Evolving County-Level Yield Volatility and Implications for Crop Insurance Risk Modeling

By Junkan Li

Accurately measuring how county-level yield risk evolves over time is central to agricultural risk manage- ment and the policy analysis conducted within USDA. Crop insurance evaluation, systemic exposure assessment, and stochastic simulation of program outcomes all depend on historical yield data to characterize the distribution of production shocks faced by U.S. farmers. Yet yield risk is not necessarily stationary: long-run productivity growth, changes in farm management and technology, shifting weather pat- terns, and structural differences across regions may alter both the volatility and tail behavior of yield out- comes. Understanding whether yield variability has increased or declined, and how these changes differ geographically, is therefore essential for ensuring that insurance simulation frameworks and policy tools are calibrated to current risk conditions rather than outdated historical regimes. This brief documents re- cent changes in county-level yield volatility across major crops and highlights their implications for crop insurance modeling and systemic risk assessment.

To investigate this issue, economists at ARPC compiled county-level yield data from 1980 to 2024 from USDA NASS for four major crops: corn, cotton, soybeans, and winter wheat. For each county, normalized yield shocks (percentage deviations from trend yield) are constructed using alternative approaches to re- moving long-run yield growth, including linear and quadratic trends, nonparametric LOWESS smoothing, and HP filtering. We divide the full sample into two sub-periods and compute volatility measures, including standard deviations and upper and lower tail behavior, for each county across the two periods.

Figure 1: Declining Volatility in County-Level Yields Across Major Crops

Source: Agricultural Risk Policy Center (ARPC), using data from the U.S. Department of Agriculture, National Agricultural Statistics Service (USDA NASS).

Figure 1 presents boxplots of county-level relative volatility changes between the later and earlier periods. The measure is reported in log form, so that positive values indicate increasing volatility over time, while negative values indicate declining volatility. For corn, cotton, and soybeans, the distributions are shifted below zero, and even the upper quartiles fall below zero across detrending methods, indicating that yield volatility has generally declined from the earlier to the later period for a large share of counties.

In contrast, winter wheat shows much weaker movement over time, with distributions centered closer to zero and no comparable downward shift, suggesting little systematic change in volatility for that crop. The consistency of the corn, cotton, and soybean patterns across detrending methods indicates that the volatility decline is robust to reasonable trend specifications. From an insurance modeling perspective, these results suggest that yield risk may not be stationary over long horizons, and simulation frameworks calibrated using earlier high-volatility decades may overstate contemporary dispersion in yield shocks.

To examine how these patterns vary geographically, Figure 2 maps county-level relative changes in yield volatility for corn (top row) and soybeans (bottom row), alongside corresponding yield growth patterns. The spatial results indicate that volatility changes are geographically clustered rather than randomly distributed. The most pronounced declines in volatility are concentrated across the Upper Midwest and central Corn Belt, including parts of the Dakotas, Minnesota, Iowa, and Illinois, where yield outcomes have become more stable in the later period. In contrast, counties with relatively higher or increasing volatility are more prevalent in peripheral and climate-exposed regions, including portions of the western and southern Plains (eastern Colorado, western Kansas and Nebraska, and Oklahoma), as well as some counties along the Southeast Coast and in Texas. Overall, these patterns suggest that yield risk has evolved unevenly across space, with broad stabilization in core production regions alongside persistent or rising uncertainty in more vulnerable areas.

Figure 2: Spatial Patterns of Changes in Yield Volatility and Yield Growth for Corn and Soybeans

Source: Agricultural Risk Policy Center (ARPC), using data from the U.S. Department of Agriculture, National Agricultural Statistics Service (USDA NASS).

Figure 2 also highlights a clear geographic association between long-run yield growth and declining volatility. Counties experiencing stronger yield gains over time, particularly in the Upper Midwest and central Corn Belt, also tend to exhibit the largest reductions in yield volatility in the later period. In contrast, regions with more modest yield growth often show smaller declines in volatility or localized increases in dispersion. This positive spatial relationship suggests that productivity improvements and yield stabilization have occurred jointly but unevenly across the agricultural landscape, underscoring important regional differences in the evolution of production risk that are relevant for crop insurance modeling and systemic exposure assessment.

The observed pattern of declining yield volatility alongside stronger yield growth across core production regions may reflect a combination of technological progress and evolving climate conditions. In northern areas such as the Dakotas and Minnesota, recent warming and longer growing seasons may have contributed to improved yield potential and greater stability, while more severe water scarcity in parts of the western Plains, including Colorado, may help explain more persistent volatility in marginal regions (EPA, 2021). At the same time, adaptation through production technologies may also play a role. Expanded irrigation practices in regions such as the Mississippi Delta provide greater control over moisture conditions and can reduce downside yield risk (Hrozencik and Aillery, 2021), while increasing adoption of precision agriculture technologies, including GPS guidance systems across the Upper Midwest and central Corn Belt, may improve input management and reduce production uncertainty (McFadden et al., 2023). Al- though this brief does not apply rigorous causal inference to identify mechanisms, these climate and technology factors provide plausible explanations for the joint emergence of higher productivity and smoother yield outcomes over time. Future research is needed to more formally disentangle these channels and assess their implications for agricultural risk modeling and insurance program design.

In summary, the findings of this brief indicate that county-level yield volatility has generally declined over time for major crops such as corn, cotton, and soybeans, while remaining more stable for winter wheat, with substantial geographic heterogeneity across regions. These results suggest that yield risk is not stationary over long horizons, and that calibration of crop insurance simulation frameworks should account for evolving trends in volatility and tail exposure rather than relying mechanically on long historical windows. Failure to adjust for these shifts may overstate contemporary risk and distort projections of indemnity exposure and systemic program costs. Incorporating time-varying yield risk measures and updated shock distributions will improve the accuracy and policy relevance of insurance evaluation. Future re- search is needed to better understand the drivers of these changes and their implications for program design under evolving climate and production systems.

References

EPA. (2021). Seasonality and Climate Change: A Review of Observed Evidence in the United States. U.S. Environmental Protection Agency, EPA 430-R-21-002. www.epa.gov/climate-indicators/seasonality-and- climate-change

Hrozencik, R. Aaron and Marcel Aillery (2021). Trends in U.S. Irrigated Agriculture: Increasing Resilience Under Water Supply Scarcity, EIB-229, U.S. Department of Agriculture, Economic Research Service.

Li, Junkan and Francis Tsiboe (2025). A Horse Race Comparison of County-Level Crop Yield Prediction Methods. ARPC Brief 2025–14. Agricultural Risk Policy Center, North Dakota State University.

McFadden, Jonathan, Eric Njuki, and Terry Griffin (2023). Precision Agriculture in the Digital Era: Recent Adoption on U.S. Farms, EIB-248, U.S. Department of Agriculture, Economic Research Service