Assessing Cost-effectiveness of the Conservation Reserve Program (CRP) and Interactions between the CRP and Crop Insurance

I. INTRODUCTION

Conservation effectiveness when facing resource constraints is a central theme in the environmental conservation literature. Cost-benefit analysis has been advocated as a basic tool to enhance conservation effectiveness (Ando and Mallory 2012; Murdoch et al. 2010; Arrow et al. 1996). This article focuses on the Conservation Reserve Program (CRP), the largest land retirement program supported by the U.S. federal government. It shows that the CRP’s effectiveness can be significantly enhanced by appropriate use of cost-benefit targeting and by accounting for interactions between the CRP and the federal crop insurance program (FCIP), the largest U.S. agri-cultural commodity subsidy program.

The CRP is a voluntary program that pays farmers to retire environmentally sensitive land from active production and plant it with grass or trees for a contract period of 10 to 15 years (Stubbs 2013). The program was first authorized in the Food Security Act of 1985, and enrollment reached 33.9 million acres in 1990, which accounted for about 8% of the country’s total cropland. Enrollment in the CRP ranged from 30 to 37 million acres between 1990 and 2011 but has declined to about 23.5 million acres as of November 2015 due to the favorable market environment for cropping and the subsequent reduction of the CRP acreage cap in the 2008 and 2014 farm bills. The current annual program outlay is about $2 billion (Farm Service Agency 2014). The CRP is generally considered to have succeeded in providing multiple environmental benefits including those to soil, water, wildlife, and other natural resources (Feather, Hellerstein, and Hansen 1999; FAPRI-UMC 2007; Wu and Weber 2012).

The CRP currently faces serious challenges. In light of strong demand for agricultural commodities as well as pressure to reduce the federal budget deficit, the 2014 farm bill provided legislation to reduce the maximum amount of land that can be kept in the CRP from 32 million acres to 24 million acres by 2018. Commencing about 2007, a higher crop price environment also implies that farmers will require higher rental payments to enroll their land into the CRP, which will result in higher program outlays and hence dull the program’s political appeal (Hellerstein and Malcolm 2011). These new challenges heighten the need for a CRP enrollment mechanism designed to achieve maximum environmental benefit for any given federal government budget outlay.

In this article we examine the environmental and budgetary implications of different CRP enrollment designs in the context of the program’s interactions with the FCIP. The FCIP insures more than 280 million acres of land, with an annual average liability worth about $95 billion in 2015 (Risk Management Agency 2016). It supports farmers mostly through premium subsidies, which have averaged about 60% of total premiums paid in recent years. It also includes significant amounts of administration and operating costs paid to approved private insurance companies (Shields 2015). The FCIP is predicted to cost about $8.9 billion per year over 2013–2022, making it the most expensive agricultural commodity program. It has been heavily criticized and is under intense scrutiny (USGAO 2012, 2015; Goodwin and Smith 2014).

The CRP and the FCIP can interact. The enrollment design currently used for the CRP as of 2015 is targeted at removing cropland from production if the land performs sufficiently well according to the program’s Environmental Benefits Index (EBI). Throughout this paper we refer to the EBI design used in the CRP as of 2015 as the current EBI design.¹ Factors entering the current EBI include wildlife impacts, water quality, air quality, erosion propensity, and carbon sequestration potential. Enrollment cost factors are also included where, ceteris paribus, land that commands higher rental payment will perform worse on the index and so is less likely to be accepted for enrollment. Omitted from the current EBI, however, is the crop insurance subsidy reduction that would occur were the land removed from production.² To underscore the omission’s significance, in some regions average crop insurance premium subsidies amount to significant percentages of CRP rents and even cropland cash rents. Table 1 shows that in North Dakota the state-average crop insurance subsidies were about 70% of CRP rent and 50% of land cash rent over 2008–2014. Even in Iowa, where land is fertile and has high rent, the average crop insurance subsidy can be as high as 20% of average CRP rent in some years. This indicates that some land could be taken out of production at very small budgetary cost to the government. That is, when crop insurance subsidies on a land tract are close to its CRP rent, then the saved crop insurance subsidies through enrolling the land into the CRP can largely offset its CRP payment.

Furthermore, the omission of crop insurance subsidies in the calculation of the EBI may also affect the geographical configuration of enrolled acres. A cursory inspection of any CRP enrollment map shows high concentration of enrolled acres in the southern Corn Belt, the eastern Dakotas, Montana, the southern Great Plains, and parts of the Palouse (Farm Service Agency 2013). These are in the main marginal cropland regions where CRP enrollment costs are low and benefits may be high. Regions with more marginal land tend to have higher insurance premiums per acre, which implies higher premium subsidies per acre given that subsidy is proportional to premium (USGAO 2015). Therefore, all else equal, the incorporation of subsidy savings will increase the competitiveness of offers from marginal land to be enrolled in the CRP.

These observations highlight the importance of taking crop insurance subsidies into account when designing CRP enrollment policies. Although program fund sources differ, federal taxes spent and saved have equal weight when calculating the budget deficit. Moreover, were saved crop insurance subsidies included in the index, then incentives for optimal land allocation would be strengthened because the inclusion would mitigate potential dissonance across the suite of agro-environmental policies.

Large literatures have examined the efficiency of the CRP and the FCIP separately, whereas interactions between these two programs from a CRP enrollment design perspective have received little attention. Since the CRP’s inception in 1985, the efficiency of its land enrollment designs and associated environmental and economic impacts have attracted scrutiny. Reichelderfer and Boggess (1988) and Ribaudo (1989) argue that the CRP enrollment design used in the first nine signups basically maximized acres enrolled.³ Babcock et al. (1996, 1997) compare three enrollment designs under a budget constraint: (1) enrolling land with highest benefit-to-cost ratio; (2) enrolling land with lowest costs among eligible land (akin to enrollment design for the first nine signups); and (3) enrolling land with highest environmental benefits among eligible land regardless of costs. Wu, Zilberman, and Babcock (2001) study how these alternative designs are preferred by different interest groups. However, none of these studies have considered the cost-effectiveness of the current EBI design. Although Hellerstein, Higgins, and Roberts (2015) focus on the cost-effectiveness of the current EBI design, they mainly investigate effects of setting a maximum CRP payment rate (i.e., bid cap) on the cost-effectiveness of the CRP by applying auction theory and economic experiments, which is not a focus in this paper. On the other hand, most of the considerable body of research on U.S. crop insurance has focused on issues related to product design, rate setting, and farmers’ participation decisions (e.g., Goodwin 2001; Sherrick et al. 2004; Norwood, Roberts, and Lusk 2004). A few studies have focused on the land use impacts of the FCIP (e.g., Young, Vandeveer, and Schnepf 2001; Goodwin, Vandeveer, and Deal 2004; Lubowski et al. 2006; USGAO 2007; Miao, Hennessy, and Feng 2014, 2016). Goodwin and Smith (2003) investigate the offset effect of crop insurance on CRP-induced soil erosion reduction. However, none of these studies consider interactions between CRP enrollment design and the FCIP.

To our best knowledge, this article is the first study that examines the cost-effectiveness of the current EBI design while accounting for the interactions between the CRP and the FCIP. We show that, although it tries to balance environmental benefits with rental costs, the current EBI is not cost-effective. The current EBI can be interpreted as an effort to maximize net benefit per acre, targeted where benefits measured in index points are assumed to be commensurate with land rental rates. By contrast, a cost-effective enrollment criterion requires benefit-cost ratio targeting so that environmental benefit per dollar spent is maximized. Benefit-cost ratio targeting is widely used for analysis of programs and regulations (e.g., Mishan and Quah 2007). Therefore, we identify a cost-effective EBI and examine how crop insurance savings can be included in the current EBI design and also in this cost-effective EBI. We then simulate the environmental and budgetary consequences of alternative EBI designs by using contract-level CRP offer data in Signup 26 (conducted in 2003) and Signup 41 (conducted in 2011) across the contiguous United States. Data were obtained from the Farm Service Agency of the U.S. Department of Agriculture. We select these two signups because crop prices, and hence crop insurance subsidies, were much higher in 2011 than in 2003. The large differences between insurance subsidies in these two years provide an opportunity to assess the impacts of incorporating crop insurance subsidies over a wide range of values.

The results based on the two signups suggest that adopting a cost-effective enrollment design and incorporating crop insurance subsidies into the EBI would result in significant increases in CRP acreage, total environmental benefits, and savings on crop insurance subsidies, while leaving government outlay unchanged. For example, had the cost-effective targeting enrollment design been adopted and insurance subsidies been accounted for in Signup 41, then the government could have enrolled about 50% of offered acres in that signup at zero real cost, defined as nominal CRP payment minus saved insurance subsidies. Under cost-effective targeting, CRP acreage and payments would increase in the Great Plains and the southeastern states but would decrease in the Midwest.

II. CONCEPTUAL FRAMEWORK

In this section we model the CRP’s enrollment mechanisms. To meaningfully model and assess the mechanisms, we need to first understand the goals that the CRP is intended to achieve. When the CRP was first established, in 1985, its major goal was to reduce soil erosion. For the first nine signups between 1986 and 1989, the program focused on quickly enrolling acres, not on maximizing environmental benefits, and its enrollment was consistent with maximizing total acreage for a given budget outlay (Reichelderfer and Boggess 1988; Ribaudo 1989). Commencing in 1990, multiple environmental factors were introduced and the concept of the EBI was used in order to balance environmental gains with program costs. For general Signups 10 through 13 over 1991–1995, land enrollment was intended to maximize environmental benefits per dollar of cost where the specific design of the EBI was not disclosed (Osborn 1993; Jacobs, Thurman, and Marra 2014).

The current EBI design was created after the Federal Agriculture Improvement and Reform Act of 1996. In this design, CRP rental payments requested by farmers are added to environmental components after a linear transformation. For all enrolled acres, funding is mandatory, i.e., it is not subject to annual appropriation. For each CRP offer, the FSA assigns an EBI value to it based on the offer’s environmental benefits and rental payment requested by the landowner. Then all offers are ranked according to the EBI, and offers with EBI score no less than the cut-off EBI value will be enrolled into the CRP. Here we focus on the efficiency of the current EBI design.

Let EEBI denote the environmental benefits of an offer, and r_k denote the rent per acre requested in an offer for land tract k.⁴ The EBI [1] where f (rk) is a function of the requested rental rate, and the extra bonus points are a relatively small number reflecting whether the offer requests cost share or how much the requested rental rate is below the weighted average soil rental rate (WASRR) of the offered land. Under the current CRP implementation, f (r_k) is set to be [2] where parameters a and b are determined by the program administrator. Table 2 presents values of these two parameters, the maximum possible points for the cost components, EEBI, EBI, and the cut-off EBI points for acceptance in general signups over 1997–2011. For the general sign-up in 1997, a = 190 but has been 125 since then. The value of b has increased gradually from 165 in 1997 to 220 in 2011. In addition, total available points for environmental benefits (i.e., maximum EEBI) accounted for 67% to 73% of total EBI points over 1997–2011 (Table 2). Such large percentages suggest that the current EBI puts more emphasis on environmental benefits than on program costs.

The EBI formulation in equations [1] and [2], as explained above, gives positive incentives for providing environmental services and negative incentives for requesting a high rental rate. Thus, it is an attempt to balance environmental benefits and program costs. A notable feature of the EBI in equations [1] and [2] is that it is a linear combination of costs and environmental benefits. We refer to this form of benefit targeting as “pseudo cost-effective targeting.” The term “pseudo” is used here because benefits and costs are not necessarily measured in a common unit and the EBI is not necessarily a true measure of net benefits. Before demonstrating that pseudo cost-effective targeting is consistent with an optimization problem to be described below, we first introduce some notation. Define by k∈ Ω ≡ {1,2, …, K } the kth parcel of land offered to the CRP. Let l_k denote the size of parcel k. The set of all subsets of Ω is written as P(Ω). Environmental benefits arising from land retirement amount to e_k per acre for parcel k. Therefore, if set h ∈P(Ω) is placed in the CRP, then environmental benefits amount to Σ_k∈h^ek^lk. The optimization problem that induces the currently employed CRP enrollment design (i.e., pseudo cost-effective targeting) can be written as follows:

III. SIMULATION APPROACH AND DATA

Consistent with the four optimization problems laid out in the previous section, we simulate four different scenarios. These are the Baseline scenario, under which the EBI is kept at the status quo (in line with OP1); Scenario 1, under which the EBI in the Baseline is modified to include crop insurance subsidies (i.e., OP1′); Scenario 2, under which the cost-effective EBI is employed without including crop insurance subsidies (i.e., OP2); and Scenario 3, under which the cost-effective EBI is employed and crop insurance subsidies are included (i.e., OP2′). We calculate the EBI under each scenario based on each problem’s enrollment criteria, specifically, [11] where a and b are cost parameters as in equation [2], and c represents extra bonus points.

Once we have calculated EBI points for each scenario, the offered CRP parcels can be enrolled into the CRP in descending order by each parcel’s EBI until the acreage or budget constraint is reached. We are interested in the environmental benefits, acreage, payments, and saved crop insurance subsidies obtained from the CRP under each scenario. Specifically, when h*∈P(Ω) is the selected set of parcels under a scenario, then total environmental benefits under that scenario are Σ_k∈h*l_ke_k, total acres enrolled are Σ_k∈h*l_k, total insurance subsidy savings are Σ_k∈h*l_ks_k, total CRP nominal payments are Σ_k∈h*l_kr_k, and total CRP real payments amount to Σ_k∈h*l_k×(r_k-s_k).⁶ That is, we refer to real payment as the difference between total CRP nominal payments and total saved crop insurance subsidies.

In the simulation we use contract-level CRP offer data from Signups 26 and 41 across the contiguous United States, as obtained from the FSA. These two signups occurred in 2003 and 2011 with 71,073 and 38,677 offers, respectively. The signup data available to us include detailed contract-level information such as EBI for each land tract offered to the CRP, rental rate requested by land owners, and whether an offer was accepted. Table 3 provides some summary statistics for the two signups, from which we can see that the average CRP rent between these two signups is around $50/acre. We also find that under each signup the average CRP rent required by farmers is only slightly lower than the average WASRR, the cap for the CRP rent payment.

Crop insurance premiums and premium subsidies are jointly determined by coverage type, coverage level, and crop price. Since the CRP offer dataset does not include crop insurance information for a CRP offer, a matching mechanism is developed to map each CRP offer onto an insurance subsidy. Specifically, we employ a simple polynomial curve fitting approach to map each CRP offer onto an insurance subsidy. To capture heterogeneous relationships between CRP rent and crop insurance subsidy across states, we fit a polynomial curve state by state. For each state in the dataset we first regress county-level crop insurance subsidy on county-level WASRR using data available in the state. New Mexico data are pooled with Texas data, while Wyoming data are pooled with Nebraska data because New Mexico and Wyoming have few counties enrolled in the CRP. Here we use WASRR instead of CRP rental payment because WASRR is a better measure of land productivity and hence a better measure of the land’s insurance prospects.

Suppose we estimate an Nth degree of polynomial curve [12] where S^ij and R^ij are county-level insurance subsidy and county-level WASRR in county i of state j, respectively; and , is the estimated coefficients for state j.In this study we set N=4.Setting N=3 or N = 5 yields only a negligible difference in terms of predicted insurance subsidies. In order to avoid negative insurance subsidy predictions for a CRP offer, we utilize the natural logarithm transformation of the dependent variables and independent variables in equation [12]. Upon obtaining [12], for each CRP offer we insert its WASRR value into equation [12] and estimate the offer’s projected insurance subsidy.⁷

To test for the robustness of this matching approach, we also perform an alternative matching approach in which unit-level, instead of county-level, insurance subsidy information obtained from RMA is used. We show that insurance subsidy predictions and simulation results from these two different approaches are close. See Item A in the online supplement for details.⁸

County-level insurance subsidies for the contiguous United States in 2003 and 2011 were obtained from Summary of Business Reports and Data from the U.S. Department of Agriculture’s Risk Management Agency.⁹ This dataset includes information on each insured crop such as insurance type, coverage level, insured acres, total premiums, and total subsidies. We first remove crops that are ineligible for CRP enrollment and then take the weighted average (using net insured acres as weights) of total subsidies to obtain each county’s average subsidy per acre. Table 3 presents the average crop insurance premiums and subsidies in years 2003 and 2011 based upon the matching approach described above. We can see that average crop insurance subsidy in 2011 was almost quadruple that in 2003 (i.e., $31.2/acre vs. $8.6/acre).

IV. SIMULATION RESULTS AND ANALYASIS

In Table 4, we present summary simulation results where Baseline and Scenario 1 are constrained by the actual enrollment acreage that occurred under the two signups (2 million acres for Signup 26 and 2.8 million acres for Signup 41), whereas Scenarios 2 and 3 are constrained by the actual level of CRP real payment (defined as CRP rental payment minus crop insurance subsidies saved by CRP enrollment) that occurred under the two signups (i.e., $95.6 million for Signup 26 and $50.2 million for Signup 41).¹⁰ Under Baseline (i.e., the actual CRP enrollment outcomes of the two signups), the saved crop insurance subsidy in Signup 41 is significantly higher than that in Signup 26 ($84.1 million vs. $16.9 million), although the enrolled acres are similar (2.8 million acres vs. 2 million acres). The reason for this large difference in insurance subsidy savings is that, as shown in Table 3, the average crop insurance subsidy is much higher in 2011 than in 2003.

To facilitate exposition, we specify four comparisons. Comparisons I through III compare Scenarios 1 through 3 with Baseline, respectively; Comparison IV compares Scenarios 2 and 3. When comparing the scenario outcomes we focus on (1) CRP enrollment acreage, program payment, and saved crop insurance subsidies, (2) environmental benefits from the CRP as measured by the environmental component of the EBI (i.e., physical environmental benefits, labeled as EEBI), and (3) geographic patterns in CRP enrollment changes under different designs.

Geographic Patterns in CRP Enrollment Changes

When compared with the Baseline scenario, Scenarios 1 through 3 result in noticeable geographical patterns for the changes in CRP enrollment (see Figures 1 and 2).¹² The Great Plains and southeastern United States would gain CRP acreage and payments, while the Midwest would see reductions. Accordingly, the Midwest would gain more commodity revenues and crop insurance subsidies under Scenarios 1 to 3 than under Baseline. These regional differences in CRP enrollment are important because of implications for welfare distribution and regional politics, and also because spatially distinct environmental and natural resources are concerned.

• Download figure
• Open in new tab
• Download powerpoint

FIGURE 1

Acres Enrolled into the Conservation Reserve Program (CRP) in Baseline and CRP Acreage Changes under Comparisons I, III, and IV (Signup 26)

• Download figure
• Open in new tab
• Download powerpoint

FIGURE 2

Acres Enrolled into the Conservation Reserve Program (CRP) in Baseline and CRP Acreage Changes under Comparisons I, III, and IV (Signup 41)

Table 4 also presents the amount of acres that change status under the four comparisons. That is, these acres either change from being accepted to being rejected or vice versa. Incorporating saved insurance subsidies into EBI design can result in change of status for a significant percentage of acres. In Signup 41, for instance, incorporating saved subsidies into the current EBI design would cause 10.4% of acres to change status (Comparison I in Table 4) whereas incorporating saved subsidies into the cost-effective EBI design would cause 9% of acres to change status (Comparison IV in Table 4). The largest status changes are observed under Comparison III, where 46.8% and 24.5% of total offered acres would change enrollment status in Signups 26 and 41, respectively. Figures 1 and 2 show changes in CRP acres under Comparisons I, III, and IV for Signups 26 and 41, respectively.¹³ The patterns of CRP acreage changes between these two signups are similar. CRP enrollment will increase in areas with high subsidy-to-rent ratios, such as the Great Plains and southeastern states. As discussed in Section I, these are in the main marginal cropland regions where CRP enrollment costs are low and environmental benefits may be high. What makes these locations marginal for cropping and environmentally sensitive often also renders them poor crop insurance prospects, which indicates high insurance premiums and subsidies.

By contrast, under Scenarios 1–3 the Midwest would lose CRP acreage relative to Baseline because cropland in this region requires higher CRP rental rates and receives lower crop insurance subsidies when compared with cropland elsewhere. For example, in Iowa the average CRP rent and insurance subsidy in 2011 are $128.1/acre and $27.3/acre, respectively, whereas in North Dakota these two numbers are $36.2/acre and $28.8/acre, respectively (Table 1). Notice that the CRP acreage change across regions is not significant in Comparison I because the total enrolled CRP acreage in Scenario 1 and Baseline are the same. However, large differences exist between cost-effective targeting EBI and current EBI, as shown by Comparison III (Figures 1 and 2).

Cost-effective targeting would overwhelmingly favor low-rent regions whether or not crop insurance subsidies are incorporated. Our analysis suggests that the central Corn Belt would have an even lower enrollment rate were cost-effective targeting used. Furthermore, the southern Iowa and northern Missouri border region would become less competitive compared with other regions based on EEBI points and rental cost information submitted in Signups 26 and 41. Our assessment may raise concerns that CRP enrollment under the cost-effective targeting criterion would deemphasize productive regions with erodible land, for example, southern Iowa and northern Missouri. Our finding has bearing on the ongoing debate regarding whether it is better to spare land, a common view on the role of CRP, or to share land, as with environmental initiatives on working land (Green et al. 2005). We find that a CRP enrollment design favoring concentration of the CRP on the Great Plains is a regional land-sparing policy.

Simulation Results When Constraints Vary

In the above analysis we fixed the acreage constraint for Baseline and Scenario 1 at the actual enrolled acreage in Signups 26 and 41 and fixed the budget constraint for Scenarios 2 and 3 at the actual CRP real payment in the two signups. In this subsection we vary these constraints to study whether similar conclusions hold regarding the efficiency of the four EBI designs under different constraint levels. Here we use Lorenz curves to depict effectiveness of the four EBI designs as the CRP budget changes, with a focus on three aspects of CRP enrollment outcomes: acreage, crop insurance subsidies saved, and total EEBI (see Figure 3).¹⁴ The horizontal axes in Figure 3 represent the proportion of total CRP real payment that can enroll all CRP offers in a signup. The vertical axes in the left, middle, and right columns are proportion of total acreage, proportion of total subsidies saved, and proportion of total EEBI achieved among all CRP offers in a signup, respectively.

• Download figure
• Open in new tab
• Download powerpoint

FIGURE 3

Proportion of Acreage, Subsidy Saved, and EEBI when CRP Real Payment Varies

From Figure 3 we can see that across all scenarios and signups the EBI design under Scenario 3 (i.e., cost-effective EBI design accounting for saved insurance subsidies) is the most efficient design in terms of total acreage enrolled, subsidy saved, and total EEBI associated with enrolled acreage. For Signup 26, the difference in enrollment acreage and environmental consequences between incorporating insurance subsidies into the EBI design and not doing so is insignificant because crop insurance subsidies in that signup are small. However, from the upper middle graph in Figure 3 we can see that incorporating insurance subsidies into the EBI design can increase subsidy savings, especially when enrollment is constrained at a budget less than 50% of total payment that can enroll all offers.

For Signup 41, the efficiency of EBI design can improve significantly when cost-effective EBI design is adopted and when insurance subsidies are accounted for in EBI (see the three graphs in the lower panel of Figure 3). Given the large magnitude of insurance subsidies in 2011, it is not surprising that about 35% of the total offered CRP acreage in Signup 41 can be enrolled at zero real payment under Scenario 2, as depicted by the vertical part of the Lorenz curve under Scenario 2 for Signup 41. If we further incorporate saved crop insurance subsidies into the cost-effective EBI design, then about 50% of offered CRP acreage can be enrolled at zero real payment. Similar conclusions hold for subsidies saved and total EEBI. These findings reveal the potential in efficiency improvement when cost-effective targeting EBI design is adopted and when insurance subsidies are accounted for in the design.

V. ON MOTIVATING THE CURRENT EBI DESIGN

The CRP has 30 years of history since its inception in 1985. Like many other government programs, economic efficiency is often only one factor that influences program design and implementation. The history of the CRP and current institutional constraints all play key roles. In this section we briefly discuss the development of the CRP enrollment mechanism and try to interpret its evolution, including omission of crop insurance subsidies, from the perspectives of transactional costs and of political economy. Briefly, because of transactional costs, political pressures, and many other reasons, EEBI may not be an accurate measure of a CRP offer’s environmental benefits. Therefore, we further analyze the performance of OP1 (i.e., the current EBI design, or pseudo cost-effective targeting) and OP2 (i.e., cost-effective targeting) when EEBI does not accurately measure the environmental benefits of a CRP offer.

As discussed in Section II, for the first nine CRP signups between 1986 and 1989, enrollment procedures were consistent with maximizing total acreage enrolled for a given budget. Specifically, any eligible CRP offer would be accepted if the offered rental payment was lower than a rent ceiling, which was determined after bids were submitted. After the 1990 farm bill, an EBI design based upon benefit-cost ratio was created in order to improve CRP enrollment efficiency. According to Ribaudo et al. (2001, 15), “each [EBI] factor was divided by a term representing the estimated government cost of enrolling the bid, and standardized so that each term had the same mean and standard deviation.” For general Signups 10 through 13, CRP enrollment was based on this EBI design but specifics were not publicly disclosed (Osborn 1993; Jacobs, Thurman, and Marra 2014).¹⁵

Commencing with the 1996 farm bill, the EBI underwent significant changes culminating in the current 2015 EBI design as modeled in this paper. The EBI design based on benefit-cost ratio in the early 1990s was discontinued. In the current EBI design, benefit data are aggregated in a linear manner, where linear aggregates are known to have certain agreeable contract design properties (Carroll 2015). CRP rental payments requested by farmers are added to environmental components after a linear transformation. Moreover, three environmental factors (wildlife habitat, water quality, and soil erosion) are assigned the same weights. In what follows we briefly discuss the drivers behind these changes.¹⁶

First and foremost, as discussed above, the EBI design prior to the 1996 farm bill was a “black-box” where the lack of transparency often frustrated farmers and the public (Jacobs, Thurman, and Marra 2014; Hamilton 2010, 35). The FSA also realized that this lack of transparency might impede program efficiency (Hamilton 2010, 33). Moreover, according to Hamilton (2010), the FSA struggled with assigning a dollar value to a CRP offer’s environmental benefits. Hamilton (2010, 33) notes, “In earlier discussions of the EBI, there had been proposals to try and divide benefits by costs. The difficulty of quantifying the dollar value of benefits meant that eventually cost was simply incorporated as another ranking factor, with a low bid price treated as a desirable by assigning the parcel higher ranking points.” However, as we have discussed in Section II, adding the cost factor to the EEBI implicitly assumes that the EEBI and the cost factor are measured in comparable units, whereas dividing the EEBI by costs merely assumes that measured benefits and costs scale in proportion. Therefore, the difficulty of quantifying the CRP’s dollar value of environmental benefits does not justify abandoning the benefit-over-cost approach.

The goals of reducing transaction costs and avoiding errors when measuring environmental benefits have also significantly affected EBI design (Hamilton 2010). CRP enrollment decisions are centralized, whereas EBI factors are measured and calculated by local FSA offices, most of which had little information technology support in the early 1990s (Hamilton 2010). Given the large number of offers within a signup period, typically a few weeks, local FSA offices have to process large amounts of information. Therefore, the current EBI design might be a good rule of thumb that uses approximations to achieve most of a “first best” solution where first-best may not take account of bounded rationality and transaction costs.¹⁷ According to Ralph Heimlich, influential in EBI design during the 1990s, “a more ambitious evaluation proposal would have failed because it could not have been implemented in a timely and economic fashion, but the scheme we developed only needed readily available secondary data, data on the parcel itself, … and could be done quickly and cheaply” (Hamilton 2010, 30). The advent of automation did not immediately annul the benefits of computational ease. According to Hamilton (2010, 36), “a portion of the later changes in the 2002 iteration of the EBI came from the need to simplify and automate the process. Categories not easily automated at the time were sometimes dropped. … They removed the points based on distance to water and wetlands because it was the source of much human error. Some people measured distance from the center of a lake and others started from the closest shore.”

Political pressures are another factor that shapes the current EBI design. When the FSA designed the EBI, conservation and wildlife organizations as well as other interest groups were brought into the discussions (Hamilton 2010, 31). The fact that wildlife habitat, water quality, and soil erosion factors receive a common weight in EBI design may indicate that it is an equilibrium outcome under these political pressures, where equal division is a common equilibrium in nonmarket bargaining situations (Andreoni and Bernheim 2009). Intense pressure from agricultural, environmental, and other interest groups was mediated through the political process with intent to have the U.S. Department of Agriculture adjust EBI specifications (Hamilton 2010, 30). Jacobs (2010) has found a positive correlation between a state’s political representation on congressional committees that influence the CRP and the state’s CRP acreage and payments.

In the 2014 farm bill crop insurance subsidies are linked with conservation compliance programs such as the highly erodible land conservation provision (Sodbuster), wetland conservation (Swampbuster), and grassland conservation (Sodsaver) (Orden and Zulauf 2015). However crop insurance program participation has never been included in CRP enrollment mechanisms. Several practical reasons may explain this omission. First, premium subsidy is a portion of the premium that depends directly on crop price levels and so may vary across years. Since CRP contracts typically have a duration of 10–15 years, it is difficult to quantify premium subsidy savings for a CRP offer. Second, some land eligible for the CRP may not have a crop insurance record. According to Shields (2015), about 83% of total crop acreage was covered under crop insurance in 2011. But land might be enrolled into future subsidized federal crop insurance programs even if the tract has no historical record of enrollment, so quantifying premium subsidy savings from CRP enrollment is problematic.

A further complication is that crop insurance premium subsidies have not been presented as a pure transfer to farmers, but rather as a safety net in the event of a contingency outside the course of normal events (Coble and Barnett 2013). Transfers are readily seen as a government redistribution endeavor rather than as the provision of a public good or redress of market failure. Premium subsidies have been seen as support to farmers in order to make crop insurance affordable (Coble and Knight 2002). In addition to the added cost to growers who use subsidized insurance, the crop insurance industry and agribusiness in general may dislike any codification of a transfer view of premium subsidies, such as their inclusion as a cost in EBI design.

In sum, lobbying pressure and intent to reduce transaction costs have significantly affected the EBI’s design. As we have discussed above, difficulty in assigning a dollar value to environmental benefits should not be a valid reason to abandon a benefit-over-cost approach to EBI design. Perhaps, due to methodology issues, use of a dollar value for environmental benefits is unrealistic so that some form of EEBI will always be an approximation for the environmental benefits a CRP offer provides. Innovations in information technology suggest that the technology difficulties that rationalized EBI design a generation ago no longer exist. Indeed, although insurance premium quotes for specific land parcels are all unique and the underlying ratesetting methodology is involved, growers can now obtain close estimates online at the U.S. Department of Agriculture Risk Management Agency website.

There are many other reasons that the current EEBI score system may not represent the true environmental benefits even in the absence of political pressure and transaction costs. Perhaps the greatest difficulty lies in assessing the exact environmental benefits of a CRP offer. The assessment would have to first identify and quantify all physical benefits associated with a CRP offer, which in itself is not a trivial task. This is evidenced by the fact that signups may differ on identified benefits. For instance, carbon sequestration was not included in the EEBI for early signups. Signup 41 accounted for benefits from “pollinator habitat,” whereas Signup 26 did not. Quantification might further involve the demand for environmental services. In general a benefit should receive a larger weight when more people can enjoy it. Yet most factors in the current EEBI are not weighted by population.

Given that the EEBI in use will perhaps never precisely measure the environmental benefits of a CRP offer, one may ask if there is an EBI design choice that is less prone to measurement errors. Therefore, in the following subsection we examine the robustness of the EBI designs under OP1 and OP2 with respect to potential EEBI measurement errors.

Implications When EEBI Scores Do Not Measure Actual Environmental Benefits

The actual environmental benefits of a CRP offer are unknown and so are the true measurement errors. Consequently we need to make assumptions about the properties of measurement errors before we analyze the performance of EBI designs derived from OP1 and OP2 to compare their robustness to measurement errors. Actual measurement error may have a complicated format because, as we have discussed above, the current EEBI score system may deviate from the true environmental benefits in many aspects. For simplicity, however, we consider two types of measurement errors: additive error (denoted by ϵ) and multiplicative error (denoted by ξ). We admit that this approach is quite ad hoc, and it is used only for illustrative purposes.

Let EEBI^corr stand for the EEBI after measurement error is corrected. That is, EEBI^corr is the actual environmental benefit of a CRP offer. Therefore with (1) additive error, EEBI^corr is defined as EEBI^corr ≡ EEBI + ϵ, and (2) multiplicative error, EEBI^corr is defined as EEBI^corr≡EEBI⋅ξ. The additive correction for EEBI posits a missing environmental benefit that adds to total environmental benefit, following the current EBI design, which implicitly assumes that various benefit components are additive. The multiplicative format posits missing variation in a factor that multiplies other environmental benefits. For example, if a social planner puts an equal weight on each person’s valuation for CRP benefits, then the population that have access to a tract of CRP land could be one of such multipliers. In what follows we utilize the county-level population to construct ϵ and ξ. In Item B of the online supplement,¹⁸ we present a more general simulation approach where ϵ and ξ are randomly drawn from certain probability distributions. We find that the basic conclusions regarding the robustness of OP1 and OP2 still hold under this more general approach.

County-level population data are obtained from the U.S. Census Bureau. We use 2003 and 2011 population data for Signups 26 and 41, respectively. Since we do not have specific location information on offered CRP land within a county or the geographical distribution of a county’s population, we assign the same value to ϵ and, separately, to ξ to for CRP offers within the same county. That is, for any CRP offers i and j that are in the same county, we assume ϵ_i =ϵ_j and ξ_i = ξ_j. Suppose a county’s population is at the nth percentile among all U.S. counties. Then for any CRP offer i in this county, we let ϵ_i = n, following the spirit of current EEBI design (i.e., each factor of wildlife, water quality, and soil erosion accounts for 0–100 points).¹⁹ For the multiplicative correction, we let _ξ = n/50 indicating the EEBI of CRP offers in the county with median population is not affected by the multiplicative correction.

In this section we consider optimization problems OP1 (equation [3]) and OP2 (equation [5]). Because here we focus only on environmental benefit measurement errors, we do not consider subtracting the saved crop insurance premium subsidies from the rent of a CRP offer as described in optimization problems OP1′ and OP2′. Let OP1^corr and OP2^corr denote OP1 and OP2 when optimizing on corrected EEBI, EEBI^corr. That is, OP1^corr and OP2^corr are the same as OP1 and OP2, respectively, except that in OP1^corr and OP2^corr environmental benefit e_k= EEBI^c_k^orr for CRP offer k. Recall that in OP1 and OP2, however, e_k= EEBI_k. We are interested in (1) the total actual environmental benefits measured by EEBI^corr under each optimization problem OP1, OP1^corr, OP2, and OP2^corr, (2) the robustness of OP1 and OP2 to environmental benefit measurement errors, and (3) county-level enrollment changes after environmental benefit measurement has been corrected.

Table 5 summarizes simulation results based on the enrollment mechanisms derived from the four optimization problems (i.e., OP1, OP1^corr, OP2, and OP2^corr). Recall that OP1 has an acreage constraint whereas OP2 has a budget constraint that is equal to the CRP real payment under OP1. From Table 5 we can see that for the same CRP outlay, total EEBI^corr under OP2 is always greater than that under OP1 across the two signups and the two environmental benefit measurement corrections. Under a given signup and a given measurement correction, total EEBI^corr under OP1 (respectively, OP2) is smaller than that under OP1^corr (respectively, OP2^corr) because OP1 and OP2 optimize total EEBI instead of total EEBI^corr. Total EEBI^corr achieved by OP1^corr is not directly comparable with that achieved by OP2^corr because CRP acreage and payment differ under the two optimization problems.

View this table:

• View inline
• View popup

TABLE 5

Simulation Results Based on Additive Correction and Multiplicative Correction

Regarding robustness with respect to measurement errors, we find that OP2 outperforms OP1 across the two signups and the two types of measurement correction. Under the additive correction, for both signups total EEBI^corr achieved by OP1 is about 96% of that achieved by OP1^corr. However, total EEBI^corr achieved by OP2 is above 99% of that achieved by OP2^corr. OP2 also shows greater robustness than OP1 under multiplicative measurement errors. For example, in Signup 41, OP2 reaches 98.2% of total EEBI^corr achieved by OP2^corr, whereas OP1 reaches only 87.1% of total EEBI^corr achieved by OP1^corr.

A possible explanation for this result is as follows. Notice that an error term affects the CRP enrollment outcomes through affecting CRP offer ranking. One way to measure how including an error term will affect CRP offer ranking is to calculate the correlation coefficient between the EBI values before and after including the error term. Intuitively, the higher the correlation coefficient, the smaller the effect will be. Table S1 in the online supplement²⁰ presents such correlation coefficients for Signups 26 and 41 under both additive and multiplicative corrections. For each signup and under each type of measurement errors, the correlation coefficient for OP2 is larger than that for OP1, indicating that including the error term will have smaller impacts on CRP offer ranking under OP2 than under OP1. However, one should note that since the robustness is determined by the distributions of error terms and of measured EEBI, as well as the construction of EBI under an optimization problem, we cannot conclude that OP2 is always more robust than OP1 with respect to measurement errors. Here we have considered only two measurement error specifications. Under some other measurement error specifications, OP1 may be more robust than OP2. We leave investigations in this direction for future research.

Figure 4 includes maps depicting changes in county-level CRP acreage under Signups 26 and 41 upon correcting for our posited additive measurement error.²¹ Across the four maps in Figure 4, we find that CRP acreage in the Great Plains counties would decrease, whereas CRP acreage in eastern U.S. counties would increase under both OP1 and OP2 were the additive correction to be applied to the EEBI. This is intuitive because counties in the eastern United States are generally more populous than those in the Great Plains (see Figure S1 in the online supplement²²). The same result holds under the multiplicative correction for the measurement errors (Figure 5). In Figure 4, the acreage changes shown on the two maps on the right side have smaller magnitude than those on the two maps on the left, which indicates that OP2 is more robust to additive measurement errors than is OP1. A similar finding holds under multiplicative correction (see Figure 5).

• Download figure
• Open in new tab
• Download powerpoint

FIGURE 4

Changes of Acres Enrolled into the Conservation Reserve Program after Measurement Errors Are Corrected Assuming Additive Correction

• Download figure
• Open in new tab
• Download powerpoint

FIGURE 5

Changes of Acres Enrolled into the Conservation Reserve Program after Measurement Errors are Corrected Assuming Multiplicative Correction

Figure S2 in the online supplement²³ presents four maps of the county-level CRP acreage change when comparing enrollment under OP2^corr with that under OP1 (i.e., the Baseline scenario). We find that under the additive correction (see the two maps on the left side of Figure S2), the CRP acreage changes are quite similar to those under Comparison III in Figures 1 and 2. That is, when compared with the Baseline scenario (i.e., enrollment under OP1), enrollment under OP2^corr would reduce CRP acreage in the Midwest. This indicates that the effect of the additive correction on enrollment acreage is dominated by that of switching from OP1 to OP2. However, the effect of multiplicative correction dominates that of switching from OP1 to OP2. As a result, many counties on the Great Plains would lose some CRP acreage (see the two maps on the right side of Figure S2).

VI. CONCLUDING REMARKS

Strong demand for food and biofuels in recent years has increased the pressure to draw more land into agricultural production. On the other hand, both the CRP and the FCIP are receiving intense scrutiny as legislators seek to tighten federal budget outlays. The purpose of this article is to promote a better understanding of how EBI design affects the cost-effectiveness of the CRP as well as interactions between the CRP and the FCIP, with a focus on the budgetary and environmental impacts. Given that the 2014 farm bill will gradually reduce total acreage enrolled in the CRP, it is likely that competition to enroll in the program will strengthen so long as CRP rental payments are not too low when compared with cash rental rates. Consequently, choice of enrollment design will likely play a more important role in future signups. Accounting for savings on crop insurance subsidies, perhaps through a soil attribute proxy for yield variability, can assist in managing total program tax dollar costs while also better screening land for its most efficient use.

Based on CRP enrollment data from Signups 26 and 41, as well as crop insurance subsidy data in corresponding years, we simulate the impacts of including saved crop insurance subsidies into EBI calculations under different EBI designs. We show that there is significant potential to improve upon the CRP’s enrollment design and that large regional redistributions of enrolled CRP acres would result.

Several simplifying assumptions in our analysis may result in biased estimates of budgetary and environmental impacts upon incorporating crop insurance into EBI design. We do not consider administration and operating (A&O) costs associated with crop insurance. The government also subsidizes crop insurance by paying A&O costs, which amounted to $1.4 billion in 2014, compared to $6.3 billion in premium subsidies (Shields 2015). If enrolling land in the CRP reduces some of these A&O costs, our results would underestimate the budgetary impacts of incorporating crop insurance subsidies into the CRP enrollment formula. In addition, although CRP payment per acre is fixed during a contract’s life, once an offer is accepted into the CRP, the saved insurance subsidy may vary across years because of fluctuating crop prices or changes in crop insurance policy. For simplicity our analysis has assumed away this variation in insurance subsidies. We expect that the insights arrived at would not be affected were we to relax this assumption. A further caveat is that some land offered to the CRP may not have crop insurance coverage and hence would not receive crop insurance subsidies. This fact will render an overestimation of the budgetary impacts of incorporating crop insurance subsidies into EBI design. Moreover, in our analysis we do not consider how farmers’ bidding behaviors may respond to a change in EBI design and hence affect the efficiency of the CRP (Jacobs, Thurman, and Marra 2014; Vukina et al. 2008). Relaxing these simplifying assumptions may be of interest for future research.

Acknowledgments

This study is partially supported by Ducks Unlimited. We thank two anonymous referees for helpful suggestions. We also thank Alex Barbarika for providing data and insights along the way. All opinions expressed and any errors are those of the authors.