Get in the Zone: The Risk-Adjusted Welfare Effects of Data-Driven vs. Administrative Borders for Index Insurance Zones

Agricultural index insurance seeks to protect producers against negative shocks that are common across a prespecified area, i.e., an index insurance zone. Often, administrative boundaries are used to delineate such index insurance zones. However, administrative boundaries may not reflect relevant variations in yield over space, which can be costly for policyholders as well as the public, especially since agricultural insurance is often heavily subsidized. Increased availability of finely resolved geospatial data on agronomic conditions coupled with machine learning approaches to identify similarities promises the ability to reduce losses associated with index insurance by identifying more homogeneous zones. In this work, we examine the changes in welfare impacts of a hypothetical area-yield index insurance when redrawing zone boundaries on the basis of relevant observed agronomic conditions. Drawing upon crop cut data from over 10,000 maize fields in Kenya from 2016-2020 combined with satellite-based estimates of agronomic conditions, we examine the changes in expected utility to assess the value of data-driven and administrative insurance zones. When keeping the number of insurance zones equal to the number of administrative zones, we find that data-driven zones may offer only slightly higher risk reduction value than administrative zones. If no set number of zones are prespecified, the data-driven approach offers a flexible approach to identify an optimal number of zones that balances costs and performance. This approach can help inform program design as well as impact evaluations, as it further sheds light on trade-offs between the costs of ground sampling and zone size that can inform how to design and evaluate new programs in resource-constrained environments for maximum impact.