets. The minimum quantity to trade on an exchange is one option, i.e tons cotton lint. The third restriction is the premium. The premium depends on the level of protection (the strike price and price volatility). If the market is volatile and the strike price is high, the option premium will be high. As example, in December 2010 the cotton lint price was US$1.91/kg. The premium to have the right to sell cotton in one year at the same price was US$0.196/kg, i.e percent of the strike price. 71

72 5. Delivery channels The distribution and marketing of index-based insurance products face several challenges. First, distribution costs have to be minimized. There are different options: i) establish partnerships with organizations that are already engaged in providing service in rural areas and thus are trusted by farmers or ii) link the product with another transaction. This strategy would be even more efficient if insurance could be linked with another financial transaction. Second, the timing of the premium payment should consider the cash flow formation at the farm level; in particular, the premium payment should take place during the cash inflow period. Third, due to the low awareness and literacy levels, training and information dissemination should be paid significant attention. Finally, telecommunication and other technologies should be used to reduce administrative costs. In the Mozambican cotton sector, the best-established transaction for smallholders is selling cotton to the concessionary ginner at the end of the growing season. At this time, a cash flow takes place, and farmers can finance the premium. The second most frequent transaction is acquiring input, primarily free seeds offered by ginners. Since ginners are the principal counterparty in both transactions, they seem to be an interesting partner to reach farmers (Figure 29). Figure 29: Transactions and main relationships within the cotton supply chain in Mozambique. Product delivery The insurance product could be bundled with input in an input package. By purchasing improved seeds, the farmer pays a premium for a weather index insurance product. If weather impedes the optimal development of the crop and the index threshold is reached, insurance payouts are triggered. It is also possible to sell Weather index or area-based yield insurance bundled with a input package 72

73 this package with credit to farmers. The credit repayment usually occurs at the end of the season. Eventual insurance payouts are used to repay part or the whole loan. With this solution, two major problems must be anticipated. First, farmers obtaining a loan to purchase seeds can sell their cotton to another party (side-selling) and default on their loan independent of weather risks. Second, the seed market is biased in Mozambique. The input package will compete with free seeds. Bank and microfinance institutions are also developing activities in rural areas. A strategy for securing loans with index insurance would allow financial institutions to reduce the risk of default in the case of adverse weather events. The financial institution can pay the premium to the insurer when the institution grants a loan to a farmer. If the index threshold is met, the insurer pays out part or the whole loan to the bank. However, weather insurance enhances only existing agricultural supply chains and businesses and does not create them. This strategy can be applied only if the bank is interested in lending in the agricultural sector. Actually, most cotton farmers do not have access to credit. They obtain low-quality input from the ginneries. Insurance can also be marketed as a standalone product. To reach an individual farmer, we can follow the example of local telecommunication providers. Instead of contracting each client, telecommunication providers engage local agents at low cost to sell telecommunication coupons. A solution would be to sell coupons as a lottery: One can buy coupons at the beginning of the season and receive a payment if the weather variable exceeds the threshold. The experience in Malawi showed that standalone products (as opposed to credit-bundled products) had no takers. Thus, premiums will always have to be pre-financed through loans to achieve a sufficient willingness to pay by farmers. Henceforth, this solution seems to be limited by factors such as understanding, acceptance and trust. Weather index or are-based a yield insurance bundled with a loan Weather index or area-based yield insurance as standalone product 73

74 6. Conclusions Due to the high exposition and the low capacity to manage production and price risks, cotton growers in Mozambique experience a high level of vulnerability. Ex ante risk management strategies can help reduce farmers vulnerability Recommendations Production risk Based on the findings of this report and international experiences, the most appropriate scheme for smallholder cotton growers is an index insurance product bundled with a loan or an input package. This is the best way to take advantage of the actual structure of the cotton supply chain in Mozambique. As index, area-based and number of days of crop drought stress generate similar performance. The principal advantage of the weather index is the choice of which peril should be insured (rainfall deficit or excess). Market risk To shift the risk from vulnerable farmers to less vulnerable ginneries, price setting should take place before the growing season starts. That way, knowing the price in advance, farmers and the whole supply chain can fully profit from good years and limit their exposure in bad years. Option contracts are effective tools for shifting the risk from ginneries to international markets. Option contracts can also be used to hedge against exchange rate fluctuations. Here, the main limiting factors are production uncertainty and the education of ginneries managers Development of weather index insurance: next steps Some limiting factors have to be resolved before launching a pilot project (Table 8). These obstacles are based on and summarize international experiences and are described in detail below. 74

75 Table 8: Road map for resolving the existing gaps and implementing index insurance in the Mozambican cotton sector Prerequisite Object Situation met partially met not met Risk assessment Risk assessment conducted by the World Bank (2010) X Pilot area Montepuez, Ribaue (proximity of a weather station) X Target crop Cotton (economic importance) X Description of the production function Information about the typical farm Data availability and access (including phenological steps) Production function is available (including water requirement) X Size, farm structure, ownership, types of technology X Map of weather stations distribution Map available X 25 years of historical weather data (daily rainfall, temperature, evapotranspiration) Dense weather station network Monthly regional price at farmgate for the different qualities of cottonseed 25 years of production and yield statistics at aggregated levels (districts and provinces) Data available, evapotranspiration is missing Not necessary to conduct a pilot project but necessary to scale up Minimum prices fixed by the IAM Data available at the provincial and national levels, 15 years available for the main producing districts No farm level data, necessary to develop a contract trusted by producers and by the insurance sector X X X 25 years of production statistics at the farm level, including - Sown Area X - Harvested Area X - Production X Contract design and premium Contract designed using aggregated yield data X Viability and value of the contract The contract reduces the insured s income variability X Identification of the Financial institutions are the most appropriate clients of index insurance clients in Mozambique. X Ability to reach the Ginning companies are well integrated in the supply chain and can serve X X Procedure proposed and role of the Governement (Gov.) and the private sector (PS) Guarantee future access to meteorological data (Gov.), check the security of stations used (Gov.), measurement extension to evapotranspiration (Gov.) Increase the number of weather stations in areas with a high density of cotton producers (Gov.) Gather historical yield data at the farm level near existing weather stations and new weather stations (Gov./PS). 75

76 client Identification of all potential stakeholders in Mozambique Ability to provide education and training Functioning input and credit market Enrolment of a strong local insurer partner Identification of the insurer s risk taking capacity Clearance from the insurance regulator Overview of all initiatives in agricultural risk management in Mozambique Establishment of a coordination framework Source: Authors as a delivery channel for loans bundled with an insurance product. Insurers (insurance companies, insurance association); Reinsurer (reinsurance companies, hedge funds); Agribusinesses and financial partners (agricultural banks, rural service organizations, non governmental organizations [NGOs]; MFIs, input suppliers, agribusiness companies); Farmers (farmer associations, cooperatives); Government departments (meteorological service, insurance regulator, Ministry of Finance, Ministry of Agriculture, planning ministries, research and specialist institutes); Donors (technical assistance, financing key investments). X Actually no index insurance expertise among the actors X Credit and input market experience serious constraints X There is interest in index insurance among local insurers (Emose, Impar and Hollard Moçambique). One (Hollard) has international experience. It is important to find a strong local partner to overcome initial set up problems and barriers. The insurer partner should be capable of taking on the risk of the pilot project. There are actually no specific regulations for index insurance in Mozambique. X X X X X Provide training to all potential stakeholders in Mozambique (Gov.) Before making a link with a loan and/or a supply of improved seeds, development constraints in these markets should be addressed (PS). Locally involved insurers expressed the need to develop a reinsurance strategy. Drought risk is systemic and can affect large regions. Information asymmetry should be small enough to interest international reinsurance companies (PS). In all cases, it is necessary to make sure that the regulator has approved the product (Gov./PS). Map out the different efforts undertaken as contributions toward risk management in Mozambique, indicating the institutions involved; the aim of the exercise; the progress, achievements of the work and the need to follow up on the previous initiatives. (Source: TORs from IAM) (Gov.) Develop and propose a coordination framework, defining the potential role of each stakeholder, and characterization of its intervention in the process of developing a risk management strategy for the cotton sector (Source: TORs from IAM) (Gov.) 76

77 Basis risk There are only two districts (Ribaue und Montepuez) where adequate yield data and weather data are simultaneously available. Additionally, in each district, there is only one weather station. Moreover, the weather data contain some missing values. To develop a weatherbased insurance market, the weather station network must be extended to allow the measurement of rainfall and other critical weather indicators at locations near cotton growers. During this project, collecting yield historical data at the household level was not possible. The use of the farm data is however a prerequisite for developing an insurance scheme trusted by the insurance and reinsurance sector. Accordingly, an important task is to collect reliable historical data on yields at the household level located near existing weather stations. This can be achieved by encouraging ginning companies to transmit their data on individual farm production and yields. The second important task is to install new weather stations in areas with a high density of farmers and improve the collection of farmers yield data in these areas. That way, basis risk can be minimized, and trust in the insurance scheme can be built. Uniform production conditions Spatial uniform production conditions are necessary to guarantee the reliability of index insurance. Microclimates can dramatically increase basis risk. In Cabo Delgado, only one weather station records data that can be used for insurance purposes. Thus, the uniformity of weather patterns within the province cannot be assessed. In view of the correlation in yield data between the districts, we can assume that adverse weather events generally affect the whole province. Installing supplementary weather stations could provide more information about this issue in the future. In Nampula, on the contrary, district yields are poorly correlated with provincial yield. Either the production systems are heterogeneous or weather patterns are not uniform within this province or both. Education Educating the target population about insurance, its function and benefits is a key component in developing a sustainable risk management strategy. Education determines in large part farmers willingness to pay for an insurance product. Actually, the ability to provide education and training to farmers is limited by their low reachability and the lack of knowledge of the different actors along the supply chain. NGOs, the government and ginners can play a key role in educating farmers about the potential benefits of risk management strategies. Functioning markets Linking weather index insurance products with a financial or an input acquisition transaction will not resolve institutional limitations in rural financial markets and agricultural input/output markets. Before making a link with a loan and improved seeds, for example, development 77

78 constraints in these markets should be addressed. Under current conditions, competing with free, low-quality seed, an input-insurance package has hardly any opportunity to meet farmers demand. However, in transitioning to liberalized input markets, weather insurance can be a good tool for reducing farmers and financial institutions exposition to weather risks, and this improve credit access in rural areas. 78

79 References Books, reports and journal articles ASSOCIATION DU FONDS DE LISSAGE DU BURKINA FASO (2007). "Règlement du Fonds de Lissage". Burkina Faso BACHKE, M. E. (2009). "Are farmers organizations a good tool to improve small-scale farmers welfare?" II Conferencia do IESE Dinamicas da Pobreza e Padrões de Acumulação em Moçambique. BARNETT, B. J., C. B. BARRETT, ET AL. (2008). "Poverty Traps and Index-Based Risk Transfer Products." World Development36(10): BARRETT, C. B., B. J. BARNETT, ET AL. (2007). Poverty Traps and Climate Risk: Limitations and Opportunities of Index-Based Risk Financing. Paper prepared for the Policy Roundtable on Climate Risk, Poverty Traps and Index-Based Financing, hosted by the International Research Institute for Climate and Society. C. University. BENNETT, B., K. (2006). Introduction to Cotton Options. Texas Cooperative Extension, The Texas A&M University System. Options/IntroductionToCottonFutures_2006/IntroductionToCottonFutures_2006.pdf BINGEN, J., A. SERRANO, ET AL. (2003). "Linking farmers to markets: different approaches to human capital development." Food Policy28(4): BOKUSHEVA, R. (2004). Crop Insurance in Transition a Qualitative and Quantitative Assessment of Insurance Products. D. P. N. 76. Halle, Germany, Institute of Agricultural Development in Central and Eastern Europe BRYLA, E. AND J. SYROKA (2007). "Developing Index-based Insurance for Agriculture in Developing Countries. Sustainable Development Innovation Brief." Policy Integration and Analysis Branch of the Division for Sustainable Development(Issue 2). COLLIER, B., J. SKEES, ET AL. (2009). "Weather Index Insurance and Climate Change: Opportunities and Challenges in Lower Income Countries." Geneva Pap R I-Iss P34(3):

80 DUFLO, E., M. KREMER, ET AL. (2011). "Nudging Farmers to Use Fertilizer: Theory and Experimental Evidence from Kenya." The American Economic Review101(6): FAOSTAT (2011). World Production. FAO. Rome, GINÉ, X., L. MENAND, ET AL. (2010). A Case Study of the Indian Rainfall Index Insurance Market. Policy Research Working Paper 5459, World Bank. GLEN, J., D. (1999). How Firms in Developing Countries Manage Risk. GLOBALAGRISK (2009). Designing Agricultural Index Insurance in Developing Countries: A GlobalAgRisk Market Development Model Handbook for Policy and Decision Makers. Lexington, KY. GOODWIN, B. K. AND O. MAHUL (2004). "Risk Modeling Concepts Relating to the Design and Rating of Agricultural Insurance Contracts " World Bank Policy Research Working Paper3392. GUIRKINGER, C. (2011). Group insurance for cotton producers in Mali. FONDATION POUR LES ETUDES ET RECHERCHES SUR LE DEVELOPPEMENT INTERNATIONAL. University of Namur, Belgium. HAZELL, P., J. ANDERSON, ET AL. (2010). The Potential for Scale and Sustainability in Weather Index Insurance for Agriculture and Rural Livelihoods. Rome, International Fund for Agricultural Development. HAZELL, P. B. R. (1992). "The appropriate role of agricultural insurance in developing countries." Journal of International Development4(6): HELLMUTH, M. E., D. E. OSGOOD, ET AL. (2009). Index insurance and climate risk: prospects for development and disaster management. International Research Institute for Climate and Society (IRI), Columbia University. C. a. S. N. 2. New York. 80

81 HESS, U. AND J. SYROKA (2005). Weather-based Insurance in Southern Africa: The Case of Malawi. Washington, DC, The International Bank for Reconstruction and Development / The World Bank. IAM (2008). Strategic Action Plan for Cotton Development. Maputo. IBARRA, H. AND J. SKEES (2007). "Innovation in risk transfer for natural hazards impacting agriculture." Environmental Hazards7(1): INGC (2009). Climate Change Report: Study on the Impact of Climate Change on Disaster Risk in Mozambique. M. report. Mozambique. JAFFEE, S., J. DANA, ET AL. (2008). The International Task Force on Commodity Risk Management in Developing Countries: Activities, Findings and the Way Forward, World Bank. KELLNER, U. AND O. MUSSHOFF (2011). "Precipitation or water capacity indices? An analysis of the benefits of alternative underlyings for index insurance." Agricultural Systems104(8): LEVIN, T. AND D. REINHARD (2007). Microinsurance aspects in agriculture Munich Re Foundation LOTSCH, A., W. DICK, ET AL. (2010). Assessment of Innovative Approaches for Flood Risk Management and Financing in Agriculture. Washington, DC, The World Bank. MAHUL, O. AND C. J. STUTLEY (2010). Government Support to Agricultural Insurance :Challenges and Options for Dvelopping Countries. Washington DC, The World Bank. MALAWI METEOROLOGICAL SERVICES (2011). Meteorological Station Network.Blantyre, MEUWISSEN, M. P. M., R. B. M. HUIRNE, ET AL., EDS. (1999). Income Insurance in European Agriculture. European Economy. Luxembourg. 81

82 MEZE-HAUSKEN, E., A. PATT, ET AL. (2009). "Reducing climate risk for micro-insurance providers in Africa: A case study of Ethiopia." Global Environmental Change19(1): PRIOVOLOS, T. (1991). "Experiences with commodity-linked issues." Commodity Risk Management and Finance, Oxford University Press. REDHEAD, K. (1996). Financial Derivatives: An Introduction to Futures, Forwards, Options and Swaps. London, Prentice Hall. ROTH, J. AND M. MCCORD (2008). Agricultural Microinsurance: Global Practices and Prospects. R. Berold. RUOTSI, J. (2003). Agricultural Marketing Companies as Sources of Smallholder Credit: Experiences, Insights and Potential Donor Role. R. t. IFAD. Rome, International Fund for Agricultural Development. RUTTEN, L. AND F. YOUSSEF (2007). Market-based price risk management: An exploration of commodity income stabilization options for coffee farmers, International Institute for Sustainable Development (IISD). SKEES, J. AND J. HARTELL (2009). Pre-Feasibility Analysis: Index-Based Weather Risk Transfer in Mali. SKEES, J., A. MURPHY, ET AL. (2007). Scaling Up Index Insurance, What is needed for the next big step forward?, MICROINSURANCE CENTRE, LLC with GLOBALAGRISK, INC. SKEES, J., J. R. BLACK. AND. B. J. BARNETT. (1997). "Designing and Rating an Area Yield Crop Insurance Contract." American Journal of Agricultural Economics79(No. 2): pp SMITH, V. H., H. H. CHOUINARD, ET AL. (1994). "Almost Ideal Area Yield Crop Insurance Contracts." Agricultural and Resource Economics Review23(1). SYROKA, J. (2009). Micro and Meso Level Weather Risk Management: Deficit Rainfall in Malawi, The World Bank. 82

83 SYROKA, J. AND NUCIFORA, A. (2010).National Drought Insurance for Malawi, The World Bank. TADROSS, M., AND ALEX LOTSCH (2008). Weather Risk Management in Mozambique. Technical note on current and planned weather stations and their potential viability for designing and monitoring agricultural weather risk insurance. Washington, D.C, Technical Paper. The World Bank. TIA (2009/2010). Trabalho de Inquerito Agricola (TIA). M. o. Agriculture. Maputo. TSCHIRLEY, D. L., C. POULTON, ET AL. (2010). "Institutional Diversity and Performance in African Cotton Sectors." Development Policy Review28(3): WFP (2007). Final Report on the Ethiopia Drought Insurance Pilot Project, World Food Programme. WORLDBANK (2005a). Managing Agricultural Production Risk: Innovations in Developing Countries Washington, DC The World Bank WORLDBANK (2005b). "Value-Chain Analysis for Strategic Sectors in Mozambique." Global Development Solutions. Background Paper prepared for the World Bank. WORLDBANK (2006). Weather Risk Management: an Ethiopian Pilot, The World Bank. WORLDBANK (2007). Mozambique Competitiveness Report. Study on Competitive Commercial Agriculture in Africa, Agriculture and Rural Development Unit.. Washington, D.C. WORLDBANK (2008). Quality and Marketing of Cotton Lint in Africa. Africa Region Working Paper Series No Washington DC. WORLDBANK (2010). "Mozambique - Cotton Supply Chain Rapid Risk Assessment ". 83

84 Annexes Annex 1: Main agro ecological zones and farming systems in Mozambique. Source: Instituto Nacional de Investigação Agronómica.... B Annex 2: Administrative map of Mozambique with weather stations... E Annex 3: Summary of Index-based Risk Transfer Products in Lower Income Countries... F Annex 4: Production of the 8 main crops at the national and provincial level (from 2002 to 2010). Source: INE.... H Annex 5: Area cultivated of the 8 main crops at the national and provincial level (from 2002 to 2010). Source: INE... J Annex 6: Development of production and yield of seed cotton per provinces from 1980 to Source: IAM... M Annex 7: Minimum prices for seed cotton (1st and 2nd quality) from 1981 to Source: IAM... N Annex 8: Development of production and yield of seed cotton per districts from 1990 to Source: IAM... O Annex 9: Development of production and yield of seed cotton per districts from 2000 to Source: IAM... O Annex 10: Summary of export quantities from 1980 to 2010 at the national level. Source: IAM Annex 11: Terms of reference for designing risk management tools in agriculture in Mozambique with cotton as pilot crop. Source: IAM Annex 12: TIA database. Source TIA Annex 13: Local market prices of the main food crops from Source: SIMA Annex 14: Meteorological data from 29 weather stations in Mozambique. Source: INAM A

85 Annex 1: Main agro ecological zones and farming systems in Mozambique. Source: Instituto Nacional de Investigação Agronómica. Zone Zone description Farming system description Region (R1) inland Maputo and south Gaza Region (R2) coastal region south of the Save River Region (R3) center and north of Gaza and the west Inhambane The inland Maputo and south Gaza region is a small area covering the Inland strip of Maputo Province and the southern inland of Gaza Province. The major part of the region is under 200 meters altitude; the land of Namaacha reaches 500 meters altitude. Rains are concentrated from November to March and the season characterized by great irregularity with respect to the beginning, duration, and quantity of precipitation. Rain can occur in this region during the cool season. During the growing period a moderately warm temperature dominates (20-25 C). With the exception of the region of Pequenos Libombos, Moamba, Limpopo valley, Incomati and Umbeluzi rivers, the soils are sandy or sandy loams. The coastal region south of the Save River is an extensive area from southern Maputo Province to northern Inhambane Province that has one of the highest population densities in the country. There is a warm rainy season between November and March in most of the region, not including an area adjacent to the coast where rain can start in October and last until April. Rains can occur during the cool season, which has particular benefits for cassava and cashew nut. With the exception of alluvial land and certain low zones, the soils have a sandy texture. The central and northern parts of Gaza and the west Inhambane region consist of a vast interior zone with little population. It is one of the most arid regions of the county with an annual rainfall of only mm, concentrated Farmers cultivate land all year. During the rainy season they produce maize, cowpea, peanuts and cassava. The most preferred soils for cassava and groundnut are of light texture. Given the short duration of the growing season, short-cycle crop varieties are normally used. Sweet potato is grown on the lowest land and along watercourses and where moisture is retained. This region has large areas of pasture with a rural population that traditionally raises cattle and goats. In the region, important areas of irrigation exist that could be increased in the medium term. The most important annual crops are maize, cowpea, groundnut, sweet potato, and cassava. Depending on the type of land, maize/cowpea and cassava/groundnut are the dominant crop systems. Due to the limited availability of land, there is a tendency to intercrop all four crops. The bush fallow rotation is in decline due to land shortage, and the fallow period has been reduced from 20 years to 5 years with 3 years of cropping. Without the use of fertilizers where conditions allow, it can be expected that land productivity will decrease significantly. The production of cashew in this region is one of the most important sources of income for the rural population. Local farmers can earn million metical. The low areas and the river valleys are important for rice production. Family farmers also have small holdings of cattle and goats. Considering the duration of the crop growing period, short-cycle varieties and techniques of moisture conservation are important requirements to ensure an B

86 Region (R4) medium altitude region of central Mozambique Region (R5) low altitude region of Sofala and Zambézia Region (R6) Semiarid Region of the Zambézi Valley and Southern Tete Province Region (R7) Médium Altitude Region of Zambézia, between November and February. Given the lack of soil moisture, sorghum and millet are also grown in the region. Maize has limited potential. The medium altitude region of central Mozambique is a region that includes land between 200 and 1,000 meters above sea level located in the Provinces of Sofala and Manica. It has an annual rainfall of 1,000-1,200 mm concentrated between November and March. The crop growing period is days. The majority of soils are light, with some occurrence of heavy soils. The average temperature during the crop growing period is C. The low altitude region of Sofala and Zambézia embraces a strip of land on the coast of variable width that extends from the south of Sofala to Pebane district in Zambézia province. Depending on the topography, soils have a sandy texture alternating with regions of heavy texture (fluvisols and vertisols). In general the region has moderate to high annual rainfall (1,000 mm-1,400 mm) and a corresponding evapo-transpiration range. The rainy period starts in November and ends between March and May, depending on the area. The medium altitude region of Zambézia, Nampula, Tete, Niassa and Cabo Delgado is vast and includes the land between 200 and 1,000 meters in altitude (subplanaltic, low planaltic, and midplanaltic) in the interior of Zambézia, Nampula and southern Cabo Delgado and Niassa. The annual rainfall and potential evapotranspiration of the region ranges from areas above 25 C (classified as warm region) and others with temperatures of C (moderately warm). The texture of the soils varies from sandy to clay, consistent with the topography. The medium altitude region of Zambézia, Nampula, Tete, Niassa and Cabo Delgado is vast and acceptable degree of food selfsufficiency for the rural population of this region. Main crops are maize, sorghum, cassava, and cowpea. In the more moist areas, farmers cultivate sweet potato and rice. This region has good potential to produce cotton. It is a region with moderate to high population. In heavy soils, rice cultivation is dominant. In regions of well-drained soils, maize, sorghum, millet, cassava, and cowpea are found in association depending on the availability of land and water. Cashew and cotton are important cash crops in the farming systems. Production estimates vary along the area, i.e., maize production of kg/ha, sorghum and millet of 500 to 750 kg/ha, cassava of 6,000-8,000 kg/ha. Basically there are two types of cropping systems that differ by being dominated by maize or sorghum. Cassava is widely cultivated; cowpeas and groundnuts are other important crops. In the eastern-most part of the region cashew is very important, in almost the entire region there is high potential for cotton production that has been practiced over several decades. This is an agricultural area with important human and agroecological potential. Typical maize yields average 1,000 kg/ha, while sorghum production averages 750 kg/ha. Basically there are two types of cropping systems that differ by being dominated by maize or sorghum. C

87 Nampula, Tete, Niassa and Cabo Delgado Region (R8) coastal littoral of Zambézia, Nampula and Cabo Delgado Region (R9) north interior region of Cabo Delgado Mueda Plateau Region (R10) high altitude region of Zambézia, Niassa, Angonia and Manica includes the land between 200 and 1,000 meters in altitude (subplanaltic, low planaltic, and midplanaltic) in the interior of Zambézia, Nampula and southern Cabo Delgado and Niassa. The annual rainfall and potential evapotranspiration of the region ranges from areas above 25 C (classified as warm region) and others with temperatures of C (moderately warm). The texture of the soils varies from sandy to clay, consistent with the topography. The coastal littoral of Zambézia, Nampula and Cabo Delgado consists of a strip of land of varying width on this coast from Pebane in Zambézia to Quionga in Cabo Delgado. The average temperature during the growing season is greater than 25 C. The annual rainfall ranges is 800-1,200 mm, and the evapo-transpiration rate is 1,400-1,600 mm. Sandy soils, with heaver soils in the lowest areas. The northern region of Cabo Delgado includes the plateau of Mueda and Macomia and the surrounding areas of more than 200 meters altitude. The annual rainfall is 1,000 mm-1,200 mm, and the annual evapo-transpiration potential is 1,200 mm -400 mm. The rains are concentrated between December and March and are normally regular. The soils are loamy to sandy texture, with heavier soils occurring in the lowest areas. The high altitude region of Zambézia, Niassa, Angonia- Maravia includes land above 1,000 meters, notably in the planaltic regions of Lichinga, Angonia, Maravia, high Zambézia, Serra Choa, Manica and Espungabera. The annual rainfall is greater than 1,200 mm and average temperature during the period is C. The soils are principally ferrasols. Cassava is widely cultivated; cowpeas and groundnuts are other important crops. In the eastern-most part of the region cashew is very important, in almost the entire region there is high potential for cotton production that has been practiced over several decades. This is an agricultural area with important human and agroecological potential. Typical maize yields average 1,000 kg/ha, while sorghum production averages 750 kg/ha. The production system is characterized by the production of cassava and millet. In the low areas, rainfed rice is cultivated. Cashew has great importance for income for family farmers. The dominant crop in the production system is maize. Sorghum, cowpeas, cassava, and sesame are also cultivated. Cashew is an important source of income. Apart from maize, common beans and potatoes are the main crops. Given the high levels of rainfall, erosion, and the loss of soil fertility are important problems. Finger millet is also cultivated in the area and has important potential as a food and cash crop. Source: Instituto Nacional de Investigação Agronómica. D

88 Annex 2: Administrative map of Mozambique with weather stations = Weather stations In red: all districts where district yield data is available E

89 Annex 3: Summary of Index-based Risk Transfer Products in Lower Income Countries Country Risk Event Contract Structure Bangladesh Drought Index insurance linked lending Caribbean Catastrophe Risk Insurance Facility China Hurricanes and earthquakes Low, intermittent rainfall to Index insurance contracts with risk pooling Index insurance Ethiopia Drought Index insurance Ethiopia Drought Index Insurance Ethiopia Drought Weather Derivative Index Measure Rainfall Indexed data from NOAA and USGS Rainfall storm count Rainfall Rainfall and day Satellite and weather data Target User Smallholder rice farmers Caribbean country governments Smallholder watermelon farmers WFP operations in Ethiopia Smallholder farmers Status In development. Pilot launched in Implemented in 2007 Implemented in Shanghai only in June Includes a 40% premium subsidy USD 7 million insured for Policy not renewed for 2007 due to lack of donor support pilot, currently closed due to limited sales. NGO Implemented 2007 Honduras Drought Rainfall In development India Drought and flood Index insurance linked to lending and offered directly to farmers. Kazakhstan Drought Index insurance linked MPCI program Kenya Drought Weather Derivative Mali Drought Weather Derivative Malawi Drought Index insurance linked lending Mexico Mexico Natural disasters impacting smallholder farmers, primarily drought Major earthquakes Index insurance to to Index-linked CAT bond Rainfall Smallholder farmers Rainfall Medium and large farms Satellite and weather data Satellite and weather data Rainfall Rainfall, wind speed, and temperature Richter scale readings Began with pilot in Now index insurance products are being offered by the private sector and the government In development NGO Implemented 2007 NGO Implemented 2007 Groundnut farmers who are members of NASFAM. State governments for disaster relief. Supports the FONDEN program. Mexican government to Pilot began in policies sold in 2006 pilot season. $7000 in premium volume. Pilot began in Available in 26 of 32 states. Currently 28% (2.3 million ha) of dry land cropland is covered Introduced in CAT bond provides up F

90 Mexico Mexico Mongolia Drought affecting livestock Insufficient irrigation supply Large livestock losses due to severe weather and index insurance contracts Index Insurance Index insurance Index insurance with direct sales to herders Morocco Drought Index Insurance Nicaragua Drought and excess rain during Peru Flooding, torrential rainfall from El Niño Index insurance Index insurance Peru Drought Index insurance linked lending Senegal Drought Index insurance linked area-yield insurance Tanzania Drought Index insurance linked lending Thailand Drought Index insurance linked lending Ukraine Drought Index Insurance Vietnam Flooding during rice harvest Source: World Bank 2010 Index insurance linked lending to to to to to Normalized Difference Vegetation Index Reservoir levels Area livestock mortality rate Rainfall Rainfall ENSO anomalies in Pacific Ocean Area-yield production index Rainfall and crop yield Rainfall Rainfall support FONDEN. Livestock breeders Water users groups in the Rio Mayo area to USD 160 million. Index insurance coverage up to USD 290 million. Launched in Sum insured USD 22.5 million across 7 states. Insured 913,000 cattle. Proposed Nomadic herders Second pilot sales season of pilot completed in 2007; 14% participation Smallholder farmers Groundnut farmers Rural financial institutions Cotton farmers Smallholder farmers Smallholder maize farmers Smallholder farmers No interest from market due to declining trend in rainfall Launched in Proposed Proposed Proposed Pilot implementation in Pilot implementation in Rainfall Smallholders Implemented in 2005, currently closed due to limited sales River level The state agricultural bank and, ultimately, smallholder rice farmers In development, a draft business interruption insurance contract is being considered by the state agricultural bank G

91 Annex 4: Production of the 8 main crops at the national and provincial level (from 2002 to 2010). Source: INE. Evolução da produção por CAP (1999/2000) provincia Cultura UND Total milho/mize TON 1'180'432 mapira/sorgum TON 193'112 feijãomanteiga/butt er beans TON 124'290 Gergelim/Sesame TON 0 mandioca/cassava TON 2'496'212 BatataDoce/sweet potato TON 196'646 tabaco/tabacco TON 0 algodão/cotton TON Cultura UND Niassa Cabo Delgado Nampula Zambezi a Tete Manica Sofala Inham bane Gaza Maput o Total milho/mize TON 175'233 85' ' ' ' '822 76'091 18'455 66'921 21'769 1'114'772 mapira/sorgum TON 11'101 24'903 43'441 15'808 7'515 19'388 15' '318 feijãomanteiga/butt er beans TON 14' '736 11'668 2' '683 Gergelim/Sesame TON 110 2'184 6' '909 1' '856 mandioca/cassava TON 58' '948 1'194'484 1'110'467 45' '419 81' '059 89'748 49'522 3'455'278 BatataDoce/sweet potato TON 34'888 11'944 21' ' '823 48'720 22'772 6'191 24'003 21' '950 tabaco/tabacco TON 8' '136 4'179 25'635 2' '568 algodão/cotton TON '512 51'636 8'722 14' ' ' Cultura UND Niassa Cabo Delgado Nampula Zambezi a Tete Manica Sofala Inham bane Gaza Maput o Total milho/mize TON 159'636 92'654 89' ' ' ' '789 16'686 56'450 7'622 1'178'792 mapira/sorgum TON 10'369 46'031 25'575 23'710 11'902 32'189 38' ' '821 feijãomanteiga/butt er beans TON 17' '852 9'402 2' '853 Gergelim/Sesame TON 113 3'459 4' '618 2' '587 mandioca/cassava TON 284' '911 2'696'409 1'826' ' ' ' ' '665 37'818 6'547'298 BatataDoce/sweet potato TON 93'784 7'746 17' ' ' ' '302 10'469 71'653 15' '158 tabaco/tabacco TON 19' '515 5'419 19'418 3' '076 algodão/cotton TON 0 19'281 28'850 6'980 6' ' ' Cultura UND Niassa Cabo Delgado Nampula Zambezi a Tete Manica Sofala Inham bane Gaza Maput o Total milho/mize TON 121'748 80' ' ' ' '199 52'651 18'013 40'818 10' '536 mapira/sorgum TON 6'596 30'477 16'710 12'103 9'256 22'242 16' '533 H

92 feijãomanteiga/butt er beans TON '306 Gergelim/Sesame TON 363 6'314 7' '260 2' '088 mandioca/cassava TON '800 BatataDoce/sweet potato TON '000 tabaco/tabacco TON 21'630 3'117 5'461 3'741 42'685 3' '842 algodão/cotton TON '654 54'862 9'760 10'358 1'481 7' ' Cultura UND Niassa Cabo Delgado Nampula Zambezi a Tete Manica Sofala Inham bane Gaza Maput o Total milho/mize TON 222' ' ' ' ' ' '489 32' '090 29'265 1'395'474 mapira/sorgum TON 13'080 25'878 32'719 14'721 27'377 45'460 39'614 2' '759 feijãomanteiga/butt er beans TON 19' '303 9'509 11'508 3' ' '628 Gergelim/Sesame TON 261 4'610 8'727 1'335 1'390 1'673 2' '561 mandioca/cassava TON 120' '432 1'631'304 2'993'065 32' ' ' ' '880 70'548 6'658'710 BatataDoce/sweet potato TON 57'311 8'625 13' ' '767 96' '826 3'915 47'982 31' '249 tabaco/tabacco TON 23'547 4'454 4'273 31'066 28' '066 algodão/cotton TON '277 29'669 4'946 24'415 6'041 18' ' Cultura UND Niassa Cabo Delgado Nampula Zambezi a Tete Manica Sofala Inham bane Gaza Maput o Total milho/mize TON 103'820 85'655 93' ' ' '935 96'837 29'049 60'941 10'891 1'133'910 mapira/sorgum TON 7'743 17'747 21'166 13'994 22'039 43'837 36'243 3' '872 feijãomanteiga/butt er beans TON 16' '749 14'529 12'441 3' ' '516 Gergelim/Sesame TON 300 4'000 6'000 1'000 2'000 2'000 4' '300 mandioca/cassava TON '000 BatataDoce/sweet potato TON '000 tabaco/tabacco TON 11' '000 5'000 16'000 1' '500 algodão/cotton TON 1'000 24'000 11'000 9'000 16'000 17'000 15' ' Cultura UND Niassa Cabo Delgado Nampula Zambezi a Tete Manica Sofala Inham bane Gaza Maput o Total milho/mize TON 104'000 87'000 65' ' ' ' '000 43' '000 39'000 1'167'000 mapira/sorgum TON 7'000 20'000 9'000 15'000 16'000 11'000 33'000 3'000 1' '000 feijãomanteiga/butt er beans TON 22' '669 15'868 3' ' '536 Gergelim/Sesame TON 1'135 5'494 14' '465 3'011 12' '696 mandioca/cassava TON 427' ' '700 1'814'140 30' ' ' ' '414 42'531 4'054'597 BatataDoce/sweet potato TON 51'311 6'453 1' ' '440 46'480 91'503 4'524 49'685 38' '039 tabaco/tabacco TON 14' '290 24' '272 algodão/cotton TON 1'481 26'771 7'546 10'526 15'042 1'078 7' '693 I

93 Annex 5: Area cultivated of the 8 main crops at the national and provincial level (from 2002 to 2010). Source: INE Evolução das áreas CAP (1999/2000) cultivadas por provincia Cultura Und Total milho/mize ha 1'294'420 mapira/sorgum ha 235'223 feijãomanteiga/butt er beans ha 61'058 Gergelim/Sesame ha 0 mandioca/cassava ha 641'670 BatataDoce/sweet potato ha 47'207 tabaco/tabacco ha 0 algodão/cotton ha Cultura Und Niassa Cabo Delgado Nampula Zambe zia Tete Manica Sofala Inhamb ane Gaza Maputo Total milho/mize ha '459'254 mapira/sorgum ha '350 feijãomanteiga/bu tter beans ha '949 Gergelim/Sesam e ha '932 mandioca/cassav a ha '817 BatataDoce/swee t potato ha '515 tabaco/tabacco ha '310 algodão/cotton ha ' Cultura Und Niassa Cabo Delgado Nampula Zambe zia Tete Manica Sofala Inhamb ane Gaza Maputo Total milho/mize ha mapira/sorgum ha feijãomanteiga/bu tter beans ha Gergelim/Sesam e ha mandioca/cassav a ha BatataDoce/swee t potato ha tabaco/tabacco ha algodão/cotton ha Cultura Und Niassa Cabo Delgado Nampula Zambe zia Tete Manica Sofala Inhamb ane Gaza Maputo Total J

94 milho/mize ha 189' ' ' ' ' ' ' ' '631 30'726 1'749'534 mapira/sorgum ha 22'384 68'503 50'481 44'441 41'957 67'455 57'372 6'686 5' '368 feijãomanteiga/bu tter beans ha 39' '116 17'743 32'949 10' ' '753 Gergelim/Sesam e ha 1'395 17'999 21'976 4'469 2'591 7'866 8' '027 mandioca/cassav a ha 32' ' ' '939 6'454 24'565 16' '171 52'416 12'997 1'038'850 BatataDoce/swee t potato ha 5'461 1'387 2'940 16'157 8'579 8'492 4'587 1'123 11'244 4'221 64'191 tabaco/tabacco ha 25' '685 3'087 43'585 5' '233 algodão/cotton ha '615 79'740 28'051 34'609 4'959 20' ' Cultura Und Niassa Cabo Delgado Nampula Zambe zia Tete Manica Sofala Inhamb ane Gaza Maputo Total milho/mize ha 153' ' ' ' ' ' ' ' '500 28'500 1'663'900 mapira/sorgum ha 36'100 60'500 65'700 43'000 51'100 67'400 67'600 9'800 4' '100 feijãomanteiga/bu tter beans ha 29' '400 13'700 33'500 9'900 2' '300 1'100 92'900 Gergelim/Sesam e ha mandioca/cassav a ha 21' ' ' '300 5'100 23'800 13'400 83'500 31'300 8' '700 BatataDoce/swee t potato ha 11'300 1'400 3'100 17'700 12'300 8'900 3'000 1'600 17'500 3'300 80'100 tabaco/tabacco ha algodão/cotton ha Cultura Und Niassa Cabo Delgado Nampula Zambe zia Tete Manica Sofala Inhamb ane Gaza Maputo Total milho/mize ha 177' ' ' ' ' ' ' ' '200 27'200 1'664'300 mapira/sorgum ha 34'400 57'100 69'600 29'700 40'900 73'400 60'200 15'700 2' '800 feijãomanteiga/bu tter beans ha 20' '200 18'000 32'900 15'200 1' ' '100 Gergelim/Sesam e ha 2'800 16'400 32'500 4'100 2'700 2'900 12' '100 mandioca/cassav a ha 27' ' ' '000 5'200 19'300 15'100 94'000 40'800 10' '100 BatataDoce/swee t potato ha 9'800 2'500 3'800 19'100 15'400 14'800 4'800 3'900 9'100 3'900 87'100 tabaco/tabacco ha 22' '300 21' '200 algodão/cotton ha 3'900 41'200 22'700 17'400 29'300 2'700 15' ' Cultura Und Niassa Cabo Delgado Nampula Zambe zia Tete Manica Sofala Inhamb ane Gaza Maputo Total milho/mize ha 196' ' ' ' ' ' ' ' '700 51'400 1'962'700 mapira/sorgum ha 27'900 50'700 62'600 52'900 38'800 42'800 88'900 12'000 7' '300 feijãomanteiga/bu ha 29' '300 18'200 43'200 7'800 2' '900 1' '400 K

95 tter beans Gergelim/Sesam e ha 5'600 24'800 38'700 5'800 6'000 7'400 30' '700 mandioca/cassav a ha 37' ' ' '000 5'200 9'200 17'300 82'100 41'900 12' '600 BatataDoce/swee t potato ha 10'100 1'200 1'200 6'400 12'900 5'600 9'700 2'500 9'100 5'700 64'400 tabaco/tabacco ha 26' '600 27' '900 algodão/cotton ha 2'200 42'400 22'800 9'700 15'700 1'000 9' '900 L

96 Annex 6: Development of production and yield of seed cotton per provinces from 1980 to Source: IAM M

97 Annex 7: Minimum prices for seed cotton (1st and 2nd quality) from 1981 to Source: IAM CAMPANHA PREÇO MÌNIMO (MZN/t) Alg. De 1ª Alg. De 2ª 1981/ / / / / / / / / / / / /94 1' /95 1' ' /96 3' ' /97 3' ' /98 2' ' /99 2' ' /00 2' ' /01 2' ' /02 3' ' /03 3' ' /04 5' ' /05 5' ' /06 5' ' /07 5' ' /08 6' ' /09 5' ' /10 8' ' /11 15' ' N

98 Annex 8: Development of production and yield of seed cotton per districts from 1990 to Source: IAM Annex 9: Development of production and yield of seed cotton per districts from 2000 to Source: IAM O

99 P