The Scottish Government

« Previous | Contents | Next »

Listen

4. Conservation Auction Design

Having reviewed conservation contract design in the previous section, we now turn to conservation auction design. Auctions can be regarded as a trading mechanism and analysed from a design perspective - that is, how these markets can be designed in such a way that they achieve specific outcomes. Outcomes commonly sought in conservation auctions are allocative efficiency, i.e. selection of participants with the highest benefit-cost ratio, and budgetary cost-effectiveness - that is, buying the most conservation benefit with a given budget of public money. We first review basic auction forms (section 4.1) to establish a benchmark for appreciating the complexity of conservation auctions. We then review the relevant theory and provide an overview of the characteristics of conservation auctions (section 4.2). Finally, we consider the design options one by one and assess the impact of different auction designs on bidding behaviour and auction outcomes (section 4.3). This section also includes an assessment of alternative bid evaluation systems and a discussion of whether conservation agencies should reveal information on bid evaluation to bidders.

4.1 Basic auction forms

There are essentially four basic auction forms: English, Dutch, first-price sealed-bid and Vickrey (Chan et al., 2003). These were initially designed as selling auctions but can also be run as procurement auctions.

• English auctions are open auctions with an ascending outcry format. The price is successively raised until one bidder remains. Bidders bid so long as the current price remains below their own valuations. As the price is raised, bidders successively withdraw from the auction in order of their relative valuations. The good is sold to the last remaining (highest-valuation) bidder at a price just above that which sees the second last bidder withdraw from the auction. The dominant strategy is to bid one's own valuation. Bidding above valuation involves the risk of winning the auction and having to pay more than one's valuation. Bidding below valuation reduces the chance of winning. Bids in the English auction thus reveal bidders' valuations.
• Dutch auctions are the reverse of English auctions, with bids announced in descending order. A bidder wins by being the first to accept an announced bid and pays that price. The name "Dutch auction" stems from the fact that this auction format has traditionally been used in the Netherlands' flower markets.
• First-price sealed-bid auctions require bidders to submit confidential bids to the seller. In contrast to the English and Dutch auctions, bidders cannot observe competitors' bids. The bidder with the highest bid wins and pays that bid.
• Vickrey auctions (named after the economist William Vickrey) have a second-price sealed-bid format. The bidder offering the highest bid wins but only pays the price of the second-highest bidder. The price is thus determined by the marginal loser's bid which is beyond the winner's control. Bidders' dominant strategy is to bid their own valuations. Any other strategy would reduce a bidder's potential gain from the auction. Bidding below valuation reduces the chance of winning the auction as it involves the risk of a bidder's bid being topped by someone else's. Conversely, bidding above one's own valuation may increase the chance of winning the auction, but the bidder may end up paying more than his or her own valuation for the good acquired. As in the English auction, bids in the Vickrey auction thus reveal bidders' valuations.

Bidding strategies in the Dutch and the first-price sealed-bid auctions can be characterised as Nash equilibria. Unlike in the Vickrey format, a bid determines not only the chance of winning but also the payment required if the bidder wins. There is thus no dominant strategy to bid one's own valuation. Instead, bidders form expectations about rival bidders' valuations and bid just high enough to win. Each bidder determines a bid as if an own valuation is the highest among all of the bidders' valuations. With this expectation, a bidder's preferred strategy is to estimate the next highest valuation among competing bidders and bid that value estimate. The Dutch and first-price sealed-bid formats thus require some guesswork on the part of bidders in order to determine their bids. This may lead to unstable auction outcomes. Assuming however, that bidders are not boundedly rational and make no systematic error in this guesswork, the winner is the one with the highest valuation.

We may therefore conclude that all auction formats select the bidders with the highest valuation (provided these bidders are fully rational). This is tantamount to saying that the allocation of goods through all auction formats is economically efficient: there will be no incentive after the auction to reallocate the traded good among bidders: none of the losing bidders would be willing to offer a price that would top the winner's valuation.

There is another outcome property of the basic auction forms that is worth mentioning: under a set of assumptions (listed in Box 1), all basic auction forms yield the same price on average. This is known as the Revenue Equivalence Theorem. This is not to say that every individual auction achieves the same price, but if the basic auction forms were repeatedly used to sell a good to a given number of bidders with stable valuations, they all produce the same average revenue. If the assumptions behind the Revenue Equivalence Theorem do not hold, particular auction formats may emerge as superior (Chan et al., 2003).

4.2 Characteristics of conservation auctions

It is obvious that auctions for conservation contracts do not fit any of the basic auction forms discussed in the previous section. Nor do they meet all of the assumptions listed in Box 1. Standard auction theory therefore offers little guidance for the design of conservation auctions. Conservation auctions differ from the basic auction forms in many respects:

First, conservation auctions are multi-item procurement auctions: government seeks to purchase (rather than sell) multiple units (rather than a single unit) of the conservation good under consideration. A conservation auction selects numerous landholders to take part in conservation, and each landholder can bid for a different activity level. Auction theory is well developed for single-unit selling auctions, but less so for multi-unit procurement auctions.

Second, the items being traded in conservation auctions are heterogeneous. Different parcels of land have different conservation value, implying that a hectare of land offered for conservation in one resource setting may be valued differently from another hectare of land under a different resource setting. Benefits vary across landholders because the same conservation activity performed on different parcels of land can deliver different conservation benefits. Landholders thus effectively bid on price (payment) and service quality.

Matters are complicated by the fact that bids may contain more than one dimension of quality. For example, a conservation scheme may target pollution control, biodiversity conservation and carbon sequestration. As these attributes come in different units of measurement, the agency needs to determine weights reflecting their relative preferences for the different attributes. The appropriate way to address the heterogeneity of bids is to devise suitable bid selection rules which combine price and quality attributes into a score for each bid (see section 4.3.6 - bid evaluation systems - for details).

Further complexity may arise from the presence of synergies in conserving adjacent areas: a parcel's conservation value may increase if adjacent parcels are also put under conservation management. Ways of capturing such synergies are discussed in section 4.3.5.

Third, conservation auctions are usually set up as repeated auctions in that tenders for the same contracts are invited in a sequence of bidding rounds. This may allow bidders to learn from the outcomes of previous bidding rounds and use this information to update their bids. The risk of this happening is quite high in 'networked' industries such as farming, where information is spread quickly through the efficient communication networks of producer organisations or lobby groups. One important design challenge is to contain the scope for bidder learning, as well as that of collusion, which can be correlated (see section 4.3.4 for details).

Fourth, there is the choice of carrying out the auction with a fixed target or, alternatively, with a fixed budget. In the first case, the number of contracts or hectares of land to come under contract is decided upon and known; the risk is with what it might end up costing. In the second case, it is the reverse: the budget is decided upon and is known; the risk is with the number of contracts or hectares that might not come under contract, that is, with the degree of effectiveness of the policy. It seems that target-constrained auctions are used where government cannot fall short of its objectives, as is typically the case with military procurement programmes. In the field of environmental policy, governments' use of the budget-constrained auction probably reflects their general political priorities. As a result, budgets are usually given for environmental procurement programs. This poses a problem to the extent that auction theory has been well developed for target- constrained ( TC) auctions, but less so for budget-constrained ( BC) auctions (Müller and Weikard, 2002). As a result, there is a gap between what is understood by economic theory and what is common practice.

Fifth, there is the choice between different payment formats. In the discriminatory format (the most commonly used), each bidder is paid an amount equal to his actual winning bid (or, if more than one unit is offered, the sum of his or her actual winning bids). In a uniform-price auction, all units sold earn the cut-off price - either the highest accepted or the lowest rejected bid. Therefore, infra-marginal winners receive payments that are higher than the opportunity costs implied in their bids. This does not necessarily mean that the uniform price auction is less efficient, given that bidding strategies are then different; namely, this form of auction contains incentives for bidders to bid their true opportunity costs (see section 4.3.1 for details).

Sixth, the conservation agency is free to set an explicit reserve price. In the procurement case, a reserve price is an upper limit on the amount the agency is willing to pay for a unit of the conservation good being traded. This can be pre-announced or not. It is useful to distinguish between explicit and implicit reserve prices. The first is actively set by the auctioneer to enhance bidding competition. The latter is determined in the bid selection process for budget-constrained conservation auctions. In this auction format, bidders are normally selected in order of their bids (or some benefit/bid ratio) until the budget is exhausted. The last winning bidder determines the cut-off price for the auction. This cut-off price may be interpreted as an implicit reserve price in the sense that it is not actively and intentionally set by the auctioneer; it rather is determined residually by the rules of the auction.

The design characteristics discussed above are summarised in Box 2 together with a number of other descriptors characterising bidders and the management of information by the auctioneer. The next section explores the scope of auction design for addressing the issues raised in this section.

Box 2: Conservation auction descriptors

Source: Hailu and Schilizzi, 2004 © Australian Journal of Management

4.3 Auction design to achieve specific outcomes

In this section we consider the design options outlined above one by one and assess the impact of different auction designs on bidding behaviour and auction outcomes. As indicated above, auction theory is not well developed for conservation auctions and thus offers little guidance for the design of such tendering mechanisms. Most analyses therefore have relied on economic experiments, numerical simulations or, simply, intuition and economic reasoning. Where theory is available, we will review it and summarise the implications.

4.3.1 The choice of payment format: discriminatory or uniform pricing?

The payment format specifies the way of determining contract payments on the basis of bids. It is important to understand that different payment formats induce different bidding behaviours. There are essentially two possible payment rules that can be used to auction multiple contracts in a procurement setting:

(a) a discriminatory (first-price) sealed-bid tender where the n lowest bidders are accepted, receiving the payment stated in their bids, or

(b) a uniform-price sealed-bid tender, sometimes referred to as strike price auction, where all successful bidders are paid the price of the highest successful or, alternatively, the lowest rejected bid. The latter format (based on the lowest rejected bid) is referred to in the literature as a generalised Vickrey procurement auction ( e.g. Hailu and Thoyer, 2005a) because successful bidders are paid the price of the lowest unsuccessful bid. When there is a single unit to buy, then the generalized Vickrey reduces to a classical second-price sealed-bid procurement auction.

For uniform pricing, the choice between the highest accepted bid and the lowest rejected bid has implications for bidding behaviour and auction outcomes (Hansen 1988; Milgrom 1989). Practical differences between these alternative rules depend on landholders' cost profiles, the degree of bidding competition and the agency's budget. These differences have been studied in some detail by Hailu and Thoyer (2005a) using an agent-based model to overcome analytical intractabilities. Their results include the more general case when bidders ( e.g. farmers) are allowed to put up bid-quantity schedules rather than a unique bid-quantity combination ⁴. Their results have been summarised in table format in Appendix 4.

The discussion below abstracts from the different ways of setting a uniform price and focuses on major differences between discriminatory pricing and uniform pricing in the case where bidders offer only one bid-quantity combination, rather than a whole schedule of such combinations. We also implicitly refer to a budget-constrained rather than a target-constrained auction, insofar as the former are nearly always used for conservation contracts.

Impact on bidding strategies

The main difference between discriminatory and uniform pricing is its impact on bidding strategies. Under the discriminatory format, a landholder's bid determines both the chance of winning and the price received for selected activities. Bidding strategies are characterised as Nash equilibria in as much as bid formation depends not only on a bidder's own cost of conservation activities but also on his or her best guess of what the highest acceptable unit bid (or price) might be. This creates room for bidders to shade their bids above their true opportunity costs and thereby to secure themselves an information rent. (The detail of how this happens can be seen by reading Appendix 2.) The optimal bidding strategy in a discriminatory auction thus is one of overbidding: the auction will not reveal bidders' true opportunity costs. This is truer of low-cost bidders than high-cost bidders. In fact, overbidding is highest for the lowest-cost bidders, whereas the highest-cost bidders will bid closest to their true costs (though still marginally above) (Latacz-Lohmann and van der Hamsvoort, 1997; Hailu, Schilizzi and Thoyer, 2005). However, these highest-cost bidders are not those that are usually selected; rather, the lowest-cost bidders are, but they get paid well above their true costs. It is important to understand that this is due to the incentives inherent in this type of auction.

Under uniform pricing, by contrast, a bidder's (unique) bid only determines the chance of winning but not the payment received. Bidders' dominant strategy thus is to bid their true opportunity costs. Any other strategy would reduce a bidder's potential gain from the auction. Shading bids above opportunity costs involves the risk of a bidder's bid being undercut by someone else's, thus reducing the chance of winning the auction. Conversely, bidding below one's own costs may increase the chance of winning the auction, but the bidder may end up being paid less than his or her income forgone from carrying out the specified conservation activities. As in the standard Vickrey auction, bids under uniform pricing rules in the (simple-bid) multiple-unit case can thus reveal bidders' true opportunity costs. As bidders are usually not fully rational, such revelation will not be perfect, but it will be fairly close. If cost revelation is the over-arching priority, and perceptions of fairness are not a political issue, then the implication is to use uniform-price auctions.

Impact on payment

Theory provides little guidance to tell which of the alternative pricing rules can generally yield greater budgetary cost-effectiveness. The rationale for using the uniform pricing rule is that it offers incentive for bidders to enter very competitive ( i.e. low) bids in the hope of increasing their chances of being selected, but on the other hand, under this system successful bidders stand to receive more than the value of their bid. Consider Figures 2A and 2B for a comparative assessment of the cost-effectiveness of the two payment formats under a fixed budget constraint.

With uniform pricing, bids are in principle (if bidders are fully rational) equal to bidders' true opportunity costs, and winners are paid a price equal to the lowest rejected bid. This strike price, p _U, is determined by the auction outcome. In Figures 2A and 2B, the agency budget thus is reflected by the area p _U* X _U (or O EC X _U), where X _U is the units of service which can be bought within the budget. Budgetary cost-effectiveness under uniform pricing ( CE _U) thus is:

Figure 1A: Cost-effectiveness of uniform pricing ( UP) versus discriminatory pricing ( DP) when UP is less cost-effective than DP in a budget-constrained auction: areas OECX_U = OABX_D, but XU < X _D

Figure 1B: Cost-effectiveness of uniform pricing ( UP) versus discriminatory pricing ( DP) when UP is more cost-effective than DP in a budget-constrained auction: areas OECX_U = OABX_D, but X _U > X _D

Under discriminatory pricing, bidders shade their bids above opportunity costs as shown by curve AB. Budgetary cost-effectiveness ( CED) thus is represented by the area underneath the bid curve AB divided by XD, the units of service bought under discriminatory pricing:

Whether CE_D > CE_U or vice versa is an empirical question. Figure 1A has been drawn such that the discriminatory pricing format outperforms the uniform pricing scheme: more units of service ( X _D versus X _U) are bought with the same budget (area OABX_D = area OECX_U). However, the more bidders shade their bids under the discriminatory pricing rule (resulting in an upward shift of curve AB), the more its advantage diminishes relative to the uniform pricing scheme. This is depicted in Figure 1B. As a result, payment equivalence in multi-item auctions cannot be taken for granted.

Within the empirical literature reviewed, we found no studies that explicitly address the choice of payment format for conservation contract auctions. Instead, lessons are drawn from numerous studies of auctions for selling electricity in electricity markets or financial instruments (for example, bonds). These studies provide mixed results on comparing revenues in multi-item auctions. These are taken here to imply mixed results on comparing payments in auctions of conservation contracts.

• Goswami et al. (1996) explain that bidder communication before an auction facilitates collusive strategies under the uniform pricing rule, but competitive strategies under the discriminatory pricing rule. Their results show that uniform pricing is revenue inferior to discriminatory pricing.
• Tenorio (1993) obtains opposite results, but states that uniform pricing can attract more bidders to participate, thereby increasing bidding competition and auction revenue compared with discriminatory pricing. The above empirical results, obtained from selling auctions, most likely apply to procurement auctions too.
• In a study of the US Treasury experience (selling bonds), Malvey and Archibald (1998) conclude that the uniform price auction will produce greater revenue on average, but at the cost of greater uncertainty in any given one auction. They thus cast the choice between uniform and discriminatory in terms of a mean-variance trade-off in seller's revenue.
• Fabra et al. (2002) extend their search to the more sophisticated auctions of electricity markets in California and the UK, where bidders tender whole supply functions (schedules of quantity price pairs). Their work suggests that the previous recommendations to use discriminatory auctions depend on the fact that electricity supply functions were modelled in a continuous manner, and that when a more realistic discrete, multi-unit auction model is used instead, the ranking in terms of social welfare of uniform versus discriminatory auctions becomes inherently ambiguous.

From the above studies, there seems to be no general clear-cut answer to the question about which pricing rule should lead to higher budgetary cost-effectiveness in conservation contract auctions. Given this conclusion, emphasis should be given to practical considerations in choosing the appropriate payment format.

Practical considerations

While uniform pricing looks potentially attractive in theory, it has a number of potential drawbacks in practice.

1. It exposes bidders to greater risk in as much as not only the acceptance probability is unknown but also the value of the bid. If landholders are risk-averse, greater risk may act as a deterrent to participation. This contradicts Tenorio's (1993) observation that uniform pricing can attract more bidders to participate.

2. Owners of the least productive, marginal land (and low opportunity costs) would benefit disproportionately from the higher price. The strike price reflects the amount of compensation required by operators of more productive land. This price clearly overcompensates those who farm marginal land, creating the impression that landholders are 'overpaid'. By contrast, discriminatory pricing has the popular appeal of not paying landholders more than what they bid at auction.

3. Uniform pricing may discourage farmers with productive land to participate because they do not anticipate any realistic chance of being selected. There is also an equity argument here: efficient farmers may find it 'wrong' that their less efficient colleagues with marginal land receive the price deemed appropriate for the efficient farmers. This may act as an additional deterrent to participation.

4. Uniform pricing is more complex and more difficult to comprehend than the discriminatory pricing rule. This may act as a barrier particularly to those who are not familiar with bidding situations. On the other hand, it may increase the risk of collusion from the few who do understand the rules and are able to spot loopholes.

As a result, the choice between discriminatory and uniform price auctions remains a controversial one, dependent on the specific implementation context. However, on balance, the discriminatory payment format appears to be the more appropriate payment rule for conservation auctions, except when there are reasons to believe that bidders will considerably shade their bids (as in Figure 1B). This is likely to happen when the information asymmetry is very substantial and the agency knows it has a very wide knowledge gap compared to bidders; that is, it knows very little about their production conditions. In the limit, however, especially if collusion is expected, an auction should not even be considered.

4.3.2 Reserve prices and reserve quantities

A reserve price strategy is a key element of auction design. McMillan (1994) attributes the failure of the New Zealand spectrum rights auction to the lack of a reserve price. In that second-price auction, the winner bid NZ $7 million but paid the runner-up's bid of NZ $5,000. This outcome could have been avoided had the government set a minimum reserve price that the winner had to pay. In the context of conservation auctions, a reserve price is an upper limit on the amount the conservation agency is willing to pay for a unit of the conservation good being traded. This can be pre-announced or not. There are two potential reasons why conservation agencies should consider setting a reserve price.

First, a reserve price adds to the risk that bidders might lose an auction by bidding too high. It thus increases bidding competition, enabling the auctioneer to capture some of the information rent that would otherwise accrue to the winning bidders.

Second, it may act as a signal of the agency's (or society's) maximum willingness to pay for conservation services, thus representing the demand side of the 'market' in countryside benefits. The benefits of a contract and society's valuation of these benefits raise the question about whether it is worth acquiring the next unit of conservation services. Put differently, demand-side considerations require the conservation agency to take an explicit stance about when the price associated with a contract is considered 'too high' and therefore should not be successful. Viewed in this way, a reserve price can be regarded as a kink in society's demand curve for environmental benefits: at prices above the reserve price, demand is fully inelastic, that is, any quantity offered above that price will not be bought. In this way, a reserve price can contribute to an economically efficient allocation of resources in that it prevents the agency from having to buy environmental services at a price that exceeds society's valuation of the benefits.

Under which circumstances should reserve prices be used in conservation auctions? A reserve price should be considered if bidding competition is expected to be weak and if there is risk of bidder collusion. This is the case when the number of potential bidders is small or when bidders learn to 'game' the auction in multiple bidding rounds. A reserve price may indeed be an effective means of combating bidders' learning in repeated auctions. It would serve to transfer funds between bidding rounds to maximise the conservation outcomes in subsequent auctions. Having said that, one must bear in mind that a reserve price limits the winners' potential gain from an auction. This effect can discourage bidder participation and reduce bidding competition (Levin and Smith, 1994). A reserve price may also be important if one budget was used across several auctions. Then the agency would want to ensure that the highest price paid in each auction was approximately equivalent. If an agency were to allocate contracts at very high prices in one auction relative to another, it could have obtained more biodiversity by redistributing funding from the 'high-price' to the 'low-price' auction (Stoneham, 2005).

Reserve prices are less important where there is a strict budget constraint (see Myerson 1981, Riley and Samuelson 1981). As indicated above, the cut-off price in a budget-constrained auction may be interpreted as an implicit reserve price which is determined residually by the rules of the auction in conjunction with the available budget. Its effect on bidding behaviour is likely to be very similar to that of an explicit reserve price.

Should reserve prices be pre-announced? We are not aware of any research explicitly addressing this question. Experience from the Conservation Reserve Program, however, suggests that reserve prices should not be announced because they may create an anchoring bias: Reichelderfer and Boggess (1998) found that bidders in the Conservation Reserve Program _ which is a repeated auction _ revised bids from previous rounds by offering bids at the reserve price. The reserve price in this case was set as a per-hectare rate and when landholders learned this reserve price, they anchored their bids accordingly. We believe, however, that where a reserve price is used, it is important to communicate to potential bidders before the auction that such a price has been set - without revealing the precise price.

We now turn to a related issue, that of a reserve quantity, rather than reserve price. Where bidders are allowed to submit a bid covering several units of a good ( e.g. several hectares of land), and where some bidders can submit a large number of units at a relatively low price, to the point where a bidder's bid represents a large fraction of the total available budget, then the government agency can decide to put a lid on the maximum allowable bid. This is what happened (ex post) in the Auction for Landscape Recovery in Western Australia (see section 5 - case studies). One bid represented a very large fraction of the budget, and although the cost per unit area was quite competitive, this bid was not selected. The decision to have a reserve quantity seems to reflect a concern for fairness or equity across landholders, rather than a concern for cost-effectiveness or economic efficiency. It could also reflect a concern for the participation rate in the next round of auctions, if it appears that some landholders may be discouraged from participating.

4.3.3 Target-constrained ( TC) or budget-constrained ( BC) auctions?

Conservation auctions can be conducted either with a fixed budget (which is the norm) or, alternatively, with a fixed target. In the first case, the agency accepts bidders based on their benefits to bid ratios until a predetermined fixed budget is exhausted. The size of the conservation scheme in terms of hectares of land enrolled is determined as a residual from the budget and the bids offered. In the target-constrained auction, by comparison, the agency predetermines the size of the conservation scheme and accepts bids until the target is achieved. In this case, the necessary budget is not known before the auction is completed. There are no a priori reasons to believe that one auction format is better than the other except, perhaps, that the existence of a budget constraint may have the psychological effect of 'disciplining' bidders, encouraging them to bid closer to their opportunity costs than they might otherwise do.

Latacz-Lohmann and Schilizzi (2005) are the first to set out to investigate the performance of these two auction formats. Given the lack of empirical studies and theoretical guidance, the comparison was made with the use of controlled economic experiments carried out both in Kiel, Germany, and Perth, Western Australia. Both auction formats were submitted to a common experimental setup. The Perth experiment was meant to replicate the Kiel experiment, in order to check for the stability of results. See Box 3 for a summary of the experimental setup.

The key findings from the Kiel experiments are reported in Table 1. Very similar patterns emerged from the Perth experiments (not reported here).

The results suggest that the choice of auction format has only a minor effect on outcomes, at least in this setting. The BC auction performs slightly better than the TC auction in the first two rounds: the budgetary costs and opportunity costs per kg of nitrogen abated are lower. However, these differences are minute and are reversed in the third round where the TC auction outperforms its BC counterpart. Note that auction performance under both formats deteriorates significantly with repetition, indicating that bidders have, to an extent, learned to game the auction, shading their bids above opportunity costs. Information rents increase from an average of 37% of payments made in the first round to over 50% in rounds two and three. There appears to be no significant difference between the two auction formats in their robustness to bidder learning.

Table 1: Dynamic BC versus TC auction performance, Kiel

Source: Latacz-Lohmann and Schilizzi (2005)

We conclude that there are no strong reasons to suggest the preferred use of one of the two auction formats. Both perform equally well as one-shot auctions and under repetition. Practical considerations, however, call for the preferred use of the budget-constrained auction: it is the more 'natural' format in the environmental area where schemes usually have a limited budget and EU regulations limit the degree of overcompensation of farmers' opportunity costs. A target-constrained auction may be called for in situations where the agency cannot fall short of its environmental objectives.

4.3.4 Dealing with bidder learning in repeated auctions

Auctions for conservation contracts are normally designed as multiple-signup, or repeated, auctions where bids for the same contracts are invited over a sequence of years. Such a system provides scope for bidders to analyse the results of preceding bidding rounds and use this information to update their bids. Experience with the Conservation Reserve Program in the US has shown that after a few bidding rounds the average bid was almost exactly equal to the maximum acceptable payment level from preceding rounds, implying that farmers had learned the cut-off points (Riechelderfer and Boggess, 1988). Under no circumstances should the conservation agency therefore publish information about the average or the maximum accepted bid or the distribution of bids received in preceding bidding rounds. An auction is a mechanism in which the distribution of information between buyer and seller plays a crucial role. The benefits of auctions decline with the amount of information shared by the two parties.

A potential remedy to the problem of learning is to amend the rules of the auction in each bidding round. This would make the system less predictable, thus maintaining a certain degree of uncertainty among landholders. This could be achieved by targeting different groups of farmers or different regions, by changing the criteria for bid selection, by introducing different cut-off for different farm types or production systems. Whether these changes should be announced to potential bidders prior to the commencement of bidding is a controversial question which will be discussed in greater detail in section 4.3.7. In any case, the use of reserve prices is advised in multiple-signup auctions.

Another approach to the problem is one which has been implicitly used in the Countryside Stewardship Scheme (in Scotland: Rural Stewardship Scheme). Instead of asking landholders to submit a financial bid over a pre-specified contract of activities, they can be asked to choose among a menu of possible activities and indicate those which they are willing to implement for a given, pre-determined price as well as the level of each activity (time, area, effort). In this approach, which has not lead to any theoretical analysis as yet, landholders effectively bid on quantity rather than price. They thus submit management plans indicating the conservation activities they are willing to implement for a given per-hectare payment. The bid thus is not financial but takes the form of management commitments. A ranking system is applied to translate management commitments into expected conservation benefits. A total ranking score is computed based on the conservation benefit likely to be delivered by each proposal. Thus, the more conservation activities a bidder chooses from the menu at a given price, the higher his or her chance of securing a contract. The reason why learning would be more difficult and slower is that instead of having to learn a single bid price (say, a per-hectare rate), bidders must now learn a whole list of activity levels and identify combinations of different activities and their respective minimum levels sufficient to secure a contract (Latacz-Lohmann and van der Hamsvoort, 1998).

4.3.5 Capturing conservation synergies

As indicated above, the value of a conservation contract to the agency may increase if nearby lands are coming under conservation management at the same time. We call this situation 'site synergies'. This is the case with catchment programmes and with the establishment of vegetation corridors for the movement of species across the landscape and regional biodiversity enhancement. Very often geophysical or ecological characteristics of the land are the sources of synergies. For example, creating favourable conditions for rare wildlife species normally requires a co-ordinated effort by many landholders in the area. The higher the ratio of land area to land perimeter, the greater is the potential for generating conservation benefits. Likewise, conserving a wetland requires all adjoining land to be put under conservation management. For each of the adjoining lands, conservation benefits are affected by any adverse use on the other lands. In economic jargon, such boundary symmetries may be interpreted as increasing returns to adoption of conservation management. Present policies do not take account of site synergies because they concentrate on contracts between government agencies and individual landholders.

The presence of site synergies requires auction design refinements. If the agency's goal is to attain allocative efficiency, it may wish to encourage all pertinent landholders to participate in an auction so that bids with synergetic values are accepted together. However, this bid selection rule exposes landholders to the double uncertainty of winning a contract and being able to realise the synergetic value (if any) of their lands. Since the latter depends on neighbours' entry and bidding strategies, individual bidders may not consider site synergies when determining their bids. Their bids will then only depend on their own activity costs and benefit contributions, but not on neighbours' benefit contributions. These contingencies expose the agency to the risk of making high payments to individual landholders without realising full synergy values from contiguous lands if neighbouring landholders do not all win in an auction - for example, where landholders in a watershed do not bid for contracts because they are not aware of the synergies with surrounding lands or because high costs diminish their chance of winning (Chan et al., 2003).

To capture site synergies, the conservation agency needs to look beyond the design that allows only separate bids from individual landholders. One solution is to allow neighbouring landholders, or indeed a consortium of landholders in a particular region, to submit joint bids that cover sites belonging to different holdings. Such an arrangement would make bidding more flexible for landholders to the extent that they can choose to bid for conserving their own lands or make joint bids with others, or both. Landholders who submit a joint bid have the local knowledge to decide among themselves how best to share effort and payments among their members, thereby helping the agency with the funding decision (Chan et al, 2003).

The efficiency and payment properties of joint bidding are barely explored in the literature and detailed auction rules are yet to be developed. A potential problem is lack of bidding competition. If there are only a handful of conservation consortia competing against each other, there is scope for collusion. Also, transaction costs of joint bidding are likely to be high and, to the extent that these are upfront costs, may act as a deterrent to participation. Moreover, the feasibility of making long-term agreements with multiple parties would need to be explored before joint bidding can be put into practice. Auction theory has not been extended to this particular circumstance and cannot therefore offer useful guidance for the design of joint bidding schemes. The outcome properties of joint bidding schemes would need to be explored through appropriately designed economic experiments, before such schemes can be recommended for practical use. To the authors' knowledge no such experiments have been reported in the literature.

Another unresolved issue is how best to provide incentives for landholders to act together. Stoneham (2005) suggests that the conservation agency could score bids contingently: landholder A's score would be greater if he or she were successful in conjunction with landholder B. This would have to be communicated to potential bidders before bidding commences. Since landholders cannot be expected to be aware of conservation synergies, it is important to provide information to those landholders that they were providing synergy benefits. In the Scottish context, this may be relevant to the Rural Stewardship Scheme, particularly its Common Grazing measures. A more complex approach would be to provide some financial incentive that was associated with the synergy (Stoneham, 2005). Economic experiments would have to be carried out examining the appropriate structure of such incentives before they can be recommended for practical use.

4.3.6 Bid evaluation systems

The choice of an appropriate bid evaluation system is a key design issue. The main purpose of bid evaluation is to incorporate information on conservation benefits in an auction. As discussed above, conservation benefits often depend upon the resource setting. Matters are complicated by the fact that bids usually contain more than one type of conservation benefit.

The conservation agency can enhance the efficiency and cost-effectiveness of the auction by tailoring bid selection methods for different regions or different resource settings, thereby accounting for spatial variations in the potential for delivering conservation benefits. There are essentially three policy design variables that can be used for targeting conservation auctions: eligibility criteria, bidding pools, and bid discrimination mechanisms. These are considered in turn below.

Eligibility criteria

Eligibility criteria are an ex ante instrument for excluding landholders with resource settings which are deemed to generate insufficient environmental benefits if put under conservation management. Ex ante means before bids are submitted or landholders apply for the scheme. By contrast, bid selection criteria are ex post instruments, which become operative at the time landholders have submitted their bids.

Eligibility criteria can relate to any aspect or characteristic of the resource setting ( e.g. soil type, proximity to water bodies, hydrological conditions), the activities carried out on the land ( e.g. only pasture or only arable land) or its owner ( e.g. level of education, age, other demographic characteristics). They can also relate to the environmental objectives pursued by the scheme (such as particular species targeted). For example, eligibility for the US Conservation Reserve Program is limited to land classified as highly erodible.

Eligibility criteria are an effective means of directing funds towards specific groups of landholders and resource settings - those who are deemed to contribute most to the scheme's objectives. Their use is advised in particular when service quality cannot be precisely measured or if the agency is unable to verify service quality after auction. Eligibility criteria constitute certain quality thresholds that bidders are required to pass before submitting bids.

However, there are two economic criticisms of this approach. The first is that activity-based eligibility criteria may create perverse incentives for landholders to change their land use activities ( e.g. convert grassland into arable land if only arable land is eligible) in order to become eligible in future bidding rounds. The second criticism is that eligibility criteria reduce the number of potential bidders. This may affect bidding competition or increase the risk of collusion. From a bidding point of view, a system that allowed everybody to bid and then discriminated through bid selection criteria would be preferable. Finally, there is an equity issue. Ineligible landholders may find it unjust or inequitable to be denied the chance of participating in the scheme.

Bid pools

Another way of increasing the cost-effectiveness of conservation management is to use bidding pools. This is a useful exercise if the cost of implementing the stipulated conservation measures varies widely between different farm types, production systems, soil types, regions, etc. If for example, it is known that arable farmers face significantly higher costs of conservation management than dairy farmers, one may assign bids to two separate bid pools, one for arable farmers and one for dairy farmers.

The conservation agency would expect bids from arable farmers to be higher on average and would accept bids up to a higher level than in the dairy farmer pool. Such a system effectively limits competition between different farming sectors and encourages farmers with higher opportunity costs to tender bids. This system is particularly useful if it is intended to attract a balanced mix of land and farm types into the scheme rather than 'buy' a large share of the least profitable land (which may not be the land that generates the greatest environmental benefits). A potential disadvantage of bid pools is that they can reduce the number of bidders (in each category) to an extent that bidding competition suffers.

Bid discrimination mechanisms

The agency selects bids by comparing payments either to activity levels or expected conservation benefits. Activity-based bid selection must be used if the transformation function that maps conservation activities into environmental outcomes is not well known. In the simplest case, bids are compared simply in terms of payment per unit area of land. In the absence of any information on the transformation function, land area is taken to represent activity level, which is a proxy measure for the amount of conservation benefits. In other cases, benefit indicators can be used to quantify site-specific effects of conservation activities.

For example, in the Australian Bush Tender pilots, expressions of interest were first called for from eligible landholders; then government officers visited the farms and the proposed land areas up for tender. Ecological data were collected on these areas and analysed by scientists to compute a Biodiversity Benefits Index ( BBI). ⁵

Bid selection can be a complex task if multiple objectives are pursued or if bids contain multiple dimensions of quality. The agency not only faces the challenge of measuring the various quality attributes, it also needs to determine weights reflecting its relative preferences for the different attributes. The relationship between these attributes or objectives can be complementary, competitive or neutral. If there is a competitive relationship between objectives or quality attributes, tradeoffs must be made - higher achievement of objective A means that less of objective B is achieved, and vice versa.

To take account of such complications, a multi-criteria bid scoring system can be used to aggregate the various dimensions of quality into one figure representing an estimate of the overall conservation benefit of each bid. These aggregate measures of conservation value draw on environmental assessments conducted by ecologists and are limited by whatever data a conservation agency is able to collect. Aggregation methods differ mainly in the choice of relative weights for individual benefit variables. Examples include the biodiversity quality ( BQ) index used in the BushTender pilots and the environmental benefit index ( EBI) currently used in the Conservation Reserve Program. The BQ index is the product of the biodiversity significance score and the habitat services score. The first factor measures current conservation value and the second measures activity level. Bids are ranked by the ratio of the BQ index to the payment. The EBI ranks bids according to their contribution to each of the six programme objectives. Bids are assessed on the basis of indicators (proxies) which relate either to certain observable attributes of the land (such as proximity to water bodies or to erodibility) or to certain land management practices (such as land cover type). Each attribute or land management practice carries a score. This scoring rule allows the multiple attributes of quality to be aggregated into one figure which can then be compared to payments. ⁶

Box 4, quoted from Chan et al. (2003), depicts the structure of the bid selection problem in a multidimensional auction.

Using Data Envelopment Analyse to rank bids

One way to estimate a value-for-money frontier as the one shown in Box 4 is to apply Data Envelopment Analysis ( DEA) to the quality attributes of the bids so as to compute a value-for-money score for each bid. DEA is a sophisticated approach to measuring and comparing the efficiency of different units where there are multiple incommensurable inputs and outputs. In the public sector, the method has been applied to measure and compare the efficiency of schools, hospitals, government departments, etc.; it has also been extensively used to compare the relative efficiency of farms and industries. Appendix 5 sets out the formal structure of DEA-based bid ranking models.

The DEA approach is tantamount to adopting the landholders' point of view on the weighting function and presenting their bids in the most favourable light in comparison to competing units - rather than applying a common set of weights across all units. For example, a farmer submitting a bid for a parcel of land adjacent to a lake would wish a high weight to be placed on water quality (if this is one of the objectives of the scheme) and little weight on other criteria. Indeed every bidder will wish to place much weight on the quality attributes that his or her bid is best poised to deliver. DEA thus recognizes the legitimacy of the argument that different units value different outcomes differently and therefore adopt different weights.

DEA effectively constructs a value-for-money frontier as the one shown in Box 4. The frontier has as many dimensions as there are quality attributes. All bids which have been assigned a value of one lie on the frontier and thus offer good value for money. This is not to say that all bids with a score of one will necessarily be selected: the value-for-money frontier does not, in itself, take account of the agency's preferences for different quality attributes. It rather is based on the concept of technical efficiency, not taking into account the agency's preference weights for each quality component.

One can easily extend the DEA approach to compute scores that reflect the agency's relative valuation of different quality attributes. This is equivalent to the use of the concept of economic efficiency in production economics. Scores are then computed relative to point k in the figure in Box 4. Point k not only represents good value for money (because it lies on the frontier), but also offers the mix of quality attributes preferred by the agency, given its preference weights (represented by line CD). Point m also offers good value for money (it also lies on the frontier) but, to the agency's liking, contains too much of B1 and too little of B2. Applying the concept of economic efficiency, it would therefore be assigned a score of less than one (although it still is technically efficient, i.e. on the frontier). The concept of economic efficiency thus allows one to include the agency's preference weightings in the bid selection process.

There is standard DEA software ( e.g. Coelli et al., 1998) which can easily be adapted to the specific case of bid ranking. The information needed to construct the value-for-money frontier comprises, for each bid, estimates of the different quality attributes and the bid amount. All quality attributes must be readily observable and not subject to manipulation or misrepresentation by the bidders. For all bids not on the frontier, the programme indicates by how much the bid would have to be lowered until it lies on the frontier. In order to take account of the agency's preferences, one can input preference weights (output prices) and click on economic efficiency to come up with appropriate scores for bid selection. Bids are then ranked by these efficiency scores, and the offers with the highest scores would be selected until the budget is exhausted.

DEA is a sophisticated method of bid evaluation. It is less transparent and more difficult to explain to bidders than alternative bid ranking systems. From a practical perspective, it offers the advantage of generating virtually incontestable bid rankings: If a bid turns out to be bad value for money (even when the most favourable weights have been chosen from the bidder's perspective), then there is a strong argument for rejecting the bid. DEA effectively 'polishes' each bid (by choosing optimal weights) before subjecting it to competition with rival bids. This minimises the scope for complaints and appeals by unsuccessful bidders.

Using the DEA technique to weight multidimensional bids was suggested in a study that Latacz-Lohmann (2001) did for the SEERAD auction for decommissioning fishing vessels in Scotland, and it was used in preliminary work analysing the auction data by Schilizzi and Latacz-Lohmann (2005a). Work along these lines is also being carried out by Peter Bogetoft in Denmark (Bogetoft and Nielsen, 2004). Critics may argue that the flexibility in the choice of weights for individual bids may allow a bid to appear good value for money, but that this is to do more with the choice of weights than with the bid's inherent value for money. Policy administrators may also argue that DEA gives away their judgment of the 'right' weights to the bidders. Both criticisms are not tenable if the concept of economic efficiency is applied. This allows the policy administrators' judgment of appropriate weights to be included in the bid selection process, ensuring that bids which appear to offer good value for money also match the agency's preferences. Another criticism might be that the agency can, after the bids have been put in, manipulate its preference function ( i.e. change the slope of the CD line) so as to distort the selection in one way or another. However, the procedure would, even if only ex-post, make such a choice transparent. Whether such information should be communicated ex-ante, that is, in advance, is our next question.

4.3.7 Information hidden versus information revealed to bidders

Prior to landholders forming their bids, the government agency needs to decide how much information about what it values in the auction it is going to reveal to them. As it happens, the choice is not straightforward, as there are things to be said for and against each option. It is best to quote from a draft report by Stoneham et. al. (2005, p.32) regarding their experience with BushTender in Australia:

"One of the most interesting design issues with the BushTender pilot contracts was the extent to which information was made known to landholders prior to formulation of their bids. For the pilot auctions some of the information about the biodiversity metric was withheld from landholders: they knew the Habitat Services Score, which revealed the value of their management actions, but not the Biodiversity Significance Score, which reflected the ecological value and potential of their chosen land plots. As noted earlier, this strategy was supported by the experimental findings of Cason et al. (2003). Although the strategy to withhold information was adopted for cost-effectiveness reasons, other considerations suggest that full disclosure of information about biodiversity significance may be appropriate. In the short-run, withholding some information limits the scope for landholders to extract information rents from the auction. Clearly, if a landholder knew that she had the only remaining colony of some plant or animal, she would be able to raise her bid well above opportunity cost, compared with a situation where this information were not known. The alternative strategy also has merit in that (i) the information rents that accrue to landholders would influence land markets and encourage investment in nature conservation; and (ii) landholders would know exactly what scarce biodiversity assets they have and could self-select into the auction process, that is, there may be a better matching between government priorities and the bidders in an auction."

The choice as to what to reveal to bidders ex ante therefore depends on how the government agency weights its different policy objectives; for instance, the relative weight between short term cost-effectiveness in terms of budget outlay and the longer term investment incentives created for landholders.

Chan et al. (2003) argue that the agency's optimal information policy depends upon who holds information about the significance of environmental assets that exist on farm land. On the one hand, landholders can have private information about the environmental impact of their production, such as the potential effects on rare or threatened species on particular tracts of land. On the other hand, the conservation agency can know better than landholders the ecological significance of their lands and the match of policy objectives with land characteristics. We consider both cases in turn.

Bidders with private information on service attributes

If bidders hold private information on service quality, transparency in the bid selection process is essential for efficient contracting. The scoring rule and the relative weights put on different quality attributes should be announced to potential bidders. Bidding competition then effectively ranks bidders by their cost structures. The winning bids then reflect the agency's preference ordering, contributing to allocative efficiency. The agency overpays for quality because bidders who offer high quality will tend to shade their bids above opportunity costs, exploiting their information advantage. If however service quality cannot be precisely measured or if the agency is unable to verify service quality after auction, then there is no sense in laying open the scoring rules or indeed in trying to score bids based on service quality. In these circumstances, the agency should require bidders to pass certain quality thresholds (eligibility criteria) before submitting price-only bids (Cripps and Ireland, 1994; Chan et al., 2003).

Agency with private information on service attributes

Often the agency has superior information about the environmental significance of certain activities on particular parcels of lands - for example because only the agency has the know how to measure environmental impacts. Furthermore, landholders may not have all the relevant information about government priorities and are unlikely to understand how this information might influence subsequent contracts. In these circumstances, the agency's information policy plays an important role in determining bidding behaviour and auction outcomes.

Box 5 quotes from Chan et al. (2003) two stylised cases to illustrate the importance of the information setting to auction outcomes. Thus, if bidders know their own quality estimates and the agency's preferences, they will strategically avoid price competition and emphasise quality competition, thereby capturing a larger share of information rents. Each bidder effectively chooses to offer a particular mix of service attributes - the one that best matches the agency's preferences. They thus compete directly only with those offering the same quality mix, thereby reducing price competition. By contrast, lack of information on the agency's preferences compounds the guesswork that bidders face in formulating their bids: not knowing the agency's preferences, bidders face increased uncertainty in their own quality assessments. The greater this uncertainty, the less they are able to supply favourable but costly service attributes in order to increase their chance of winning a contract. Price competition becomes the driving force of the auction, preventing bidders from capturing large information rents.

The choice between disclosing or concealing information on the agency's quality assessment rules also has important implications for allocative efficiency, i.e. selection of the participants with the highest benefit-cost (or quality-cost) ratios. If bids are positively correlated with quality-cost ratios, the auction selects low-cost, high-quality participants. This is likely to be the case if bidders are well informed about the agency's preferences and its quality assessment procedures. If, by contrast, bidders are unable to assess the quality of their bids, they have to determine their bids on the basis of guesswork involving random choices. The less bidders are able to assess the quality of their bids, the more their bidding strategies will be driven by strategic motives and randomness, and the less they will be driven by quality-cost considerations, leading to greater efficiency distortions (Chan et al., 2003).

Empirical evidence

Cason et al. (2003) used laboratory experiments to examine bidder behaviour in a discriminatory price conservation auction when the regulator reveals to landholders the environmental benefits estimated for their alternative projects, compared with when this information was not revealed. These experiments indicate that when bidders did not know the value of output, their bids tended to be based on the opportunity costs of land-use change. By contrast, when bidders were given information about the significance of their biodiversity assets, they tended to raise bids and appropriate some information rents. Total abatement was lower and seller profits were higher when landholders knew their projects' environmental benefits. This performance difference arises from landholders' ability to condition their offers on their projects' environmental quality when the regulator reveals quality information. Sellers in this treatment clearly made higher offers for high-quality projects, since they knew that high quality gave these projects priority in the bid selection process. This strategic incentive to raise bids for high-quality projects also led to greater bid variance in this treatment. As a consequence, some high-quality projects that should be undertaken on efficiency grounds could not be funded within the given budget, so less environmental benefit was acquired through the auction when landholders learned their projects' environmental quality. The authors conclude that concealing quality information will improve regulatory efficiency.

Vukina et al. (2004) arrive at similar conclusions in their analysis of bids from the Conservation Reserve Program ( CRP) auctions. The CRP pays farmers to remove land from production and put it to a conservation use. An interesting aspect of these auctions is that winners are determined by a combination of low bids and environmental scores of individual plots (see section 5.2 for details). The results indicate that farmers condition their bids on the strength of their environmental scores and that they consistently value those environmental improvements which are concentrated locally such as reduced soil erosion, while they place less emphasis on those benefits which resemble public goods such as air quality and wildlife habitat.

Conclusion

By concealing information on service quality, the agency may succeed in reducing contract payments and gaining a larger share of information rent. Any payment savings for the agency, however, come at the potential cost of allocative efficiency losses. A trade-off exists between efficiency and cost-effectiveness. Extracting a higher share of information rents from bidders, by itself, achieves little in promoting efficient use of resources.

In a quality-concealed auction, bidders have an incentive to acquire information on quality assessment so that they can avoid price competition and increase their share of information rents. If bidders can infer quality measures from previous auction results, then the agency's information advantage degenerates. As a counter tactic, the agency can secretly alter the weights attached to each dimension of quality from year to year. Bidders will eventually perceive such secret changes in the way of assessing quality to be a source of uncertainty in their quality assessments, encouraging them to submit closer to opportunity costs (Chan et al., 2003).

A reasonable compromise might be to publicise the quality criteria but not the weights attached to individual attributes. On the one hand, knowledge of the package of attributes that the conservation agency intends to purchase is necessary to attract the 'right' landholders and give them some guidance to formulate bids that match the agency's preferences. On the other hand, too detailed knowledge of the criteria and the weights for each attribute may encourage bidders who offer high-quality packages to shade their bids above opportunity costs.

« Previous | Contents | Next »