Ethiopian Project on Interlinking Insurance with Credit for Agriculture

The idea of the project is to test whether providing the private bank with WII on its loans can release credit resources for production by smallholders, and whether smallholders are willing to pay for the combined cost of credit and insurance. EPIICA is piloting the sale of WII tied with short-term production credit. To that end, a baseline survey was conducted in early 2011 in 120 rural Kebeles (villages or farmer associations) in four zones of the Amhara region in Ethiopia . The choice of the Kebeles was non-random but instead was designed on the basis of informed opinion of NISCO as to where in the Amhara region the market for WII has best potential. Households within the selected kebeles were randomly sampled to participate in the study; in each village 18 cooperative households and 2 households that are not a member of the primary cooperative were selected. Because fertilizers are procured exclusively through primary cooperatives and their upper level zonal Cooperative Unions (CUs) in Ethiopia, it is anticipated that cooperative households may display a higher propensity to uptake additional fertilizer if risk concerns can be ameliorated.

The core of the project is a rainfall index insurance product pioneered by Nyala Insurance Company. This product is based on rainfall as measured at 8 different rainfall stations in Amhara, operated by the National Meteorological Agency (NMA). Farmers could purchase insurance prior to the planting season for a specified crop and quantity of land, and for an amount up to the total cost of inputs for the size of land planted to a given crop. For each crop and location an index has been designed based on local meteorological conditions and crop water requirements that predicts partial or total crop failure based on realized rainfall measurements. If the index indicates a partial crop failure, the insurance makes a partial payout. If the index indicates total crop failure, the insurance will make a full payout. The intent is that farmers are insuring the cost of their inputs (not the total value of the harvest), and so the payouts are designed to reimburse farmers for lost input investments, and thereby enable them to take more risk in the use of fertilizers and improved seeds. If the farmers take a loan to purchase the inputs, then the insurance basically acts as collateral, for default that can be caused by weather events as seen in the index, to the Bank for the amount lent.

The EPIICA study is a state-of-the art scientific pilot, which, after having designed the relevant weather indices and loan products, implements a Randomized Controlled Trial (RCT) to generate an experimental control group and thereby to measure the impact of the program on farm productivity and household welfare, as well as the institutional profitability of Nyala and Dashen. The study is a two-armed trial, comparing two distinct ways of implementing weather insurance. The first of them, the Standalone product, offers only an index insurance product to farmers in 17 kebele cooperatives. Here, farmers can purchase insurance (although the costs of the insurance premiums may be financed by the cooperatives in some cases) and receive payouts if inadequate rainfall is measured at the nearest rainfall station. The second is the Interlinked product, in which the project provides insured credit to cooperative farmers in 17 additional kebeles. For the Interlinked product, farmers will first purchase insurance from Nyala, and are then eligible to access credit from Dashen Bank for an amount up to the sum insured by the index product. On Interlinked loans the insurance contract is written over to Dashen as a form of collateral - for default that can be caused by weather events as seen in the index.Dashen Bank has accepted to use this type of insurance contract as an innovative form of partial collateral in this pilot. If the weather is good, and hence the rainfall index does not trigger insurance payments, the farmers must repay this loan plus interest. If the weather is bad, then the loan will be partially or completely forgiven based on the insurance payout that will be made directly from Nyala to Dashen

2012-05-01

2014-09-01

The primary intermediate outcome for the study is enhanced use of improved inputs, namely fertilizer and HYV seeds. These are then pathways to the intended final outcomes, which is an increase in agricultural productivity (yield/hectare), increased ag sales and income, and improved household outcomes (consumption, schooling, and household asset ownership, decreased vulnerability). We will also examine how the implementation of the insurance changes cooperative-level outcomes such as number of members, number and type of services offered to members, and relationship with second-tier cooperative unions.

The study started with a sample of 120 Primary Cooperatives in Amhara. This sample of kebeles has been tracked for two years in a detailed household, cooperative, and kebele-level survey conducted by the Ethiopian Economics Association. We selected a sample of 20 households per kebele (18 of them cooperativized farmers, and 2 non-cooperativized) and have tracked them in two rounds of surveys, conducted in Jan-Feb of 2010 and 2011. Two factors have reduced this sample size to the group of 62 that we are now pursuing. First, detailed data work in the field revealed that some of the originally selected regions did not suffer from severe deficit rainfall risk and hence were not good candidates for the insurance product we have developed. Secondly, Swiss Re (the international reinsurance company that Nyala utilizes to reinsure part of its risk exposure) has agreed to provide reinsurance cover for only 7 out of the 17 stations that cover the study area because of incomplete data at the other stations (Nyala will provide insurance at one additional station in North Wollo without reinsurance). We have therefore restricted the sample to consist of only those kebeles that face deficit rainfall risk and are located within 15 km of the 8 project rainfall stations (20 km in the flatter terrain of West Gojam; statistical analysis shows that the correlation between yields and rainfall in stations more than 20 km away from the farmers is unacceptably low). The number of farmers covered by this pilot remains large; roughly 15,260.

The research design therefore is an RCT comparing 17 kebeles (villages) offered standalone insurance with 17 kebeles offered interlinked insurance and credit, where the counterfactual for both is formed by 28 control kebeles offered neither product. The impact of the study will be tracked using two different survey instruments. One of these is a household survey tracking 20 households per kebele (18 cooperative members, two non-members), and the other is a survey conducted at the level of the primary (kebele) cooperative.

As a part of the research design, as well as randomizing the placement of the treatment within study kebeles, we distributed voucher coupons to study households via a lottery system. The amount of vouchers distributed ranges from 0 to 500 birr, and the presence of this variation in the effective price of the insurance allows us to estimate the demand curve for insurance in a statistically rigorous fashion. In the first two years these were distributed in a manner that allowed farmers to get a very small insurance coverage for free by using only the voucher; in the third year the vouchers were reframed so that they cover a fraction of any premium payments but cannot make up the entire purchase amount.

Both the village level randomization and the individual randomization of insurance vouchers was done in office by a computer.

The unit of randomization for the core intervention is the kebele (village), so it is a cluster randomized trial. The unit of randomization for the price vouchers is the cooperative member.

Yes

120 clusters tracked in initial study, only 62 clusters (kebele/villages) remain in the final sample that is eligible for insurance with the reinsured stations and will be tracked for the outcomes of the study.

1,240 households.

17 villages standalone treatment, 17 villages interlinked treatment, 28 villages serve as control.