Tendering and bidding for access: a regulator's guide to auctions.

Abstract:

Tenders and traditional auctions are increasingly being used in regulation; as a consequence it is essential regulators have an understanding of the basics of auction theory. There is no simple prescription for the design of an auction or tender that will fit all situations; moreover, the costs from inappropriate auction and tender design can be high. For example, while a second-price sealed bid auction is efficient in the absence of collusion, if participants can collude then such auction formats can result in very low revenue. Low revenue can also result with a second-price auction when there are few bidders and no reserve price.

Keywords:

TENDERS; AUCTIONS, REGULATION.

1. Introduction

Tendering and traditional auctions have become increasingly common in the context of regulated industry. Examples under the Australian Competition and Consumer Commission's (ACCC) charter include tenders for airport freight handling, airport services, allocation of gas distribution rights and harbour towage services. (1) These examples are relatively small scale when compared with the potential use of auctions in other areas of interest to the ACCC such as auction of airport landing slots; auctioning of access to local loops for internet, cable and telephony; auction of mobile telephone spectrum; and even auctions of airports themselves! The application of auction theory to regulated firms is very new, and there has been very little written in this area. Therefore, much of the analysis in this paper will draw on existing auction theory, and make preliminary conclusions concerning their application to regulated entities. An understanding of the basics of auction theory is essential for understanding how regulated firms might react in the face of different auction formats. To that end, we present a summary of some of the main conclusions from auction theory. The aim of this paper is not to create a recipe book for regulators on the application of auctions, but to introduce concepts and improve understanding of how auctions might work in a regulated environment.

2. Overview

Tendering for the purpose of supply and the selling of access to competitors are important areas of concern for regulators. It turns out that the lessons from auction theory are applicable to both of these areas; there are some general conclusions that can be made regarding tendering when the regulated firm is a cost-minimiser. However, the main conclusion of this paper--and indeed of the economics profession--is that auctions are very sensitive to context; that is, the best form of an auction when regulated firms are involved depends on the market environment, for example, what information participants have, how this information is generated, how many participants there are, whether there are concerns of collusion in the auction, and so on. Different types of auctions perform well in different circumstances, and the consequences of choosing the wrong form can be dramatic. (2)

The consequences of a poor understanding of auctions have been disastrous in some cases. Unfortunately two particularly bad examples come from Australasia. The auction of cable TV rights in Australia allowed parties to make multiple bids, and included no penalty for withdrawal of these bids. The strategy adopted by bidders was to make a series of bids ranging from low to high, withdraw at will from winning high bids, leaving their lower bids in with a chance of winning. The actual winner purchased a cable licence at a very low price, and subsequently sold it soon after for a very large profit. (3) The second example was the use of a Vickery or second-price auction design to sell radio, television and mobile telephone licences in New Zealand. This auction awards the licence to the highest bidder who then pays the second-highest price. Unfortunately, in some of these auctions few bidders participated. As the government had not imposed a reserve price, the winners of these auctions paid prices well below their bids; revenue was very low as a result. (4)

Overall, there are two main points that this paper makes. First, given the increased popularity of auctions, particularly in regulated environments, an understanding of the predictions of auction theory is essential for policy makers. Second, particular institutional and market details can dramatically influence the outcome of any auction or tender process. It is critical, therefore, that policy makers consider these factors when designing an auction.

The paper is set out as follows. In section 2.1, we introduce some key questions that arise for regulators in the areas of tendering and access. In section 3, we look more closely at the economics of tendering and access. Section 3.1 examines the key questions as they arise for tendering to illustrate the importance of the regulated firms objectives. We argue that if the regulated firm has cost minimization as an objective, then regulators need not concern themselves with the format of the tender. However, if this is not the regulated firm's objective, then there is a case for regulators or government to be concerned about tender design. In section 3.2 we conduct a similar exercise for sale of access by a regulated firm. Section 4 provides an overview of different types of auction and provides some of the basic lessons from the existing theory. Section 5 explores the preferred design of an auction in different circumstances, particularly in the presence of collusion. Section 6 concludes the paper.

2.1 Key Questions on Tendering and Access

Here we consider providers of goods and services in both the private and public sectors. Let us take the case of a buyer who advertises for potential sellers to submit bids to supply a good or service, and selects from among the pool of applicants. For example, in the case of architectural tenders, sellers submit a basic plan for the building, and the buyer selects the bid that has the best mix of price, functionality and aesthetic characteristics for their tastes. Tendering is most effective when it takes place between a substantial number of buyers and sellers, with repeated interactions over time: for example, if there are many buyers wanting buildings constructed, and many architectural firms to bid for this work. In this way, tendering works as a matching process between buyers and sellers under reasonably competitive circumstances. However, tendering also occurs in the context of imperfectly competitive, regulated industries, and under circumstances where markets are thin and information is imperfect. Consider regulated airports tendering for supply of freight handling and other services, or regulated port authorities that ask for bids from potential ship-towing companies. Concern about market failure is reflected in the fact that regulation or government ownership itself may have been imposed because of fears of adverse effects from imperfect competition.

Tendering raises two important questions for regulators:

1. Does the regulated or government firm have the incentive to run the tender in the most efficient manner, or does regulation (or government ownership) give perverse incentives?

2. Is favourable treatment of bidders in tendering inefficient?

A related set of questions arises in the case where a regulated or government firm provides access to an asset that is an essential input for numerous firms. For example, an airport can ration access by airlines to a new airport terminal, an airport can ration access to landing slots, or a gas pipeline owner can ration capacity to transmission or distribution companies. Questions that arise in such cases include:

1. Will the firm provide access at efficient prices, that is, what is the best way to ration access?

2. What will be the effect of rationing access on the structure of the market?

3. What is the interaction between access to the asset and merger policy?

It turns out that the theory of auctions is the primary area of economics that covers both the case of tendering and rationing of access. This will be explained in section 4 of this paper. Below, we argue that tendering introduces only limited concerns for regulators when the objectives of the firm are correct. However, if firm objectives diverge from what is appropriate which depends on the nature of the regulation that is imposed on the firm--then there is an argument for some part of government to be involved in designing rules for tendering.

3. The Economics of Tendering and Access Pricing

We identified two emerging areas of concern for regulators in the introduction. In this section, we discuss them in more detail, and present insights that economic theory and practice bring to bear.

With tendering, there is a single buyer who wants to find the best of many potential sellers for some good or service. With access, there is a single seller, who would like to sell access to a subset of many potential buyers. Although they can both be considered as auctions (and we will explain why below in section 4), these situations can be quite different in their implications for regulators. We examine them in turn below.

3.1 Tendering

Consider the first question flagged in the introduction: Does the regulated or government firm have the incentive to run the tender in the most efficient manner, or does regulation (or government ownership) give perverse incentives?

For purposes of comparison, suppose first that firms are unregulated profitmaximizers. With tendering, a profit-maximizing unregulated firm that wants to buy a good or service of known quality would ideally wish to have full information on each of the seller's costs of production. It would then contract with the seller who has the lowest cost of production. This follows directly, since by selecting the firm with the lowest production cost, the buyer's costs of production are minimized. Suppose instead that the buyer does not have full information on sellers' costs of production. The buyer's objective is still the same: to somehow select a lowest-cost seller. However, there is an additional informational constraint on the buyer's problem. It turns out that the buyer can devise a method for minimizing the expected cost by offering a menu of cost and quantity combinations that leads sellers to reveal their costs by their selection of a particular contract. (5) This is more complicated than the full information case, but the lesson is the same: as long as the buyer's objective is to maximize profits--and therefore to minimize cost--it will end up with the lowest-cost seller (at least in expected terms subject to informational constraints). For a practical illustration of this idea, consider the following example. Suppose that an airport is an unregulated profit maximizer, and wishes to find a company to supply freight-handling services. It will clearly choose the company that provides the services at the lowest cost. Now suppose instead that the airport doesn't know the costs of its potential service providers. The airport still has the objective of minimizing cost--expected cost--and does so subject the informational constraints it faces.

If a regulated firm faces the same objectives as our hypothetical unregulated firm, there would be no cause for concern to regulators provided some key assumptions are met. This is because by selecting the lowest (expected) cost supplier of services subject to informational constraints, the firm is able to charge a lower price to customers. In other words, the firm's actions in choosing a supplier are completely compatible with the objective of keeping customer prices low. (6)

The crucial aspect of the argument above was that the firm's objectives were correctly aligned with the goal of keeping customer prices low. In other words, it is crucial that the firm wants to minimize the cost of the service or good for which it tenders. Now suppose that the firm is regulated. In this case, of course, the form of regulation determines the firm's objectives. Suppose the firm is subject to rate of return regulation, that is,

allowed revenue = (operating costs) + depreciation + (allowed rate of return) x (rate base).

The cost of inputs provided by an outside supplier enters this equation as part of operating costs. A firm that faces this objective clearly does not have an incentive to exert effort to keep its operating costs low, since costs are returned to the firm regardless of their magnitude.

What can the regulator do about this situation? The traditional response has been to audit the firm on operating costs to check that they are sufficiently low before being allowed as part of revenue. Another method would be to prescribe the method by which suppliers are selected. This approach will work quite well if there is a simple objective measure of the service that is being provided, and there are no significant informational constraints. For example, if the contract is to supply a widget--and the quality of a widget can be easily and objectively measured--then all the regulator needs specify is that the supplier of the lowest-priced widget be selected. Of course, if widget quality can be objectively measured, and a record of the bids of potential widget suppliers is kept, then an audit will work well too. (In fact, the advantage of an audit is that it can be random and occur after the fact. It may therefore save regulatory enforcement costs).

If the quality of the good or service being supplied is difficult to measure, or if there are informational constraints, then the basic story is the same. Either some kind of audit, or prescription of a method for selecting a supplier, will address the problem. The only difference with the case above (where quality could be easily measured) is that the selection method will be more complicated. Below, in our discussion of the theory of auctions, we explain the issues of tender design when firm objectives are not correct (as in the situation just described), and there is a cause for intervention by government.

We argued above that rate-of-return regulation does not give the regulated firm the correct incentives to choose an appropriate supplier. In general, as the type of regulation changes, so do the incentives of the regulated firm to select an appropriate supplier. Consider price-cap regulation. In principle, price caps are a form of what are called 'fixed price' contracts. Simply speaking, the firm receives a price p for each unit of output it provides that is fixed and hence independent of current firm decisions. The firm must therefore pay costs associated with production. (7) Firms that are subject to such regulation clearly have the same objective as the private firm we examined above. Therefore, firms that are subject to true price-caps will have an objective that is compatible with consumer welfare.

There has been some recent debate as to whether price-cap regulation behaves as a fixed-price contract in practice. Regulators have an incentive to tighten regulation if the firm appears to be making too much profit, and the regulated firm has an incentive to make excessive claims for cost pass-through. To the extent that price-cap regulation does not give the correct incentives in practice, it may be necessary for the regulator to have a role in prescribing the form of selection of suppliers.

We repeat the general principle that can be drawn from this discussion: if the regulated firm faces the full cost savings associated with choosing a supplier, then there is no need for regulators to be concerned. However, if the regulated firm's incentives are different--and the degree of difference will depend on the type of regulation--there is an argument for the regulator to be involved in the supplier selection process, or in auditing the firm for the prudency of the costs it incurs. Exactly how the regulator should intervene will depend on the type of regulation and more generally on informational and other constraints. This discussion in answer to the first question is summarised as: If a regulated firm has the same incentives to minimize cost as an unregulated firm, no government intervention in the tendering process is required. If, however, a regulated firm does not face the same incentives to minimize costs (e.g. under rate-of-return regulation) there may be a case for intervention.

The second question of interest flagged in the introduction was: Is apparently favourable treatment of bidders in tendering inefficient?

It turns out that even if a firm faces the correct objectives, it may be optimal for it to give some preference margin to incumbents (or others). Thus, the observation of apparently favourable treatment is not necessarily a cause for concern by the regulator provided the firms' objectives are to minimise cost. For purposes of illustration only, consider the example of architectural firms that bid to design buildings in the private sector. Suppose that at stage 1 of the project, the firm Architecture Inc. was chosen as the best supplier. Stage 2 of the project--for example, another building that complements the first--could well require a firm that has a working knowledge of the stage 1 building, good relations with the building company and union, a knowledge of the constraints of the landscape etc. That is, Architecture Inc. has made certain investments at stage 1 that are specific to the stage 2 project. If Architecture Inc. is not awarded stages 1 and 2 of the project at the outset, it has an incentive to choose levels of these project-specific investments that are too low. Alternatively, it might be desirable to award a firm stage 1 and introduce an up-front preference margin for stage 2 at the beginning of the entire scheme.

With regulated firms the argument is similar. For example, a supplier of towage services to a port authority may tailor its capital to the particular port. Similarly, a supplier of freight-handling services to an airport may make investments that are specific to that particular airport. An example would be investments in streamlining business dealings with airlines and airport employees, machines that are specific to the airport (e.g. that are located at the airport (8)), or home locations decisions near the airport by trained employees and managers. The particular type of such investments is not important (and we are not suggesting that our simple illustration represents reality). What is important is the extent of specific investments that must be made. In general, the greater the importance of specific investment, the longer the term of the contract that should be awarded. In cases of more extreme investments, it might be optimal for the airport itself to own the service supply company. (9)

In the context of a buyer bidding for supply, this argument supports the notion of giving an incumbent a bidding margin when tendering for supply, instead of long-term contracts or ownership. For example, the incumbent could be given a 10% margin over its competitors in the bid. In this way, if the incumbent becomes slack, it is still open to competition. However, it also has an incentive to make specific investments that would not be made in the absence of such a margin. It is important to note, however, that this system will only work if the supplier believes that the buyer will commit to awarding a bidding advantage. The buyer clearly has an incentive to undermine this contract after the seller has made the appropriate investments. Such undermining would be inefficient, since a firm would anticipate the problem, resulting in the same incentive to under-invest as when there was no preference margin. (10)

The practical content of this argument depends on the value of specific investments that suppliers must make, and the sensitivity of their incentive to supply such investments. This will vary across different cases. If there are very valuable and sensitive specific investments in a particular instance, this argues for longer-term contracts or margins of preference in shorter-term contracts. Finally, it is worth pointing out that a preference margin for an incumbent might appear to be unfair or favourable treatment. As mentioned, the reason for preference is to compensate the incumbent for prior specific investments, and indeed encourage that firm to make the investments in the first place. However, if preference margins are given, authorities need to be very careful to avoid the appearance of corruption. The above discussion is summarized in the following remark: Favourable treatment to incumbents in a tender process may not be inefficient if it provides the incumbent incentives to make important specific investments.

When an incumbent or a particular party are given favourable treatment, there is the potential for corruption between that party and the auctioneer; for example, in a tender the auctioneer could inform the incumbent of the lowest competing bid so that the favoured party can make the highest bid that will be accepted. The gain can then be split between the party and the auctioneer. Several points arise in the face of potential corruption. First, favourable treatment should only be used when it is required to ensure important specific investments are made. Second, if the initial bidding process is competitive and corruption free, the initial bidding price will incorporate a margin for the anticipated subsequent gain that results from any preference margin. For example, return to the architecture project mentioned above. The project has two stages, each with a tendering process. In the first stage the project goes to the lowest bidder, however in the second stage the incumbent will receive a preference margin. Anticipating the potential windfall in the second period, parties will lower their first-period bids. Third, some auction or tender designs deal better with collusion between an auctioneer and a bidder; this issue is discussed in section 5.2.

This discussion raises the question as to how important is the competitiveness of the tendering system, in particular given that we argued that the competitiveness of the bidding process, or competitiveness ex-post (in the form of long-term contracts or ownership) would optimally be restricted to the extent that specific investments are present. We also argued earlier that it is not the competitiveness of the tender per-se, but rather the objective of the buyer that matters. Even if there are very few parties bidding for the supply of a good or service, or if these parties are corrupting the system by colluding, a buyer with a cost-minimizing objective will be doing its best to keep costs low, and this will therefore lead to the best outcomes for consumers (subject to the constraints of few sellers). There is no need for additional regulation of the buyer, although there may be some need for government to impose penalties if collusion can be detected. On the other hand, if there are few substitutes for the inputs to be supplied (e.g. there is some kind of bottleneck service), and there are very few or only one bidding companies, then a separate argument for regulation of the industry that supplies the input can be made.

3.2 Bidding for Access

With access issues, there is a single supplier and many potential buyers (as opposed to a single buyer and many potential suppliers in the ease of tendering). If the supplier's costs are known, then it is a simple matter to regulate access: just set the access price equal to the marginal cost plus a premium to cover a reasonable return on capital. (11) However, when the seller's costs are not known, and buyers' valuations of access are not known--as might typically be the case--then such simple regulation is not possible. In this case, the economic theory and practice of auctions provides useful insights.

A typical auction that might be familiar is the auction of art by Christie's. The owner of the painting wishes to get the highest price. Potential buyers want to pay the lowest price possible. In this case, there can only be one winner, since there is only one painting. Consider the case of a franchise for the distribution of gas. The pipeline owner is the supplier. The good itself is an amount of capacity of the pipeline. The buyers are gas distribution companies that wish to have capacity so they can deliver gas to their franchise area. An auction could be used whereby distribution companies bid for gas capacity much like art lovers bid for the painting.

Consider the first question on access from the introduction: Will the firm provide access at efficient prices, that is, what is the best way to ration access?

There is an analogous argument to the one we used for tendering in this case, though unfortunately much more fanciful! Consider a firm that is perfectly regulated. It charges price equal to marginal cost, and its capacity is chosen optimally. Suppose such a regulated firm auctions access to its capacity to competing firms, and its reserve price is set equal to its valuation of the capacity. Firms that are unwilling to pay the price for bundles of access will not be able to enter the market. However, firms that do pay the auction-determined price will enter the market, and will have to charge the same price as our unrealistically-well-regulated monopoly. As long as the auction allocates capacity efficiently, there is no cause for concern by the regulator of the practice of auctioning capacity. The regulated firm's objective will be to obtain the highest price for its capacity, and no less than its reserve price (which will reflect the normal return that the ideally regulated entity can earn).

The reader is (we hope) sceptical of the relevance of this example. Regulation does not operate perfectly. Regulated monopolies have objectives that diverge from social objectives. The tendency is for an owner of capacity to charge too high a price, and restrict entry of competitors. To achieve this with an auction, the regulated firm could set a high reserve price. In this way, such a firm can increase its own prices, because competing firms have either been deterred from entry, or have higher costs. Alternatively, the regulated firm might find it desirable to restrict the capacity it sells in the auction in order to achieve the same goal. An argument can be made, however, that if the regulatory environment is optimal, including the ex-post market structure--subject to informational constraints of the regulator, and both the reserve price and capacity have not been excessively manipulated--then the regulated firm's use of an auction to ration capacity will be desirable. This is because (like the fanciful example of the perfectly regulated firm) the optimally regulated firm that aims to get the highest price for its capacity subject to its regulation-determined opportunity cost (equal to the reserve price) will be rationing (known and optimal) capacity to highest valued users, who then compete ex-post with the regulated firm.

To the extent that regulation works well, then, auctions will work well to ration capacity, provided reserve prices and the capacity for sale are chosen appropriately. If regulation does not work as well as we have described, then as with tendering (often referred to as 'procurement' in the economics literature), there is an additional role for government to intervene in its design (on top of monitoring reserve prices and capacity for sale). An understanding of the basic functioning of an auction is crucial in these circumstances. This is summarized as: Provided regulation provides efficient incentives, auctions are an appropriate mechanism to ration capacity.

These examples are situations where the government is directly selling assets to private firms. However, the lessons for the design of auctions cross over to the auctioning of access to capacity put into place by regulated firms. Regulators must be aware of the fact that auctions often need to be designed very carefully, and with attention to the details of the specific market involved. (12) Indeed, the experience with auctions has led one prominent expert in the field to declare strongly that the devil is in the details. (13) For this reason, and to address the questions asked in the introduction, our discussion presents an overview of the theory of auctions. Such a basic understanding is essential for an awareness of potential pitfalls.

4. An Introduction to the Economics of Auctions

Here we explain the main insights gained from the theory of auctions. This summary is brief and pertains specifically to the application of the lessons of auction theory to regulated firms. We make the assumption that the goal of the buyer (e.g. a government or the regulated firm involved in procurement) is to pay the lowest possible purchase price (subject, clearly, to certain specifications, which can include technical requirements or quality standards). Recall that this objective is desirable for consumers of the product sold by the regulated firm, since this ensures lower prices. If the regulated firm does not have cost minimization as its objective, then our assumption is that the regulator wishes to design the tender to receive the lowest price.

Equivalently, we assume that when the regulated firm is a seller, its goal is to maximize revenue from the sale of assets, or of access to assets. (Recall that this is a reasonable objective if regulation, reserve prices and capacity are optimal.) In the case where regulation distorts the firm's incentives, the regulator or some other government body will need to be involved in the design of the auction. In this case, we assume that the goal is to maximize revenue, subject to setting an appropriate reserve price for the regulated firm's capacity, and that there is an appropriate amount of capacity being auctioned. Setting reserve price and capacity appropriately ensures that the price of access will not be too high. With this constraint, maximizing revenue ensures that the regulated firm gets a reasonable return on its capital.

In this part of the paper we restrict ourselves to auctions in which the regulated firm is the seller and there are several potential buyers. It is important to point out that all the results also apply, with appropriate adaptations, to the case where the regulated firm is the purchaser and there are various potential sellers. In other words, the theory of auctions is general enough to encompass the case of tendering and procurement that we discussed above. (14) This is explained in detail in section 4.1.

Below, we will discuss the results on the expected revenue that auctions generate. In the case of tendering, these results translate into the expected cost that the regulated firm pays, and thus directly into the prices that consumers of the regulated product will pay. The goal of the regulator in keeping expected cost down translates one for one to the equivalent goal with access of maximising expected revenue (subject to reserve price constraints etc).

Note that auctions are universally used in: privatisation processes at the local, state and federal levels; in the assignment of subscription television licenses and cellular telephone concessions; (15) and in the sale of dollars or government bonds by the central banks. (16) We will sometimes explain auctions as if the central character is a privatising government, rather than a regulated firm. The principles are the same.

4.1 What is an Auction?

An auction is simply a method to allocate scarce resources. Other methods include sale at a fixed price--just like at a supermarket or store, with no bargaining between the buyer and seller--and direct negotiation--such as the sale of a house or car. Unlike a fixed-price sale, in an auction the price is established by the interaction among the potential buyers.

To sum up, an auction is a mechanism to establish a price at which supply is equal to demand in a situation in which the seller does not know the value the various buyers attribute to the object to be sold. In an auction, the seller and each potential buyer know their own values or estimates for the object, but do not necessarily know those of the other participants. The seller wishes to obtain the highest price possible and the buyers want to win the auction paying the lowest price.

As an illustration, consider figure 1 below:

There are three bidders, and one seller whom we call the auctioneer. The seller is in possession of an object for which the potential buyers bid. An auction mechanism is a process of allocating the object. Thus, the 'seller' could be an airport that wants to choose a company to supply freight-handling services. The 'object' is then the right to provide airport-landing services. The price of the object is negative, in the sense that bidders are paid for their provision if they win the auction. All of the technical analysis that apply below to traditional auctions are therefore also directly applicable to tendering.

4.2 When Should a Seller Resort to an Auction?

An auction should be used to sell an asset for which there is no established market, that is, when the market is thin. For example, auctions can be used to sell unique objects, such as a painting or shareholding control of a company owned exclusively by the government, or in the sale of various objects, such as public bonds, subscription TV licenses, access to port facilities, airport terminals, or access to gas pipelines by distribution and transmission companies. In emerging economies, they are necessary due to the absence of prices that can serve as an appraisal basis for a company to be privatised. We discuss privatisation in some of the examples below: note that sale of a whole asset is similar to sale of access to an asset, such as airport terminals, telecommunications capacity etc. The difference is that there is one unit in the case of privatisation, and many units for sale in the case of access.

An auction is more flexible than a fixed-price sale--where there is no shortterm margin for variation in the price and where each customer is charged the same price--and quicker than direct negotiation--where the process of offer and counteroffer can be quite drawn out. In terms of tenders, for example, as the government is generally a large buyer, it would pay more if buying at a fixed price with no quantity discount. The alternative of bargaining directly with the supplier, however, leaves room for corruption. Since an auctioneer must follow determined rules, the ability for a corrupt government employee to favour a determined company is reduced.

Auctions protect uninformed sellers. Instead of having to predict the market price, a government employee who has insufficient information about the potential value of the asset to be sold can simply let the competitors determine the final price.

Besides this, the seller--or a regulator in the case of asset access--chooses or can influence the auction rules. It can create an auction to sell something unique that has never been sold before. For instance, we have already mentioned the use of auctions for cellular telephone and PCS frequencies in the United States, United Kingdom and other countries.

Finally, auctions are generally efficient from an economic standpoint: the asset auctioned tends to go to the bidder that attributes the greatest value to it. In this fashion, a well designed bidding process can ensure that the most competitive player emerges victorious.

4.3 What Types of Auction are There?

There are various classifications of auctions. For example, we can distinguish between open outcry auctions where the participants make their bids in public and sealed-bid auctions--where bids are submitted in sealed envelopes. We can also distinguish between ascending- and descending-price auctions. Finally, we can differentiate auctions by the number of objects involved--a single item or many. Regardless of this diversity, there are four basic types of auctions for a single object: open ascending price, or English; open descending price, or Dutch; sealed highest-bid; and sealed second-highest bid, or Vickrey auction.

The English auction is open, with rising bids. The auctioneer announces the minimum bid (or reserve price) and the participants, through gestures or vocally, express their bids. The winning buyer is the one whose bid is not topped by any other. Despite its apparent simplicity, it is really quite complex, for there is a continuous flow of information during the course of bidding.

The Dutch auction is open but with descending bids. The auctioneer announces a high starting price, which is reduced successively by an automatic indicator. The winning buyer is the one who yells 'stop' or who makes the indicator stop (in the automatic version), and the price paid is that marked on the indicator. This format is used in the sale of flowers in Holland, fish in England, Israel and Australia, and foreign currency in some African countries.

In a first-price sealed-bid auction, the participants submit sealed bids and the winner is the one presenting the best offer (either the highest for purchasing or the lowest for supplying the good or service). This type of auction is generally used for the purchase of goods or services by the government--a public tender is nothing more than a descending-bid auction where the winner is the one offering the good or service at the lowest cost.

In a second-price sealed-bid auction--also known as a Vickrey auction--the winner is the participant with the highest bid, but she pays the equivalent of the second-highest (or highest unsuccessful) bid. The contribution of Vickrey was to propose an auction in which the bid of each participant only affects the probability of winning but not the profit in the event of winning. This type of auction is in truth not very common, being used, for example, in the sale of rare stamps and more recently in some business-to-business auctions conducted on the Internet.

Other auction formats include all-pay, where all participants pay their bids and the winner is the highest bidder, and hybrid auctions, which involve combinations of different formats, such as a rising-bid auction where the top two bidders are invited to take part in a highest-price sealed-bid 'runoff.' A good example of a hybrid auction was that employed in the sale of the companies resulting from the break-up and privatisation of Telebras--the Brasillian telecommunications monopoly.

Besides this, when various objects are sold simultaneously or sequentially, the auctions used are modelled on the basic formats described above. Government bonds are commonly sold through discriminatory price auctions, an extension of the first-price sealed-bid auction for the case of multiple identical objects. In this format, participants submit sealed bids consisting of price and quantity pairs. The auctioneer ranks the bids from highest to lowest price and allocates the bonds desired at the various prices to each bidder until all objects are sold.

4.4 Basic Auction Models

The main element in modelling auctions is the presence of asymmetrical information. If there were complete information, that is, the seller knew exactly the values attributed by each potential buyer of the object being sold, it would be trivial to design an auction to extract the highest price. The seller would only have to approach the buyer attaching the greatest value to the object in question and make a 'take it or leave it' offer at any price between the first- and second-highest values. Consider figure 2 below:

[FIGURE 2 OMITTED]

Again, there are three bidders and one seller. Each participant (including the seller) has a personal valuation of the object to be sold that is private information represented by [s.sub.i]. Thus, the bidders do not know each other's valuations, and do not know the seller's valuation. Similarly, the seller does not know the bidders' valuations. Private information means that the bidders have an incentive to bid strategically. For example, suppose buyers make open bids. Once a party bids, the other parties receive information about the value of the object. It might be that bidders' valuations are independent. In this case, one party's bid may reveal to the other parties some information on their valuation. Such information will be useful for them in constructing their own bids. However, each bidder is aware of this process and may have an incentive to distort their bid in order to gain strategic advantage. It is the strategic processes that underlie bidding with asymmetric information that can make auctions difficult to design.

There are three basic auction models. In the private-value model, each participant knows its valuation of the object (or objects) up for bid, but this value is only known to that bidder. As an illustration, consider an auction of a racehorse, and assume that the potential buyers are stable owners who want the horse strictly for their own pleasure and thus are not interested in a future sale (or stud fees). The value of the horse for each bidder is simply a function of each one's individual taste, that is, the valuation bidder A gives to the horse is independent of the other participants' valuations. This does not mean that the bids are independent: each individual must predict how the others will behave and respond accordingly. Note that the winner's curse does not apply here since there is no uncertainty as to the true value of the object--which in this case is different for each participant.

In the pure common-value model, the true worth of the object is the same for all players, but they each have distinct information regarding this value. For example, consider an auction for the right to explore a mineral deposit. The bidders will only know the size and quality of the mineral deposit after exploration. Each one has access to different geological information (signals) about the quantity and quality of the deposit, and based on these signals formulates an estimate. In this case a certain buyer would change its estimate if it discovered the signal of one of its competitors. In the private-value case, if one bidder discovered the preferences or information of another, it would not change its estimate (its behaviour in the auction, however, would change). Regarding the terminology employed, we refer to any model where there is a common component among the individual values as being a common-value model. In regulatory contexts, the object being sold are access to infrastructure. Since such access generates returns to the company through sales to final customers, auctions of access to infrastructure are part of the common-value model.

Finally, in a general model that takes in both previous models, each participant receives a private signal (information), although the worth this bidder attributes to the object is a general function of all signals. In the case of the sale of infrastructure access, these signals could include the state of demand for the product, the cost of labour or materials of the participants in the industry, or expected changes in the regulatory regime. (17)

4.5 Bidding Strategies

We now analyse the bidding strategy in each auction format. We start with the descending-bid or Dutch auction. Although the description of this auction suggests treating it as a dynamic 'game', the problem each bidder must resolve in choosing his optimal bid is essentially static in nature. Each bidder must choose a price at which he will stop the indicator, as long as no other bidder already has done so. The bidder choosing the highest price wins and pays that price. In this fashion, the descending auction is the strategic equivalent (18) of a highest-price sealed-bid auction, and hence participants will formulate their bids in similar form in these two auctions.

With private values, staying in a rising-bid openoutcry (or open English) auction until the price reaches the individual's private value is dearly a dominant strategy, that is, a bidder will remain until the price reaches a point where he no longer cares if he wins the object or not. The second-highest bidder will drop out when his value is reached, and the bidder attaching the greatest worth to the object will win and pay the equivalent of the second-highest value. (19)

We now consider the strategies for formulating bids in a sealed second-highest bid, or Vickrey-auction. Vickrey's argument is that, in this type of auction, a dominant strategy is for a bidder to offer an amount equal to the value he attaches to the object. As the winner makes a payment equal to the second-highest bid, there is no advantage to a bidder to offer a bid that is not equal to their value for the item; in fact, any other bid can only make the bidder worse off. For example, if they bid less than their true value, they decrease their chance of winning, while bidding more than their value creates the possibility of paying more than their value for the item. (20)

Since making a bid equal to the value is a dominant strategy for the participants in a Vickrey auction, the winner will be that individual with the highest valuation, who will pay the equivalent of the second-highest valuation--in the same fashion as in an ascending-price auction. For this reason, we sometimes refer to a rising-price auction as a second-price auction. (21)

An important element in choosing a strategy for taking part in an auction where the valuations have a component in common, is the presence of the winner's curse: each bidder must recognize that he will be awarded the object only when receiving the highest signal--the most optimistic information on the value of the object--when the bidders behave symmetrically, that is when individuals with similar information submit similar bids. The consequence of ignoring the fact that winning means having the most optimistic information on the value of the object is that the winner may pay more for the object than its true worth. This means that in equilibrium, the participants should adjust their bids downward. This phenomenon has been widely documented in many auctions.

Since the equivalence between descending and first-price sealed-bid auctions for a single object is general, and since ascending and second-price sealed-bid auctions are also equivalent under some conditions (and have similar properties under very general conditions), from this point on we will only distinguish between first- and second-price auctions. This information is useful, since it enables us to simplify our analysis of auctions that at the surface seem quite different.

4.6 Which is the Best Type of Auction for the Seller?

The question of the best type of auction for the seller is particularly relevant in the case of privatisation--that is, sale of government-owned assets to the private sector. (22) Note also that these auctions will also be best when a firm that is optimally regulated sells capacity to competitors, as long as the reserve price and capacity are appropriate.

Note that privatised finns are almost always subject to regulation ex-post, since one reason why these assets were in public hands in the first place was to address concerns of monopoly pricing. As long as the regulation governing the soon-to-be-private firm are efficient--that is, deliver the highest social welfare, and the reserve price is set appropriately--it is clearly optimal for the asset to be sold at the highest possible price that the government can get. However, if regulation of the future private firm was flawed, for example, if the firm was able to charge prices that are too high, then the government would certainly be able to raise more revenue, but this would be at the expense of efficiency later on. The auctioned object in this case must be defined as an optimally regulated set of assets in order for revenue maximization from sale to be desirable. We will now consider the particulars of revenue arising by auction. As noted elsewhere, this could apply to auction of access; similarly it could be tendering to provide a particular service (in which case minimising cost is the objective of the regulator).

In the early 1960s Vickrey (1961, 1962) demonstrated a special case of one of the most celebrated results of the theory of auctions, the Revenue Equivalence Theorem, later generalized by Myerson (1981), Harris and Raviv (1981), and Riley and Samuelson (1981): Suppose that the potential buyers of an object are riskneutral (23) and that their private information on the value of the object is generated independently, based on a common distribution function. Then any auction format (tendering process') generates the same revenue for the seller where: (i) the object is assigned to the individual with the highest signal; and (ii) the participant with the lowest signal receives an expected gain of zero.

Note that the above result is applicable both to the private- and common-valuation models, as long as the signals are independent. In the example of the sale of an artwork (in the private-valuation model) this restriction implies that how much a potential buyer likes the object is independent of how much the other bidders like it. In the case of an auction for the right to explore a mineral deposit, this hypothesis requires that the geo-technical data received by the bidders be independent.

In this form, with independent signals and risk-neutral players, all traditional auctions (ascending, descending, first-price sealed-bid, second-price sealed-bid) generate the same expected revenue for the seller. Besides this, other less traditional auctions, such as the all-pay described previously, generate the same expected revenue.

While it may seem that 'all auctions are the same' in terms of expected revenue, it must be noted that these results hold under very special circumstances. It is worth noting that, when particular assumptions are relaxed, not all auction formats are equivalent, in particular:

1. When bidders are risk averse, a first-price auction generates more revenue than a second price-auction.

2. When bidders signals are correlated (specifically affiliated), an ascending auction generates higher expected revenue than a second-price auction, which in turn generates higher expected revenue than a first-price auction. The optimal reserve price is also lower.

3. With (non-drastic) asymmetric bidders, a first-price auction generates more expected revenue than an ascending auction, although it is not allocatively efficient.

4. When the seller is able to obtain royalties from the winning bidder, it can gain more expected revenue.

5. Selection of Auction or Tender Format

Clearly, the design of auctions or tenders is no simple matter. Therefore, if government has a hand in the design of either auction or tenders, a great deal of care needs to be exerted. This section explores the implications for how different auction formats perform in different circumstance, in particular in the presence of collusion between bidders and between a bidder and the auctioneer and when there are multiple-good auctions. As noted in the previous section, with auction design the objective relates to obtaining the highest possible price from the bidders; equivalently, with a tender the objective is to minimise the cost of the service or product required. In this way, the predictions of auction theory relate to both the design of tendering processes as well as to auctions themselves.

5.1 Collusion Among Bidders

A central concern of auction modellers, whether for selling assets or purchasing goods and services, is the possibility of collusion among bidders. (24) Ignoring for the moment the division of spoils amongst colluders, note that in a second-price auction the optimal plan involves a very high bid by a predetermined winner, while the other bid zero. No bidder has an incentive to deviate from this strategy. In a first-price auction, on the other hand, the winner designated by the group will make a very low bid, with all the others bidding zero. In this case there is a strong incentive for a bidder to break the deal to obtain the object at a low price and keep the surplus all to herself. (25)

It is important to determine if there is indeed an efficient mechanism, compatible in terms of incentives, which colluders can use to designate the winner and divide the resulting profits of the collusion when the bidders possess private valuation information. In truth, such a mechanism consists simply of the use of an informal auction among the members of the cartel prior to the holding of the real auction. The optimal policy of the auctioneer to prevent such collusion is strategic use of the reserve price. (26)

Thus, we have, A first-price auction (tender process) is less prone to collusion among bidders than a second-price auction. If there is suspected collusion, the auctioneer should use the reserve price (maximum price) strategically.

5.1.1 Collusion Over Time Repeated auctions are more likely to lead to collusive equilibria. Moreover, this phenomenon is more likely to be present in repeated ascending and uniform price auctions than in repeated sealed-bid and discriminatory auctions. In ascending multi-object auctions, the agents can use the initial stages, when bids are still low, to signal the final allocation of the objects. (27) On reaching a consensus, the prices stop rising and remain at lower levels than would be the case with no collusion. In simultaneous first-price sealed-bid auctions, unless there is 'cooperation' from another competitor, there is no chance for signalling or retaliation.

Uniform-price sealed-bid auctions for homogenous objects are, in turn, vulnerable to implicit collusion. Since the unit price is determined by the lowest winning bid, the agents can reach tacit agreement to divide the market at favourable prices. In contrast, in discriminatory auctions, since each participant pays the value of its own bid for the quantity it obtains, a firm cannot use its bid as a veiled threat (seeking a collusive equilibrium) without risking having to pay.

Although some collusive tactics are illegal or could be made so, it is inherently easier to deal with this type of problem through the auction design rather than trying to apply complex rules that often wind up restricting the flexibility of agents without necessarily producing efficient results.

Another important focus of competition policy is to ensure that entry is not overly difficult. An auction with few bidders will tend to reduce the expected revenue of the seller, as well as being potentially inefficient.

In ascending auctions, the presumption that the bidder with the highest valuation will win can discourage the participation of relatively weaker competitors. This phenomenon can be aggravated by the costs of formulating bids. In both the independent-private-value and common-value models, potential bidders (who could even win) often wind up being no shows. In the common-value model this problem is exacerbated due to fear of the winner's curse. When an object has similar value to all bidders, each participant must be careful in bidding, realizing that her chance of winning increases along with her rising estimate of the object's true worth. Going up against a competitor who is assumed to have some advantage tends to heighten this realization, leading to a more cautious stance (since beating a strong opponent would suggest the possibility of having overbid). Aware of this, the bidder with the advantage can reduce her bid, meaning that her signal discourages others. In summary, the 'weak' agents become overly cautious, when participating at all, resulting in the strong agent winning most of the time, paying less than otherwise would be the case and thus reducing the seller's revenue.

Therefore, in ascending auctions, small advantages in valuation or even in reputation by a strong participant can dramatically alter the outcome. This creates incentives to invest in creating a reputation that can discourage the entrance and participation of rivals.

In sealed-bid auctions, since each participant makes only one 'best and final offer', the result is more uncertain. The firm with an advantage has no way to change its strategy (start with lowball bids that are increased only if necessary). Since it intends to profit, its bid will not be as aggressive as it would be in an ascending auction. This opens the door for the participation of comparatively 'weaker' competitors, who encounter a relatively reduced winner's curse.

In synthesis, a sealed-bid auction can serve not only to attract more firms, but also can lead to more satisfactory results for a given number of firms. But there is a tradeoff in efficiency--a sealed-bid auction is less likely than a second-price auction to assign the object to the agent with the highest valuation.

While ascending auctions, as discussed, can have problems with collusion and predatory activity, they tend to allocate the objects to the agents with the highest valuations, the definition of efficiency for an auction. They also permit the bidders to learn about the valuations of their opponents during the auction, which makes the agents more comfortable in their own valuations. This can imply increased revenue for the seller, absent collusion and predation.

This quandary of tradeoffs has led to the proposal of new auction formats. Klemperer (2002) argues in favour of a hybrid he calls the Anglo-Dutch format. With this mechanism, the auctioneer starts with an ascending auction until there are only two bidders willing to pay the current price. These two agents are then called to make a final best and last sealed bid, of not less than the current price, and the winner pays the value of its bid. The author's argument is that an Anglo-Dutch auction performs better in terms of expected revenue and can resolve some problems of incentives. The probability of selling to the agent with the highest valuation is greater relative to a pure sealed-bid auction since the Anglo-Dutch model reduces the number of participants entitled to bid in the first-price auction. Furthermore, each of the two finalists has had a chance to learn about the valuations of its opponent and also about the value of the object by observing the bidding in the first phase.

The Anglo-Dutch model eliminates the final stage of an ascending auction, in which there is only one excess bidder. This is the stage at which an ascending auction is particularly vulnerable to collusion and predatory activities. Sealed-bid auctions, in turn, induce some uncertainty about which of the two finalists will win, which attracts entrants who know they will have a chance to attain the final stage.

Another hybrid mechanism with interesting properties was used in Brazil to privatise the companies created from breaking up the national telephone monopoly Telebras. (28) This auction, which we will simply designate the Hybrid, also combines a sealed-bid and ascending auction, but in the opposite order. Each participant submits a sealed bid. If the highest bid is greater than the second-place one by a previously determined margin or percentage, that bid wins outright. If not, all participants whose bids are within that range of the highest bid take part in a final ascending auction, with the best previous sealed bid being the reserve price.

Dutra and Menezes (2002) developed a model that captures some of the characteristics of this Hybrid auction. They modelled a situation in which three risk-neutral agents compete in a two-stage auction: a first-price auction followed by a Vickrey auction when there are bids sufficiently near the highest first-price auction bid. In a situation where the potential buyers have just as much private as common valuation information on the object in question, the authors showed that this hybrid model could generate more revenue than any standard auction. The reason is that this auction can be seen as a Vickrey auction with an endogenously determined reserve price. Additionally, unlike the optimal auction, this hybrid auction is ex-post efficient.

The first-price auction, which serves as the first stage of this hybrid mechanism, attracts bidders. Based on the chance of winning outright in the first stage (since the Vickrey portion is only contingent), weaker participants may bid more aggressively. Faced with this knowledge, a strong competitor is also led to raise its bid. However, the desire to profit (which will cause a strong competitor not to bid as high as would be the case in a pure second-price auction), allied with the belief it has a good chance of winning the second-price auction, soften this tendency (a very aggressive stance would wind up reducing the incentive to enter).

Summarizing, because it employs a second-price auction, this hybrid retains the desirable property of allocative efficiency (laboratory experiments attest to this property, as shown by Dutra & Menezes 2000). In turn, the first-price auction in the first stage has the desirable properties of discouraging collusion and predatory behaviour. In this fashion, the final price of the first stage can easily be higher than

would result from a pure first-price auction, even in the event the Vickrey auction does not occur.

Extending this format to multiple-object auctions, one can well argue that the sealed phase would hinder tacit collusion. In summary, we argue that hybrid auction mechanisms, such as the Anglo-Dutch and the Hybrid used in the Telebras case, often combine the best of ascending and sealed-bid auctions.

5.2 Corruption Between Bidders and Auctioneer

Besides collusion among bidders, another concern is collusion between a group of bidders and the auctioneer, more properly called corruption. The auctioneer is an agent of the seller and thus the possibility of corruption between the auctioneer and bidders exists. What can we say about the expected revenue from traditional auctions in the presence of corruption?

Consider a first-price auction in which the auctioneer approaches the winner and offers him the chance to reduce his winning bid (let us say 1 cent above the second-highest bid) in exchange for a bribe. With independent private values and risk-neutral bidders, this corruption will reduce the expected revenue, although in this case there be no allocative loss because the object will still go to the highest bidder. Now suppose the auctioneer approaches the loser and offers him a chance to raise his bid (say 1 cent above that of the official winner). In this case there is loss of revenue and also the possibility of allocative inefficiency, because the object can be awarded to an individual who does not give it the greatest valuation.

Ascending auctions are less susceptible to corruption. Since the optimal strategy for each bidder in this type of auction is to tender a bid equal to his valuation, the individuals do not gain in revising this bid. (29)

Hence, we have: With a corrupt auctioneer, ascending auctions generate greater expected revenue than first-price auctions, with independent private valuations and risk-neutral bidders.

Of course, a similar conclusion can be made about a tender process: under the same conditions, a descending-price tendering process generates a lower expected price than a first-price tendering system.

5.3 Auctions of Various Objects

The previous discussion focused on auctions of a single object. In the case of selling more than one good, simultaneously or sequentially, the auction literature is still in flux. An exception is the case in which each bidder wants only one of several identical objects. In this context, the Revenue Equivalence Theorem continues to hold with independent private valuations. (30)

Another special case already resolved in the literature is the extension of optimal auction theory to the case where potential buyers have negatively sloping demand curves for a certain number of identical objects. In this case the insights of optimal auction theory as applied to a single object also apply to several objects, with appropriate adaptations. (31)

The literature on single-object auctions has established that, under general conditions, an ascending auction is efficient in the case of unidimensional bidder signals, even in the presence of asymmetries and common components in the individual valuations. (32) However, as mentioned, this is an emerging area. Without going into technical detail, a summary of findings is as follows:

The Revenue Equivalence Theorem and the results regarding the optimal auction can be extended to the analysis of auctions of various identical objects with independent private values, and symmetrical and risk-neutral bidders, as long as the demands are unitary. More generally, there is a consensus in the literature that the conventional formats of multiple-object auctions cause allocative inefficiency. An ascending auction format will be efficient, although it can result in lower expected revenue for the seller.

5.4 Conclusion

An auction can be seen as one of the purest forms of the market in operation, in which supply matches demand through a price determined by interaction between buyers and the seller. The theory of auctions--one of the most successful applications of mathematical economics and game theory--can assist in the choice or creation of an auction appropriate for the sale of goods for which there are no fixed markets and for government purchases of products and services, seeking to minimize its costs. Auctions protect uninformed sellers and inexperienced negotiators, as long as they understand the rules of the game. Besides this, auctions limit the opportunity for favouritism in government procurement.

The principal lesson from auction theory for government purchases of goods and services is that the format can affect the expected revenue and allocative efficiency. Furthermore, we have shown that there is no general rule that can be used to obtain the highest revenue (or lowest cost). Each case must be analysed to determine which format is best, based on the desired objectives (maximizing revenue, minimizing cost, allocative efficiency, minimizing collusion among bidders or the possibility of corruption between bidders and the auctioneer, etc). Nevertheless, some general lessons can be discerned.

When the bidders have independent information regarding the value of the object, as would be natural, for example, in the purchase of specialized consulting services, and the bidders are risk-neutral with symmetry among them, the auction format does not matter--any auction mechanism guarantees the same expected revenue. However, in the presence of asymmetries among the interested firms, a first-price auction minimizes the expected cost to the government. With symmetric but risk-averse firms, a first-price auction also is the most appropriate to minimize cost, but can result in allocative inefficiency. The chance for collusion among bidders suggests the choice of a first-price auction as well, which should be used in this case with the establishment of a reserve price.

In cases of independent information, we can again indicate the first-price auction as the most adequate. The only exception regards the possibility of corruption between bidders and the auctioneer. To minimize this chance, an ascending or second-price auction is most indicated.

We have argued that with affiliated values, risk-neutral and symmetric bidders, as is natural in the purchase of common goods and services, we have seen that the ascending auction, among all conventional formats, generates the highest expected revenue for the seller (or minimizes the expected cost to the government). With risk-aversion, the ranking of auction formats in terms of expected revenue is ambiguous. Nevertheless, an ascending auction elicits more information revelation than a first-price auction.

When there is a possibility for the government to observe, even if imperfectly, some information on the value of the object for the winner such as in the auction of oil exploration rights, the expected revenue can be augmented through a system of royalties.

Finally, when various objects are being sold, such as government bonds, the theory of auctions as yet offers no definitive answer as to the best format. It has been shown, however, that conventional formats, such as the discriminatory auction and uniform price auction, can result in allocative inefficiency. A specific recommendation suggests the use by the government of ascending formats.

The above discussion suggested the performance of various different auction formats depends on the particular environment in which they are conducted. Table 1 summarizes the discussion for single-unit auctions. In the tables a '+' indicates that the auction type performs well in the given aspect; a '-' indicates the format performs badly with respect to that desired criterion. Table 2 summarizes the performance of different auction format in auctions for multiple goods

6. Final Comments and Policy Summary

This paper has presented an informal discussion of the literature on auction theory, and applied this to the case of regulated firms tendering for inputs or selling capacity, and similarly to governments that seek to privatise assets. From the discussion, it should be clear that there is a role for the government in the design of auctions when regulated firms are involved. In the case of tendering, it must be determined that the regulated firm has cost minimization as its incentive. In some cases, favourable treatment of incumbents is desirable.

When capacity is being auctioned, the situation becomes more complicated. If it can be determined that the reserve price is not too high, and capacity has not been unduly restricted, then a well regulated firm can be left to choose an auction format that maximizes expected revenue. However, if the regulated firm does not have the correct objectives, there is a need for intervention not only with regard to capacity and reserve price, but also for the appropriate design of the auction. Auction design is no simple matter. Unfortunately, there is no way around this: any way that is chosen for allocating objects can be considered as some form of auction. It is better to understand the complexity of the environment than resorting to an arbitrary allocation. As a final point, it is worth making somewhat of a caveat. If there are problems in an industry, such as market power, it is difficult for auctions in themselves to relieve these structural issues; other regulatory techniques may be required to address these more fundamental problems of market environment.

Table 1

Single-Item Auctions With Private Values

                                First-price      Ascending--Unsealed
                               (sealed bids)             Bids

Collusion (between buyers)     + (use reserve)   -
Collusion (auctioneer-buyer)   -                 +
Collusion (over time)          -                 -
Encourage buyer entry          +                 -(private and common
                                                  values)
Efficiency                     -                 +

Table 2

Multiple-Good Auctions

             First-price   Discriminatory   Sequential   Simultaneous
               uniform     price auctions     auction      auction
               price        (homogeneous     ascending   First-price
            (homogeneous       goods)          bids      (sealed) bids
               goods)

Collusion        -               +              -               -
(between
buyers)

We thank the Australian Competition and Consumer Commission (ACCC) for providing funds for this project. The views expressed are solely those of the authors and do not reflect those of the ACCC. We thank an anonymous referee for useful comments.

(1.) The Productivity Commission (2002) found that tendering for the right to provide harbour towage services could result in more efficient pricing and outcomes, particularly if the the interests of port authorities and towage users are closely aligned (pp. xlv-xlvi), Also see Gans and King (2003) who show that direct customer contracting will not result in a perfectly contestible outcome if there is a complimentary input provider with market power.

(2.) Supply contracts and access could be allocated in an ad-hoc fashion, but this would simply be equivalent to allocating these scarce resources through particularly badly designed auctions.

(3.) Ucom paid AUD$117 million for its licence, AUD$95 million less than its initial bid; it sold the licence shortly after for a AUD$21 million profit (McMillan 1994, p. 149).

(4.) In one auction a firm won bidding NZ$100,000 but paid NZ$6, the second-highest bid; in another auction a firm bid NZ$7 million and paid NZ$5,000. In the auction for a television licence in a small city a university student bid NZ$1 and no one else bid, resulting in a price of $0. Revenue for the auctions of mobile phones was NZ$36 million, compared to the NZ$240 million estimated (McMillan 1994, p. 148).

(5.) See, for example, chapter 7 of Laffont and Tirole (1993).

(6.) There is a caveat to this argument if the regulated firm has monopsony power in the market for services. If such power is extreme, then intervention in the tender design may be warranted to protect sellers rather than customers.

(7.) In some cases, cost pass-through is allowed, however.

(8.) Alternatively, the airport may own the machinery.

(9.) See Joskow (1987).

(10.) in another context, Menezes and Monteiro (2001) suggested a reason that a domestic firm receive a preference margin could be the lower loyalty of a foreign firm, reflected in a greater chance it will not fulfull its contractual obligations.

(11.) A more sophisticated version of this would be to require the supplier to implement a peak-load pricing formula where if demand is less than capacity, then price equals operating and maintenance costs, but if demand exceeds capacity at this price, then the price is adjusted upwards to just ration demand to capacity. See Rees (1976) for a discussion of the optimality of this pricing scheme.

(12.) Indeed, auction design theory has developed quite rapidly over the last few years precisely as a response to actual problems. For example, when auctioning access, the actual outcome of the auction might influence the future interaction among participants. This interdependence has yet to be fully explored by theorists but a preliminary analysis is undertaken by Jehiel and Moldovanu (2000).

(13.) See Klemperer (2003).

(14.) In procurement or tendering, there is one buyer and many potential sellers. The buyer could, for example, call out prices starting from zero, and sellers make bids as the price rises. The first seller to accept a price wins the auction.

(15.) The recent sale in the United States of the frequency spectrum to be used for the next generation of cellular phones--the PCS (Private Communication Services)--raised more than $20 billion for the government coffers. A similar sale in the United Kingdom brought in the equivalent of over $ 25 billion.

(16.) Note, auctions have also been employed in privatizations in Eastern Europe, in the sale of public debt in over a hundred countries, and in the electric power market created by deregulation in many nations.

(17.) Algebraically, in a model with n potential buyers, participant i who receives a signal [S.sub.i] would have a value [V.sub.i] ([S.sub.l], ..., [S.sub.i], ..., [S.sub.n]) if he knew the private signals of all the other n-1 buyers. In the private-value model, [V.sub.i] ([S.sub.l], ..., [S.sub.i], ..., [S.sub.n] is a function only of [S.sub.i]. In the pure common-value model, [V.sub.i] ([S.sub.l], ..., [S.sub.i], ... [S.sub.j], ..., [S.sub.n]) = [V.sub.j]([S.sub.l], ..., [S.sub.i], ..., [S.sub.j], ..., [S.sub.n]) for all [S.sub.i], ..., [S.sub.n]. Sale of access to infrastructure is clearly a case of common values, since the information of all participants--market conditions, regulatory uncertainty etc. would affect an individual buyer's future profitability, and hence that buyer's valuation of the access.

(18.) This is to say that the set of strategies available to a bidder is the same in both auctions. Any strategy chosen generates the same expected profit in both types of auction as a function of the bids of the other players. For a formal definition of strategic equivalence, see Myerson (1991).

(19.) In reality, the winner will pay a bit more than the second-highest value since the auctioneer normally increases the bids by discrete increments. Nevertheless, for simplicity we assume that the price announced by the auctioneer increases continuously.

(20.) To explore this further, consider the problem of one of the bidders, say bidder number 1, who attributes a value v1 to the object. Assume that the other bidders are making bids at random, and also that bidder 1 offers b1 > v1, regardless of the bids of the others. If the maximum bid among the other players is lower than v1, then bidder 1 wins the object with a bid equal to b1; and thus also equal to v1. If the highest bid of the others is greater than b1, then bidder 1 loses when he would also lose with a bid equal to r1. Nevertheless, if the highest bid of the other players is higher than v1 but lower than b1, bidder 1 wins the object but pays more than his value for it, wasting money. In this fashion, a player does not gain anything by bidding higher than his own valuation, but can lose. An analogous argument is valid to discard the possibility of a player's bidding lower than his valuation. This was Vickrey's insight in proposing an auction where the winner does not pay his bid, but instead the second highest.

(21.) However, this strategic equivalence between the two auction formats is only valid with private valuations or if there are only two bidders. When the individual valuations have a common component or there are more than two bidders, the participants have the chance to update their estimates of the object's worth after observing other participants dropping out, and condition their behaviour on this information. In this case, bidding strategies are truly dynamic, as opposed to the Vickery auction where there can be no updating of estimates,

(22.) Achievement of the highest possible revenue means that the government can levy lower taxes on other activities, and hopefully in so doing improve the efficiency of the economy.

(23.) Various factors affect the revenue generated by an auction. Among the most important, we can mention the behavior of bidders in relation to risk: bidders can be risk-averse, risk-neutral or risk-loving. Suppose that an individual has to choose between the certainty of receiving $1,000 and a coin toss where he either receives $0 or $2,000. A risk-neutral individual is indifferent, a risk-averse person will prefer to receive the $1,000 and a risk-lover will take a chance on the outcome of the toss.

(24.) For an excellent rundown of the circumstances under which there can be collusion in auctions and on the methods to detect this, see Hendricks and Porter (1989).

(25.) See Robinson (1985).

(26.) See McAfee and J. McMillan (1992).

(27.) Klemperer (2003) reports an episode occurring during a spectrum auction in the United States in 1997. U.S. West was competing aggressively against McLeod for Rochester, MN (License 378). Although the majority of bids were in exact thousands of dollars, U.S. West entered bids of $ 313,378 for Waterloo, IA, and $ 62,378 for Marshaltown, IA, where McLeod was high bidder, apparently trying to punish it. Thereafter, McLeod stopped bidding on Rochester. For a detailed report of this episode, see Cramton and Schwarz (2000).

(28.) Although several regional 'baby Telebras' were auctioned off, this Hybrid can be considered as involving a single object due to prohibitions on cross-holdings.

(29.) See Menezes and Monteiro (2000).

(30.) See Menezes (1998).

(31.) For more details on the conditions for validity of this result, see Maskin and Riley (1989).

(32.) See Maskin (1992).

References

Cramton, P. & Schwarz, A. 2000, 'Collusive bidding in the FCC spectrum auctions', Journal of Regulatory Economics, vol. 17, no. 3, pp. 229-52.

Dutra, J. & Menezes, F.M. 2000, 'An experimental investigation of a hybrid auction mechanism', Mimeo, Fundacao Getulio Vargas.

Dutra, J. & Menezes, F.M. 2002, 'Hybrid auctions', Economics Letters, vol. 77, no. 3, pp. 301-97.

Gans, J.S. & King, S.P. 2003, 'Contestability, complementary inputs and contracting: The case of harbour towage', mimeo.

Harris, M. & Raviv, A. 1981, 'Allocation mechanisms and the design of auctions', Econometrica, vol. 49, no. 6, pp. 33-64.

Hendricks, K. & Porter, R. 1989, 'Collusion in auctions', Annales D'Economie e de Statistique, vol. 0, no. 15-16, pp. 217-30.

Jehiel, P. & Moldovanu, B. 2000, 'Auctions with downstream interaction among buyers', RAND Journal of Economics, vol. 31, no. 4, pp. 768-91.

Joskow, P. 1987, 'Contract duration and relationship-specific investments: the case of coal', American Economic Review, vol. 77, no. 1, pp. 168-85.

Klemperer, P. 1998, 'Auctions with almost common values', European Economic Review, vol. 42, no. 3-5, pp. 757-69.

Klemperer, P. 1999, 'Auction theory: a guide to the literature', Journal of Economic Surveys, vol. 13, no. 3, pp. 227-86.

Klemperer, P. 2002, 'What really matters in auction design', Journal of Economic Perspectives, vol. 16, no. 1, pp. 169-90.

Klemperer, P. 2003, 'Why every economist should learn some auction theory', in Advances in Economics and Econometrics: Theory and Applications, ed. M. Dewatripont, L. Hansen & S. Turnovsky, Cambridge University Press, Cambridge.

Laffont, J.J. & Tirole, J. 1993, A Theory of Incentives in Procurement and Regulation, MIT Press, Cambridge and London.

Maskin, E. 1992, 'Auctions and privatization', Privatization, ed. H. Siebert, J.C.B. Mohr Publishers, pp. 115-36.

Maskin, E. & Riley, J. 1989, 'Optimal multi-unit auctions', The Economics of Missing Markets, Information, and Games, ed. F. Hahn, Oxford University Press, Oxford.

McAfee, P. & McMillan, J. 1992, 'Bidding rings', American Economic Review, vol. 82, no. 3, pp. 579-99.

McMillan, J. 1994, 'Selling spectrum rights,' Journal of Economic Perspectives, vol. 8, no. 3, pp. 145-62.

Menezes, F.M., 1998, 'Auctions of identical objects with single-unit demands: a survey', Revista de Econometria, vol. 18, no. 2, pp. 309-40.

Menezes, F.M. & Monteiro, P.K. 2000, 'Corruption and auctions', Mimeo, Fundacao Getulio Vargas.

Menezes, F.M. & Monteiro, P.K. 2001, 'Porque favorecer firmas nacionais?', Revista Brasileira de Economia, vol. 55, no. 4, pp. 453-66.

Myerson, R. 1981, 'Optimal auction design', Mathematics of Operations Research, vol. 6, pp. 5873.

Myerson, R. 1991, Game Theory: Analysis of Conflict, Harvard University Press, Cambridge.

Productivity Commission 2002, Economic Regulation of Harbour Towage and Related Services, Inquiry Report no. 24, Canberra.

Rees, R. 1976, Public Enterprise Economics, Weidendeld and Nicolson.

Riley, J. 1988, 'Ex post information in auctions', Review of Economic Studies, vol. 55, no. 3, pp. 409-30.

Riley, J. & Samuelson, W. 1981, 'Optimal auctions', American Economic Review, vol. 71, no. 3, pp. 381-92.

Robinson, M, 1985, 'Collusion and the choice of auction', RAND Journal of Economics, vol. 16, no. 1, pp. 141-45.

Vickrey, W, 1961, 'Counterspeculation, auctions, and competitive sealed tenders', Journal of Finance, vol. 16, no. 1, pp. 8-37.

Vickrey, W. 1962, 'Auctions and bidding games', Recent Advances in Game Theory, 15-27, The Princeton University Conference, New Jersey.

Flavio M. Menezes ([dagger])

Rohan Pitchford ([double dagger])

Andrew Wait ([section])

([dagger]) School of Economics, Australian National University, ACT 0200. Email: [e-mail]

([double dagger]) Asian Pacific School of Economics and Management, Australian National University, ACT 0200. Email: [e-mail]

([section]) Economics, University of Sydney, NSW 2006. Email: [e-mail]

(Date of receipt of final transcript: June 23, 2003. Accepted by Joshua Gans, Area Editor.)