Abstract
This paper examines how auctions for ecosystem services introduce adverse selection, limiting cost-effectiveness. Induced-value experiments test theory and examine the extent of adverse selection in these markets. In comparison to the baseline of either doing nothing or the optimality of an externality-correcting tax, a discriminatory reverse auction and two screening contracts are tested. In limited budget situations, the auction achieved the lowest social surplus, while screening contracts that rely upon the government observing the development value yield the highest relative social surplus. These results are important because recent environmental policy trends are focused on expanding fiscally costly reverse auctions. (JEL Q57, Q58)
I. Introduction
Landowners’ decisions about how to use their land have implications beyond the nominal boundary of the pricing system. Urban sprawl, congestion, erosion, groundwater depletion, and carbon emissions are representative of cost-shifting behavior, while benefit-conferring behavior arises from amenities such as scenic views, open-space, habitat preservation, and the provision of carbon sequestration. Such externalities preclude efficient resource allocation. Decentralized markets are unlikely to achieve efficiency in the face of overwhelming transaction costs, particularly because each impact involves very large numbers of relatively low stake beneficiaries and victims.
Over the past 30 years, efforts to incent landowners to abate cost shifting and to perpetuate benefit provision have proliferated. In 2010 alone, the U.S. Department of Agriculture planned to spend $1.8 billion on rental payments under the Conservation Reserve Program (CRP), and yearly CRP enrollments exceed 30 million acres (USDA 2009). U.S. states and counties spent over $2 billion to protect over 1 million acres with permanent agricultural conservation easements in a recent 25-year period (American Farmland Trust 2010), while the entire set of open-space referenda authorized $31 billion between 1998 and 2003 (Trust for Public Land 2009). In addition, 1,667 private U.S. land trusts protected 37 million acres through 2005 (Aldrich and Wyerman 2006). Beyond these expenditures, the most recent U.S. Farm Bill (2008-2012) allocates $11.7 billion for conservation, while the European Union plans to spend €35.4 billion between 2007 and 2013 on its "agri-environmental" programs.1 China’s Sloping Land Conversion Program, perhaps the largest program in the world, has an estimated budget of $48 billion (Xu et al. 2010). Future carbon sequestration offset markets would likely be even larger.
Given the significant transfer of tax dollars to landowners through conservation programs, it is important to understand how well these programs perform. The common use of reverse auctions to select participants in these programs ostensibly allows buyers to price discriminate among heterogeneous owners, achieving procurement in a fiscally responsible manner. The basic economics seems simple. Institutions either create new conservation markets or alter prices within existing land markets to promote conservation ends. The efficiency arguments supporting such interventions are well known: (1) they internalize externalities, (2) they bring land use decisions into the market economy, and (3) reverse auctions can easily incorporate funding constraints.2 This paper argues that because of program budget constraints, conservation markets are subject to an extreme adverse selection problem in which some landowners should optimally be excluded from the market entirely, but hidden information precludes this outcome. In addition, by effectively assigning property rights to externalities to privately informed owners, conservation markets create informational rents for owners that might be avoided by alternative conservation programs.
Drawing on the spirit and intuition of Akerlof (1970), our analysis demonstrates how heterogeneity among landowners introduces hidden information that makes conservation markets vulnerable to adverse selection that prevents these markets from effectively altering land use decisions.3 In particular, existing conservation markets transfer large dollar amounts from taxpayers and nonprofits to landowners who were most likely to deliver the desired ecosystem services in the absence of a policy. Two factors help drive this problem: (1) the use of discriminatory reverse auctions that allow landowners to "name their price,” and (2) relatively small program budgets, which ensure that only a small subset of potential sellers will be selected.
We explore the nature of adverse selection in conservation markets, both theoretically and experimentally. A simple model of conservation auctions explains an equilibrium in which constrained budgets generate the adverse selection phenomenon that is replicated in the lab. Induced-value lab experiments are used due to the hidden information nature of this problem that makes it difficult to measure empirically in existing environmental markets. The efficiency of the discriminatory reverse auction and two sets of screening contracts are analyzed and compared to the outcomes that occur in the absence of any conservation policy or with a price instrument (an impact fee).
We believe this is the first experimental examination of adverse selection in conservation auctions and is the first to compare the performance of discriminatory auctions to settings with no policy, second-best screening contracts, or an impact fee. The experimental results show that adverse selection can be replicated in the lab, suggesting it is likely a problem in real-world ecosystem markets. The results also show that the second-best screening contracts outperform the discriminatory auction under a range of assumptions about the deadweight loss of government expenditures.
There is an extensive literature investigating incentives in government procurement settings. In most of this literature the incentive problem arises because the government purchaser of services cannot observe private information about the service provider. This typically results in either an adverse selection problem in which low-cost producers have an incentive to mimic high-cost producers, or a moral hazard problem in which the producer has no incentive to adopt cost-cutting measures once the contract has been awarded. Laffont and Tirole (1993) analyze many of these issues. More recent examples include those of Bajari and Lewis (2009), who investigate highway construction procurement contracts when the government cares about the time taken to complete the project, and Asker and Cantillon (2010), who analyze optimal procurement mechanisms and the relative performance of commonly used auction and bargaining mechanisms in an environment with multidimensional private information when the government cares about both price and quality.
The incentive procurement approach also has been applied to a variety of environmental issues. Lewis (1996) surveys this literature and demonstrates how incentive regulation can create firm-specific incentives to invest in pollution control when the costs and benefits of pollution control are privately known. Applications to agricultural and natural lands include those by Wu and Babcock (1996), Smith (1995), Ferraro (2008), and Canton, De Cara, and Jayet (2009). These papers all adopt a similar approach to address asymmetric information. The informed agents have private information about a specific attribute, such as the cost to the agent of providing environmental services (e.g., reduced pollution or land erosion, habitat protection, etc.). These attributes vary monotonically across agents, and the uninformed principal attempts to design a mechanism that incents agents to provide a level of conservation services that is strictly monotonic in the unobserved attribute. As we demonstrate below, one critical difference between the existing models and the setting we consider is that in our setting some landowners have a strong preference for conserving their land and will provide ecosystem services in the absence of any conservation policy. As a result, the optimal incentive structure is not strictly monotonic in the unobserved attribute. This non-monotonicity exacerbates the adverse selection problem in a way that is particularly detrimental to the auction mechanism.
The existing literature has compared the performance of reverse auctions with fixed-price payments to owners. Conservation auctions may seem preferable to fixed payments because they “solve” some information problems (by paying heterogeneous owners by their individual offers) and because they conserve scarce financial resources (by forcing owners to compete for limited budget expenditures). Evidence provided by Horowitz, Lynch, and Stocking (2009) and Messer and Allen (2010) suggests that these auctions outperform fixed-price approaches and provide more acres of conservation. Schilizzi and Latacz-Lohmann (2007) use experiments to compare auctions to fixed-price conservation, and find that the advantages of auctions erode as the conservation buyers and sellers interact over time. While these studies provide some support for the use of reverse auctions in conservation markets, they do not assess whether adverse selection occurs in these markets. Empirically assessing adverse selection poses a particularly difficult problem for economic analysis because of the inherent unobservability of each owner’s private information. Lab control thereby offers advantages. For example, in a conservation setting somewhat different from the one we consider, Cason and Gangadharan (2004) assume the conservation buyer is better informed about the benefits of conservation efforts than the landowner, and they present lab results showing that a discriminatory auction generates greater surplus than a uniform-price auction.
II. The Model
We present a simple model with asymmetric information, which highlights the incentive problems inherent in purchasing ecoystem services from privately informed owners when a buyer of ecoystem services (hereafter referred to generically as the ES buyer) has a limited budget. The ES buyer could be thought of as a government conservation agency, a for-profit ecosystem service aggregator, or a nongovernmental conservation organization. For structure, we focus on a specific type of conservation program that purchases permanent conservation easements.4 Although conservation easements are legally recognized rights, typically governements and nonprofit conservation groups are the only buyers for such lesser rights in land. Purchase of conservation easements is a leading policy thought to attenuate the negative externalities of development at the urban fringe and in rural areas at a lower cost than outright fee-simple acqusition of the parcel. This problem, popularly termed “sprawl,” is analyzed by Burchfield et al. (2006).
Consider a market in which N individuals own land that is currently undeveloped and provides a positive externality S in its undeveloped state.5 Each parcel has a private development value σiD in a competitive commercial real estate market, where D is a common development parameter and σi reflects the ease of developing parcel i. The value of σi is known by the owner and developer but is not easily observed by the ES buyer. The value O1D represents the price the owner can obtain by selling the parcel to a developer. In its undeveloped state, the property provides the owner a positive land return f that is common to all parcels, as well as a private land utility of h(z,Vi) where z is a utility parameter common to all owners and v, is an owner-specific utility parameter that is private information. The value f can represent, for example, the capitalized farm income or lease value of the land, while h(z,vi) reflects the owner’s personal preference for preserving the land.6 Assume h is twice continuously differentiable with hz > 0, hv > 0, hvz > 0, hzz < 0, and hvv > 0. The assumption hvz > 0 implies that the marginal benefit from an increase in the parameter z is greater for an owner with a higher private value v;. Following standard practice in the screening literature, the utility to the owner from holding an undeveloped parcel is linearly separable in the monetary payment f and the private land utility: Ui = f + h(z,vi). We assume that it is possible for the ES buyer to change the owner’s private utility through changes in z, which impose the same cost on the ES buyer regardless of the parcel but that have a different marginal impact on the landowner’s private utility depending upon vi. Each parcel is characterized by a pair {σi,vi}. The ES buyer knows the set of {σi,vi} pairs but is unable to determine the particular parcel to which a specific {σi,vi} pair is assigned—each landowner’s {σi,vi} is private information.
In the absence of any program to purchase conservation easements, a landowner will retain the property in its undeveloped state if f + h(z,vi) ≥ σiD and will sell to a developer if σiD > f + h(z,vi). Thus,
[1]defines a boundary such that in the absence of an ES buyer, a property characterized by {Oi, vi} that lies above this boundary will be sold to a developer and any parcel below the boundary will be preserved. Given the externality S generated by a preserved parcel, it is socially optimal for a parcel to be preserved if f + h(z,vi) + S ≥ σiD and to be developed if σiD > f + h(z,Vi) + S. This implies a second boundary
[2]such that it is socially optimal for any parcel characterized by a pair {σi,vi} that lies above this boundary to be developed and for any parcel that lies below this boundary to be preserved.
These boundary conditions, presented in Figure 1, classify each parcel into one of three categories.7 We define type I landowners (or parcels) as those satisfying 
type II parcels as those satisfying 
, and type III parcels as those satisfying 
. It is socially optimal for type II and type III parcels to be preserved and for type I parcels to be developed. However, in the absence of a preservation program, the socially optimal outcome will only occur for type I and type III parcels. The type II parcels, for which 
will be sold to a developer despite the social optimality of preservation. These parcels are a primary target of programs purchasing conservation easements. Let nI denote the number of type I parcels, nII the number of type II parcels, and nIII = N − nI ‒ nII the number of type III parcels.
Experimental Types and Theoretical Predictions without a Program
Allocation Using a Reverse Auction Mechanism with a Budget Constraint
In this subsection we consider the use of a discriminatory reverse auction to purchase conservation easements when the ES buyer has a limited budget. Under the auction mechanism the ES buyer announces the amount E allocated to the auction and invites the N landowners to participate. Each participant submits an offer bi that he is willing to accept for selling the easement. If the offer is accepted, then the ES buyer precludes development and the owner is paid bi and receives a final utility of Ui = f + bi + h(z,vi). If the offer is not accepted, then the owner retains development options and receives utility Ui = f + h(z,vi). As with auctions used by government conservation agencies as well as the process of for-profit aggregators searching for the lowest-cost providers, the ES buyer accepts the lowest offer first and continues accepting offers in ascending order as long as the sum of offers is less than E. The ES buyer rejects the first offer such that the sum of the accepted offers plus the lowest offer not yet accepted exceeds E. All higher offers also are rejected.
Owners for whom σtD ≤ f + h(z,vi) (the type III owners) are willing to sell to the ES buyer at any price, while type I and II owners will only sell to the ES buyer instead of a developer at an offer bi such that bi + f+ h(z,vi) ≥ σiD or bi ≥ σiD - f - h(z,v) Defining 
as the lowest offer that would induce owner i to sell an easement to the ES buyer, we have 
We assume that each owner owns one parcel and knows his own {σi,vi} pair as well as the overall set of {σi,vi} pairs.8 Thus, in equilibrium all winning owners must submit the same equilibrium offer 
, satisfying 
where nw is the number of winning biddders. If two owners with offers bi and bj, where bi> bj, are both successful in the auction, then owner j could have submitted a higher offer and still been successful, so bj was not an equilibrium offer. The claim that 
follows from the fact that if 
then any one of the winning owners would also win with an offer 
for some small ε > 0. In addition, the type III owners all submit successful offers in any equilibrium. Because the type III owners are willing to sell at any offer b > 0, they will always submit the equilibrium bid b.
Given the known set of {σi,vi} pairs, we can order owners according to the minimum amount 
each would accept in the auction where 
for the nIII type III owners. Ordering owners from the lowest to highest minimum acceptable offers, we have 
.
Proposition 1:
If
then the equilibrium offer , and only the nIII type III owners, who would not have developed their parcel in the absence of a program, sell easements to the ES buyer.If
where k is an integer satisfying , then and the first k of the type II owners as well as all of the nIII type III owners sell easements to the ES buyer.If
, where k is an integer satisfying, then all of the type III and type II owners and the first k of the type I owners sell easements to the ES buyer.If
then the equilibrium bid is , and all owners sell easements to the ES buyer.
Proposition 1 illustrates the adverse selection problem inherent in the use of an auction to purchase conservation easements. Any auction equilibrium that induces the type II owners to sell also attracts all of the type III owners, who would not develop their property in the absence of an auction. The type III owners drive up the minimum expenditure required by the ES buyer to preserve parcels (of type II owners) that would be developed in the absence of an ES buyer. Furthermore, any increase in the budget amount E will increase the winning offer received by the type III owners; the increase in E will not necessarily lead to an increase in the number of parcels preserved. An example of this is provided below in the discussion of the experimental design.
Because the externality S generated by each parcel is known, one might conclude that adverse selection should not be an issue. In fact, an efficient allocation could be achieved by offering each landowner an amount S to preserve his parcel. Alternatively, in the auction setting this can be achieved with a budget of E = (nIII + nII)S. However, in the face of budget constraints that limit E, the adverse selection problem arises despite the fact that S is known, because individual landowner preferences are not observable.
Screening Mechanism
In this section we consider the use of a screening mechanism to improve upon the auction mechanism in the effort to preserve undeveloped land. It is well known that screening mechanisms are ideally suited for addressing adverse selection problems (see Riley 2001 for a survey of the literature). We consider screening contracts of the form {p, t} where p ∈ (-∞,∞) is a direct payment to the owner (or payment from the owner to the ES buyer if p < 0), and t ∈ (-z, ∞) is an adjustment to z that impacts the owner’s personal value of undeveloped land (i.e., preservation value). In particular, given the fixed parameters f and p, if an owner agrees to sell an easement to the ES buyer in return for a contract {p,t}, then the owner’s utility is 
. Examples of contractual clauses that would positively impact an individual’s preservation value (and imply a positive t) include government- funded wetlands preservation or other wildlife habitat restoration or creation on the owner’s parcel or on surrounding land, while a negative t value could be generated by requiring public access to the parcel.
It is helpful to compare our setting with the standard hidden knowledge framework. As in the standard model, our assumptions on h ensure that 
so the Spence-Mirrlees single-crossing condition holds. This condition implies that the marginal utility of enhancing the owner’s private land value through t relative to the marginal utility of the monetary payment p (which is equal to 1) is increasing in vi. The Spence-Mirrlees condition also implies that owners can be separated via contracts satisfying the standard participation and incentive compatibility constraints. However, the conservation program we describe differs from the standard model in two important ways.
First, the value of inducing a particular owner to participate in the program is not monotonically increasing in vi. This contrasts with a standard screening model in which an employer, for example, would like to hire all workers with a skill level that exceeds some minimum threshold. In our setting the total surplus gained by inducing an owner to participate in the preservation program depends upon the owner’s type. To see this, fix σ sufficiently large and define 
so that 
for 
and define 
so that 
for 
Owners for whom 
are type I owners whose property should be developed. Ideally, such an owner should be excluded from the program regardless of his vi. Owners for whom for 
are type II owners. The gain in surplus from inducing type II owners to participate in the program is f+ h(z,vi) + S - σi - D, which is nonnegative and increasing in vi. Finally, if 
then the owner is a type III owner. The gain from inducing a type III owner to participate in the program is zero because type III owners never sell to the developer. Similarly, because the development value paramenter σi can differ across parcels, it may be optimal for the ES buyer to induce some owners with a given private preservation parameter v to participate in the program while simultaneously excluding other owners who have the same private value. In particular, for a given v, an owner for whom σi is sufficiently large so that σiD > f + h(z,v) + S is a type I owner who should sell to a developer, and an owner for whom σi is sufficiently small so that σiD < f + h(z,v) is a type III owner who should not be paid for ecosystem services. The optimal incentive contracts should be attractive only to the remaining owners, the type II owners. Thus, it is generally optimal to induce some owners with a given private value to participate in the program, while excluding others with the same private value.
A final complication is the fact that the each owner has two alternatives to participation in the conservation program: selling to a developer or retaining the parcel. As a result, the reservation utility levels that determine the participation constraint for each owner (each {σi,vi} pair) do not all correspond to the owner’s utility under the contract {p = 0, t = 0} (i.e., to the owner’s utility under no contract). Because the type I and type II owners sell to a developer in the absence of a conservation program, the reservation utility level for these owners is σiD, while the reservation utility for type III owners is f + h(z,vi). The fact that each owner has two outside options exacerbates the adverse selection problem identified in the auction mechanism analysis earlier in this section. In particular, restricting our attention to contracts with t = 0 (as is the case with the auction), inducing a type II owner to participate in the program requires p > 0. However, any contract with p > 0 and t = 0 provides every type III owner a level of utility that exceeds his reservation utility, so all of the type III owners also accept the contract.
For a given owner i the reservation utility level from not accepting a contract is
[3]Ideally, the contracts would offer both the type I and type III owners a utility less than 
while inducing type II owners to sell easements. Let B denote the set of owners who choose to sell to a developer, let C denote the set of owners who accept one of the contracts in return for selling easements, and let R denote the set of owners who refuse both the developer’s offer and the contracts. The ES buyer’s objective is to choose a set of m different contracts {{p1,t1}...,{pm,tm}} that maximizes total welfare W, where
subject to
where 
is the contract choice that maximizes owner i’s payoff given that owner i accepts a contract and 
is the cost to the ES buyer of implementing the contract 
. The first three constraints are incentive compatibility and participation constraints that characterize the action that each owner will adopt as a function of the owner’s private information. If i ∈ B, then the owner prefers selling to a developer; if i ∈ C, then the owner maximizes utility by accepting one of the m contracts offered by the conservation buyer; and if i ∈ R, then the owner rejects both the offer from the developer and the contracts. The final constraint is a budget constraint, requiring that the cost of contracts accepted by owners not exceed the expenditure limit E.
It is clear that a screening mechanism at least weakly dominates an auction because the ES buyer can always adopt the contract set 
, which offers to purchase easements at the equilibrium bid 
for the auction. This contract would result in the same outcome achieved by the auction. Thus we have
Proposition 2: Screening contracts dominate the discriminatory auction as a mechanism for purchasing conservation easements in order to preserve undeveloped parcels.
While it is easy to see that screening contracts at least weakly dominate the auction, whether screening strictly dominates the auction is unclear. The Spence-Mirrlees condition implies the ES buyer can exploit differences in each owner’s private land preferences h(z,vi) with contracts that separate different owners. However, the cost of the separating contracts may exceed the benefit they generate. Whether screening contracts strictly dominate the auction mechanism depends on the cost to the ES buyer of substituting the contract parameter p for t relative to the owners’ marginal rate of substitution between these two parameters.
Proposition 3: Screening contracts strictly dominate the discriminatory auction as a mechanism for purchasing easements if 
at the contract {p = b, t = 0} for at least one type III owner.
Proof: Consider the contract space {p,t}. For increases in t the slope of the buyer’s isocost defined by 
is 
, and the slope of owner i’s indifference curve through the contract 
is 
. Under the auction all type III owners participate and are paid the equilibrium offer 
. If 
at the contract 
for at least one type III owner i, then that owner’s indifference curve is steeper at the point 
than the buyer’s iso-cost, so there is an alternative contract with 
and t > 0 on the same indifference curve (so it still induces owner i to participate), which is on a lower iso-cost for the buyer. ■
If the conditions of Proposition 3 are met, then the screening contracts reduce the impact of the adverse selection problem by reducing the payment made to the type III owners. In addition, following standard arguments in the screening literature it can be shown that, if a contract is accepted by a type II owner with parameters 
, then all owners with parameters satisfying 
also will accept one of the offered contracts. However, more general statements are not possible about the extent to which the set of optimal contracts will separate different owners and whether optimal screening contracts will further reduce the adverse selection problem by excluding type III owners from participating in the program. Such outcomes depend upon the exact specifications of the cost function c(p,t), private land value h(z,vj), and the distribution of {σi, vi;} pairs.
A Simple Taxation Mechanism
The adverse selection problem associated with mechanisms to purchase conservation easements arises because these mechanisms shift the lower boundary defined by equation [2] upward and are therefore attractive to both type II and type III owners. As an alternative to purchasing development rights, a policy that reduces the seller’s return σiD to 
such that
and
also would ensure efficient allocation of each parcel. These equations imply that
and
This is satisfied by any policy such that 
or 
. Because this policy is independent of the owner’s private preservation value vi, any policy that reduces the private development value vi of the owner by the amount SID will result in the efficient development and preservation outcomes for all parcels.
A simple tax can be implemented to achieve this result. In particular, let ti be the tax rate imposed on the sale price if owner i’s parcel is sold to a developer. The owner’s return from sale is gxD(1 − ti). Setting the tax rate so that this return equals the seller’s return 
achieved under the optimal policy yields 
Setting 
and solving for ti yields the optimal tax rate 
. Noting that σiD is the price the developer is willing to pay for the parcel, it follows that the optimal total tax 
for any parcel i is S. A “flat” tax of S replicates the “flat” externality, and taxing S on any sale to a developer results in only those parcels for which the value of the parcel in development exceeds the total social value from preservation being sold to a developer. Such a tax will successfully screen owners by causing each owner to internalize fully the external cost of selling to a developer.
Of course, this is just the well-known Pi-govian tax, but the extent to which it delivers optimal results in the face of adverse selection is not broadly known. In the conservation easement market, programs effectively assign property rights for development to owners. However, paying owners for providing external benefits to society creates an adverse selection problem in which the ES buyer pays some owners for services that would have been provided even in the absence of a program. This adverse selection problem arises because the conservation market assigns property rights to the parties with private information. Given a limited budget allocated to make payments to owners for this surplus, the holders of private information are able to extract informational rents, and the resulting outcome is not economically efficient. The inefficiency is greatest under the auction mechanism. It can possibly be mitigated, but not eliminated, using screening contracts. One might assume that this problem an be overcome by following Coase’s (1960) prescription of assigning rights to the highest-valued user, particularly in the face of significant transactions costs. Unfortunately, this is not possible in the conservation setting because the highest value use of some properties is in development, while for others it is in conservation, and information about these values is privately held. However, the adverse selection problem can be overcome by assigning rights based on publicly known information. In particular, because the externality S is known, if rights to the social surplus are assigned to the state, and owners (or developers) are penalized for destruction of the known social surplus, then the socially efficient outcome can be achieved through a tax in the amount of the social surplus at risk of destruction.9 This result obtains even if the externality S is parcel specific, as long as the externality is known so the impact fee can be adjusted accordingly.10
III. Experimental Design
This section describes the economic experiments11 used to test the theoretical predictions developed in the previous section. A total of 120 participants were recruited from undergraduate economics and business classes to participate in this experiment. All of the sessions were conducted in the Experimental Economics Laboratory for Policy and Behavioral Research at the University of Delaware. Subjects were assigned to a computer station at random. Each computer station was equipped with privacy screens to ensure confidentiality of participant decisions and outcomes. The choices were made using Excel spreadsheets programmed with Visual Basic for Applications. Each experimental session lasted approximately 1.5 hours, and participants earned approximately $25. At the start of each session, participants received written instructions and listened to an oral description of the experiment protocols accompanied by a PowerPoint presentation to ensure consistency. Participants were welcomed to ask questions of the administrators before the experiment began, but no communication between participants was permitted.
Each experiment session consisted of 12 participants. In each round there were 12 different induced {σi,vi} pairs. An induced-value experiment was required to provide lab control over the hidden information {σi,vi}, which reflects each owner’s opportunity cost of supplying environmental services. Each experimental session consisted of 65 rounds where the first 5 rounds served as training (without financial payment) using the Status Quo treatment.12 The subsequent 60 rounds consisted of 12 rounds of each of the five treatments described below, which meant each individual participated in each treatment as each of the 12 seller types. This design created 7,200 individual-level observations (120 participants X 60 rounds) and 600 group-level observations (10 sessions X 60 rounds). To control for potential order effects, the seller type in each round was randomly selected so that participants did not know which type they would be in the subsequent rounds, nor were there any duplicates of types. In addition, the five treatments were presented in a Latin squares design.
Each of the five treatments provided participants with a set of induced values relating to their ownership of one parcel of land. Values represented the stream of benefits that would accrue to this seller type. Transaction costs were considered part of the stream of benefits represented by the set of induced values and therefore were not expressed separately. Decisions in one round had no impact on the values or choices in subsequent rounds.13
Experimental parameters were set to reflect the land, development, and private preservation values with 12 landowners distributed symmetrically over (σ, v) space with σi ∈ (0.25,0.5,0.75) and vi ∈ ($0;$1,000,000; $2,000,000;$3,000,000). This resulted in 12 different owner types, presented in Table 1 and illustrated in Figure 1. Parameters were selected to reflect local land development and conservation markets. The typical development value for an acre of agricultural land varies widely depending upon property characteristics. However, the average price per acre is about $7,000, and the typical capitalized revenue from agricultural use, including federal subsidies, is $3,000.14 Thus, the “development increment” averages $4,000. This increment captures the opportunities foregone if development is precluded and thus approximates the market value of a conservation easement. The preservation value to the public reflects the positive externalities of perpetuating agricultural and natural land uses. Estimates of S vary widely with location and land use, but $10,000 per acre is a relatively conservative estimate from stated preference studies reviewed by Bergstrom and Ready (2009). In the experiment, we consider a market in which the parameters for the 12 owners are known, but the mechanism designer is unable to observe the parameters of any specific landowner.
Owner Types and Surplus
We assign a current land use value of f= $3,000 and a development parameter of D= $20,000 to all parcels. The values of D and σ imply that the developer will have a value of $15,000 for the land of owners 1-4; $10,000 for the land of owners 5-8; and $5,000 for the land of owners 9-12. As can be seen in Figure 1, parcel (or owner) 1 is a type I parcel that should be sold to a developer, parcels 2, 3, 4, 6, and 9 are type II parcels that should be preserved but are sold to a developer in the absence of a conservation program, and parcels 7, 8, 10, 11, and 12 are type III parcels that should be preserved and are preserved in the absence of a conservation program. In the analysis that follows, a subscript is attached to each parcel number identifying whether it is type I, type II, or type III.
The social surplus of each parcel is set at S = $10,000, the parameter z = $40, and the private preservation utility is 
. This implies that private land preservation utility values range from 0 to $10,954.45, with an average of $6,555.82, and that an owner’s overall utility 
from his parcel ranges from $3,000 to $13,954.45.
Status Quo Treatment
In the Status Quo treatment, participants of each seller type were shown a set-price offered by a developer and the utility from retaining the parcel. The set-price σiD can be thought of as the competitive price a developer pays an owner for land. In the experiment instructions, the developer was referred to generically as the “set-price” buyer. The administrator served as the set-price buyer, and the posted prices were pre-set to correspond to the theory outlined in this paper. Each seller had the choice of accepting this set-price or retaining their parcel and earning the induced “ownership return” of 
. After all participants had entered decisions, the administrator’s computer recorded these decisions and provided participants with their next set of induced values. The participants also learned of the number of parcels purchased by the set-price buyer and the lowest, highest, and average prices paid. This information was displayed on the participants’ spreadsheets. These procedures repeated until all 12 rounds of this treatment were completed.
One anticipates that the status quo generates suboptimality, wherein all type II owners sell to the developer even though optimality requires that they do not. In the development setting, this reflects the concept of “sprawl," or land development that destroys positive externalities or creates negative externalities.
Auction
In the auction treatment, participants again received a set of induced values for retaining their parcel and received a set price from the development buyer. As in the Status Quo treatment, each participant could accept the developer’s set-price or retain her parcel. In this treatment, each participant also had the alternative to try to sell an easement for her parcel by submitting an offer to sell to the ES buyer (referred to generically as the “Auction Buyer" in the experiment instructions). A discriminatory auction was used. The offers to sell easements were ranked from lowest to highest, and the ES buyer would buy the lowest-priced easement and continue up the list of offers until the budget was exhausted. Funds remaining in the ES buyer budget after the round did not carry forward into the next round.
The three alternatives available to each participant in the auction treatment yielded the following returns as a function of the participant’s {σi,vi} values. Retaining the parcel yielded a payment of 
selling to a developer generated a payment of 
and submitting an offer bi in the auction yielded a payment of 
if the offer was accepted and a payment of bi
if the offer was rejected. As demonstrated above, the equilibrium offers depend upon the budget. Recall from the previous section that the minimum offer required to induce an owner to sell a conservation easement is obtained by subtracting the owner’s utility 
from retaining her parcel from the development value σiD. This value is negative for owners 7III, 8III, 10III, 11III, and 12III (the five type III owners) because these owners will never sell to a developer. Consequently, the surplus generated by inducing these owners to sell easements is zero. The surplus generated by inducing a type II owner, who would sell to a developer in the absence of a conservation program, to sell equals the difference between the total surplus 
generated by preserving the land and the development value σiD. At this point our surplus calculations assume that any payment from the ES buyer to an owner represents a transfer of wealth that has no impact on total surplus. This assumption is relaxed in our discussion of the results in Section IV where we consider possible waste in revenue collection and expenditure. The minimum acceptable bid and the gain in surplus from inducing the participant to sell a conservation easement in the auction are presented in Table 2. Owners are ordered by minimum acceptable bid.
Auction Minimum Acceptable Bids and Surplus by Type
Using Table 2 we can determine the expenditure E needed to include each owner as a successful bidder in the equilibrium of the auction. A minimal expenditure is needed to induce owners 10m, 11iii, 12iii, 7iii, and 8m to participate, because they are willing to sell easements at any bid greater than $0. Participation by the next owner 6ii requires a minimum payment of $675. Because of the adverse selection problem, the type III owners also all must receive a payment of $675 before owner 6ii will be successful in the auction equilibrium. This implies a total expenditure of E ≥ $4,050 (6 X $675) is needed for theauction to induce owner 6jj to sell an easement. Including the next lowest offer of $1,046 from owner 4n requires a total expenditure of E ≥ 7 * $1,046 = $7,322. The increase in total expenditure of $3,272 generates an additional surplus of $8,954. Including the next lowest offer from owner 9II requires an expenditure of E = 8 *$2,000 = $16,000, or an increase of $16,000 - $7,322 = $8,678, and generates an increase in surplus of only $8,000. Adding landowner 9II to the set of successful bidders in equilibrium requires an increase in the auction expenditure that exceeds the benefit. In our experiment, we limit our auction budget to $8,535 so that only the seven lowest owners (10III, 11III, 12III, 7III, 8III, 6II, and 4II, as shown in Table 2) submit successful offers in the predicted equilibrium. Note that this exceeds the minimum budget of $7,322 required to induce these owners to sell easements in the predicted equilibrium. This budget number was selected to maintain equal total budgets across the auction and screening treatments. If participants closely adhere to the equilibrium offers, then the ES buyer will have spent more than needed to capture the $8,954 in surplus gained by inducing owners 6ii and 4ii to sell an easement. However, because the payment from the ES buyer to an owner is simply a transfer of resources, this overspending will have no impact on total surplus. If participants in the experiment overbid relative to the predicted equilibrium, then the budget of $8,535 makes purchase of development rights through the auction more likely than a budget of $7,322. Similarly if participants underbid relative to the predicted equilibrium, then the higher budget amount may support purchase of development rights from additional parcels (which should be preserved). In either case, adopting the higher budget amount of $8,535 would bias the results of the experment in favor of the auction, thereby further substantiating any evidence indicating the auction is subject to adverse selection. Participants were aware that the ES buyer had a budget of $8,535 to purchase as many parcels as possible.
After the participants made their decisions, the administrator’s computer recorded the information and informed the participants who had submitted offers to the ES buyer whether the offer had been accepted. After each round, participants learned the number of offers submitted to the ES buyer; the number of easements purchased; and the lowest, highest, and average prices paid. This information was displayed on each participant’s spreadsheet.15
Screening Mechanism I: Ease of Development Unknown
In the Screening Mechanism treatments, each owner had the same choices provided in the Status Quo treatment, as well as the alternative to choose one of several screening contracts that affected their induced values as described in the previous section. In particular, each participant could accept the set price from the developer, retain her parcel and receive the ownership return, or accept one of the contracts { p,t} and receive the resulting payoff 
.
In the first screening treatment (treatment C), participants were given the choice between three contracts. The contract C1 = {2050, - 30} offered a fixed payment of $2,050 to any owner willing to accept a reduction in z of 30. The reduction in z might be conceptualized as allowing public access on private land, which significantly reduces the private preservation utility for participants with vi > 0 but has no impact on the utility of participants 1I, 5II, and 9II, who place no value on the preservation of the land. This sort of contract would require an affirmative easement and would provide additional external benefits. We assume the cost of C1(2050 - 30) = $2,050 as requiring the landowner to allow public access imposes no incremental cost on the ES buyer. The contract C2 = {700,0} offered a payment of $700 with no change in the owner’s private preservation value. The cost of C2(700,0) = $700. Finally, under the third contract, C3 = {— 2000,26}, the ES buyer significantly improves the owner’s private preservation value in return for a payment of $2,000 from the owner. Because the third contract entails increasing the owner’s total utility through the private utility 
, the cost of this contract is not a straightforward fixed payment. We assume that the cost of this contract to the ES buyer is 1.1 times the change in total utility achieved by this contract for a participant with vi = $3,000,000. The vi = $3,000,000 participants all should prefer this contract over selling to a developer and over not participating in the program. The implies a cost of 
.16 The predicted choices of each participant and the associated cost to the ES buyer are presented in Table 3 where Cj implies the participant chooses contract, Cj, j = 1,2,3, and D implies the participant sells to a developer.
Predicted Screening Mechanism I Contract and Government Cost by Type
The total expected cost of the screening mechanism to the government is $8,535.50. Under these contracts the theory predicts that 8 of the 12 parcels will be preserved, whereas only 7 of the 12 parcels are preserved under the auction mechanism. The equilibrium with 7 of the 12 parcels being preserved with the auction mechanism also could have been achieved with a lower budget, of $7,322. Thus, the predicted outcome under the screening mechanism provides an increase in surplus of $8,000 (the surplus from preserving instead of developing the parcel from landowner 9n) at an additional cost of $1,213.5 relative to the auction.
This contract might reflect a subsidy program in which an owner who transfers an easement to the ES buyer pays $2,000 for some land improvement and receives an additional $1,228.50 subsidy from the ES buyer.
After the participants made their decisions, the administrator’s computer recorded this information and provided participants with their next set of induced values. The participants learned of the number of parcels purchased by the developer and the lowest, highest, and average prices paid, as well as the number of contracts accepted by the participants.
Screening Mechanism II: Ease of Development Known
The Screening Mechanism II treatment operated identically to the Screening Mechanism I treatment, except that the screening contracts were now based on the assumption that the ES buyer had information about the development parameter σi of each owner and could target contracts to buyers with a specific σi value to improve social efficiency.17 In this treatment, in addition to the status quo options of selling at the set price or not selling, contracts were offered to specific owners based on σi. In particular, the contract C4 = {2050, — 30} was offered to all owners with σi = 0.25; owners with σi = 0.50 were offered two contracts, C5 = {1150, — 5} and C6 = {500,0}; and owners with σi = 0.75 were offered the contract C7 = {3100,0}.
As with treatment C, after participants made their decisions, the administrator’s computer recorded this information and then provided participants with their next set of induced values. The participants learned of the number of parcels purchased by the developer and the lowest, highest, and average prices paid, as well as the number of contracts that were accepted by participants.
Impact Fee
The Impact Fee (or flat tax) treatment operated identically to the Status Quo treatment except that a tax of τ = S = $10,000 was imposed on any participant that sold to a developer. This resulted in a final utility of σiD - $10,000 if the owner sold to a developer. Owners who did not sell to a developer received the preservation utility 
. The predicted outcome from this policy is that owner 1i sells to a developer and all other participants retain their land. Like with the Status Quo treatment, after participants had made their decisions, the administrator’s computer recorded these choices and then provided participants with their next set of induced values. The participants also learned of the number of parcels purchased by the developer and the lowest, highest, and average prices paid.
The five treatments and the corresponding decisions and payoffs available to the participants are summarized in Table 4.
Participant Choices, Payoffs by Treatment, Program Expenditures
IV. Experimental Results
The driving hypothesis is that the ES buyer’s search for the lowest-cost suppliers of ecosystem services are unlikely to be cost-effective, because an information asymmetry introduces adverse selection. The data from experiments testing this theory illustrate that this hypothesis is likely to be correct. This can be seen by inspecting the Figures 2 through 6 and the summary results shown in Table 5.
Experiment Results, Individual Choices by Treatment (in Percent)
Figure 2 reflects the results from the status quo treatment A. When participants were confronted only with a developer buyer, the data replicate the problem of overdevelopment. The circle sizes and corresponding labels in the figure reflect the average number of the 12 participants (identified by {σi,vi} pairs) that accepted the developer’s offer. Many type II owners were willing to accept the developer’s offer. For instance, owners 2II, 3II, 4II, 5II, and 9II all accepted the set-price from developers over 96% of the time (Table 5). Recall that social surplus is maximized if 1I is the only owner to accept the developer’s set price. The average social surplus for each owner ({σi,vi} pair) in treatment A is depicted in Table 6 and shows that this treatment has an overall average social surplus of $16,366. Total possible social surplus, averaged across the 12 owners, is $19,722, so the status quo achieved 82.0% of social surplus. The status quo is suboptimal and offers a measure of the problems associated with externalities without a policy or program response.
Theoretical Prediction and Experimental Results from Treatment A
Experiment Results, Social Surplus by Treatment by Type ($)
Figure 3 shows treatment B, which replicates the introduction of the ES buyer seeking to purchase from the lowest-cost providers of ecosystem services. The average budget expenditure was $7,122, which is 83.4% of the available budget. The minimum budget expenditure in a round was $2,046 and the maximum spent in any round was $8,525. The standard deviation in the budget expenditures was $921.
Theoretical Prediction and Experimental Results from Treatment B
Adverse selection is rampant, with type III owners who would not develop without a program winning the auction (and thus getting ecosystem payments despite their intention to not develop). Moreover, many of the type II owners vulnerable to the developer’s offer still accept it. Hence, the net effect of the program is only a marginal improvement to the world without an ES buyer. This provides experimental evidence for the adverse selection hypothesis. Relative to the status quo treatment A, social surplus does increase from $16,366 to $17,614 (Table 6), which is 89.3% of possible surplus.
Specifically, Table 5 shows that owner 1I accepted the developer’s set-price 97.5% of the time, which is the same percentage as in treatment A. Furthermore, owners 2n, 3jj, and 5II, accepted the developers set-price more than 93% of the time despite the social efficiency of these parcels remaining undeveloped. Owner 9II accepted the developer’s set price 79.2% of the time, while selling an easement to the ES buyer only 10% of the time. The other two owners, 4II and 6II, who should not develop but tend to develop without an ES buyer, had a more mixed result as they sold easements to the ES buyer 43.3% and 72.5%, of the time, respectively, and sold to the developer 42.5% and 17.5% of the time, respectively.
Table 5 also shows that the introduction of an ES buyer using a lowest-cost strategy in a competitive easement market can lead to landowners retaining their land despite the theoretical prediction that owners should either accept the developer’s set-price or sell to the ES buyer as, on average, 12.2% of the owners retained their parcels. In most cases this was the result of the owner submitting an offer that was rejected by the ES buyer. Interestingly, the five owners with the highest percentage of retention (neither selling to a developer nor selling an easement to the ES buyer) were type III owners, such as 7II, 8II, 10II, 11III, and 12III, who in the absence of an ES buyer would not have developed. This suggests that not only were these owners active in the ES buyer market, but many of them were actively working to secure high profits by making offers significantly above their landownership returns.18 For instance, the highest level of retention occurred for owner 12II. Participants in this role retained their parcels 22.5% of the time.
The screening mechanisms in treatments C and D (Figures 4 and 5) show that contracts can successfully sort participants by types. These results demonstrate the expected improvement from screening contracts on the adverse selection found in the auction treatment. The average social surplus in treatment C, where the ease of development parameter σ was unknown was $18,307 (Table 6), or 92.8% of social surplus. Figures 4 and 5 and Table 5 show which owners accepted which contracts. In treatment C, owner 9II selected screening contract C1 = {2050, - 30} 90.8% of the time, while the other owners accepted it, on average, less than 1% of the time. Screening contract C2= {700,0} was accepted more than 90% of the time by owners 7III, 6II, 10III, and 11III. Finally, screening contract C3 = {- 2000,26} was accepted 90% of the time or more by owners 4II, 8II, and 12III. Interestingly, in treatment C, participants did not frequently make decisions that led them to lower individual payoffs as, on average, only 1.2% of participants did not sell to the developer or accept a contract.
Surplus increases when the ES buyer knows σ;. In treatment D, surplus increased to $18,881, or 95.7% of possible surplus. In comparison to treatments B and C, it is notable how three of the type III owners (10III, 12III, and 11III) retain their parcels 97.5% of the time. In other words, these owners, who were never likely to sell their land to a developer, provided the positive externality without direct payment when the ES buyer had the ability to identify the development value of each parcel and construct contracts incorporating this information.
Figure 6 shows treatment E, the impact fee. As expected, the results are stark. Almost every participant made a decision in line with social efficiency. In this treatment, owners 1I sell to a developer 97.5% of the time, and the other 11 participant types developed only 1.4% of the time (Table 5). This treatment yielded the highest level of social surplus, an average of $19,623 (Table 6) or 99.5% of possible surplus.
Theoretical Prediction and Experimental Results from Treatment D
Theoretical Prediction and Experimental Results from Treatment C
Theoretical Prediction and Experimental Results from Treatment E
Table 7 provides the results of a random-effects model that explains realized social surplus using owner types, treatments, round, and round and treatment interactions as independent variables. These models use the 7,200 observations clustering on the experiment participant, each of whom provide 60 observations (12 observations for each of the five treatments). As expected, evidence exists that the estimated social surplus associated with each different owner ({σi,vi} pair) is different than the omitted owner 1I. The results show that the social surplus for treatments B, C, D, and E are all positive relative to the status quo. Further analysis rejects the hypothesis that the treatment coefficients are the same at the 0.01 level.
Panel Data Analysis of Social Surplus
Thus, there is evidence that the introduction of an ES buyer purchasing easements through a discriminatory auction is more efficient than the status quo. However, the auction introduces adverse selection. The mechanism design contracts are more efficient than the auction because they help attenuate adverse selection. The tax is more efficient than the mechanism design contracts.
The random-effects model also includes variables designed to test whether learning over time affected the level of social surplus achieved in the experiment. The empirical results suggest that neither the round variable nor any of the four other variables including the round and treatment interaction terms were statistically significant (Table 7). This is especially important for the Treatment B X Round variable, where a statistically significant coefficient would have implied that the nature of the discriminatory auction enabled learning to be a significant influence on the overall outcome. In sum, these results suggest that the experimental design controlled for potential order effects.
Thus, there is evidence that the introduction of an ES buyer purchasing easements through a discriminatory auction is more efficient than the status quo. However, the auction introduces adverse selection. The mechanism design contracts are more efficient than the auction because they help attenuate adverse selection. The tax is more efficient than the mechanism design contracts.
The random-effects model also includes variables designed to test whether learning over time affected the level of social surplus achieved in the experiment. The empirical results suggest that neither the round variable nor any of the four other variables including the round and treatment interaction terms were statistically significant (Table 7). This is especially important for the Treatment B X Round variable, where a statistically significant coefficient would have implied that the nature of the discriminatory auction enabled learning to be a significant influence on the overall outcome. In sum, these results suggest that the experimental design controlled for potential order effects.
V. Discussion
The results from lab experiments strongly support the theoretical prediction that adverse selection is likely to exist in conservation markets that use a discriminatory reverse auction, and that a screening mechanism can sort types so as to improve cost-effectiveness with respect to the specific information asymmetry. The screening mechanisms involve large transfers and are not as cost-effective as the externality-correcting tax. This comparison to the socially optimal impact fee is an important result for policy because recent trends are focused on expanding fiscally costly government auctions rather than taxes. The ecosystem service markets considered (either the auction or screening contracts), which assign ownership rights to individual landowners, enable privately informed landowners to extract information rents when providing ecosystem services. Since the impact fee effectively assigns property rights based on the external surplus S provided by preserved land, the impact fee does not create the opportunity for individual owners to capture information rents.
For policy purposes, the various treatments have different implications for the role of government in dealing with problems of overdevelopment. In the analysis above, any transfer from the ES buyer to an owner is viewed as a direct transfer with no loss of surplus. Figure 7, considers the impact of these programs when transfers entail some degree of government waste, w. The horizontal axis of this figure extends from w = 0 (no government waste) to w = 1 (each government dollar spent on conservation wastes another dollar in, say, labor market distortions). An intermediate value, such as w = 0.2, implies that $1.20 must be raised via taxes to deliver $1.00 in government expenditure.19 In Figure 7 the surplus values on the vertical axis, with w = 0, correspond to the average surplus indicators presented in the results section above. The figure shows that the status quo treatment A is invariant to changes in w, as no conservation program exists. However, increases in waste decrease the social surplus of the auction (treatment B) and the conservation contracts (treatments C and D) because these programs involve significant transfers of wealth from taxpayers to owners of land. Figure 7 also illustrates the superiority of the impact fee (treatment E). As with treatment A, the social surplus derived from this treatment is not directly affected by government waste.
Social Surplus at Different Levels of Government Waste
Interestingly, if the concept of the double dividend from the environmental taxation literature is applied to the impact fee (treatment E), the surplus generated by this program actually increases in w. This occurs because the revenue collected to correct the externality can be used to offset other revenue sources, such as wage taxation, which are distortionary. In other words, the tax has the benefit of avoiding government waste. The concept of a double dividend can also be applied to Screening Mechanism I, which improves its general efficiency.
VI. CONCLUSIONS
This research demonstrates, both theoretically and through economic laboratory experiments, how ecosystem service markets are impacted by adverse selection. Despite spending billions of dollars, these markets may not positively alter land use patterns or other environmental outcomes very much. In particular, our results imply that these policies transfer billions of dollars from the government and nonprofits to landowners who were the most likely to deliver the desired conservation services in the absence of a policy. Among the drivers of this problem are the mechanisms used, which seek to select the lowest-cost providers of the service and in the case of discriminatory reverse auctions allow landowners to “name their price." When these selection techniques are combined with constrained program budgets, only a small subset of potential landowners will be selected, and this subset includes the owners who are least likely to sell their land to a developer in the absence of the ecosystem service buyer.
A model of ecosystem service auctions is developed to describe the adverse selection phenomenon. A screening mechanism that mitigates the incentive problem is also analyzed. Economic experiments test the screening mechanism relative to the commonly used discriminatory auction and the status quo of no intervention. Experiments closely follow the theoretical predictions by testing five different treatments (see Figures 2 through 6). Of particular interest is the finding that conservation markets are highly influenced by adverse selection. In contrast, various conservation contracts can improve on the auction by sorting the landowner types to achieve a higher level of efficiency. However, all of these policies involve government expenditures and are inferior to the implementation of a tax on development, which not only solves the externality problem underlying this question, but also resolves the adverse selection problem as well. These results strongly suggest that changes to conservation programs will improve their efficiency.
Our analysis demonstrates how conservation markets that seek to acquire parcels based on the lowest cost, such as discriminatory auctions to purchase easements on undeveloped land, introduce problematic informational asymmetries by assigning property rights to parties with private information that is relevant to the optimal use of each parcel. As a result, privately informed landowners are able to extract informational rents, and the efficiency of the conservation program is reduced. Because the standard mechanism design approach takes the informational structure and assignment of property rights as given, the screening mechanism achieves only a second-best outcome. However, our results also demonstrate how the choice of the conservation program itself can alter the assignment of property rights in a way that supports economic efficiency.
The results may be affected by what one assumes about the availability and cost of information, and this affects the practical relevance of the research. First, the reverse auction requires no information, unless the ES buyer seeks to evaluate its performance. The impact fee requires an estimate of external benefits, S. Though such estimates are provided in the economics literature in various settings, many ES buyers would likely have to commission original valuation research or benefit transfer studies to estimate S. In addition, such estimates would be less costly in settings where there was little variation and the valuation scenario and populations were more easily specified. Because the impact fee is, in effect, a tax, it faces political acceptability issues. ES buyers may find screening contracts to be an appropriate compromise— greater efficiency with less political dispute. These contracts require that the ES buyer have information about the distribution of the private preservation parameter vi. The more precise the ES buyer’s information, the more likely it will be that screening contracts will dominate the auction. Furthermore, as demonstrated in the experiments, the efficiency of screening contracts can be increased if the ES buyer also has information about the distribution of the development value parameter O1.
This research was conducted under the assumption that the value of the social benefits for keeping a land parcel undeveloped, S, is homogeneous and known with certainty. An area for future research would be to examine the case when the value of social benefits varies and is uncertain and to consider under what conditions, if any, the auction mechanism or screening contracts could yield a better result.
Finally, this research operationalizes an emerging concern related to the design of carbon offset markets. Land uses can provide carbon sequestration services, though institutional limits on aggregated services must be “additional," that is, not yet practiced (the connection between “additionality" and information asymmetry was noted by Ferraro [2008]). The practical concern is that landowners will stop practicing carbon sequestration activities for a few years in order to then sign up with an aggregator. This perverse incentive will be difficult to monitor but will create a potentially worse environmental outcome because years of sequestered carbon will be released simply to enable future participation. The research in this paper suggests that the perverse incentives associated with additionality requirements vanish when one institutes a tax program. Instead of paying one to sequester carbon, tax those who do not wish to deliver these services. Although this paper helps clarify the economics, the political economy of such a different perspective on property rights is far from clear.
Acknowledgments
Without implication, we have benefited from feedback at presentations at the Agricultural and Applied Economics Association, USDA-ERS, and the University of Delaware and from discussions on adverse selection and additionality issues from Peter Schwarz, Marca Weinberg, Dan Hellerstein, John Horowitz, and Lori Lynch. We also thank two anonymous referees for their helpful comments. Senior authorship is shared. This research was supported by grants from the University of Delaware Alfred Lerner College of Business and Economics and the USDA Economic Research Service (#58-6000-7-0089).
Footnotes
The authors are, respectively, associate professor, Department of Economics; professor, Department of Applied Economics and Statistics; associate professor, Department of Applied Economics and Statistics, University of Delaware, Newark.
↵1. U.S. calculations based on data from Claassen (2010). E.U. calculations based on data for Axis 2, Section 214, of the second pillar of the Common Agricultural Policy (EU-DGARD 2009).
↵2. Duke and Lynch (2006) question the first rationale, finding no price link between payment and externalities. Although policies may alter the behavior of landowners and lessees (Kirwan 2009), this paper conceptualizes a single decision-making unit of the capital owner and the land use decision maker.
↵3. Kirwan, Lubowski, and Roberts (2005) offer empirical evidence of heterogeneity in landowner opportunity costs of providing environmental services.
↵4. More specifically, the parties enter into a legal contract for a negative easement, which is a nonpossessory interest encumbering land by precluding development in perpetuity.
↵5. As a simplifying assumption, S is homogeneous. This limits the realism of this model in so far that some existing reverse auctions also try to incorporate benefit measures. An obvious extension of these experiments would be to develop a heterogeneous measure of S.
↵6. Although f is assumed to be homogeneous to simplify notation and the experimental parameters, heterogeneity in land returns due to differences in privately observable land characteristics could equally be seen to enter the analysis as augmenting the existing heterogeneity parameter, v.
↵7. Lewis (1996, figure 1) presents similar intuition in the context of pollution control.
↵8. This informational assumption simplifies the analysis. It also reflects the nature of many conservation easement auction settings in which bidders are often from the same agricultural community and have at least some awareness of each other’s land holdings and individual preferences.
↵9. Note that if the ES buyer faces no budget constraint, then the assignment of rights to the social surplus is irrelevant á la Coase. The auction with a budget of E = (nII + nIII)S results in an equilibrium bid of S, and with this auction budget the type II and III owners will sell easements, while the type I owners will sell to a developer. This is the socially optimal allocation. However, in contrast to the impact fee, an auction with this budget entails substantial transfers from taxpayers to the targeted landowner.
↵10. Because the informational assumptions of the model are somewhat strong, it is worth considering alternative settings. One key feature of the auction setting considered here is that the number of units sold (the parcels for which the government purchases development rights) is determined endogenously. Our assumption that the {σi,vi} are fixed is made to simplify the analysis. If the {σi,vi} pairs are, instead, drawn from a cumulative joint distribution function F, intuitively optimal equilibrium strategies would entail each landowner calculating the expected number
of winning bidders contingent upon his own realized valuation and the number (or expected number) of participants in the auction. Each owner’s equilibrium bid would then equal the budget divided by . Owners with a high private preservation value vi and/or a low development value σi still would be likely to sell development rights to the government in the auction, so the adverse selection problem would persist. However, alternative informational assumptions might alter the extent to which the signaling mechanism dominates the auction. The tax would remain the best mechanism overall unless the government is unable to determine social surplus S. Further exploration of the impact of varying informational assumptions on the efficiency ranking of different allocation mechanisms remains an area for future research.↵11. Experiment instructions are available from authors.
↵12. Auction experiments traditionally involve having the participants submit offers in several “practice” rounds before making decisions in rounds involving higher financial stakes. However, in this experiment, we included practice rounds only at the beginning of the session for the status quo treatment, as we felt that sufficiently helped participants become familiar with the computer interface. As discussed later, we tested explicitly for learning effects in all treatments and found none.
↵13. The potential exception arises in the repeated discriminative auction in treatment B. In this treatment, participants may be able to learn about the market during repeated rounds, and this may affect bidding and overall social surplus over time. As discussed later, analysis of experimental results did not reveal any significant learning effects in the auction or in any other treatments.
↵14. The most recent Census of Agriculture (USDA-NASS 2007, table 8) reports average farmland values for four development-influenced states in the Mid-Atlantic: Delaware ($10,347), Maryland ($7,034), New Jersey ($15,346), and Pennsylvania ($4,775). Agricultural use values vary widely; however, consider 2009 survey data (Delaware Agricultural Statistics Service 2010) from Kent County, Delaware, on the per-acre rent paid for irrigated cropland ($132). If this value is capitalized with a 4% discount rate, then the implied use value is $3,300. One explanation for this substantive development increment-that is, why owners do not sell land into its highest-valued use-is that owners gain nonpecuniary rents. The nonpecuniary rents to land ownership might arise from the heterogeneous utility of an agricultural lifestyle, agricultural land ownership, historical connections, family heritage, stewardship, working as one’s own boss, and so forth. The model parameterizes h(z,vi) so that it captures values in line with development increments in the Mid-Atlantic region.
↵15. It would be expected that providing participants with different sets of information regarding the market outcome would lead to different behavior. For instance, Messer, Duke, and Lynch (forthcoming) show that from the perspective of maximizing the amount of ecosystem services secured by the auction, the buyer should not provide too much information, as that can lead to increased strategic behavior, nor too little information, as that can undermine the competitive nature of the market.
↵16. This contract might reflect a subsidy program in which an owner who transfers an easement to the ES buyer pays $2,000 for some land improvement and receives an additional $1,228.50 subsidy from the ES buyer.
↵17. While there are clearly many difficulties in discerning the development value of each parcel, the ES buyer might obtain this information by conducting an appraisal of parcels prior to establishing contracts.
↵18. These results from the experimental laboratory also correspond to behavior observed in actual reverse auctions for ecosystem services (see, for instance, Kirwan, Lubowski, and Roberts 2005; Messer and Allen 2010).
↵19. Alston and Hurd’s (1990) estimate of the opportunity cost of a U.S. federal dollar is an additional $0.20 to $0.50. Also see Wu and Babcock (1996).
