Using insurance data to quantify the multidimensional impacts of warming temperatures on yield risk

Climate change significantly impacts crop production, affecting global food security. While warmer temperatures are known to negatively affect crop yields, disentangling the direct (e.g., heat stress) and indirect (e.g., soil moisture deficit) mechanisms remains challenging. Most research focuses on changes in average yields, neglecting the impact on yield variability and risk. Unexpected yield shortfalls can cause price spikes, social unrest, and economic hardship for farmers. This study addresses this gap by examining the effects of warming temperatures on yield risk, specifically focusing on the relative contributions of heat stress and drought, and the implications for crop insurance costs. The study utilizes a unique dataset of crop insurance indemnity payments – the Cause of Loss (COL) database maintained by the Risk Management Agency (RMA) – combined with data from the Summary of Business (SOB) database. This approach allows for a multidimensional assessment of warming temperature impacts, including the propensity of below-expected yield and revenue outcomes, the relative contributions of heat stress versus drought stress, and the implications for crop insurance pricing. The analysis focuses on U.S. corn and soybean production from 1989 to 2014, encompassing approximately 30,000 county-year observations.

Existing literature demonstrates the sensitivity of crop yields to climate change factors like temperature, precipitation, and CO2 concentration. Studies show that warmer temperatures negatively impact crop yields at regional and global scales, acting through both direct (heat stress) and indirect (soil moisture deficit) mechanisms. However, disentangling these effects from large-scale datasets is difficult. While previous work has focused on changes in average yields, the impact of warming on yield variability and risk has received less attention. This is crucial because yield variability leads to revenue shocks that can affect farm investments and even bankrupt farmers, impacting food supply. The growing global crop insurance sector is also significantly affected, with large indemnity payments increasing program costs. Recent studies have begun to focus on the variability of farm production in the context of climate change; however, detailed analysis of cause-specific impacts and implications for insurance are lacking. This study aims to fill this gap by leveraging the detailed cause-of-loss information provided by the RMA's dataset.

The study uses data from the RMA's Cause of Loss (COL) and Summary of Business (SOB) databases, merged to generate county-level aggregate and cause-specific loss-cost ratios (LCRs). The LCR is defined as the ratio of indemnity payments to liabilities, representing a measure of yield risk. The data cover U.S. corn and soybean production from 1989 to 2014 across approximately 1700 counties, excluding irrigated areas. The LCR data are matched with growing-season aggregates of precipitation and temperature exposure. Multivariate regression analysis is employed to estimate the relationship between LCRs and weather covariates, controlling for county-level time-invariant confounders (county fixed effects), year fixed effects, and state-specific time trends. The analysis focuses on five types of LCRs: total LCR and four cause-specific LCRs (heat, cold, drought, and excess moisture). A piecewise linear function of temperature exposure and a quadratic function of cumulative precipitation are used to capture the nonlinear relationships. The model includes county fixed effects to control for unobserved factors like soil quality, year fixed effects and state-specific trends to control for time-varying factors like changes in the federal crop insurance program. The impacts of a uniform 1°C warming are simulated using the estimated coefficients from the regression models, considering both aggregate and cause-specific LCRs. Robustness checks are conducted using different temperature variable specifications, subsets of insurance products, and additional control variables. To account for potential adaptation in crop production and insured liabilities, the analysis also incorporates a long-difference model, a cross-section model, and models with 5- and 10-year moving averages of the variables.

The analysis reveals a strong positive relationship between warming temperatures and yield risk. For corn, a 1°C increase in temperature is associated with a 32% increase in the total LCR. For soybeans, the increase is 11%. The cause-specific analysis shows that drought is the primary driver of increased losses under warming temperatures. For corn, a 1°C warming is associated with a 92% increase in drought losses and a 105% increase in heat losses, while excess moisture and cold-related losses decrease. For soybeans, drought losses increase by 59% and heat losses by 105%, while excess moisture and cold-related losses also decline. However, the increase in drought and heat losses outweighs the decrease in other loss types, leading to a net increase in overall losses. Spatial heterogeneity in the impact of warming is evident, with southern counties experiencing more adverse effects than northern counties. Robustness checks using various model specifications and accounting for potential adaptation confirm the main findings. Even with adaptation considered, warmer temperatures are shown to increase production risk and premium rates.

The findings highlight the significant impact of warming temperatures on yield risk for corn and soybeans. The increase in production risk has important implications for farm-level decision-making. Farmers, facing riskier outcomes, may reduce input expenditures or increase risk-reducing inputs, impacting overall food supply. The spatial heterogeneity in the impact of warming emphasizes the need for location-specific adaptation strategies. The study's strength lies in its ability to quantify the relative impacts of different causes of yield losses, providing valuable insights for adaptation and policy. The results indicate that drought losses are considerably larger than heat losses, implying a need to prioritize drought mitigation strategies. The increased production risk associated with warming temperatures will significantly increase the cost of operating government-supported crop insurance programs, potentially necessitating adjustments to policy design and subsidy levels.

This study demonstrates that warming temperatures increase yield risk for corn and soybeans in the United States. Drought is identified as the primary driver of increased losses. The findings highlight the importance of considering yield risk in assessing the impacts of climate change on agriculture and the need for location-specific adaptation strategies. Future research could focus on refining the analysis using higher-resolution data, exploring the interactions between temperature, precipitation, and CO2, and investigating farmer adaptation strategies in greater detail. The insights from this research can inform the design of more effective crop insurance programs and policies to mitigate the economic and social impacts of climate change on agriculture.

The study's reliance on insurance data might not capture all yield losses, as some farmers may not be insured or may not file claims for smaller losses. The cause-of-loss assignments are based on claims adjusters' and producers' interpretations, potentially introducing subjective bias. The analysis assumes unchanged precipitation and CO2 levels in response to warming, which might oversimplify the situation. Farmer adaptation is partially accounted for in robustness checks but more sophisticated modelling approaches could offer a better understanding of adaptation's role.