3. Agri-Environmental Contract Design With Limited Information
The above arguments highlight the importance of research into efficient agri-environmental contract design when information is limited and costly to obtain. Significant progress has been made in designing policy mechanisms which alleviate adverse selection ( e.g. Moxey et al., 1999; Wu and Babcock, 1996; Fraser, 1995; Slangen, 1997; Latacz-Lohmann and Van der Hamsvoort, 1997, 1998; Smith, 1995; Bystrom and Bromley, 1998). Progress on designing efficient solutions to the moral hazard problem has been slower and confined to theoretical models with little relevance to the practice of agri-environmental management (Latacz-Lohmann, 1998; Choe and Fraser, 1998, 1999; Ozanne et al., 2001; Fraser 2002, Hart and Latacz-Lohmann, 2005).
3.1 Policy design to address adverse selection
The policy mechanisms that have been proposed to address adverse selection can be broadly classified into two categories: self-selection mechanisms and auction mechanisms. These are reviewed in turn.
Self-selection mechanisms (through contract menus)
Self-selection mechanisms are set up as principal-agent ( PA) models whereby the principal (the government agency) devises different contracts for different resource settings of agents (farmers) and tailors farmers' choices in such a way as to induce farmers with a particular resource setting to choose the contract meant for them. This type of contract is difficult to design because farmers have an incentive to misrepresent their environmental characteristics to obtain a favourable combination of management prescriptions and payment rates. The self-selection constraint, however, ensures that farmers reveal through the choice of contract their true environmental characteristics, thereby rectifying the previously existing information asymmetry. The self-selection constraint simply requires that farmers with a particular resource setting (say, highly productive land) prefer the contract intended for that resource setting to all other options offered in the menu. Contract variables ( e.g. management prescriptions and payment rates) are thus to be chosen such that a contract for highly productive land provides a farmer who owns this type of land with the highest net payoff compared to all other contract options on offer. The self-selection constraint is supplemented with a participation constraint (also known as an 'individual rationality' constraint) which guarantees that farmers will be at least as well off participating in the scheme as not participating - thus in principle ensuring their participation. An agri-environmental scheme is feasible if it satisfies both the participation constraint and the self-selection constraint. When the agency uses a feasible scheme, each farmer will choose the contract option intended for him or her, thereby reducing the problem of adverse selection. The agency's problem is to determine the feasible scheme that maximises its objective function.
In Wu and Babcock's (1996) model, the agency is modelled as maximising 'social surplus from agricultural production' which is the difference between revenue from, and costs of, agricultural input use minus the damage of pollution and the 'deadweight loss' from distortionary taxes of raising government revenue (such costs are a net loss to society). The optimal solution to the agency's problem is a combination of management prescriptions (input use reductions) and payment rates for different farmer types that maximise the objective function. Under the assumptions made, the optimal payment turns out to be a single, constant payment equal to the minimum amount of money needed to induce the farmers with the best land to participate. Thus, the government can do no better than to offer a uniform payment to all farmers although they are known to differ in their compliance costs. This is tantamount to using a fixed price scheme! The optimal management prescriptions (input schedules) are found to differ from the ones that would be socially optimal in a perfect-information setting: in the presence of information asymmetry, farmers would be allowed to use more input than the socially efficient level under full information. The policy conclusion is that farmers who declare to have more productive land must be allowed to apply more inputs than farmers who declare to have less productive land. The compensation for both farmer types must be the same. If more productive land were compensated at a higher rate than less productive land, then all farmers would claim to have highly productive land, "and the program would soon be a hazard to the morals of farmers" (Wu and Babcock, 1996, p. 944).
Moxey et al. (1999) use a similar model setup and come up with similar results. Their regulator is assumed to maximise the net social welfare of a contract consisting of: a) reduced pollution damage; b) the farmer's monetary benefit (or rent) as the excess of transfer payment over the costs of abatement; and c) the deadweight loss from distortionary taxes for raising public funds. In contrast to Wu and Babcock (1996), Moxey et al. (1999) assume that the regulator has a subjective prior probability of whether a farmer is a high-productivity or a low-productivity type farmer 2. These probabilities are used as weights in the regulator's objective function. They then solve the objective function subject to two participation constraints and two self-selection constraints - one for each farmer type, respectively. The former ensure that farmers must at least be compensated for their reduction in profit; the latter removes the incentives for one type of farmer to declare themselves untruthfully to be another type of farmer. Compared to the full-information (first-best) optimum, the optimal solution to this problem features less demanding management prescriptions (less input reduction) for the high-productivity farmer and a higher transfer payment to the low-productivity farmer. The costs associated with deviating from the first-best (full-information) solution may be interpreted as the cost of information asymmetry relating to the farmer type. The 'truth-telling' second-best solution is designed to minimise this cost and represents a significant improvement on the third-best, no-information, solution of offering a single, uniform contract to a group of heterogeneous farmers. The authors stress that this improvement is brought about without costly information gathering on the part of the regulator. It is achieved through recourse to the revelation principle and the use of incentive compatible contracts.
Despite these advantages, no use seems to have been made of incentive-compatible contracts in the practice of agri-environmental management. This is not surprising given the theoretical nature of contract mechanism design and the restrictive assumptions made in the models; e.g. only two farmer 'types'. This criticism does not hold for auction mechanisms which are discussed next.
Auction mechanisms
Auctions represent an alternative mechanism by which landholders are encouraged to reveal their 'type' ( i.e. compliance costs, resource setting). Rather than relying on self-selection, auctions harness market forces to induce farmers to reveal, through the bidding process, their compliance cost to the conservation agency. Competition is the driving force behind this cost revelation: farmers are asked to bid competitively for a limited number of conservation contracts. In formulating their bids, they thus face a trade-off between a higher net gain from a higher bid and a reduced chance of winning. Producers facing competition are less likely to 'overbid' relative to their true compliance costs. The expectation thus is that competitive bidding will reduce information rents and increase cost-effectiveness. Ideally, the bid would reveal the bidder's true opportunity cost of producing the environmental good in question, thus rectifying the information asymmetry.
Tendering mechanisms have at least three advantages over single, fixed-rate payments. First, they enable participants to deal with the uncertainty about the value of the object being traded. Countryside benefits are public goods for which there are no markets. Holding an auction means that the better-informed party (the landholder) makes the first move in determining an appropriate price, while the less well-informed party (the conservation agency) retains the bargaining power by setting up rules under which the competing claims are compared and selected. In other words, bidding acts as a price discovery mechanism where prices are determined through a decentralised process which takes account of private information held by the bidders. Therefore, compared to a centrally decided, flat-rate payment, auction prices are more likely to reflect the landowners' true opportunity costs. Second, as indicated above, tendering explicitly introduces an element of competition between landholders. Bidders facing competition are more likely to reveal their true valuation in their bids rather than a strategically inflated value. Put differently, bidding reduces the scope for opportunistic behaviour resulting from informational asymmetries. Theoretically modelling has shown that optimal bids increase with both the bidder's opportunity costs and his/her expectations about the unit bid cap (Latacz-Lohmann and van der Hamsvoort, 1997). Thus, a bidder's bid conveys information about his or her opportunity cost, which is private information unknown to government. The information asymmetry is thus reduced, but not completely: indeed, the auction's cost revelation property is blurred by the fact that the bid also reflects the bidder's beliefs about the unit bid cap chosen by the agency. Third, tendering is perceived to be fair, which is politically important, making a transfer of public money legitimate. By holding an auction, the conservation agency avoids being confronted with questions about the level of pre-determined payments.
However, an auction is a complex incentive mechanism, involving a higher risk of failure than a simple fixed-rate payment. First, there is the potential problem of insufficient bidding competition. The smaller the group of potential bidders, the lower the level of bidding competition and the higher the likelihood of collusion and strategic behaviour. Second, bidding involves the risk of learning on the part of the bidders. Experience with the Conservation Reserve Program in the US have shown that bidders tend to analyse the results of preceding bidding rounds and use this information to update their bids. Finally, auctions involve high transaction costs. To the extent that these are upfront fixed costs, they may deter farmers from participating in the scheme.
The diffusion of auctions into the practice of agri-environmental management has been slow, but interest in auctions for purchasing conservation services from landholders has recently grown. At a large scale, auctions have been used only on one occasion: for allocating contracts for the US Conservation Reserve Program ( CRP) since 1986 (Babcock et al., 1996; Plankl, 1999). Interest in conservation auctions has recently increased throughout Australia, especially after the BushTender biodiversity trial auctions in Victoria (Stoneham et al., 2003). In Europe, conservation auctions are being trialled in the states of Lower Saxony and North-Rhine Westphalia, Germany. These trials are reviewed in section 5.
There is, to date, little empirical evidence about the efficiency gains of auctions vis-à-vis fixed-payment schemes. Stoneham et al. (2003) claim that the amount of biodiversity benefits acquired through the first round of BushTender auctions would have cost about seven times as much if a fixed-price scheme had been used instead. A study by Schilizzi and Latacz-Lohmann (2005a) is in the process of evaluating the Scottish 2001 fishing vessel decommissioning exercise, and preliminary results are suggesting that the gains from the auction relative to a budget-equivalent fixed price scheme are not nearly as high. This is more in line with findings reported in Latacz-Lohmann and van der Hamsvoort (1997) who simulated farmers' bidding behaviour in a hypothetical conservation programme. They found efficiency gains ranging from 16 to 29%, depending on how the auction was implemented and how winners were selected. However, White and Burton (2005) find efficiency gains between 200 and 315% for the Auction of Landscape Recovery ( ALR) pilot in Western Australia. These variations suggest that it is probably too early to make a robust assessment of the cost-effectiveness of auctions in agri-environmental management. A more comprehensive and systematic review of conservation auction performance is presented in section 6 of this report.
3.2 Policy mechanisms to address moral hazard
Moral hazard arises due to asymmetric information about compliance (actions), but imperfect information about individual farmers' compliance costs will also affect the solution: farmers with high compliance costs are more likely to cheat, as their payoff to cheating (compliance costs saved) is larger than that of other farmers, whereas the loss on discovery (a fixed fine) is likely to be constant. 3 Furthermore, choices will also be affected by farmers' estimates of the probability of detection as a cheat, and its perceived cost which depends on the fine, but also, potentially, risk aversion and other psychological factors.
A seminal paper on the implications of moral hazard on contract design is that by Grossman and Hart (1983). Since then, principal-agent theory has been applied to the problem of moral hazard in environmental policy. Latacz-Lohmann (1998) develops a decision model to analyse incentives for contract violations at the farm-level (the model is reproduced in Appendix 1). He shows that incentives for non-compliance are shaped by the farmer's compliance costs, the payment level, the detection probability and the level of fine for detected violations. The propensity to compromise conservation agreements is highest when compliance costs are high in relation to the payment level. This in turn suggests that (a) overcompensation can reduce the risk of non-compliance and thus the need for compliance monitoring and (b) that monitoring efforts might be concentrated on 'high-cost farmers', e.g. farmers with high pre-contractual land use intensities. Payments in excess of compliance costs are most likely to be cost-effective in situations where compliance monitoring is technically difficult (and thus costly) and where each contract involves only a small number of hectares. It is further demonstrated that monitoring effort and payment level are substitutes with respect to discouraging non-compliance and that, under certain circumstances, over-compensation of compliance costs can be a cost-effective means of ensuring compliance if the agency's objective is to minimise programme costs.
Choe and Fraser (1998, 1999) also assume the agency's objective is to minimise its own costs, but unlike Latacz-Lohmann (1998) they allow for risk aversion and inaccurate monitoring. They demonstrate that the optimal monitoring intensity and the incentive payment required to ensure farmer compliance may or may not be higher for risk-averse farmers than for risk-neutral farmers. However, they characterise imperfect monitoring as an inability to identify accurately whether or not a farmer has complied, rather than a failure to detect cheating.
In contrast to earlier papers, Ozanne et al. (2001) model agri-environmental contract design as a social welfare maximization problem that recognizes the trade-off between increased environmental benefits and increased costs of compliance monitoring. The solution to the problem identifies the optimal environmental standard (in terms of input abatement), incentive payment and detection probability ( i.e. monitoring effort) for each individual farmer. The authors characterise the solution to the moral hazard problem as second-best in as much as it involves less input abatement than the first-best (perfect information) solution. The effectiveness of the enforcement strategy is measured by the extent to which the second-best solution converges on the first-best solution, and it is shown that this is largely determined by the degree of farmers' risk aversion and the cost structure of the monitoring process. A key finding from their analysis is that high degrees of risk aversion result in convergence of the second-best solution to the first-best solution and thus in a reduction in the severity of the moral hazard problem. Indeed, the authors argue that the study by Latacz-Lohmann (1998), which has assumed that farmers are risk-neutral, may have exaggerated the moral hazard problem.
Fraser (2002) extends Ozanne et al.'s analysis by including production risk. Agri-environmental schemes often stipulate limited use of risk-reducing inputs such as fungicides and pesticides, increasing the variability of farm income. It is shown that full recognition of the income risk faced by farmers, where this income comprises not just policy payments but also production income, diminishes the attraction of cheating among even moderately risk-averse agents and encourages compliance as a risk management strategy. Furthermore, Fraser (2002) introduces the concept of a mean-penalty preserving increase in non-compliance risk, involving a shift in the balance of compliance instruments away from the level of monitoring and towards the size of the penalty. This concept is used to show how the moral hazard problem among risk-averse farmers can be diminished without any change in the expected penalties. It is shown that a principal who has control over both the level of monitoring and the size of the fine has the potential to reduce cheating among risk-averse agents by adjustments to these two instruments which increase the variance of farmers' income but leave the expected penalty for cheating unchanged.
It is reasonable to suppose, and consistent with observations, that the regulatory solution to the moral hazard problem will lead some farmers to participate and cheat, others will participate and comply, whereas a third group will not participate at all. Yet none of the models in the literature reviewed so far have a solution of this nature: indeed, the majority of models allow for a single farmer type, and hence the regulatory solution is one in which all farmers participate and comply. A model which does allow for different farmer types is that of White (2002), who extends the model of Moxey et al. (1999) to include both uncertainty about actual compliance and about compliance costs. However, the variation across farmers is simple enough that the regulator's solution remains (as in cases with identical farmers) to set the monitoring rates at levels such that all participating farmers comply, i.e. there is never any cheating. This is effectively a corner solution which does not fully reflect the complex regulatory problem of dealing with a range of farmer types.
Hart and Latacz-Lohmann (2005) develop a model which allows for a continuum of farmer types with differing compliance costs. The regulator thus deals with multiple heterogeneous agents and is assumed to know the distribution of compliance costs, but not the compliance costs of individual agents. The regulator can affect the payoff to compliance and cheating via both the payment and the probability of monitoring. The fine for detection is constrained to be low (in accordance with observations from agri-environmental management) and assumed not to be under the regulator's control. The regulator has a pre-determined environmental target and minimises total budgetary costs. Only a proportion of farmers are pure profit maximisers, and therefore consider cheating. The remainder are characterised as honest, and never consider the option of cheating. Equilibria with low levels of both monitoring and cheating, in accordance with actual observations, are comprehensible within this model, even when fines are constrained to be low. The honesty/dishonesty distinction has a decisive effect on policy. It is shown that, paradoxically, the total number of cheats may go up at the optimum following an increase in the proportion of honest farmers: if there are more honest agents, it is less worthwhile to perform monitoring which means that the remaining dishonest farmers are more inclined to cheat.
3.3 Payments linked to activities or outcomes?
In contracting for environmental outputs, such as increased bird numbers, reduced water pollution, or slowing of soil erosion, an uncertainty problem exists. Except for some very special cases, the relationship between a set of conservation-oriented activities and the final results at a given point in time are subject to uncontrolled factors. These may include climate, in its normal fluctuations as well as in its more or less catastrophic events, such as floods and droughts, epidemic or invasive species events (locusts, weeds, fungi, disease), and the impacts from activities of neighbouring properties or those situated upstream. Another issue is the measurability of the environmental output. Uncertainty may also lie with the observational and measurement techniques available. For instance, if a number of species is to be achieved in a given area in a given period, such as birds, insects, or fish, it may be difficult, if not impossible, to come up with precise estimates of the numbers.
For these reasons, environmental-oriented contracts have to this day been primarily, if not exclusively, predicated on the landholder's activities or conservation effort. Payment has not been linked to actual conservation outcomes, with the exception of a recent, ongoing project being carried out in Northeim, near Göttingen, Germany, as described in the Case Studies section (Rüffer, 2004; Groth, 2005). Nevertheless, the government agency usually has some control over the final outcome, however imperfect, through the specification of precise management activities, which are expected to deliver to a large extent the aimed-for environmental outcomes. The assumption is that if the landholder indeed carries out the management activities as specified in the contract, the likelihood that the expected outcome is not achieved is rather small. Of course, this would essentially be wishful thinking were it not for the fact that a large proportion of landholders do not necessarily try to renege or cheat on their contractual obligations. The downside is the fact that management activities need to be specified ex-ante, preventing the landholder from using his site-specific knowledge to optimise his effort in response to random events (such as rainfall, frost or disease). And as has been shown by Chambers and Quiggin (2001), the specification of complete contracts in an uncertain environment is extremely costly if not impossible. The landholder is thus not given any incentive for initiative and innovation, since actions have already been contractually pre-specified Indeed, any deviation from the pre-specified actions with the intention to enhance environmental output, may be regarded ex post as a contract violation, especially in situations where the output is judged to be unsatisfactory.
There is a dilemma here between two aspects. Payments made on the sole basis of effort without reference to output create little incentive for achieving the desired output. Payments made on the sole basis of output without reference to effort create a very risky situation for the landholder, insofar as many factors beyond his control, mentioned above, can intervene to perturb the relationship between effort and outcome. The responsibility of the landholder can therefore be difficult, if not impossible, to establish in court.
This dilemma between incentive creation and risk mitigation has been studied in various contexts. Classic examples include the design of crop sharing agreements between tenants and landlords as well as incentive contracts between employers and employees. If the agent ( i.e. tenant or employee) is paid solely for his or her effort, then there is hardly any incentive to work hard in the interest of the principal ( i.e. landlord or employer). Conversely, if payments are made solely on the basis of outcomes ( e.g. yields or sales revenue), risk-averse agents may not be willing to sign the contract in the first place because, given the noisy relationship between effort and output, the result-based payment scheme exposes them to substantial risk. Agents who are willing to bear that risk will demand a high risk premium in return which, in turn, may make the transaction unattractive to the principal. This example serves to illustrate a well-established fact in the contract theory literature: that it is not efficient to expose agents to the full risk of a transaction when the agent is risk-averse and the principal is risk-neutral. The solution to this type of problem is one in which the principal and the agent share the risk. This is achieved by splitting the total payment into two components: one linked to effort, and one linked to outcomes. The effort-based payment could be made once a farmer has signed a conservation contract and has implemented the stipulated conservation activities ( e.g. reduced use of fertiliser or pesticides). The outcome-based payment would depend upon some quantitative measure of the environmental benefits after a stipulated time period. Depending on the objectives of the scheme, these may be reduction below a benchmark level of nitrate concentrations in surrounding water bodies, or the number of certain bird or wildlife species targeted by the programme. Ideally, landholders would be paid a price per unit of environmental service generated, e.g. per bird or per ppm of reduced nitrate concentration. Of course, this solution only works if the environmental outputs of interest are easy to observe or measure and if the time scale is short enough, e.g. one year. Such a scheme would create an incentive for landholders to be productive and innovative in generating the contracted-for environmental goods or services, while at the same time limiting the financial risk from unsatisfactory outcomes.
An outcome-based scheme could be run by allowing landholders to choose the weights that they wish to put on the effort-based and the output-based part of the payment, respectively. These weights would act to reveal each landholder's perception of his/her potential to generate environmental outputs: if the landholder is confident that his efforts will translate into substantial environmental improvements, he or she will put much weight on the output-based instalment, and vice versa.
In practice, however, the prospects of incentive-based payment schemes are likely to be limited. First, there are likely to be measurement issues. Even if environmental outcomes can be quantified in principle, there may be disputes between the contracting parties over the method and the benchmark used for measuring environmental improvements, or disagreement as to how to deal with random external factors such as droughts or floods or natural fluctuations in bird populations. Given the complexity of ecological systems, it seems virtually impossible to control for all contingencies and eventualities in a contract. Hence the transaction costs of a fully specified contract are likely to be prohibitively high. Second, in many cases there are synergies in conserving adjacent tracts of land. For example, nitrate concentrations in a water body are affected by land use practices on adjoining land which may belong to different holdings. For each of the adjoining parcels of land, conservation benefits are affected by any adverse use on the other lands. If part of the payment is linked to environmental outcomes ( i.e. nitrate concentrations), an individual farmer is unlikely to be willing to participate because he or she does not have full control over the way in which adjacent parcels of land are being managed. On the other hand, output-based payments may encourage farmers in the aquifer to join forces in controlling nitrate emissions. To realise such synergies, the agency needs to look beyond the design that allows only contracts to be signed between the conservation agency and individual landholders. We shall return to this issue in section 4.3.5 when we discuss alternative ways of capturing conservation synergies.
An (as yet untested) alternative to output-based payments is to split the payment for management activities into several instalments, on the understanding that the second and subsequent instalments will be paid only if the first-period activities are deemed to have been carried out according to best practice. The recognition of 'best practice' management acts as a reliever of responsibility should the final outcome fail because of random external factors. Yet another alternative is to announce a repeated payment scheme, where new contracts or contract renewals will be made on a regular basis, subject to strict eligibility criteria. This creates some incentive for landholders to achieve the desired outcomes if they wish to be eligible in subsequent rounds. This also allows for more long-term outputs to be observed, if for instance renewals happen say every three years. Another (as yet untested) idea could be to associate an (auxiliary) insurance contract with the (primary) conservation contract, where the insurer might be the government agency; this would allow some transfer of risk away from the landholder, but at perhaps a prohibitive cost to the agency. Lesourd and Schilizzi (2003: chapter 7) show that indeed, for several reasons, environmental risks are not always insurable.
The authors of this report are working on a more general solution to this payment problem in the case of a single environmental output, such as number of birds or ground water nitrate pollution. The context is that of a conservation-oriented contract to be put up for tender in a competitive auction, and the problem is one where the disadvantages of payments made solely on the basis of either effort or outcome are to be overcome, or at least minimised.
Mainly because of the problem of unobservability of environmental outcomes and the risks of litigation linked to unclear landholder responsibility, it appears preferable, for the time being and until some general guidelines have been worked out, to stick to effort based-payments. These can be modulated using ad hoc schemes such as those mentioned above. However, improvements in knowledge (for example, new technology that allows lower-cost monitoring of specific environmental outcomes) may enable conservation agencies to base at least part of their payments on output.