1. Producer Demand and Welfare Benefits of Price and Weather Insurance in Rural Tanzania Alexander Sarris (FAO), Panayiotis Karfakis (Univ. of Athens and FAO) and Luc Christiaensen (World Bank) FAO, Rome, June 27, 2006

2. Plan of Presentation Background and Motivation Methodology Relevant Characteristics of households in Kilimanjaro and Ruvuma Price and rainfall risks and household perceptions of it Desirability of and variables affecting demand for price and weather based insurance The demand and welfare benefit of providing price and weather insurance Conclusions and policy implications

3. Background and Motivation Small agricultural commodity producers face many income and non-income risks Individual risk management and risk coping strategies maybe detrimental to income growth Considerable residual income risk and vulnerability Is there a demand for additional price and weather related income insurance in light of individual existing risk management strategies? What is the welfare benefit of price and weather based insurance? Is there a rationale for market based or publicly supported price and weather based safety nets

4. Methodology Use Contingent Valuation (CV) approach (ask directly about willingness to pay given amounts for specific insurance contracts) Consider change in the status quo of farmer from q0 to q1 Let indirect utility of respondent be v(p,q,y,s,  ), - p vector of prices for market goods -y is the respondent’s income -s is a vector of respondent characteristics -  is the stochastic component of utility.

5. Methodology (continued)

6. Variables likely to affect WTP Degree of risk aversion (+) Degree of consumption smoothing (-) Household vulnerability to poverty (+/-) Degree of unpredictability (variability) of future prices or incomes (+) Variance of returns of insurance contract (-) Correlation between returns to insurance and future income (-)

7. Household characteristics 1

8. Household characteristics 2

9. Percentage of households affected by various shocks between 1999 and 2003, by region and status as cash crop grower or not

10. Vulnerability to poverty is high but portion due to covariate shocks varies by region

11. Variability of nominal prices received for coffee in Kilimanjaro over the previous 10 years.

12. Variability of nominal prices received for coffee in Ruvuma over the previous 10 years.

13. Variability of nominal prices received for cashew nuts in Ruvuma over the previous 10 years.

14. Interest in minimum price coffee insurance among coffee producing households

15. Interest in minimum price cashew nut insurance among cashew nut producing households in Ruvuma. (Number of households)

16. Variables used in the Selection and WTP equations Household characteristics (e.g education). Income structure and income level variables (e.g. per capita income, wealth, shares of cash to total income, share of coffee in total income, whether cash income from coffee is important, a banana production dummy, the share of coffee input costs in total coffee production value, easy access to seasonal credit, and the Herfindhal index of cash income diversification) Variables designed to proxy for recent conditions (e.g.the level recent prices received) Variables designed to indicate the level of instability faced (e.g. the range of prices received in the last ten years, the number of years in the last 10 when coffee cash income or total income fell below 50% of normal, or whether the household perceives cash crop income as very unreliable) Variables designed to capture the importance of different coping mechanisms to shocks affecting livelihoods (used four mechanisms with respective dummies: whether in response to a shock in the past the household used own savings or other own resources, assistance from other non-household family, assistance from non- family (including friends, neighbours, NGOs, government, etc), or whether sought to find new ways to generate income. Village level effects

17. What affects the desirability for minimum price insurance? Income instability variables Household coping mechanisms

18. What affects the WTP for minimum price insurance? Kilimanjaro: bid value (-), income (-), the number of coffee trees (-), total value of wealth (+), whether cash income from coffee is important (+), Herfindhal index (+), while coping mechanism variables (-), easy access to credit (-). Predictive value is quite high, more than 70 percent correct predictions. Ruvuma coffee. Bid value (-), Importance of coffee in income (+), easy access to seasonal credit (+), share of cash to total income (-), number of coffee trees (-), past price variability (-), coping mechanism involving the use of new ways to earn income (-). Share of correct predicted values is more that 80 percent. Ruvuma cashew nuts. Bid value (-), Income (+), number of cashew trees (+), importance of cashew income (+), whether cashew income declined in the recent past (+), ease of access to seasonal credit (-), coping mechanism relating to use of new ways to earn income (-) Percent correct predictions larger than 74 percent.

19. Summary statistics of the predicted value of WTP for coffee minimum price insurance in Kilimanjaro from round 1.

20. Summary statistics of the predicted value of WTP for coffee minimum price insurance in Ruvuma from round 1.

21. Summary statistics of the predicted value of WTP for cashew nut minimum price insurance in Ruvuma from round 1.

22. Kilimanjaro coffee: Welfare benefit and cost for minimum price insurance.

23. Ruvuma coffee: Welfare benefit and cost for minimum price insurance.

24. Ruvuma cashew nuts: Welfare benefit and cost for minimum price insurance.

25. Monthly rainfall patterns observed across weather stations in Kilimanjaro are reasonably well correlated

26. Monthly rainfall patterns observed across weather stations in Ruvuma are quite well correlated

27. Households’ assessments of yearly rainfall is consistent with objective rainfall measurements

28. Average number of years in past 10 that households report rainfall as being in different ranges relative to normal.

29. Similarity between farmers’ perceptions concerning rainfall and their village average is high (index ranging from 0 (no similarity) to 1 (perfect similarity))

30. Perceptions of households concerning rainfall relative to objective rainfall incidence

31. Reasons for which households indicated they were not interested in rainfall (or drought) insurance

32. What variables affect WTP for Weather insurance? Bid values (-) Size of household (+) Per capita income (+) Education (+) Share of cash in total income (+) Use of self insurance to cope with shocks (+) Rely on family assistance to cope with shocks (-) Degree of vulnerability (-)

33. Kilimanjaro. Summary statistics for WTP for rainfall insurance

34. Ruvuma. Summary statistics for WTP for rainfall insurance

35. Kilimanjaro. Welfare benefits and cost of rainfall insurance (10% rainfall reduction)

36. Kilimanjaro. Welfare benefits and cost of rainfall insurance (1/3 rainfall reduction)

37. Ruvuma. Welfare benefits and cost of rainfall insurance (10% rainfall reduction)

38. Ruvuma. Welfare benefits and cost of rainfall insurance (1/3 rainfall reduction)

39. Conclusions and policy implications (1) Producer households are affected by a variety of shocks, and prominent among them are health and death related ones, as well as weather induced ones. Shocks induce considerable variability of incomes Most prevalent coping mechanism through own savings and asset depletion. There seems to be considerable variability in prices received for the main cash crops and incomes. This induces considerable interest in minimum price and weather based income insurance. Instability variables contribute positively to the demand for price insurance, while the existence of coping mechanisms contributes negatively, as expected. Large estimated values of individual WTP for coffee and cashew nut price insurance. Higher in Kilimanjaro than Ruvuma Considerable welfare benefits (net of costs) of minimum price insurance. Market based price insurance viable (premiums comparable to option prices in organized exchanges)

40. Conclusions and policy implications (2) Interest in rainfall insurance higher in Kilimanjaro, a richer and more exposed to rainfall shocks region Vulnerability contributes negatively to the demand for insurance, while the existence of self insurance coping mechanisms contribute positively or negatively, depending on the type of mechanism. Considerable demand for weather insurance in Kilimanjaro and higher for contracts paying out when decline in rainfall is 10% below normal. Weak demand in Ruvuma. In Kilimanjaro average WTP is about 30-55 percent of actuarially fair premium. In Ruvuma average WTP only 5-18 percent of actuarially fair premium.

41. Conclusions and policy implications (3) In Kilimanjaro for 10 percent rainfall shortfall, about 30-40 percent of households would purchase the insurance at the average WTP, insuring 40-45 percent of their total acres cultivated. The insured land would constitute 15-20 percent of total cultivated land. In Kilimanjaro, for insurance against a 1/3 rainfall shortfall, participation at average WTP would be around 25-35 percent of households, and they would insure 40-45 percent of their cultivated acres. Total area insured would be around 15-20 percent of total cultivated land. For Ruvuma and for the 10 percent rainfall shortfall, the participation at average WTP would be of only 10-15 percent of households, insuring about 20-30 percent of their total area cultivated. At actuarially fair prices, however, participation would fall to less than 10 percent of households, insuring about 30 percent of their cultivated land. Above numbers decline significantly when computed at the actuarially fair values of the contracts. Market based weather insurance not easily viable. Provision of subsidised weather insurance could reduce considerably the vulnerability of poor households