[Paper Review] On the benefits of index insurance in US agriculture: a large-scale analysis using satellite data

Matthieu Stigler, David B. Lobell|arXiv (Cornell University)|Nov 25, 2020

Agricultural risk and resilience42 references4 citations

TL;DR

This study uses high-resolution satellite-derived yield data from nearly 1.8 million US corn and soybean fields across 600 counties to simulate and compare index insurance and farm-level insurance. It finds that index insurance is demanded by 30–40% of farmers in simulations—surprisingly high—yet its benefits depend critically on the benchmark used: it outperforms no insurance in high-temporal-variance counties but underperforms farm-level insurance in the same areas due to basis risk, highlighting the importance of spatial and temporal variability in index design.

ABSTRACT

Index insurance has been promoted as a promising solution for reducing agricultural risk compared to traditional farm-based insurance. By linking payouts to a regional factor instead of individual loss, index insurance reduces monitoring costs, and alleviates the problems of moral hazard and adverse selection. Despite its theoretical appeal, demand for index insurance has remained low in many developing countries, triggering a debate on the causes of the low uptake. Surprisingly, there has been little discussion in this debate about the experience in the United States. The US is an unique case as both farm-based and index-based products have been available for more than two decades. Furthermore, the number of insurance zones is very large, allowing interesting comparisons over space. As in developing countries, the adoption of index insurance is rather low -- less than than 5\% of insured acreage. Does this mean that we should give up on index insurance? In this paper, we investigate the low take-up of index insurance in the US leveraging a field-level dataset for corn and soybean obtained from satellite predictions. While previous studies were based either on county aggregates or on relatively small farm-level dataset, our satellite-derived data gives us a very large number of fields (close to 1.8 million) comprised within a large number of index zones (600) observed over 20 years. To evaluate the suitability of index insurance, we run a large-scale simulation comparing the benefits of both insurance schemes using a new measure of farm-equivalent risk coverage of index insurance. We make two main contributions. First, we show that in our simulations, demand for index insurance is unexpectedly high, at about 30\% to 40\% of total demand. This result is robust to relaxing several assumptions of the model and to using prospect theory instead of expected utility.

Motivation & Objective

• To assess the real-world benefits and demand for index insurance in the U.S. agricultural sector using high-resolution field-level data.
• To investigate the determinants of index insurance suitability across 600 U.S. counties, particularly the role of spatial and temporal yield variability.
• To evaluate how different performance metrics—compared to no insurance or farm-level insurance—affect the perceived value of index insurance.
• To examine the robustness of results under alternative behavioral models, including prospect theory.
• To quantify basis risk in area-yield index insurance using field-level satellite data, moving beyond aggregate county-level estimates.

Proposed method

• The study uses satellite-derived yield predictions for 1.8 million corn and soybean fields across 600 counties in the U.S. Corn Belt from 1998 to 2017.
• It constructs area-based index insurance payouts based on county-level average yields, with payouts triggered when yields fall below a threshold.
• Field-level expected utility is calculated using a second-order Taylor approximation of the utility function, with risk aversion captured via the second derivative of the utility function.
• The analysis compares expected utility under index insurance, farm-level indemnity insurance, and no insurance, using both expected utility theory and cumulative prospect theory.
• Spatial and temporal yield variances are computed at the county level, and their relative importance in predicting insurance demand is assessed using relative importance metrics (lmg) in regression models.
• Basis risk is quantified as the probability that a field suffers a loss below a given yield threshold while the county index does not trigger a payout, estimated using empirical cumulative distributions.

Experimental results

Research questions

• RQ1 What is the level of demand for index insurance in the U.S. when simulated at the field level using realistic yield data?
• RQ2 How does the perceived benefit of index insurance vary when compared to no insurance versus farm-level indemnity insurance?
• RQ3 To what extent do temporal and spatial yield variability in counties affect the suitability of index insurance?
• RQ4 How robust are the results to alternative behavioral models such as cumulative prospect theory?
• RQ5 What is the magnitude of basis risk in area-yield index insurance for corn and soybeans across the U.S. Corn Belt?

Key findings

• Index insurance is demanded by approximately 30% to 40% of farmers in the simulation, indicating unexpectedly high uptake despite basis risk.
• When benchmarked against no insurance, index insurance provides the greatest benefit in counties with the highest temporal yield variability.
• When benchmarked against farm-level insurance, index insurance is least beneficial in the same high-temporal-variance counties due to basis risk.
• The choice of performance metric—especially whether comparing to no insurance or to farm-level insurance—can lead to opposite conclusions about index insurance’s value.
• Basis risk remains substantial: for corn, 24.7% of fields experience losses below the 90th percentile yield while the county index does not trigger a payout, and for soybeans, this rises to 46.3%.
• The study finds that spatial variance reduces the benefit of index insurance, while temporal variance increases it, underscoring the need to account for both in index design.