Smart, data-driven boundaries can improve crop insurance performance for smallholder farmers, according to new study

Farmers plant, weed, and hope for the right amount of rain, but droughts, floods, pests, and other surprises can wipe out their crop, no matter how hard they work. Insurance is supposed to help by softening the financial impact when yields fall short of expectations. The challenge is that someone must first detect these losses. Sending people to review claims in the field is expensive and takes time, especially in remote, hard-to-reach areas where many small-holder farmers live.

Because of these challenges, index insurance is growing in popularity. Instead of checking each farmer’s claim individually, index insurance uses indicators to say whether an area had a bad year, such as average crop yields. If that indicator shows trouble, everyone in that zone gets paid automatically. This approach makes payouts faster and much less costly to administer.

“Well-designed index insurance can offer many benefits, but its design matters,” said Ella Kirchner, a research associate with the Department of Agricultural and Applied Economics.” One important issue is how insurers decide on which area one should measure distress and issue payouts. Often, they use boundaries like counties or districts because they are convenient, but nature doesn’t always follow administrative borders.”

The December 2025 study, Get in the Zone: The Risk-Adjusted Welfare Effects of Data-Driven vs. Administrative Borders for Index Insurance Zones, published in the Journal of Development Economics, examined whether and how insurance could better match the losses farmers face if zones were drawn based on observed growing conditions instead of administrative zones.

The researchers used more than 13,000 crop-cut measurements, which refer to harvesting and weighing a small portion of a field’s crops to estimate yields, from smallholder maize fields in Kenya. Using satellite data on rainfall, temperature, vegetation, and soil properties in the locations of the crop cuts, the team created alternative insurance zones by grouping areas that experience similar agricultural conditions.

The results showed that even simple data-driven zones could often match or, in some cases, improve the value of insurance to policyholders over administrative boundaries. Even when the differences were not dramatic, the data-driven approach offers more degrees of flexibility than one based on a fixed number of administrative borders. These data-driven zones can be drawn in different sizes to better match how production conditions change across a landscape. When the authors allowed for more zones or adjusted the amount or cost of sampling within a zone, the distinctions became clearer. In other words, boundaries matter, and choosing a data collection intensity that balances costs and performance can make insurance significantly more useful to farmers.

The study also explored broader uses of satellite data in agricultural insurance. Remote sensing has been used to create the index that triggers payouts, for example, using composite vegetation measures instead of rainfall to proxy losses. This research took a different angle by showing that satellite data can also help define the zones themselves, even when reliable yield data are limited. This is especially important in many developing countries, where historical yield data are often sparse, and field measurements are expensive. Yet insurance zones should still reflect reality on the ground.

“Overall, the findings suggest that while traditional zones are convenient, they are not always the best option for farmers,” said Elinor Benami, an assistant professor in the department. “Data-driven zones offer a promising path forward, especially when insurers have the flexibility to choose the number of zones and how intensively to collect field data.”

These choices come with tradeoffs. More zones and more sampling generally improve insurance quality but also increase costs, especially compared with an Earth-observation approach. Even so, the resulting system can remain far less expensive than claims-based insurance while also delivering greater protection, especially when paired with additional smart features like conditional audits and strategic, low-cost verification – for example, phenocams or targeted field visits – when satellite indicators are most likely to misclassify outcomes. The key is finding the sweet spot where the system remains affordable while still giving farmers the protection they need.

Co-authors include Andrew Hobbs of the University of San Francisco; Michael R. Carter of the University of California, Davis; and Zhenong Jin of the University of Minnesota.

Original study: DOI 10.1016/j.jdeveco.2025.103658