Options for Improving Conservation Programs: Insights From Auction Theory and Economic Experiments
USDA spends over $5 billion per year on conservation activities, mostly through voluntary programs that pay farmers and landowners to provide environmental services.
USDA can reduce expenditures and encourage landowners to provide greater environmental services by making use of available data.
Laboratory experiments suggest that using more sophisticated enrollment mechanisms could reduce the cost of enrollment in a given program by as much as 18 percent.
For the last several decades, USDA has spent billions of dollars on conservation programs, and the Agricultural Act of 2014 authorizes spending about $5.5 billion/year through 2018, spread over a variety of programs. All of these—such as the Conservation Reserve Program (CRP) and the Environmental Quality Incentives Program (EQIP)—share an important feature: they are voluntary. However, program budgets, or program acreage caps, are often not large enough to enroll everyone interested in participating.
These USDA conservation programs use an “enrollment mechanism” to allocate resources among would-be participants: a process for eliciting offers, ranking offers, and choosing which offers to accept. The details of an enrollment mechanism, which are often the result of a mix of legislation and administrative decisions, can be critical to the ability of the program to meet its goals. They can help determine which farmland parcels are ultimately enrolled, which farmers and landowners gain financially from the program and how much they gain, how much the Government spends, and what environmental benefits are associated with the program.
While there are a number of options available for allocating program funds, auctions can be a core component of an enrollment mechanism. In particular, conservation programs can use reverse auctions—auctions where there is one buyer (for example, the USDA) and many possible sellers (for example, rural landowners). With clear rules regarding participation, the bidding process, and selection criteria, auctions can facilitate simple and transparent program enrollment. When combined with publicly available information about potential participants—such as county-level measures of farmland rental rates and soil productivity—a carefully designed auction can reduce program costs or encourage more environmentally beneficent choices. But there are other cases where auctions may not be very useful.
Auctions have advantages and disadvantages when selecting program participants
- the buyer (for example, USDA) can use available information about potential participants (such as measures of farmland productivity) when designing the auction; or
- no well-established market exists; or
- budgets do not allow funding all proposals and the buyer needs a fair and transparent way of selecting participants.
- goods being purchased have unique attributes (when each parcel has no close substitutes) a more explicit targeting mechanism that identifies desired parcels might be more suitable; or,
- when the administrative costs of the auction exceed the cost of individual negotiations; or
- when goods being purchased are either quite similar in cost or the purchaser has very good information on the sellers’ (for example, farmers) costs, more direct mechanisms (such as Government making direct offers to sellers) can be more cost effective than auctions.
When designing a conservation auction, details matter. How offers are elicited, what criteria are used for ranking and then selecting offers, what information USDA shares with farmers, and other factors all affect auction performance. For example, promoting the opportunity to participate in a program and disseminating detailed information about how the auction works can enhance competition by lowering barriers to entry for sellers who have little experience with conservation auctions. Conversely, if auction rules are unclear, or if they are overly complex, potential sellers may have difficulty determining the best offers, which can discourage participation. Ideally, the auction format clearly encourages behavior that public policy wants to encourage.
|Observable differences in costs among sellers||When cost differences across potential sellers (for example, farmers or rural land owners) can be predicted using available information, the purchaser can construct an auction that gives all auction participants a chance to make some profit, but which reduces the large profits that low-cost sellers (for example, farmers with less land that has low agricultural productivity) would otherwise enjoy.|
|Entry||Without the participation of many sellers, an auction will not be cost-effective. In extreme cases, a lack of participation can lead to under- enrollment and an auction that does not meet its primary policy goals (e.g., conservation).|
|Ambiguity||If auction rules are unclear, potential sellers could have difficulty determining their best offers. Individuals facing needlessly complex decisions will tend to forgo making them, staying with a status quo alternative; for example, farmers will keep farming land that they might be interested in retiring. Too much complexity may also limit farmers’ ability to assess the trade-offs of offering to install practices with higher conservation benefits.|
|Risk and regret||Sellers who feel they may expose themselves to risk by participating in an auction are less likely to participate, so an auction design that reduces unnecessary risks to potential sellers can encourage participation. Though it is impossible to eliminate all the associated risks, auction outcomes can be improved when risks can be decreased without harming the program’s goals.|
The Conservation Reserve Program’s use of auctions
The Conservation Reserve Program—now 29 years old—is USDA’s largest conservation program, created to take environmentally sensitive land out of production by paying farmers and landowners to grow grass, trees, and other conservation cover on eligible cropland. For almost 20 years, CRP has used a reverse-auction mechanism to select offers from farmers and rural landowners interested in participating in the program. As of October 2014, over 75 percent of CRP’s 24.1 million enrolled acres were selected through a “general signup” mechanism that has two distinguishing features. First, each offered parcel is scored using an Environmental Benefits Index (EBI) that reflects the both the environmental value from retiring a parcel and installing a proposed cover practice (such as mixtures of native grasses or trees), and the requested program payment (the rental rate). Second, the requested rental rate cannot exceed a parcel-specific maximum that is calculated using information on the parcel’s agricultural value. USDA uses a measure known as the Soil Rental Rate (the SRR) as an estimate of a parcel’s agricultural value.
The SRR is calculated using county average rental rates for non-irrigated cropland, together with a measure of a parcel’s soil productivity. The EBI score balances the package of expected ecosystem services provided by a parcel (such as erosion reduction and wildlife habitat) with the requested rental rate for the parcel. Cost effectiveness is pursued by encouraging competition among those making offers. Competition is fostered by accepting only a fraction of the offers, those that have the highest EBI scores (typically, around 70 percent of offered parcels are accepted.) Since a higher requested rental rate lowers a parcel’s total EBI score, bidders are encouraged to hold down the cost of their bids.
The bid cap is meant to limit excessive profits. In particular, the bid cap prevents parcels with lower-than-average agricultural productivity from receiving rents that are well above their agricultural value. However, the SRRs used to determine bid caps are not perfect measures of actual agricultural value. When bid caps are set too high, landowners may be paid much more than they can earn from farming the land—which can increase the overall cost of the program. When they are set too low, landowners can earn more from farming, so they are less likely to participate.
The absence of these landowners from the pool of potential participants can increase overall program costs for two reasons. First, a statistically powerful estimate of a seller’s cost (an unbiased estimate that is always close to the actual cost) can effectively eliminate excessive profits. However, it can also dissuade participation when a parcel’s agricultural value is greater than estimated—which can lead to increased overall costs. The box below illustrates a simple example; an unbiased bid cap causes half of potential participants to not submit an offer, which leads to enrollment of agriculturally productive (and hence, more expensive) parcels. Second, the bid cap may also discourage installation of more beneficial cover practices. Each offer identifies a specific parcel of land, the land cover that will be applied, and the cost. Often, by modestly increasing expenditures, a more beneficial land cover can be applied (e.g., native grasses instead of non-native mixes) that yields a higher EBI score. But a firm bid cap provides little or no incentive for participating landowners to improve the quality of their offers since doing so is likely to raise their costs while preventing a commensurate increase in rental rate. For parcels that are likely to be accepted regardless of land cover (such as parcels that are highly erodible), the increase in EBI score due to the higher quality practice may not be worth the resulting reduction in earnings (see box, "An example illustrating the interaction between non-participation and bid caps").
Alternative auction mechanisms
Can a superior mechanism be designed, one that encourages broader participation without risking excessive costs? A number of alternatives exist that may do a better job of leveraging information (such as estimates of rental rates). These include the following:
- Relaxed bid caps. When calculating a maximum bid, a simple approach is to use an estimate of the value of the offered item (such as the SRR). However, as illustrated above, an unbiased estimated value will typically be too low for half of all parcels, which will dissuade potential participants with land that has higher-than-average agricultural value. This scenario can be avoided by using a value greater than the unbiased estimated average value; for example, an estimate of the “maximum value” that each parcel is likely to have.
- Reference price auctions use a nonbinding estimate of each parcel’s value (a reference price). Offers are then ranked relative to their reference price. Thus, an offer with an asking price greater than its reference price will be ranked lower than it otherwise would be, while an offer with an asking price below its reference price will be ranked higher, even if the two requested asking prices are the same. This feature is meant to dissuade low-cost sellers from seeking too much profit, lest they lose to higher cost sellers who happen to have high reference prices.
- Quota auctions limit the number of offers accepted from any group having similar observed characteristics. For example, in a program like CRP, program administrators might limit acceptances to 90 percent of all offers from any given farm reporting district. This limit forces landowners in low-cost groups (say, districts with lower-than-average agricultural productivity) to compete with each other rather than with landowners in higher cost districts. The added competition acts to keep their bids lower than they otherwise might be.
Such alternatives entail tradeoffs. Relaxing bid caps may reduce the number of dissuaded potential participants but can lead to higher expenses if “already interested” participants increase their bids to match these higher maximums. Any plausible quota auction or reference-price auction necessitates accepting some high-cost parcels. The cost of these higher cost parcels may more than offset the savings gained by enticing low-cost sellers to lower their bids.
Therefore, when considering alternative auction mechanisms, it is wise to analyze how these mechanisms will perform. Given the complexity of conservation programs—with a variety of potential participants who have unobservable preferences and perceptions—theoretical models that can predict the performance of different enrollment mechanisms are often unavailable. Even when an applicable theory exists, its predictions may depend on the specific characteristics of the pool of potential participants. Another approach is to estimate statistical models of bidding behavior using data from existing programs. However, this requires data from programs that have implemented the auction mechanism of interest—which may be nonexistent!
Using experiments to investigate auction mechanisms
An increasingly popular analysis method is the use of economic experiments. Economic experiments are replicable, can careful control extraneous factors, and can effectively incorporate incentives. Experiments can occur in a variety of settings, from a classroom “laboratory” to the “field.” Laboratory experiments—where subjects are provided with real-money payments that depend upon their choices—are often used, as they are relatively inexpensive and can be readily tailored to specific questions. In contrast, field experiments—experiments that take place in the context of a Government program or policy—are intrinsically more realistic. But they can be very expensive and are not readily customized to focus strictly on a particular question of interest.
Findings from laboratory experiments
ERS, in collaboration with academic researchers, crafted a series of classroom experiments that considered relaxed bid caps, reference price auctions, and quota auctions. To avoid biases due to students’ attitudes about environmental issues, these were “context-free” experiments: neutral language was used, with no mention of agriculture or conservation. Each participant in the experiments had an opportunity to sell a “ticket” with a randomly assigned cost; the amount of money he or she could earn depended on whether or not the ticket was accepted by the auctioneer and the difference between the participant’s bid and the ticket’s cost. Some of the experiments focused on overall cost to the auctioneer, while others allowed for “quality improvements” that mimic CRP’s evaluation of each offer’s land cover and conservation practice choices when determining its EBI score. By offering to install more environmentally beneficial land-use improvements, participants can improve their EBI scores, but doing so usually adds to their participation costs, so there is a trade-off. The box, Experimental results: average profit rate for three different auction mechanisms illustrates the results of a series of experiments on alternative auction designs.
Statistical analysis of the results of these laboratory experiments show that alternative mechanisms can improve auction performance. Somewhat counterintuitively, using stringent bid caps—that are often below costs—can yield more expensive auctions than auctions with less stringent bid caps. And auctions that do not use bid caps at all—the reference-price and quota auctions—can further reduce overall costs. Furthermore, when “quality” improvements are permitted, these results are even stronger.
|The auction design||Findings|
|Relaxed bid caps (no quality improvements)||Auctions where 20 percent of the tickets had costs greater than their bid cap were more costly than auctions where 10 percent had costs greater than bid caps.|
|Relaxed bid caps (quality improvement)||In auctions where quality improvement is permitted, the most cost-effective auctions used a bid cap equal to the greatest observed cost.
A bid cap twice the greatest observed price was just as cost effective as auctions where 20 percent of tickets had costs greater than their bid cap.
|Quota auction (no quality improvement)||With a non-binding bid cap in place, allowing only a certain percent of each seller group to be accepted led to a 9-percent reduction in costs.|
|Quota auction (quality improvement)||The cost of achieving a quality-adjusted target was reduced by 14 percent compared to similar auctions where just a bid cap was used.|
|Reference price auction (quality improvement)||With a non-binding bid cap in place, using a reference price when determining which tickets to accept yields an 18-percent cost reduction.|
These findings show that the design of an auction mechanism does matter, and highlight the potentials of designs that do not use bid caps. However, while laboratory experiments provide a means of exploring the impacts of alternative auction design, the policy conclusions we can draw from them will always be limited. A number of simplifications were made, such as a uniform distribution of costs and a simple means of “improving quality.” Moreover, participants in each experimental session completed many rounds of each auction, giving subjects an ability to learn and adapt their bidding strategies—a learning process that may not exist in an actual conservation auction. And student participants engaging in a lab session may behave differently than farmers choosing how to manage their lands.
In much the same way that a new airplane carries passengers for the first time after a long process of design, wind-tunnel testing, and test flights, laboratory experiments offer a first opportunity to rigorously test a new economic design—a wind-tunnel test of potential improvements to the auction rules. The next logical step after laboratory testing is a field test—the equivalent of a test flight. Designs that have survived laboratory testing can be evaluated in a field experiment to ensure that the auction produces the desired effects before the auction is utilized at scale. The opportunity to implement a field experiment can often be found in the normal operation of an existing program, enabling a relatively low-cost test. For example, rather than instituting a broad change in a national program (such as the CRP’s general signup), a pilot program in a limited geographical area or for a particular conservation initiative could be used as a 'natural' field experiment. Repeated, iterative testing in progressively more realistic environments can provide the accurate information needed to implement a successful program or a successful change in program operations.
An example illustrating the interaction between non-participation and bid caps
Consider a conservation program with a goal of retiring 5 of 10 parcels. All parcels have identical environmental attributes but differ in agricultural productivity. Parcel owners maximize their earnings either by farming and earning agricultural “profits” or by enrolling their land in a conservation program at a “bid.” Their bid must be at, or below, the parcel’s “bid cap.” As illustrated below, case #3 uses a high-quality, unbiased estimate of profit to compute a bid cap (within 2 percent of agricultural profits). This results in 50 percent of potential participants having a maximum bid that is less than their profits from farming, so these farmers do not make offers. Program costs increase since more expensive parcels (that happen to have bid caps above profits) must be chosen instead of less expensive parcels (that happen to have bid cap below profits). In contrast, case #2 uses an upwardly biased and less accurate estimate (one that is between 2 percent and 5 percent above profits). This leads to all parcels being offered—with the five least costly being chosen. Case #2 yields total expenditures that are closest to the ideal, least cost case #1, where agricultural profits are matched.
Experimental results: average profit rate for three different auction mechanisms
Results across a number of auction rounds are displayed below for several different auction designs. In each auction round, 10 or more participants submit “tickets” to the buyer, and the buyer uses the auction rules to accept a fixed number of tickets. For each accepted ticket, the “profit rate” is computed as its earnings (its bid minus its costs) divided by its costs. For each round, the average of these profit rates is computed. The charts display the distribution of these average profit rates. Note that lower average profit rates suggest a more cost-effective auction design.
Options for Improving Conservation Programs: Insights from Auction Theory and Economic Experiments , by Daniel Hellerstein, Nathaniel Higgins, and Michael J. Roberts, ERS, January 2015
Natural Resources & Environment , USDA, Economic Research Service, August 2019