Abstract
Purpose
Area-based insurance plans trigger payments based on losses which may not match actual loss experience at the farm level, an issue often referred to as basis risk. The purpose of this paper is to quantify the basis risk associated with the Supplemental and Enhanced Coverage Option (SCO and ECO) crop insurance programs, and the risk reduction that can be achieved when these area-based plans are added to farmers’ risk management portfolios.
Design/methodology/approach
This study utilizes simulation techniques to build a stylized model for representative farms at the county-level for non-irrigated corn and soybean production. We model farms for each county in the 17 states included in USDA’s Crop Progress Reports for corn and soybeans, which comprise more than 90% of planted acreage for those crops. Yield and price data from the USDA’s National Agricultural Statistics Service (NASS), futures price data and insurance premiums from the Risk Management Agency are used to calibrate the simulation model.
Findings
Area-based plans may provide (1) insufficient coverage for actual losses, which is a risk management concern or (2) payments exceeding actual losses, which is a program efficiency concern given federal support for the insurance program. The risk of insufficient coverage (under-compensation) can be reduced by increasing the coverage level of the area plans, but that also increases the likelihood of support exceeding actual loss experience (over-compensation). The scale of basis risk associated with the area plans differs by region and crop due to differences in yield risk. Area plans do have the potential to provide additional risk reduction; however, risk reduction is inversely related to the level of basis risk.
Originality/value
To the best of the authors’ knowledge, this study is the first to focus on quantifying the basis risk associated with the relatively new supplemental area options (SCO, ECO) currently available in the US federal crop insurance program. It provides important insights which could inform current and future Farm Bill debates as policymakers consider modifications and enhancements to commodity and crop insurance programs. It also provides useful information to help educate farmers and other stakeholders about the use of SCO and ECO in their risk management plans.
Keywords
Citation
Tsay, J.-H. and Paulson, N.D. (2024), "Quantifying basis risk associated with supplemental area-based crop insurance", Agricultural Finance Review, Vol. ahead-of-print No. ahead-of-print. https://doi.org/10.1108/AFR-10-2023-0145
Publisher
:Emerald Publishing Limited
Copyright © 2024, Juo-Han Tsay and Nicholas D. Paulson
License
Published by Emerald Publishing Limited. This article is published under the Creative Commons Attribution (CC BY 4.0) licence. Anyone may reproduce, distribute, translate and create derivative works of this article (for both commercial and non-commercial purposes), subject to full attribution to the original publication and authors. The full terms of this licence may be seen at http://creativecommons.org/licences/by/4.0/legalcode
Introduction
Under the Federal Crop Insurance Program, farmers have both individual (farm-level) and area-based (county-level) coverage plan options. Individual plans provide coverage levels from 50% to 85% of a farmer’s guarantee while area plans can increase the farmer’s overall coverage level up to 95% of a county-based guarantee. In the 2014 Farm Bill, there were significant changes to both commodity and crop insurance programs. Most commodity programs from previous Farm Bills were discontinued and replaced with new programs [1]. Two new area supplemental, area-based insurance coverage options, the Supplemental Coverage Option (SCO) and Stacked Income Protection Plan (STAX) [2], were added to the Federal Crop Insurance Program. Starting with the 2021 crop year, another new supplemental area program, the Enhanced Coverage Option (ECO), was also introduced. Farmer uptake in the initial years of availability suggests relatively small but growing interest in the SCO and ECO programs (Paulson et al., 2022). Farmers and other stakeholders could benefit from information and analysis to better understand how the relatively new policies might impact risk exposures, and how or if they should be incorporated into the farm’s overall risk management plan.
In contrast to traditional area-based plans of insurance (e.g. Area Risk Protection Insurance or ARPI), both SCO and ECO provide add-on coverage to supplement a required underlying individual plan of insurance [3]. SCO provides a band of coverage from 86% of a county-based guarantee down to a farmer’s underlying individual plan of insurance coverage level. SCO coverage mimics that of the underlying individual policy (i.e. county revenue or yield based). Losses on SCO are triggered when actual county revenue or yield falls below 86% of the county guarantee, with a maximum loss equal to the band of coverage (i.e. 6% when SCO is combined with and 80% individual plan); the indemnity paid to an individual farmer is the county-based percentage loss applied to their individual farm’s expected crop value [insurance price times the farm’s actual production history (APH) yield]. Eligibility to use SCO is also tied to a producer’s commodity program choice [4]. This limitation, among other factors, has been associated with the relatively low initial uptake of the product.
In the 2021 crop year, ECO was made available for purchase on 31 spring-planted crops including corn, soybean and wheat. The coverage level options for ECO are either 95% or 90% down to 86% of the county guarantee. Similar to SCO, ECO coverage mimics the underlying individual policy and the county-based percentage loss is applied to the farm’s expected crop value in calculating the indemnity payment. However, use of, and eligibility for, ECO is not limited in any way by commodity program choice.
The potential for, and benefits of, area-based insurance coverage is not a new concept to the agricultural economics literature. Halcrow (1949) first proposed an area yield insurance plan, providing farmers with an indemnity only when average yields across all farms in the area fall below a critical yield. There are several advantages of area-based designs over individual plans of insurance. Area-based programs can be delivered at lower administrative costs. Moreover, area-based designs can improve actuarial fairness and decrease the potential for asymmetric information issues such as moral hazard and adverse selection (Chambers, 1989; Miranda, 1991; Smith et al., 1994; Belasco et al., 2020). Furthermore, some of this previous work has shown that the Group Risk Plan, the first area yield crop insurance plan offered in the US federal crop insurance system and now under the ARPI umbrella, can provide risk reduction at least as effectively as individual yield plans. However, individual insurance plans provide additional benefits and flexibility such as the ability to insure at sub-farm levels and coverage for prevented planting or replanting (Barnett et al., 2005). Glauber (2013) noted the participation rate of area plans had been relatively low, likely due to farmers preferring crop insurance policies that directly cover their individual losses.
Since area-based program payments are triggered based on losses at the county-level, indemnities may not always match losses experienced at the farm-level. This issue is commonly referred to as basis risk. We consider two dimensions of basis risk: under-compensation (indemnities are insufficient to cover actual farm-level losses) and over-compensation (indemnities exceed what is needed to cover actual farm-level losses).
Under-compensation results in insufficient coverage at the farm level and concerns over the effectiveness of risk protection offered by the supplemental area plans. For example, an individual farm may experience a loss due to a localized weather event while generally favorable conditions across the county result in no loss occurring at the county level. In this situation, a county-based product would not trigger an indemnity to cover the farmer’s individual loss.
Over-compensation refers to the opposite scenario where generally unfavorable conditions occur in an area resulting in an indemnity payment being triggered for a county-based product while the farmer may have experienced smaller or no losses on their individual farm. Over-compensation results in concerns related to program efficiency, particularly since the crop insurance program is supported with public resources through partial subsidization of the premiums farmers pay and administration and risk-sharing costs borne by the federal government.
More broadly, index [5] insurance programs have been common approaches considered in developing countries where the information needed to accurately design and rate individual insurance may be lacking. While index or area-based designs can be implemented and offered more easily and at lower cost, multiple studies have shown that basis risk can reduce demand for index insurance products (Giné et al., 2008; Binswanger-Mkhize, 2012; Elabed et al., 2013; Elabed and Carter, 2015; Clarke, 2016; Yu et al., 2019). Studies focused on basis risk in the context of domestic crop insurance in the USA have focused on Pasture, Rangeland, Forage Rainfall Index (PRF-RI) and similar agroclimatic-based products.
The imperfect correlation between the Rainfall Index (RI) and forage yield risk results in significant basis risk and indemnities from the index products which are often insufficient to offset realized forage production losses (Maples et al., 2016). Keller and Saitone (2022) used the Normalized Difference Vegetation Index as a proxy for realized forage yield in California. They found the probability that an insured producer suffered a loss and no indemnity was made from PRF-RI can range from 31% to 46%. Tsiboe et al. (2023) found substantially less basis risk among county-level yield products compared to localized single-covariate agroclimatic products. They attributed the reduction in basis risk may be due to area yield plan indemnities being associated with actual area-wide harvest estimates which implicitly account for the full set of seasonal factors impacting yield, while the agroclimatic products focus on a single, localized event within the growing season. Area-based measures, while still imperfect, are likely better proxies for a producer’s production than other types of indexes (Bulut, 2022).
Despite the index-based design, the use of federal RI policies has grown substantially in recent years with total insured acres increasing from 75 million in 2018 to over 297 million in 2023. As use has grown, subsidies associated with RI products have increased by over 300% over that timeframe, approaching $1bn in 2023 (RMA-USDA, 2024). Moreover, proposals from the Senate and House Ag Committees for the 2024 Farm Bill include provisions which could substantially increase farmer use of the SCO program through higher subsidy rates and coverage levels (House Committee on Agriculture, 2024; Senate Committee on Agriculture, 2024). As interest, and access to, area-based insurance designs continue to grow, farmers, government agencies and policymakers and the insurance industry can benefit from additional analysis of the basis risk associated with these products and how it might vary across commodities and regions.
The objective of this paper is to quantify the basis risk (under-compensation and over-compensation) and risk reduction when adding SCO and ECO to farmers’ risk management portfolios, focusing on dryland corn and soybean production across the major production regions of the USA. We utilize simulation techniques to develop a stylized, county-level model for representative corn and soybean farms across the USA. We define measures of both dimensions of basis risk as the probabilities of under-compensation and over-compensation, respectively. The coefficient of variation and expected shortfall (i.e. value at risk) are used to measure and assess risk reduction. While existing literature has quantified the basis risk of weather-based index insurance products (Keller and Saitone, 2022; Tsiboe et al., 2023), we contribute to the literature by quantifying the basis risk and risk reduction associated with coverage combinations which include supplemental area plans relative to benchmark farm-level coverage.
Our results illustrate the classic tradeoff between Type I (under-compensation) and Type II (over-compensation) errors. Increasing the coverage level of the supplemental area plan reduces the likelihood of under-compensation (insufficient payments from area plans to cover farm-level losses) but increases the likelihood of over-compensation (indemnity payments from the area plan will be larger than what is needed to fully cover farm-level losses). We also find differences in basis risk across regions of the USA that can be tied to differences in absolute and relative yield variability. Counties in the heart of the Corn Belt tend to have lower likelihoods of under-compensation and higher likelihoods of over-compensation compared to other corn and soybean production regions. Finally, our results show that the supplemental area plans do have the potential to provide additional risk reduction to farmers, but the extent of that risk reduction is arguably marginal and inversely related to the level of basis risk for the area.
This is one of the first analyses to focus on the basis risk associated with the SCO and ECO programs and provides important and timely insights into these program designs for the current 2024 Farm Bill debate. Further enhancing the SCO and ECO programs has been part of the discussion surrounding changes to current commodity and crop insurance programs. The consideration of basis risk, which impacts the effectiveness of area-based insurance designs in meeting risk management objectives, is critical to these discussions.
Data and methodology
This study utilizes simulation techniques to model a representative farm at the county level for the 17 states included in USDA’s Crop Progress Reports for corn and soybeans as of 2021 [6]. These 17 states represent at least 90% of the 2020 corn and soybean acreage. This section provides a brief overview of the simulation process which was implemented using the MATLAB software package.
Data used to calibrate the simulation model include county-level and state-level yields from the USDA National Agricultural Statistics Service (NASS) from 1972 to 2020, historical futures harvest prices, US Marketing Year Average (MYA) prices over the same period and projected insurance prices and volatility factors from the USDA Risk Management Agency (RMA) for the 2021 crop year. Futures harvest prices for corn and soybean are the average settlement prices on the December and November contracts, respectively, during October. For county yield data, we focus on non-irrigated yields and exclude counties with less than 30 years of data.
County yields
The NASS county-level yield data for corn and soybean were detrended to 2021 equivalents using simple linear regression with a time trend following Vedenov et al. (2004) and Paulson et al. (2008). We fit Weibull distributions to the detrended county yields and obtained estimated shape and scale parameters by using the maximum likelihood method. Five thousand draws of simulated county-level yields for each county are then generated from the fitted Weibull distributions.
Previous research has focused on assessing the statistical fit and economic implications associated with alternative yield distribution assumptions. Sherrick et al. (2004) conducted goodness-of-fit tests across a wide range of distributions that have been assumed in the literature, finding the Weibull and beta distributions to be superior for corn and soybeans. Therefore, the Weibull distribution was selected for this analysis to represent both county and farm-level yield distributions.
Farm-level yields
For our baseline scenario, we assume an average (expected) yield for the representative farm equal to the average detrended county yield. We calibrate farm-level yield variability (standard deviation) to match the crop insurance premiums for 85% Yield Protection (YP) in that county (Coble and Dismukes, 2008). The 85% YP insurance premiums for each representative farm were generated using RMA procedures. The representative farms are assumed to be 100-acre enterprise units. The use of premiums to calibrate farm-level yield variability makes use of the implicit assumption that RMA rating procedures are actuarially fair. We limit our analysis to the following practices and types: non-irrigated corn for grain and commodity soybeans with no type specified. For our baseline scenario, trend-adjusted actual production history (T-APH) yields for the farms are set to equal the average detrended county yields. The expected farm-level yields and calibrated standard deviations for yields are then used to define the method of moments parameters for the Weibull distribution [7].
Insurance prices
Following RMA and assuming lognormality, we generate distributions for corn and soybean insurance prices based on the projected insurance prices and volatility factors for the 2021 crop year. The corn projected price is $4.58/bushel and the corn price volatility factor is 0.23. The soybean projected price is $11.87/bushel with a price volatility factor of 0.19.
Yield and price correlations
We imposed rank correlation structures between the simulated county- and farm-level yields and insurance prices using the method outlined by Iman and Conover (1982). The advantage of the technique is that it preserves the marginal distributions for each random variable and simply re-sorts the random variates to achieve the target correlation structure. Target correlations between county yields and insurance prices are based on the historical correlation between state-level yields and harvest insurance prices. The target correlation between farm- and county-level yields is set at 0.8 for all counties in our baseline scenario [8].
Marketing year average (MYA) prices
We regressed historical MYA prices on historical harvest futures prices to obtain coefficient estimates and residuals. We then drew 5,000 observations from the normal distribution implied by the residuals from the simple price regression. Simulated MYA prices were then generated using the estimated regression coefficients, simulated errors and simulated harvest prices from the previous step.
Insurance indemnity and revenue distributions
The simulation algorithm generates 5,000 draws of yields and prices for a representative farm in each county with a target correlation structure based on historical price and yield relationships imposed. These draws represent 5,000 different potential outcomes or realizations for the representative farm. The outputs for each draw are simulated county-level yields, simulated farm-level yields, simulated insurance (harvest futures) prices and simulated cash (MYA) prices. We use these outputs to calculate farm and county revenue measures and insurance indemnities for the various insurance plans considered.
County revenue is defined as insurance price times county yield and is used in calculating area-based insurance indemnities. Both SCO and ECO indemnities are calculated by applying the county-level loss to the farm’s individual expected crop value (insurance price times the farm’s T-APH yield), which effectively scales the area plan payments to the farm’s yield. In our baseline, this distinction is not relevant as the representative farm’s T-APH yield is assumed to equal the expected county yield. However, we also examine alternative scenarios where the farm’s T-APH yields are assumed to be above and below the county’s expected yield as part of our sensitivity analysis.
Individual insurance indemnities are calculated using farm yields and insurance prices. Insurance premiums are assumed to be actuarially fair and set to equal the average of the indemnity distribution for each policy considered. Net insurance indemnities refer to the net payment received by the farmer and are calculated as the indemnity less the farmer-paid premium, where farmer-paid premiums are the implied fair premiums adjusted by the appropriate subsidy rate assuming enterprise units are used by the farmer.
Gross farm revenue is revenue from farm production and is calculated as the cash price times the farm yield. Net farm revenue is gross farm revenue plus net insurance indemnities for the insurance portfolio being considered.
Basis risk
The SCO and ECO are area-based policies where indemnity payments are triggered when realized county revenues or yields fall below county guarantees. Since indemnity payments are not based on direct measures of individual farm losses, it is possible that SCO and ECO could result in indemnities being paid when individual losses are not realized or indemnities being paid which are insufficient to cover actual farm-level losses. Again, this potential mismatch in loss coverage between farm and area insurance products is often referred to as basis risk (Miranda, 1991; Smith et al., 1994).
We focus on two measures of basis risk, cases where farm-level losses are not fully covered by the supplemental area plan (under-compensation) and those where the supplemental area plan triggers payments that exceed farm-level losses (over-compensation) [9]. Cases of under-compensation are of primary concern to farmers who include a supplemental area-based plan in their overall insurance coverage portfolio. Farm-level losses could occur due to localized perils that do not result in sufficient county-level losses to trigger payments. Over-compensation scenarios, due to the subsidization of federal crop insurance policies, are primarily a public policy concern. County-level losses could trigger indemnity payments that exceed farm-level losses or even if the farmer does not experience a loss at all. Indemnity payments in over-compensation scenarios represent income transfers of taxpayer dollars which exceed losses experienced by the farmer.
Under-compensation
To measure the risk of the farmer being under-compensated by an area plan combination, we compute the frequency with which indemnity payments from the coverage alternatives are less than those from benchmark coverage. This frequency measure of under-compensation is also referred to as a false negative probability (FNP).
Revenue protection with a coverage level of 85% (RP85) is assumed to be the benchmark against which actual farm-level losses are defined, while alternatives are the combinations of a lower coverage-level individual plan (80% RP) and supplemental area coverage. Our choice of benchmark was driven by the fact that RP is by far the most popular individual plan of insurance and farmers, particularly in the Midwest, tend to choose high coverage levels (Schnitkey et al., 2021a, b) [10]. Furthermore, the use of RP85 as our benchmark and comparing it to alternatives with total coverage levels that exceed 85% (86% for combinations with SCO and up to 95% for combinations that include ECO) will result in conservative estimates of the probability of under-compensation.
We compare the RP85 baseline to three alternatives: RP80 with SCO (RP80 + SCO), RP80 with SCO and 90% ECO coverage (RP80 + SCO + ECO90) and RP80 with SCO and 95% ECO coverage (RP80 + SCO + ECO95). The policy alternatives and benchmarks are shown in Table 1.
Since the introduction of the supplemental plans, some farmers may have considered lowering the RP coverage level to, for example, 80% and then using SCO and ECO to provide an overall coverage level that could range from 86% to as high as 95%. Previous research supports the magnitude of this “buydown” effect of being in the 5 to 10 percentage point (PP) range as this will tend to maximize the benefit of higher subsidy rates on the supplemental area plan options while also balancing the loss in risk protection in shifting to an area-based program for a portion of the farmer’s overall coverage (Bulut and Collins, 2014). Moreover, buydown strategies can also achieve total premium costs that are comparable to what the producer would pay for their benchmark policy and previous work has shown evidence for budget considerations in farmers’ insurance choices (Bulut, 2018).
For example, in 2021, RP80 with SCO was shown to generally result in some premium cost savings relative to RP85 for most corn and soybean farm situations across Illinois and other Midwestern states. RP80 with SCO and ECO90 generally resulted in slightly lower to slightly higher farmer-paid premium costs compared with RP85 while providing a higher overall coverage level of 90%, albeit based on the county-level trigger. Using ECO95 increases the total coverage level to 95% but also generally results in a total premium cost that is considerably higher compared with RP85 coverage (Schnitkey et al., 2021a; Tsay et al., 2021a, b).
Other adjustments in producers’ underlying individual plans of insurance may also be made as the use of the supplemental area plans is considered. For example, shifting from Revenue Protection (RP) to Revenue Protection with the Harvest Price Exclusion (RPHPE) at a similar coverage level can also result in premium savings which can then be put toward a supplemental plan to increase the overall coverage level for a producer’s insurance portfolio. Farmers with higher cost structures may find this strategy preferable to achieve a higher revenue guarantee even with some basis risk associated with the supplemental plan’s band of coverage (Schnitkey et al., 2022). The menu of options available to farmers in terms of packaging policies and choosing coverage levels to define a coverage portfolio is extensive and complex. We limit our analysis to comparing the most popular style of coverage (revenue coverage with the harvest price increase offered through RP) used by major crop producers to alternative combinations that involve what has been suggested to be the most likely form of buydown for individual coverage as the supplemental options are added.
Over-compensation
We also quantify the probability of over-compensation as the frequency of indemnity payments for the policy alternatives being greater than indemnity payments from a benchmark coverage, referring to this measure as a false positive probability (FPP). For FPP, we utilize three different benchmark policies: RP85, RP90 and RP95. RP85 remains the benchmark to assess over-compensation for the RP80 + SCO alternative. For the alternative combinations with SCO and ECO90 or ECO95, we construct RP90 and RP95 benchmarks, respectively.
The use of RP85 as the benchmark loss measure for these alternatives would be inappropriate given the alternatives result in total coverage levels of 90% and 95%. If a farmer chooses to add ECO to their insurance coverage portfolio, they are paying for additional coverage resulting in an overall coverage level of 90 or 95%. Therefore, for the purpose of measuring over-compensation, we define our benchmarks as individual policies with the same overall coverage level as those alternatives. While RP90 and RP95 are not available plans of insurance in the current crop insurance program, our simulation approach allows us to easily calculate indemnities (and implied fair premiums) for such policies. The policy scenarios for the false positive probability analysis are also shown in Table 1.
Risk reduction
We also analyze revenue risk in the context of combining individual coverage with supplemental coverage options versus individual coverage on its own. Two different summary statistics of the farmer’s revenue distribution are reported to quantify the risk reduction of alternative crop insurance choices: the coefficient of variation (CV) and the expected shortfall (ES).
The first risk reduction measurement is CV, which can be expressed as follows:
To provide a comparison of tail-risk, or downside risk, we also analyze the ES measure. ES is the conditional expected value of the revenue distribution in the tail for the worst
Results
Examples of farmer-paid premiums, implied from the simulations in our baseline scenario, for the coverage combinations considered are given in Table 2. Champaign County (low yield risk) and Monroe County (high yield risk) in Illinois are presented as specific examples. Implied premiums for each policy combination are calculated as the average indemnity across the simulations.
Again, total farmer-paid premiums for RP80 + SCO will typically be lower than for RP85. This is the case for the representative farms in Champaign and Monroe counties of Illinois. Farmer-paid premiums for RP85 and RP80 + SCO + ECO90 are very similar for Champaign County, while the RP80 + SCO + ECO90 premium is cheaper per acre than RP85 for Monroe County. Premium costs further increase for the RP80 + SCO + ECO95 combination, exceeding the costs of RP85 coverage alone for both corn and soybeans in Champaign County and for corn in Monroe County. The RP80 + SCO + ECO95 combination is actually a few dollars per acre cheaper than RP85 for soybeans in Monroe County. Areas and crops with higher farm-level yield variability relative to the county will tend to see larger potential premium savings in substituting away from individual coverage and replacing with supplemental area coverage [11]. This illustrates the tradeoffs among farm- and county-level yield variability, premium costs and coverage that must be considered by farmers, with reality providing a very large and seemingly complex list of potential options.
The probability of simultaneous losses (indemnities triggered from both the alternative coverage combination and benchmark coverage of RP85) ranges from around 35% to 45%, on average, across most states considered for both corn and soybeans [12]. These simultaneous loss probabilities do not vary significantly across alternative coverage combinations in our baseline scenario, nor in our sensitivity analysis scenario that considers lower farm-county yield correlation. Simultaneous loss probabilities actually tend to be on the lower end of the range for states in the Corn Belt region, and higher in other areas. While this may suggest greater basis risk in the Corn Belt, decomposing the mismatch in losses between cases of under-compensation versus over-compensation provides a much more nuanced interpretation.
The risk of under-compensation is shown to be lower in the main Corn Belt region, whereas the risk of over-compensation, at least in terms of frequency, is greater in the Corn Belt. Comparing across the two crops considered, the risk of under-compensation (over-compensation) tends to be greater (lower) for soybeans across most areas. Additional risk reduction potential associated with the alternative coverage combinations also tends to be greater for corn than soybeans. Again, we attribute this to the larger amount of farm-level yield variability compared with county-level yield variability that exists for soybeans (refer to footnote 10).
Below, we present summaries for FNPs (or under-compensation basis risk) and FPPs (or over-compensation basis risk) from our simulation model under different policy scenarios. We also calculate average indemnity payment differences between the coverage comparisons when under-compensation or over-compensation occurs. Lastly, we show the results associated with the risk reduction measures of the CV and ES at 10%.
Indemnity payments and under-compensation of farm losses
To quantify the risk of under-compensation, the indemnity payments from the policy combinations using the supplemental area plans are compared against indemnities for RP85 coverage. RP85 is assumed to be the benchmark comparison, and the true measure of whether farm-level losses occurred. Figure 1 shows our FNP estimates for all counties in the 17 states included in our analysis. These represent the probability that alternative coverage indemnities are less than those from RP85 or the frequency with which a farmer might be under-compensated for realized losses exceeding 15% of expected revenue. The maps in Panels (a), (c) and (e) are FNPs for corn, while the maps in Panels (b), (d) and (f) are FNPs for soybeans. The darker color shades in the maps indicate larger estimated FNPs.
Counties located in the heart of the Corn Belt – southern Minnesota through northern and central Iowa, Illinois, Indiana and western Ohio – have lower FNP estimates. In these regions, farmers tend to choose coverage levels on individual plans at or above 80% due to lower production risks and lower relative premium costs than areas outside the heart of the Corn Belt (Schnitkey et al., 2021b). Major production areas outside the Midwest are associated with higher FNP estimates or greater risk of being under-compensated for realized losses. Estimates of FNPs decline across all regions when ECO90 or ECO95 are added to the alternative coverage combination.
For corn, the average probability of RP85 triggering an indemnity is 45% across all included counties. The average FNP estimate for RP80 + SCO is 37%. The average probability of no indemnity from the RP80 + SCO combination when RP85 would trigger payments is 3.7%. Average FNP estimates decline to 28% and 12% when ECO90 and ECO95, respectively, are added to the alternative coverage plan. The probability of no indemnity from the coverage combinations with ECO when RP85 would trigger payments falls to 3% (with ECO90) and 1.7% (with ECO95) across all counties.
For soybeans, the average probability of a farm loss (RP85 triggering an indemnity) is also approximately 45% across all counties. The average FNP for RP80 + SCO is 42% and declines to 34% and 26% when ECO coverage is added at 90% and 95%, respectively. The average estimated probability of no payment from RP80 + SCO when farm losses occur is 3.4%. When ECO coverage is added this declines to 2.8% with ECO90 and 1.8% with ECO95. The marginally higher FNP estimates for soybean, compared with corn, are attributed to the higher farm-level yield variability relative to the country for soybeans. Lower relative yield variability at the county level for soybeans implies fewer instances of losses being triggered by the supplemental area plans relative to corn and smaller losses when they occur, resulting in a higher frequency of cases where a farm may be under-compensated.
Figure 2 shows an average measure of the difference in indemnity payments ($/acre) triggered by the alternative coverage combinations to those from the benchmark coverage, conditional on under-compensation occurring [13]. The darker color shades in the maps indicate larger payment differences or a larger level of average under-compensation. Similar to Figure 1, panels (a), (c) and (e) in Figure 2 are for corn while panels (b), (d) and (f) are for soybeans.
For corn, the average level of under-compensation between for RP80 + SCO is about $9.60 per acre. The average level of under-compensation declines to $7.00 and $4.40 per acre when ECO90 and ECO95 are added, respectively. For soybeans, the average level of under-compensation of RP80 + SCO is similar to that for corn at $9.70 per acre. Average under-compensation declines to $7.60 and $5.20 per acre with ECO90 and ECO95, respectively. Counties in the Corn Belt region tend to have lower estimated levels of under-compensation by the area coverage combinations for both corn and soybeans, implying lower basis risk in terms of both the frequency and level of under-compensation of realized farm revenue losses.
Indemnity payments and over-compensation of farm losses
To estimate the probability of over-compensation of farm losses, or FPPs, the frequency of simulated outcomes where indemnities from the coverage combinations exceed those from the single individual policy benchmarks (RP85, RP90 and RP95) were computed. Figure 3 illustrates the estimated FPPs (i.e. payments from the alternative policy combinations exceed realized farm losses). Panels (a), (c) and (e) are estimated FPPs for corn; Panels (b), (d) and (f) report estimated FPPs for soybeans. Midwest counties in the Corn Belt region tend to have the highest estimated FPPs, and the probability of over-compensation tends to increase as the higher coverage level supplemental plans are added to the representative farms’ overall coverage portfolios.
For corn, the RP80 + SCO combination over-compensates losses with an average probability of 17%. The probability the RP80 + SCO combination triggers a payment when no farm-level loss occurs (RP85 indemnities are zero) is 8.6%. Indemnities from RP80 + SCO + ECO90 exceed farm-level losses with a 17% probability. The probability of over-compensation increases to 23% for the combination with ECO95. The probability of indemnities from the ECO90 and ECO95 combinations when no farm loss occurs are 10% and 12.7%, respectively. For soybeans, the likelihood of over-compensation from RP80 + SCO is 11%. This increases to 12% for the ECO90 combination and 18% for ECO95. Indemnities from RP80 + SCO when no farm level loss is realized occur with a 6.5% probability. This increases to 8.4% for ECO90 and 11.7% for ECO95. Lower frequencies of over-compensation for soybeans compared with corn can also be tied to the difference in relative county-yield variability. Lower county yield variability for soybeans makes it less likely that area-based payments will be triggered, reducing the likelihood of farm losses being over-compensated by the alternative coverage combinations.
Figure 4 provides the average level of over-compensation, taken as the average difference in indemnities conditional on farm losses occurring. On average, the level of over-compensation from RP80 + SCO is about $4.40 per acre across corn counties. This increases to $7.70 and $13.60 per acre for the ECO90 and ECO95 combinations. For soybeans, over-compensation averages $2.60 per acre for RP80 + SCO and increases to $4.80 and $9.20 per acre for ECO90 and ECO95.
Sensitivity analysis
In addition to our baseline scenario, we also provide a summary of FNP and FPP results for a range of alternative scenarios which adjust key simulation parameters. Specifically, we consider a lower correlation between farm- and county-level yields (0.4 vs 0.8 in the baseline), farm-level APH yields which differ from the county expected yield (high and low APH scenarios with farm APH yields set at 125% and 75% of the county expected yield used in the baseline, respectively), and alternative insurance price volatility levels (high and low price volatilities set at the maximum and minimum volatilities experienced for corn and soybeans from 2011 to 2021). State-level average FNP and FPP estimates for the RP80 + SCO vs RP85 comparison are provided in Columns 2–6 of Tables 3 and 4 [14].
Assuming a lower correlation between farm and county yields increases the risk of both under- and over-compensation. The probability of under-compensation tends to increase each state-level average by less than 1 pp for both corn and soybeans. Estimated FPPs increase by larger amounts, ranging from 2 to 6 pp for both crops. Increases in both FNP and FPP estimates are expected as a lower correlation between farm and county yields increases the potential mismatch between the county-based loss trigger and farm-level experience.
The high APH yield scenarios result in slightly smaller FNP estimates and slightly larger FPP estimates while the opposite is true for the low APH scenario. Reductions (increases) in FNP estimates range from 2 to 4 pp (3–5 pp). Increases (reductions) in FPP estimates are smaller in magnitude, not exceeding 2 pp on average for most states. These results are also as expected as both SCO and ECO are designed to scale indemnity payments to the farm’s individual APH yield. Losses are determined, on a percentage basis, using the county-based trigger. The percentage loss factor is then applied to the farm’s expected crop value. Thus, higher (lower) APH farms will have their payments scaled up (down). Larger payments for high APH farms will tend to reduce cases of under-compensation and increase cases of over-compensation while the opposite is true for low APH farms.
Finally, the high and low price volatility scenarios result in somewhat mixed effects on FNP and estimates relative to the baseline scenario. The FNP estimates with high price volatility decline, on average, for all states while the low price volatility scenario results in a mix of FNP estimate changes. The impacts are relatively small, with changes from the baseline being less than 1 pp for all states and both crops. The impact of the price volatility scenarios on FPP estimates is larger, with increases in FPP estimates ranging from 3 to 6 pp for the high price volatility scenario and from 5 to 11 pp reductions in FPP estimates for the low price volatility scenario. Higher (lower) price volatility will tend to increase (reduce) the likelihood and potential size of both farm- and county-level losses. Since the same price measure is used for the individual area plans it is somewhat intuitive that this does not have a large impact on the FNP measure of basis risk. The larger changes to FPP are also intuitive since higher price volatility would scale up payments from coverage combinations utilizing the supplemental area plans in outcomes with price-driven revenue losses. The lower variability in area/county yields relative to farm yields implies higher county yields would tend to offset low prices less for the coverage combinations (i.e. RP80 + SCO) than farm yields might for individual coverage (i.e. RP85).
Risk reduction
In addition to quantifying basis risk, we also assess whether the supplemental area plan alternatives compare with the RP85 benchmark in terms of risk reduction. We use two measures of risk: CV and ES for the net farm revenue (gross revenue plus net insurance indemnities) distributions.
Figures 5 and 6 present the risk comparisons using CV and ES at 10%, respectively. Counties in green suggest the area plan combinations provide greater risk reduction compared to the benchmark policy of RP85. Red counties indicate areas where the area combination results in more risk than the benchmark policy of RP85.
From Figures 5 and 6, relative risk reduction improves as higher coverage levels for the supplemental area plans are included. In terms of CV, the RP80 + SCO combination provides higher relative risk reduction than RP85 in very few soybean counties and for some corn counties mainly concentrated in the Corn Belt region. Adding ECO to the alternative coverage plans improves risk reduction across more counties, with the majority of areas achieving lower net revenue CVs with ECO95.
When measuring risk using the 10% ES (i.e. downside risk), we find most individual and area plan combinations for corn result in more risk reduction than RP85. For soybeans, the coverage level on the supplemental area plan needs to be at the 90% level (i.e. ECO95) in order to obtain significant risk reduction relative to RP85 on its own.
Managing tail revenue risk is exactly what revenue insurance is designed to do. Individual coverage, such as RP85, should effectively cut the left tail of the revenue distribution off at 85% of the farm’s expected revenue less the farmer-paid insurance premium. Reducing the individual portion of coverage results in a lower cutoff point at 80% of expected revenue less the farmer-paid premiums, which can only reduce a tail risk measure such as ES. Whether adding the supplemental area coverage plans to the lower individual coverage can result in a net improvement depends on the tradeoff between basis risk and the net increase in overall coverage level. The 86% overall coverage level of RP80 + SCO provides a relatively small increase from the 85% coverage level of RP such that the negative basis risk effect could dominate. This is what the results indicate to be the case for most soybean counties until the ECO95 policy is added to the coverage portfolio being compared with RP85. For corn, the basis risk effect dominates the higher coverage levels for some counties when comparing RP80 + SCO to RP85, but the higher coverage level begins to dominate for all counties once the ECO option is added to the coverage portfolio.
Conclusions and policy implications
The introduction of the SCO and ECO insurance plans following the 2014 and 2018 Farm Bills has offered producers new, and increasingly complex, crop insurance options. Debate and the formation of proposals for program modifications for the 2024 Farm Bill are underway, and the continuation of support for the strong US federal crop insurance program is widely supported by commodity advocacy groups and policymakers. Proposals to further strengthen supplemental area-based coverage options, through increased subsidy support for SCO in particular, have been part of this process [15].
These area-based insurance plans trigger payments based on county loss experience, creating the potential for basis risk which refers to the potential mismatch between the loss measure used by the area-based products and realized farm-level losses. We use a stylized county-level simulation model to quantify the basis risk associated with the SCO and ECO for the major corn and soybean production areas in the USA. Specifically, we consider two aspects of this potential mismatch: cases where the coverage combinations which utilize the supplemental area plans under-compensate for farm losses and cases where farm losses are over-compensated by the area plan combinations.
Our results show that basis risk varies regionally based on relative levels of production or yield risk. The likelihood of individual and area plan combinations resulting in indemnities that are insufficient to cover actual losses at the farm-level are lowest in the main corn belt region for corn and soybeans; the risk of under-compensation increases in areas with greater yield variability. Proposals which further strengthen supplemental area-based programs have the potential to provide greater risk reduction benefits to farmers in areas of lower basis risk. Furthermore, subsidy incentives which make area-based designs more attractive for all potential buyers, could result in inferior insurance coverage scenarios from a risk management perspective, with lower individual coverage and greater area-based coverage. This could be particularly problematic for producers in areas of higher basis risk.
We also find the probability of excessive area-plan payments – those which exceed actual farm-level losses – is higher in the main corn belt region for corn and soybeans. The potential for over-compensation represents another pitfall of the basis risk associated with area-based insurance, this time at the expense of taxpayers in the form of relatively inefficient subsidization of the supplemental area policies.
While the supplemental area plans have provided farmers with the ability to add coverage to an underlying individual plan of insurance, farmers need to be aware of the potential mismatch in coverage introduced by using these area plans, particularly when it involves adjustments to what they cover with individual plans of insurance. At the same time, this study shows supplemental programs which are partially supported by taxpayers through premium discounts may also result in an inefficient allocation of resources compared to individual-level crop insurance with the same coverage level. This is an important consideration for policymakers and those involved in the development of policy proposals focused on risk management improvement for agriculture moving forward.
The results from this study are also important and relevant for extension education efforts to help inform stakeholders about existing programs and the implications of their use. Educating producers about basis risk and illustrating how it can vary across crops and regions, can help extension educators provide better information to aid in making risk management decisions.
Our use of a stylized simulation model approach leads to a number of limitations and caveats. While we examine a number of alternative scenarios to assess the impact of changes to modeling parameters and assumptions, further sensitivity analysis could provide additional results that are even more refined to the characteristics of specific regions, crops and farm situations. We focus on the non-irrigated production of corn and soybeans in the USA; our approach could be extended to examine other crops and production types and practices. We utilize a specific parametric distribution assumption for crop yields (Weibull) which could be relaxed to consider other parametric and non-parametric forms. Consideration of additional crop insurance policy and coverage level choice scenarios provides another opportunity for further extensions. We leave these as directions for continued research on farmers' crop insurance choices and the role of alternative policy options, such as the supplemental area plans, in their risk management portfolios.
Figures
Benchmark and alternative policy combinations for coverage scenarios considered
Scenario | Benchmark policy | Alternative policy combination |
---|---|---|
Under-Compensation | ||
Coverage scenario 1 | RP85 | RP80 + SCO |
Coverage scenario 2 | RP85 | RP80 + SCO + ECO90 |
Coverage scenario 3 | RP85 | RP80 + SCO + ECO95 |
Over-Compensation | ||
Coverage scenario 4 | RP85 | RP80 + SCO |
Coverage scenario 5 | RP90 | RP80 + SCO + ECO90 |
Coverage scenario 6 | RP95 | RP80 + SCO + ECO95 |
Source(s): Table created by authors
Implied farmer-paid premiums for insurance coverage scenarios, selected county examples
Illinois county | RP85 | RP80 + SCO | RP80 + SCO + ECO90 | RP80 + SCO + ECO95 | |
---|---|---|---|---|---|
Corn | Champaign | $21.54 | $14.41 | $22.45 | $36.16 |
Monroe | $36.82 | $24.04 | $30.89 | $42.23 | |
Soybean | Champaign | $19.02 | $13.51 | $20.38 | $32.14 |
Monroe | $36.14 | $22.55 | $26.40 | $33.11 |
Note(s): The implied farmer-paid premiums equal average indemnities less the premium subsidy. The representative farms have T-APH yields of 195 bu/acre for corn and 62 bu/acre for soybean (Champaign) and 158 bu/acre for corn and 45 bu/acre for soybean (Monroe)
Source(s): Table created by authors
FNP estimates by scenario, RP80 + SCO vs RP85, state-level averages
FNP difference = Alternative – Baseline | ||||||
---|---|---|---|---|---|---|
Baseline FNP | Low yield correlation | High APH | Low APH | High price volatility | Low price volatility | |
Corn | ||||||
Illinois | 0.30 | 0.013 | −0.02 | 0.04 | −0.004 | 0.000 |
Indiana | 0.31 | 0.008 | −0.02 | 0.04 | −0.006 | 0.001 |
Iowa | 0.25 | 0.010 | −0.02 | 0.03 | −0.006 | 0.003 |
Kansas | 0.45 | 0.003 | −0.03 | 0.04 | −0.003 | −0.002 |
Kentucky | 0.39 | 0.015 | −0.03 | 0.04 | −0.003 | 0.003 |
Michigan | 0.36 | 0.002 | −0.03 | 0.03 | −0.004 | 0.006 |
Minnesota | 0.31 | 0.005 | −0.02 | 0.04 | −0.005 | 0.005 |
Missouri | 0.41 | 0.004 | −0.03 | 0.04 | −0.004 | −0.002 |
Nebraska | 0.39 | 0.005 | −0.03 | 0.04 | −0.003 | 0.005 |
North Carolina | 0.41 | 0.002 | −0.03 | 0.04 | −0.003 | 0.001 |
North Dakota | 0.44 | 0.001 | −0.03 | 0.04 | −0.003 | 0.000 |
Ohio | 0.33 | 0.005 | −0.03 | 0.04 | −0.005 | 0.004 |
Pennsylvania | 0.41 | 0.003 | −0.03 | 0.05 | −0.004 | 0.002 |
South Dakota | 0.42 | 0.002 | −0.03 | 0.04 | −0.005 | 0.001 |
Tennessee | 0.39 | 0.004 | −0.03 | 0.03 | −0.004 | 0.002 |
Texas | 0.48 | 0.000 | −0.04 | 0.04 | −0.003 | −0.003 |
Wisconsin | 0.38 | 0.003 | −0.03 | 0.05 | −0.005 | 0.006 |
Soybeans | ||||||
Arkansas | 0.51 | 0.000 | −0.03 | 0.03 | −0.002 | 0.000 |
Illinois | 0.34 | 0.007 | −0.03 | 0.04 | −0.003 | −0.003 |
Indiana | 0.36 | 0.004 | −0.03 | 0.04 | −0.004 | 0.001 |
Iowa | 0.33 | 0.004 | −0.03 | 0.04 | −0.004 | 0.001 |
Kansas | 0.50 | 0.000 | −0.03 | 0.02 | 0.000 | −0.003 |
Kentucky | 0.44 | 0.000 | −0.03 | 0.05 | −0.003 | −0.001 |
Louisiana | 0.51 | 0.000 | −0.02 | 0.03 | −0.002 | 0.000 |
Michigan | 0.44 | 0.000 | −0.03 | 0.04 | −0.003 | 0.001 |
Minnesota | 0.40 | 0.001 | −0.03 | 0.03 | −0.003 | 0.002 |
Mississippi | 0.49 | 0.000 | −0.04 | 0.05 | −0.002 | −0.001 |
Missouri | 0.47 | 0.000 | −0.03 | 0.05 | −0.002 | 0.000 |
Nebraska | 0.42 | 0.001 | −0.03 | 0.04 | −0.003 | 0.001 |
North Dakota | 0.48 | 0.000 | −0.04 | 0.05 | −0.004 | −0.003 |
Ohio | 0.40 | 0.000 | −0.03 | 0.04 | −0.004 | 0.001 |
South Dakota | 0.45 | 0.000 | −0.03 | 0.04 | −0.003 | −0.001 |
Tennessee | 0.44 | 0.000 | −0.03 | 0.05 | −0.002 | 0.002 |
Wisconsin | 0.46 | 0.000 | −0.03 | 0.04 | −0.002 | 0.000 |
Note(s): FNP estimates are the frequency with which the RP85 benchmark indemnities exceed alternative coverage indemnities
Source(s): Table created by authors
FPP estimates by scenario, RP80 + SCO vs RP85, state-level averages
FPP difference = Alternative – Baseline | ||||||
---|---|---|---|---|---|---|
Baseline FPP | Low yield correlation | High APH | Low APH | High price volatility | Low price volatility | |
Corn | ||||||
Illinois | 0.20 | 0.024 | 0.01 | −0.01 | 0.06 | −0.11 |
Indiana | 0.19 | 0.021 | 0.01 | −0.01 | 0.06 | −0.11 |
Iowa | 0.21 | 0.024 | 0.01 | −0.01 | 0.05 | −0.10 |
Kansas | 0.14 | 0.045 | 0.02 | −0.01 | 0.05 | −0.08 |
Kentucky | 0.17 | 0.035 | 0.01 | −0.01 | 0.05 | −0.09 |
Michigan | 0.16 | 0.031 | 0.01 | −0.01 | 0.04 | −0.09 |
Minnesota | 0.18 | 0.025 | 0.01 | −0.01 | 0.05 | −0.09 |
Missouri | 0.16 | 0.043 | 0.02 | −0.01 | 0.05 | −0.09 |
Nebraska | 0.17 | 0.014 | 0.01 | −0.01 | 0.05 | −0.10 |
North Carolina | 0.14 | 0.055 | 0.02 | −0.01 | 0.04 | −0.06 |
North Dakota | 0.13 | 0.047 | 0.02 | −0.01 | 0.04 | −0.06 |
Ohio | 0.18 | 0.028 | 0.01 | −0.01 | 0.05 | −0.10 |
Pennsylvania | 0.15 | 0.041 | 0.02 | −0.01 | 0.05 | −0.08 |
South Dakota | 0.15 | 0.036 | 0.02 | −0.01 | 0.05 | −0.09 |
Tennessee | 0.16 | 0.035 | 0.02 | −0.01 | 0.05 | −0.09 |
Texas | 0.11 | 0.062 | 0.02 | −0.01 | 0.03 | −0.05 |
Wisconsin | 0.16 | 0.022 | 0.01 | −0.05 | 0.04 | −0.09 |
Soybeans | ||||||
Arkansas | 0.08 | 0.038 | 0.01 | −0.01 | 0.03 | −0.05 |
Illinois | 0.15 | 0.024 | 0.00 | −0.01 | 0.05 | −0.09 |
Indiana | 0.14 | 0.024 | 0.00 | −0.01 | 0.04 | −0.09 |
Iowa | 0.14 | 0.030 | 0.00 | −0.01 | 0.04 | −0.09 |
Kansas | 0.09 | 0.049 | 0.01 | 0.00 | 0.03 | −0.05 |
Kentucky | 0.11 | 0.040 | 0.01 | −0.01 | 0.03 | −0.06 |
Louisiana | 0.07 | 0.056 | 0.00 | −0.01 | 0.02 | −0.03 |
Michigan | 0.11 | 0.038 | 0.01 | −0.01 | 0.03 | −0.06 |
Minnesota | 0.11 | 0.040 | 0.01 | −0.01 | 0.04 | −0.06 |
Mississippi | 0.08 | 0.050 | 0.01 | −0.02 | 0.02 | −0.05 |
Missouri | 0.10 | 0.041 | 0.01 | −0.01 | 0.03 | −0.06 |
Nebraska | 0.12 | 0.025 | 0.00 | −0.01 | 0.04 | −0.08 |
North Dakota | 0.09 | 0.051 | 0.01 | −0.02 | 0.03 | −0.04 |
Ohio | 0.12 | 0.034 | 0.00 | −0.01 | 0.04 | −0.07 |
South Dakota | 0.10 | 0.044 | 0.01 | −0.01 | 0.03 | −0.06 |
Tennessee | 0.11 | 0.045 | 0.01 | −0.01 | 0.03 | −0.06 |
Wisconsin | 0.10 | 0.045 | 0.01 | −0.02 | 0.03 | −0.06 |
Note(s): FPP estimates are the frequency with which the alternative coverage indemnities exceed RP85 benchmark indemnities
Source(s): Table created by authors
Notes
The Agricultural Risk Coverage (ARC) and Price Loss Coverage (PLC) commodity programs were created by the 2014 Farm Bill, replacing the Average Crop Revenue Election (ACRE) program and counter-cyclical payments. The ARC program is a revenue-based program which provides options for coverage of county-based or farm-level revenues. The PLC program is a price-based program where the payments are issued when the MYA price is less than a fixed reference price. ARC and PLC are farm commodity programs intended to cover prolonged, multi-year declines in revenue or price which differs from the intra-year risks covered by crop insurance guarantees set annually based on current market prices.
STAX was specifically designed for cotton producers; however, policymakers are currently considering modifications to SCO to make it more similar to STAX as part of the 2024 Farm Bill. This would potentially create a “STAX-like” option for producers of a broader range of crops.
Specifically, a farmer must have underlying coverage through a RPHPE or YP plan to be eligible for SCO or ECO.
It is available on acres where the farmers chose PLC, but acres covered by the ARC program are not eligible. For the 2014 Farm Bill, commodity program choice was made once and effective for all years covered by the Farm Bill. In 2018, these rules were relaxed so that commodity program choices can be changed each year, creating more flexibility and potential for SCO use from year to year.
We refer to index programs as a broader category of products which are designed around a loss trigger which differs from but is a proxy for, the direct risk or loss measure being insured. Area-based insurance programs, such as the SCO and ECO programs which are the focus of this paper, are viewed as examples of index products where the index is an aggregated area measure of the direct risk being covered (i.e. the county revenue index as a proxy for farm-level revenue).
Corn states include Illinois, Indiana, Iowa, Kansas, Kentucky, Michigan, Minnesota, Missouri, Nebraska, North Carolina, North Dakota, Ohio, Pennsylvania, South Dakota, Tennessee, Texas and Wisconsin. Soybean states include Arkansas, Illinois, Indiana, Iowa, Kansas, Kentucky, Louisiana, Michigan, Minnesota, Mississippi, Missouri, Nebraska, North Dakota, Ohio, South Dakota, Tennessee and Wisconsin.
The method of moments approach here refers to setting the two Weibull shape parameters to define a distribution whose first and second moments (mean and standard deviation) match the target expected yield and yield variation value for each representative farm.
We chose 0.8 as the level of correlation between farm and county yields in our baseline scenario as it is near the upper bound of farm-county yield correlations cited in Barnett et al. (2005). Use of the 0.8 correlation level likely results in estimates which understate the true level of basis risk in some areas and likely overstates the amount of risk reduction that could be achieved with the supplemental area plans being analyzed. We also include a summary discussion of results from a scenario with a lower level of county-farm yield correlation (0.4, closer to the lower bound of the range reported in Barnett et al., 2005) in our sensitivity analysis section.
The aspect of basis risk that we refer to as under-compensation is also analogous to a “false negative” or Type I error, while over-compensation can be thought of as a “false positive” or Type II error (Elabed et al., 2013).
Farmers in regions outside of the Midwest tend to elect lower average coverage levels. The RP85 concept as a benchmark measure for actual farm-level revenue losses still applies to areas that elect lower coverage. However, for regions where producers have historically elected lower individual plan coverage levels, considerations of policy alternatives which combine supplemental area plans with lower coverage level individual plans (i.e. RP75 + SCO + ECO) to RP85 would provide basis risk measures that are better tailored to producers’ actual experience. In general, both measures of basis risk (under-compensation and over-compensation) would be expected to increase as the individual coverage level declines and supplemental area plans cover a wider band of the overall coverage.
The ratio of farm-level to county-level yield variability in our simulations is, in general, lower for corn than for soybeans and lower for states in the main Corn Belt region of the Midwest US compared with other areas. For example, the average yield variability ratio in Illinois is 2.19 for corn and 3.08 for soybeans. The average ratios in Nebraska are 4.58 (corn) and 5.14 (soybeans) and in North Dakota are 3.28 (corn) and 4.25) soybeans. Yield variability ratios, average at the state level, are provided in an Appendix (Table A.V.) and available from the authors upon request.
Probabilities of simultaneous losses, averaged at the state level, are available from the authors upon request. They are included as Tables A.VI and A.VII in Appendix to the manuscript.
The difference between indemnity payments from the alternative coverage combinations and the benchmark policies was found to be statistically significant for the vast majority of representative farms (counties) for both corn and soybeans. RP85 indemnities and total indemnities from the alternative combinations with ECO coverage were not statistically different for a small number of counties in southern MN and in a band extending from eastern Nebraska to Iowa, Illinois, Indiana and western Ohio. Statistical significance was tested using the Wilcoxon Rank Sum test. Results are available from the authors upon request. Maps illustrating the statistical significance of indemnity payment differences are provided in Appendix in figures A.1 and A.2.
Comparisons of estimated FNPs and FPPs for the alternative coverage combinations (those that included ECO) and the benchmark coverage follow similar patterns and can be provided by the authors upon request. We include county-level maps (Figures A.3 through A.10) and tables with the state-level average FNP and FPP estimates (Tables A.I. through A.IV.) in Appendix.
This was signaled well ahead of the formal proposals released by House Ag and Senate Ag (2024). See, for example, https://www.agri-pulse.com/articles/20171-stabenow-eyes-new-funding-for-farm-bill-crop-insurance-expansion
The supplementary material for this article can be found online
References
Barnett, B.J., Black, J.R., Hu, Y. and Skees, J.R. (2005), “Is area yield insurance competitive with farm yield insurance?”, Journal of Agricultural and Resource Economics, Vol. 30, pp. 285-301.
Belasco, E.J., Cooper, J. and Smith, V.H. (2020), “The development of a weather‐based crop disaster program”, American Journal of Agricultural Economics, Vol. 102 No. 1, pp. 240-258, doi: 10.1093/ajae/aaz021.
Binswanger-Mkhize, H.P. (2012), “Is there too much hype about index-based agricultural insurance?”, Journal of Development Studies, Vol. 48 No. 2, pp. 187-200, doi: 10.1080/00220388.2011.625411.
Bulut, H. (2018), “US farmers' insurance choices under budget heuristics”, Agricultural Finance Review, Vol. 78 No. 1, pp. 152-172, doi: 10.1108/AFR-02-2017-0009.
Bulut, H. (2022), “Area or index-based insurance plans, review and a look ahead”, Crop Insurance Today, October 2022.
Bulut, H. and Collins, K. (2014), “Designing farm supplemental revenue coverage options on top of crop insurance coverage”, Agricultural Finance Review, Vol. 74 No. 3, pp. 397-426, doi: 10.1108/AFR-08-2013-0032.
Chambers, R.G. (1989), “Insurability and moral hazard in agricultural insurance markets”, American Journal of Agricultural Economics, Vol. 71 No. 3, pp. 604-616, doi: 10.2307/1242016.
Clarke, D.J. (2016), “A theory of rational demand for index insurance”, American Economic Journal: Microeconomics, Vol. 8 No. 1, pp. 283-306, doi: 10.1257/mic.20140103.
Coble, K.H. and Dismukes, R. (2008), “Distributional and risk reduction effects of commodity revenue program design”, Review of Agricultural Economics, Vol. 30 No. 3, pp. 543-553, doi: 10.1111/j.1467-9353.2008.00429.x.
Elabed, G. and Carter, M.R. (2015), “Compound-risk aversion, ambiguity and the willingness to pay for microinsurance”, Journal of Economic Behavior and Organization, Vol. 118, pp. 150-166, doi: 10.1016/j.jebo.2015.03.002.
Elabed, G., Bellemare, M.F., Carter, M.R. and Guirkinger, C. (2013), “Managing basis risk with multiscale index insurance”, Agricultural Economics, Vol. 44 Nos 4-5, pp. 419-431, doi: 10.1111/agec.12025.
Giné, X., Townsend, R. and Vickery, J. (2008), “Patterns of rainfall insurance participation in Rural India”, The World Bank Economic Review, Vol. 22 No. 3, pp. 539-566, doi: 10.1093/wber/lhn015.
Glauber, J.W. (2013), “The growth of the federal crop insurance program, 1990-2011”, American Journal of Agricultural Economics, Vol. 95 No. 2, pp. 482-488, doi: 10.1093/ajae/aas091.
Halcrow, H.G. (1949), “Actuarial structures for crop insurance”, Journal of Farm Economics, Vol. 31 No. 3, p. 418, doi: 10.2307/1232330.
House Committee on Agriculture (2024), available at: https://agriculture.house.gov/uploadedfiles/detailed_summary_final.pdf
Iman, R.L. and Conover, W.J. (1982), “A distribution-free approach to inducing rank correlation among input variables”, Communications in Statistics - Simulation and Computation, Vol. 11 No. 3, pp. 311-334, doi: 10.1080/03610918208812265.
Keller, J.B. and Saitone, T.L. (2022), “Basis risk in the pasture, rangeland, and forage insurance program: evidence from California”, American Journal of Agricultural Economics, Vol. 12282 No. 4, pp. 1203-1223, doi: 10.1111/ajae.12282.
Maples, J.G., Brorsen, B.W. and Biermacher, J.T. (2016), “The rainfall index annual forage pilot program as a risk management tool for cool-season forage”, Journal of Agricultural and Applied Economics, Vol. 48 No. 1, pp. 29-51, doi: 10.1017/aae.2016.3.
Miranda, M.J. (1991), “Area‐yield crop insurance reconsidered”, American Journal of Agricultural Economics, Vol. 73 No. 2, pp. 233-242, doi: 10.2307/1242708.
Paulson, N.D., Babcock, B.A., Paulson, N.D. and Babcock, B.A. (2008), “Get a grip: should area revenue coverage Be offered through the farm Bill or as a crop insurance program?”, doi: 10.22004/AG.ECON.42458.
Paulson, N., Schnitkey, G., Coppess, J., Zulauf, C. and Swanson, K., “SCO and ECO insurance program use”, farmdoc daily (12):118, Department of Agricultural and Consumer Economics, University of Illinois at Urbana-Champaign, August 9, 2022.
RMA-USDA (2024), “Summary of business files, risk management agency, United States Department of Agriculture”, available at: https://rma.usda.gov/en/Information-Tools/Summary-of-Business
Senate Committee on Agriculture (2024), available at: https://www.agriculture.senate.gov/imo/media/doc/rural_prosperity_and_food_security_section-by-section.pdf
Schnitkey, G., Paulson, N., Swanson, K. and Zulauf, C., “RP, ECO, and SCO tradeoffs”, farmdoc daily (11):29, Department of Agricultural and Consumer Economics, University of Illinois at Urbana-Champaign, March 2.
Schnitkey, G., Paulson, N., Zulauf, C. and Swanson, K., “Crop insurance coverage levels and ECO use on corn”, farmdoc daily (11):71, Department of Agricultural and Consumer Economics, University of Illinois at Urbana-Champaign, May 4.
Schnitkey, G., Paulson, N., Swanson, K. and Zulauf, C., “Consider RP with harvest price exclusion in conjunction with the enhanced coverage option”, farmdoc daily (12):24, Department of Agricultural and Consumer Economics, University of Illinois at Urbana-Champaign, February 22.
Sherrick, B.J., Zanini, F.C., Schnitkey, G.D. and Irwin, S.H. (2004), “Crop insurance valuation under alternative yield distributions”, American Journal of Agricultural Economics, Vol. 86 No. 2, pp. 406-419, doi: 10.1111/j.0092-5853.2004.00587.x.
Smith, V.H., Chouinard, H.H. and Baquet, A.E. (1994), “Almost ideal area yield crop insurance contracts”, Agricultural and Resource Economics Review, Vol. 23 No. 1, pp. 75-83, doi: 10.1017/S1068280500000435.
Tsay, J., Paulson, N. and Schnitkey, G. (2021a), “Supplemental area insurance and basis risk measures: part I”, farmdoc daily (11):72, Department of Agricultural and Consumer Economics, University of Illinois at Urbana-Champaign, May 5.
Tsay, J., Paulson, N. and Schnitkey, G. (2021b), “Supplemental area insurance and basis risk measures: part II”, farmdoc daily (11):73, Department of Agricultural and Consumer Economics, University of Illinois at Urbana-Champaign, May 6.
Tsiboe, F., Tack, J. and Yu, J. (2023), “Farm‐level evaluation of area‐ and agroclimatic‐based index insurance”, Journal of the Agricultural and Applied Economics Association, Vol. 2 No. 4, pp. 616-633, doi: 10.1002/jaa2.77.
Vedenov, D.V., Miranda, M.J., Dismukes, R. and Glauber, J.W. (2004), “Economic analysis of the standard reinsurance agreement”, Agricultural Finance Review, Vol. 64 No. 2, pp. 119-134, doi: 10.1108/00214660480001158.
Yu, J., Vandeveer, M., Volesky, J.D. and Harmoney, K. (2019), “Estimating the basis risk of rainfall index insurance for pasture, rangeland, and forage”, Journal of Agricultural and Resource Economics, Vol. 44 No. 1, pp. 179-193.