Incorporating satellite data in crop insurance in Kenya: analyzing the sources of basis risk

Abstract:

Farmers experience large weather and price risk in agriculture, which can have detrimental consequences on their production and technology adoption. Offering insurance products is a promising pathway to mitigate this risk, yet traditional indemnity-based schemes are difficult to implement in developing countries, due to the high costs associated to measuring yields and difficulty to assess individual risk. As an alternative to individual-based indemnity, index insurance delivers indemnities based on a region-specific index. This has the advantage of lowering monitoring costs, easing the valuation of risk and can be implemented using satellite remote sensing techniques. A main drawback however of index insurance is that by de-linking indemnity from individual yields, farmers can experience a loss yet no receive any indemnity, a situation called basis risk.

Abstract In this paper, we analyze how satellite data can be used to improve index insurance products in Kenya. The first contribution from satellite data is to analyze sources of the basis risk. Following the decomposition by Elabed et al (2013), we estimate two components of basis risk into 1) an idiosyncratic uninsurable part, and 2) an insurable yet mismeasured part, the design risk. Understanding the relative importance of each component is crucial to identify bottlenecks in the development of index insurance, yet very few studies have done it in practice. Owing to the exhaustiveness of our satellite-derived yield dataset covering 30 regions in Kenya, our study provides an idea setup to learn about the sources of basis risk. In a second step, we investigate ways to reduce each component of basis risk. The idiosyncratic basis risk arises from a lack of homogeneity of the insurance zone, and hence we investigate how optimizing zone design can reduce idiosyncratic risk. The design risk arises from imperfect prediction of yields, and we investigate how alternative products such as satellite-based or weather based instruments deteriorate the usefulness of index insurance compared to a benchmark of perfectly observed yields.