Climate Change: The Fiscal Risks Facing The Federal Government/Crop Insurance

1. Crop Insurance



The Federal fiscal burden of providing subsidized crop insurance to American farmers could increase by billions of dollars each year by late-century due to the effects of climate change. The figure depicts estimated percentage increases in total premium subsidies in 2080 in an unmitigated climate change scenario compared to a future characterized by historical weather patterns. Estimates are graphed for three crops under five global change models.[1]


Climate Change and Crop Insurance edit

Climate change is already affecting agricultural production and negative impacts are, on average, expected to grow more severe over the course of this century. Some effects may be positive—higher levels of carbon dioxide in the atmosphere tend to increase plant growth (so-called “CO2 fertilization”) and water-use efficiency. However, negative effects from increased extreme heat and drought, more intense precipitation and soil erosion, growing stress from disease and pests, shifting soil moisture and water availability for irrigation, and higher concentrations of ozone are generally expected to outweigh positive effects, reducing yields and increasing uncertainty for producers (Melillo et al., 2014; Marshall et al., 2015).

The Federal Government provides subsidized crop insurance to American producers to cover yield and revenue losses due to natural causes (weather, fire, disease, and wildlife) and market price changes. In 2015, more than 1.2 million individual policies were issued. These policies covered more than 120 crops across nearly 300 million acres, for a total Federal liability of more than $102 billion. Three crops—corn, soybeans, and wheat—account for two-thirds of insured acres and roughly three-quarters of total premium costs. By law, crop insurance premiums must be “actuarially fair”—calibrated to match the value of total expected losses on insured acres. However, the Federal Government currently pays for almost two-thirds of crop insurance premiums on average, at a cost of more than $6 billion in 2015.

Climate-related production shocks like drought are the dominant driver of crop insurance program indemnities (Wallander et al. 2013). Climate change could affect the Federal Government’s crop insurance subsidy costs in a number of ways—most clearly by increasing the riskiness of crop production due to the impacts of shifting weather patterns and climate disruptions on yield, or the impacts of climate-related production challenges at home and abroad on crop price volatility. However, in some instances crop vulnerability could also decline due to the physiological response of crops to higher CO2 levels. Mean production levels could also increase or decrease, affecting the total liabilities covered by the crop insurance program.

Risk Assessment edit

Modeling conducted by the USDA Economic Research Service (ERS) for this assessment indicates that unmitigated climate change[2] could increase annual crop insurance premium subsidy costs for corn, soybeans, and wheat by 40 percent by 2080 compared to a projected reference scenario characterized by historical weather patterns. This estimate is the average premium subsidy increase across the five GCMs used by USDA for the assessment. It assumes the average portion of total premiums paid by the government does not change over time, which implies that current law and current average coverage rates are both held constant. In a mitigation scenario that assumes some GHG reductions, the average projected cost increase for the crop insurance program across the five GCMs is about 23 percent.

The absolute fiscal impact of such an increase will depend largely on the total liabilities insured by the program in 2080, a product of future trends in agricultural productivity and global crop demand. In the 2080 “no climate change” reference scenario, the gross revenue for corn, soybeans, and wheat is $223 billion, compared to $122 billion in 2012. This modeling baseline is consistent with an annual growth rate of approximately 1 percent for both crop yields and demand.[3] In the 2080 reference scenario, the total premium subsidy for these crops is $10.6 billion, which, relative to the $5.4 billion actual subsidy in 2012, mirrors the increase in total revenue.

Given this baseline, the fiscal impact of modeled increases in premium subsidies would be $4.2 billion each year in the unmitigated climate change scenario, the equivalent of approximately $1.0 billion each year in today’s economy. Three of the five GCMs produce estimated increases between $2.4 billion and $4.9 billion. HadGEM and GISS results provide upper and lower bounds at $9.3 billion and $17 million, respectively, the equivalent of approximately $2.3 billion and $4 million, respectively, in today’s economy.

The GISS model yields weather patterns with significantly smaller increases in temperature and significantly more precipitation than HadGEM given the same emissions pathway. The GISS model also provides weather data at a coarser spatial resolution, but the possible effect of differences in resolution on modeled yield variability, if any, still needs to be explored. Excluding GISS results would push the total estimated increase in premium subsidies from $4.2 billion to $5.2 billion. In addition, if global crop demand growth is appreciably higher than assumed, the upper bound could reach into the tens of billions each year.

The increase in subsidy costs across the GCMs is driven by an increase in total premiums predominantly due to both higher yield risk in most regions and higher price risk faced by all producers due to climate change, as well as an increase of about 5 percent in the value of production compared to the reference scenario. However, a few additional factors also affect costs. First, to maintain expected profits despite shifting climate conditions, the land allocation model predicts that risk-neutral producers will adapt in some cases by expanding acreage into higher risk areas that produce a higher expected return under climate change but also have higher premium rates per dollar of revenue insured. Second, as explored in greater detail in Marshall et al. (2015), climate change leads to less irrigated area in most regions. This result tends to increase premium rates as dryland production is generally costlier to insure per unit of production (because irrigation is itself a form of insurance). Finally, acreage shifting between crops due to climate change in some areas may have the effect of reducing producers’ crop diversification, thereby increasing the risk of total revenue losses.

Changes in Mean and Standard Deviation of Calibrated Soybean Yield by REAP Region for RCP 8.5

The dot plot above shows the percent change in mean yield and percent change in the standard deviation of U.S. soybean yield in the unmitigated climate change scenario compared to the reference scenario. Each dot represents a modeling region, and the size of the dot corresponds to the number of acres in production in that region. Dots above the 45 degree line have an increasing coefficient of variation (CV), a measure of variability per unit of crop production insured (standard deviation divided by mean). CV is highly correlated with premium rates. The plot clearly shows that far more of the regions are above the 45 degree line than below, indicating that yield variability (as indicated by CV) increases in most cases in the unmitigated climate change scenario.

Illustrative Shift in Yield Distribution

For this study, the simulated changes in means and standard deviations are calibrated to historical yields to preserve risk that is unrelated to weather and climate. Note that this calibration procedure involves a number of important and untested assumptions about future crop yields and, in particular, the nature of idiosyncratic (non-weather-related) yield risk. The calibration procedure is discussed in the Technical Supplement accompanying this report.

Since the crop insurance program insures against expected yield (or revenue), shifts in mean yields can be as important in changing yield risk as shifts in yield variability. Most regions see both a reduction in mean soybean yield and an increase in the variability of yield, which leads to increases in production risk (shown in the illustration above). In some regions, yield variability actually declines, but proportionately less so than mean yield, resulting in a net increase in risk. Some regions also see a reduction in risk (below the 45 degree line), including some regions where the standard deviation of yield increases but average yields increase proportionately more, which leads to a decline in risk.

This assessment builds on prior ERS modeling of climate change impacts on crop yield, cost, and production nationwide (Marshall et al., 2015) by estimating not only mean yields and prices but also yield risk and price risk when producers optimize planting decisions based on expectations but are exposed to weather variability—both as observed historically and as affected by various climate change scenarios. ERS then estimates total premiums and premium subsidies for revenue protection policies—the most popular insurance product for producers of major field crops.

For more information about the biophysical and economic crop production and acreage allocation models used for this assessment, see Marshall et al. (2015). For more information about the modifications made to these models for this assessment and the premium estimation methods, see the Technical Supplement accompanying this report.

Key Limitations and Uncertainties edit

The cost to the Federal Government of the crop insurance program over the course of this century will depend upon many factors, including climate change. Market conditions and technology will determine the total value of production. For example, a combination of strong demand growth and strong crop yield growth that continues historical trends would result in higher gross revenues, which in turn imply higher liabilities and therefore higher premiums and associated subsidies. The design of the insurance program and farmer participation decisions will also determine program costs. This assessment isolates the impact of climate change by assuming baseline levels of demand and supply growth, and holding program design and farmer participation decisions constant.

Estimates of the increase in crop insurance premiums due to climate change vary considerably across GCMs, reflecting sensitivity to variable climate change projections (e.g., changes in regional temperatures and precipitation patterns). In addition, the impacts of climate change on crop yield risk vary significantly by region; yield risk even decreases in some regions in the climate change scenarios. However, there is strong agreement across the GCMs that climate change will increase both price risk and yield risk in aggregate at the national level. The GCMs also demonstrate a high degree of consistency with respect to the direction of change in yield risk within regions. In particular, yield risk is increasing in much of the Corn Belt across GCMs, and decreasing in a portion of the Northern Plains. ERS also found reasonable consistency between the biophysical crop model and two alternative econometric crop yield models estimated on the same baseline weather data. Finally, while there is a fairly wide spread in fiscal impact estimates across GCMs simulations, four of the five models produce climate outcomes under which total premiums increase on the order of billions of dollars each year.

In addition to uncertainty stemming from the GCMs, the biophysical and economic crop production and acreage allocation models have several limitations that could cause estimates to be too high or too low. First, the models may underrepresent the full impact of climate change. The models capture the direct effects of changing temperature and precipitation patterns and CO2 fertilization, but the crop production results are calibrated to hold constant the effects of other climate-related impacts on crops such as those due to pests, disease, exacerbated ozone concentrations, and the frequency of certain kinds of storms such as tornadoes, hurricanes, and flooding. The timeframe used to simulate weather conditions (40 years) was selected to capture the 30-year return frequency of major droughts, but may not provide a good measure of extreme risk—such as changes in the probability of a 1-in-100 year or 1-in-1,000 year mega-drought. The models also do not place constraints on irrigation water supply, even though ERS has found that irrigation water supply will decline significantly in some regions due to climate change (Marshall et al., 2015); irrigated acres currently represent roughly 15 percent of total insured acres for principal field crops.

Second, the models do not capture changes in global crop prices due to climate-related events outside of the United States. For example, a decline in wheat production abroad due to rising temperatures could put upward pressure on global wheat prices, increasing the value of the insured wheat crop and associated crop insurance premiums in the United States.

Third, the models likely underrepresent the potential for adaptation by producers and the agricultural sector in general. For example, although crop productivity is assumed to increase year over year in both the reference and climate change scenarios due to general technological advancement, the possibility for technological improvements that may affect resilience to climate change is not represented. Some adaptive responses could reduce yield risk. For example, a considerable body of current research is focused on improving crop drought tolerance. However, as seen in the modeling results, other adaptive responses could actually increase yield risk in exchange for higher expected (mean) returns. The models assume that producers are risk-neutral and make decisions only to maximize expected profits.

Finally, the models do not consider changes in crop insurance subscription or coverage levels. Preliminary analysis suggests two potentially offsetting effects. On the one hand, an increase in risk may prompt farmers with crops that are not currently insured (roughly 15 percent of nationwide planted acreage of principal field crops in 2015) to purchase some level of coverage. This effect would increase total premiums and premium subsidies. On the other hand, increases in risk raise the actuarially fair price of insurance, which may induce farmers to purchase lower levels of coverage to reduce their total premium expenditure. This effect would reduce total premiums and premium subsidies.


  1. The Hadley Centre Global Environment Model (HadGEM), Community Climate System Model (CCSM), Canadian Earth System Model (CanESM2), Model for Interdisciplinary Research on Climate (MIROC), and Goddard Institute for Space Studies model (GISS) are global change models from the framework of models used by the IPCC to assess future changes in climate conditions in different emissions scenarios.
  2. The unmitigated climate change scenario modeled is the IPCC’s Representative Concentration Pathway (RCP) 8.5, in which emissions continue to rise throughout the century, causing radiative forcing to increase by 8.5 W/m2 relative to preindustrial levels. The mitigation scenario modeled is RCP 4.5, in which emissions peak around 2040 and then decline.
  3. Note this baseline is a projection for modeling purposes only and is not an official forecast.