Auctions
Theory and Possible Applications to Economies in Transition
  • 1 https://isni.org/isni/0000000404811396, International Monetary Fund

A major effort is taking place in many parts of the world to establish market-oriented institutions, a development that is particularly evident in the context of the transforming economies in Eastern Europe and the republics of the former Soviet Union. Against this background, this paper assesses various auction techniques to price and allocate government securities, refinance credit, foreign exchange, and state assets in the context of privatization programs. Before making our recommendations on the appropriate format for auctioning these items, the paper explains basic auction formats and assesses the advantages and disadvantages of these formats drawing on the existing, and mostly theoretical, literature.

Abstract

A major effort is taking place in many parts of the world to establish market-oriented institutions, a development that is particularly evident in the context of the transforming economies in Eastern Europe and the republics of the former Soviet Union. Against this background, this paper assesses various auction techniques to price and allocate government securities, refinance credit, foreign exchange, and state assets in the context of privatization programs. Before making our recommendations on the appropriate format for auctioning these items, the paper explains basic auction formats and assesses the advantages and disadvantages of these formats drawing on the existing, and mostly theoretical, literature.

I. Introduction

Auctions play an important role in economics. In their most basic form, they are simply one of the ways in which various commodities and financial assets are allocated to individuals and firms, particularly in a market-oriented setting. Examples of items that are commonly auctioned include original art, livestock, fresh fish, used cars, construction contracts, and a range of financial assets such as government securities, central bank refinance credit, foreign exchange, and equity shares. These items exhibit considerable diversity, but a common denominator among them is that the valuation of each item varies enough to preclude any direct and straightforward pricing schedule.

The use of auction mechanisms to guide price determination and the allocation process has been associated with the major effort taking place in many parts of the world to establish market-oriented institutions. This development is particularly evident in the context of the transforming economies in Eastern Europe and the former republics of the Soviet Union. Examples of where auctions are particularly relevant include financial instruments such as Treasury (or government) securities, refinance credit, and foreign exchange, and state assets in the context of privatization.

As the literature on auctions is extensive, one of the main goals of this paper is to survey and draw on this literature in order to shed some light on the advantages and disadvantages of various auction techniques. 1/ A second main goal is to assess the application of alternative auction techniques to the four items mentioned above. We emphasize that the implications of theoretical models do not easily carry over into real world settings. Therefore, our recommendations on the appropriate auction technique for different assets depend heavily on individual country circumstances. Indeed, the array of country experience would suggest that there is no clearcut answer to the question of what is the appropriate auction technique. In this connection, it is instructive to note that in looking at the individual country experience across a range of countries--both industrial and developing--techniques used to auction similar items vary considerably.

This experience includes several countries that in auctioning government securities have awarded them at whatever price was bid (e.g., France, Germany, Japan, and the United Kingdom), while others have charged a single, market-clearing price (e.g., Denmark and Switzerland). 2/ Some countries have recently changed the way they run their auctions. In this regard, Mexico shifted from a multiple-price to a uniform-price approach in 1990 for Treasury bills, and then in 1993 shifted back. 3/ Italy, in 1991, switched from uniform to multiple pricing in its local-currency Treasury bill auctions, while Treasury bonds and ECU-denominated bills are auctioned on a uniform price basis. Currently, in the United States, the Treasury is experimenting with uniform price auctions for some government securities while also running discriminatory auctions.

Relevant examples are also found for foreign exchange and refinance credit. Several countries have conducted discriminatory price or Dutch auctions for refinance credit (Romania) and for foreign exchange (Bolivia, Ghana, Jamaica, and Zambia). Other countries have used uniform price auctions for refinance credit (the former Czechoslovakia) and for foreign exchange (Guinea, Nigeria, and Uganda). 4/5/ Double auctions have been used for foreign exchange (Romania).

Section II explains some basic auction formats. The terminology employed in financial circles and in the academic literature can be at odds with each other; we therefore try to clarify in this section some of the confusion that may arise. In Section III, we review the theoretical results of the auctions literature, looking into the comparisons of various auction techniques from both the bidders’ and the sellers’ perspectives. Subjects such as the “winners curse,” the revenue equivalence theorem, and the linkage principle are discussed, in addition to efficiency considerations and the extent to which different auction techniques may provide incentives for collusion. Section IV focuses on the practical and policy implications of our theoretical review. Here, we discuss the four applications of the various mechanisms for auctioning different items, pulling together the material relevant to the decisions that countries need to make in deciding how to auction those items. A summary of the analysis in the paper is contained in Section V.

II. Types of Auctions

An auction is simply an allocative mechanism. Since auctions can play a valuable role in the price discovery process, they are most useful and used in situations in which the item being auctioned does not have a fixed or determinable market value or where the seller is uncertain about the market price. These situations typically involve some degree of informational and cost asymmetries in the sense that economic agents differ in their access to, and evaluation of, information pertaining to the auctioned commodity. Auctions may be for a single object or unit--as is typical in the proceedings of such well-known auction houses as Sotheby and Christies--or a “lot” of nonidentical items, as in land tracts (as is the case in many countries) or used cars (such as in the wholesale U.S. automobile market). Alternatively, auctions may be for multiple units of a homogeneous item such as gold or Treasury securities.

Our concern in this paper is mostly with auctions for multiple, homogeneous assets, such as government securities, refinance credit, and foreign exchange. The theoretical literature relies frequently, however, on the assumption of a single unit being auctioned. Partly for this reason, the definitions given below cover both single and multiple unit auctions. Of course, auctions of single units are also relevant in economics. For example, countries’ efforts to privatize may involve auctioning a single item, such as a public enterprise.

Following the pioneering work of Vickery (1961), we can distinguish between four primary types of auctions, classifying each according to its corresponding institutional arrangement. Each classification entails its own set of rules that affect the bidding strategies of the auction participants and therefore the outcome of the auction itself. The four basic types of auctions are:

1. English auction

This type of auction is also referred to as an ascending-price auction and is commonly seen in the art world. This is perhaps the most familiar form of auction. Starting with a low first bid or a specified reservation price, that is, a price below which the item will not be sold, the auctioneer solicits increasingly higher bids. As the bid price is increased, there are normally fewer willing takers. The process continues in the case of a single item until that item is “sold!” to the last and highest bidder for the amount bid (see Figure 1a). In an auction involving multiple units, the process continues until arriving at a price at which the fixed amount supplied at auction is just matched by total demand.

Figure 1
Figure 1

Auction Formats

Citation: IMF Working Papers 1993, 012; 10.5089/9781451842838.001.A001

Source: Reinhart (1992)

2. Dutch auction

This type of auction is also referred to as a descending-price auction. It gets its name from the technique used in the Netherlands to auction produce and fresh flowers. The bidding commences at a high level and is progressively lowered until a buyer claims the item being auctioned by shouting “mine!” (see Figure 1b). Practitioners have long since streamlined the proceedings through the use of an automated clock with a hand running counterclockwise through progressively lower prices. Any buyer can stop the descent by pressing a button when an acceptable price is reached. When multiple units are being auctioned, there are normally more willing takers as the price declines; this process continues until arriving at a price whereby the fixed amount supplied is just matched by total demand.

3. First-price auction

This type of auction is an example of a sealed-bid auction, as opposed to the above two formats which are examples of open auctions. The term “first price” is commonly used when referring to the sale of a single item. The highest bidder is awarded the item at a price equal to the amount bid. When multiple units are auctioned at the same time, this procedure is called a discriminatory auction. The sealed bids are sorted from high to low and the auctioned items are awarded at the highest bid prices until the supply is exhausted. Thus, the auction discriminates between bidders in the sense that they can pay different prices according to the amount they bid (see Figure 1c). The terminology “first-price” or “discriminatory” auction follows the academic literature. In the financial community--and here is where one source of confusion may arise--this type of auction is referred to as an English auction; an exception being in the United Kingdom where it is called an American auction. This type of auction is also referred to as a multiple-price (or in some cases multiple-yield) auction.

4. Second-price auction

This type of auction is also a sealed bid auction. When a single item is auctioned, the highest bidder is awarded the item at a price equal to the highest unsuccessful bid, thus the name second price. The multiple unit extension of the second-price, sealed- bid auction is referred to as a uniform-price auction (or competitive auction), since all winning bidders receive the auctioned items at the same price (see Figure 1d). Here, some confusion in terminology also arises from the use of the term “second price” or “uniform price” auction because in the financial community these auctions are referred to as “Dutch auctions,” although this would appear to be a misnomer. 6/ This type of auction is also referred to as a marginal-price auction.

A final type of auction worthy of mention is termed a double auction. Using this format, both sellers and buyers submit bids, which are then ranked from highest to lowest to generate demand and supply profiles. From these profiles, the maximum quantity exchanged can be determined by matching sell offers starting with the lowest price and moving up, with demand bids starting with the highest price and moving down. The “equilibrium” price may, however, be indeterminate using this methodology. 7/ An example of a double auction is the market clearing mechanism in organized exchanges such as the stock exchange and commodity markets, where a specialist matches bid and ask prices in a “specialist’s book,” making the market for a particular security traded on the exchange.

In addition to the institutional arrangements governing the workings of a particular type of auction, a second aspect of classifying different auction mechanisms concerns how each bidder values the item(s) on the auction block. Economists customarily distinguish between “private-value” auctions and “common-value” auctions. The former term refers to objects acquired for personal consumption with no primary motive to resell. The bidder therefore has a personal maximum that he would be willing to pay, quite independent of the valuations of rival bidders. If this is the case, we speak of the bidder as displaying “independent private values.” A frequently cited example is an object of art purchased for personal pleasure rather than for profitable resale.

The same painting can, however, be purchased to be resold. The bid is then predicated both on personal valuation and that of prospective buyers in the secondary market. This situation is referred to as the common-value assumption because each bidder places the same value on the object--that is, each one tries to estimate what the object is ultimately worth on the basis of the same objective standard. This common value may be an unobservable variable at the time of the auction, such as would be the case when a government security is purchased to be resold later in the secondary market.

In general, all auctions are “correlated value auctions,” a category that includes the common-value and private-value auctions as polar examples. This concept of correlated values captures the notion that in each auction situation, bidders’ values are to some extent related to each other, or correlated. Milgrom and Weber (1982) use the term “affiliation” to express the same idea more precisely. However, as Rasmusen (1989) points out, “as always in modeling, we must trade off descriptive accuracy against simplicity, and private value versus common value is an appropriate simplification” (page 246). We retain this distinction throughout the paper.

For the four basic types of auctions just defined, Table 1 summarizes the rules associated with each institutional arrangement (see Rasmusen (1989)). 8/ It also outlines some simple aspects of the bidder’s strategy, which are implied by the rules and payoffs to the bidder.

Table 1.

Selected Characteristics of Different Types of Auctions

article image

III. Auction Theory

Since auctions follow well-defined rules, they can be viewed as “games,” making the application of game theory an appropriate paradigm to gaining an insight into their dynamics from modeling strategic behavior. 9/ But to gain these insights, the theoretical literature relies on a number of simplifying assumptions. Although these assumptions allow one to derive key results, they make the application of auction theory to real world settings an exercise to be undertaken with caution. In addition to reviewing some common assumptions, this section discusses auction strategy, first from the bidder’s and then from the seller’s perspective, before turning to the issues of economic efficiency, and the incentives to collude under different auction designs.

The assumptions most commonly used, depending on the context are:

  • (1) Bidders are risk neutral. 10/

  • (2) The independent private-value assumption applies; or the common value-assumption applies.

  • (3) The bidders are symmetric, that is, they use the same distribution function to estimate their valuations, implying bidders cannot discern differences among their competitors.

Following the earlier theoretical literature, it is initially assumed that one item is being auctioned.

1. The bidder’s perspective 11/

(a) Private-value assumption

In comparing the English and second-price auctions, if, as is fairly common, the former has a standard bid increment, in the limit as this increment becomes small, the two institutions result in the same price and allocation, or, using the terminology of game theorists, in the same normal form. Similarly, the Dutch auction is strategically equivalent to the first-price, sealed-bid auction since there is a one-to-one mapping between the strategy sets and the equilibrium of the two games. In both of the latter formats, no relevant information is revealed in the course of the proceedings, only at the conclusion of the auction when it is too late for any bidder to act upon or change a bid. In the first-price format, the bid is relevant only if it is the highest. Likewise, in the Dutch format, the stopping price or bid is irrelevant unless it is the highest (i.e., it stops the price descent with the winning bid).

(b) Common-value assumption

Under this assumption the equivalence between the English and the second-price auction does not hold, although between the Dutch and first-price auctions it continues to hold. 12/ See Milgrom and Weber (1982) and Smith (1987).

What optimal strategies evolve in the course of the competitive bidding process under the common-value assumption? Take the example of competitive bidding for a construction contract. Here, the contract is awarded to the lowest bidder. Assume that bidders are identical except that their valuations are based on the information to which they (differentially) have access. In calculating his bid, each player faces a trade-off between the probability of winning the contract and the expected profit if he does win. If all contenders specify their bids by adding a markup to their estimated costs, the winning bid will have the lowest estimate of project costs and will, on average, be too low. In the case of auctions for art/mining rights, etc., where the highest bidder wins, the winning bidder is faced with the realization that his assessment of value exceeded all other bidders’ assessments. That is to say, he wins the auction but loses by decreasing his expected profit! This contrary observation is termed the “winner’s curse.” The best-known study in economic literature of this phenomena is by Capen, Clapp, and Campbell (1971) on bidding for offshore mining rights auctioned by the U.S. Government. One implication of the “winner’s curse” is that inexperienced bidders earn, or profit, less than expected since such a bidder is more likely to place the highest bid when he has overestimated the value of the item. One would be disconcerted, to say the least, that one has outbid 20 experts! Experienced bidders are, of course, aware of the winner’s curse and factor it into their calculations.

The winner’s curse has several implications for optimal bidding strategies. In a first-price auction, for example, the winner by implication can expect a lower profit when he attempts to resell, since competing bidders display a lower valuation of the object. Being aware of this possibility, bidders are likely to “shade” their bids below their own estimates in an effort to move closer toward the market consensus. 13/ Ceteris paribus, as the number of bidders increase, it is prudent to bid more conservatively, since the range of the distribution of bids, and thus the highest bid, is likely to be larger the greater the number of bidders. Thus, the winner’s curse is reinforced as the number of bidders increases, creating a greater shading of bids below their true estimate.

Second, as more and more information is available about the auctioned item, the gap between the highest bid and the “true” value of the item decreases. The winner’s curse is therefore muted, as more information is made available about the value of an auctioned item. With the curse muted, it is optimal for bidders to be less conservative in their bids; this implies that, as more information is available, bidding will become more aggressive and the selling price will, on average, be higher.

Milgrom (1989) provides a useful summing up up to this point: “The most important lessons to be learned…are that the returns in bidding come from cost and information advantages, that naive bidding strategies can squander these advantages and that bidders without some advantage have little hope of earning much profit, but could with a little bit of carelessness suffer large losses.” (page 6).

2. The seller’s perspective

(a) Private-value assumption

Under specific assumptions, the theoretical literature demonstrates that all four basic auction types will yield the same expected price and revenue to the seller. 14/ This central result in auction theory, termed “the revenue equivalence theorem” (Vickery, 1961), assumes that bidders display symmetric and independent private values in auctions that are free of distortions and in which only a single unit is sold. The theorem does not imply that every realization of the game, independent of the auction type, will yield the same price and revenue, only that the expected price and revenue are the same. The revenue equivalence theorem does imply, however, that the specific auction format chosen by the seller in this stylized theoretical world is not that crucial, with each format yielding on average the same payoffs to the seller.

A construct termed the “revelation principle” is used to prove a number of theoretical propositions in auction theory, including the revenue equivalence theorem. It describes the optimal mechanism from the seller’s point of view. 15/ The term “mechanism” as used in this context acts as a black box, i.e., as a process that takes bids as inputs, and as output produces the winning bidder and the price he is required to pay. Thus, each of the auction forms can be viewed as a mechanism. In a direct mechanism each bidder is simply asked to report his personal valuation of the item. A mechanism is termed “incentive compatible” if the auction is structured in such a way that it is in the bidder’s interest to state honestly his personal valuation of the object--for example, if the proceedings require each bidder to state his valuation and the object is awarded to the bidder with the highest valuation. Under the assumption of private value, this is precisely what occurs in the first-price, sealed-bid auction. Each bidder is optimizing when he submits his bid, and the revelation principle designs the payoff structure so as to make it optimal to be honest.

Note that the revelation principle is a purely theoretical construct and few, if any, resource allocation procedures used in practice are direct incentive-compatible mechanisms. Its main application is to facilitate the search for a resource-allocation mechanism that is optimal, subject to the constraints of asymmetric information. 16/

(b) Common-value assumption

Under the set of common value assumptions we see different results, and also move closer to some of the kinds of auctioned items we are concerned with, such as government securities, where assets are acquired with the intention of profitable resale in secondary markets. More specifically, it can be shown that the revenue equivalence theorem does not necessarily hold under the assumption of common values when, in determining a bid, an individual bidder faces common uncertainties such as energy prices, pollution considerations and changing consumer tastes, that might impinge on possible resale values. In these circumstances, Milgrom and Weber (1982) demonstrate that the expected revenue from selling a single object in either of the four auction formats can be ranked from highest to lowest as follows:

  • (1) The English ascending-price auction

  • (2) The second-price, sealed-bid auction

  • (3) Tied: The Dutch auction and the first-price, sealed-bid auction

The rankings clearly illustrate the advantage of increased information. As an English auction proceeds, it reveals information about rival bidder valuations and permits a dynamic updating of the personal valuation by any individual bidder. This updating tends to cause more aggressive bidding with the end result of raising the seller’s revenue. A first-price (discriminating) auction awards the object to the highest bidder. Thus, other bidders will be placing a lower value on the object, which in turn reduces the profit the winning bidder can hope for in the resale market. In response, bidders in first-price auctions will tend to shade their bids well below their estimates, resulting in reduced revenue for the seller. The same reasoning applies to the strategically equivalent Dutch auction. In contrast, in the second-price (uniform) sealed bid format, the winner pays the bid of the next highest bidder. Hence, bidders would tend to raise their bid above that of a first-price auction bid, secure in the knowledge that they will not be disadvantaged if rival bidder valuations are much lower.

As we have seen, the theoretical analysis deals with bidders who demand only one indivisible unit of the commodity being auctioned. If bidders want more than one unit--as in the case of the government securities market--and are allowed to submit bids for different quantities at different prices, i.e., demand functions, then the above results need not hold. In particular, Maskin and Riley (1989) show that in the independent and private-values model the unit demand assumption (that each buyer wishes to purchase at most a single unit) is crucial for revenue-equivalence results. The theory for the situation where this assumption does not hold has not been fully worked out, but it is conjectured that the economic logic of the arguments for the single object environment will carry over. There is no proposition, however, that states that the revenue rankings given above will hold when the unit demand assumption is relaxed. 17/ Hence, on purely theoretical grounds one cannot assert that a particular auction format is superior to another. Indeed one cannot overemphasize that the nuances and details of the auction under consideration are exceedingly important in any evaluatory context.

A number of theoretical studies have also suggested that uniform pricing is revenue superior to discriminatory pricing in running an auction. 18/ The crux of the matter is that in using a uniform price format, the winner’s curse is muted or dampened, owing to the linkage of the final auction price to the highest losing bid. Put simply, the essence of the “linkage principle” is that auctions yielding the highest payoffs to the seller are those that are most effective in undermining the benefit to bidders of holding private information, thus transferring some of the profits from bidder to seller. As Milgrom (1987) puts it, “privacy is undermined by linking price to information other than (but correlated with) the winning bidder’s private information.” (page 4)

In any auction format, the seller can influence bids, and hence the final payoffs, by revealing information about the auctioned object. Intuitively, an individual bidder’s expected profit is highest when he can exploit to his advantage information asymmetries. That is, the bidder has access to useful information about the object’s “equilibrium” value that is not held by other auction participants. In general, more accurate information about the item’s “equilibrium” value mitigates the effect of the winner’s curse and hence the price dampening effect of bidder caution. 19/ Thus, the seller’s optimal strategy is to reveal all available information and to link the price to any exogenous indicators of value. If sellers adopt a policy of revealing information, price becomes linked to the seller’s information; this undermines the winner’s surplus value, siphoning off some portion to the seller. 20/

3. Efficiency considerations

The theoretical literature on auctions puts less emphasis on economic efficiency than on other aspects of the various auction formats such as their revenue generating potential. Nevertheless, it is extremely important in our view to underscore the efficiency of using auctions. 21/ Available evidence indicates that auctions, in the absence of distortions, function efficiently--that is, they ensure that resources accrue to those that value them most highly and that sellers achieve the maximum value for their asset. It can be shown on theoretical grounds that there exists an equilibrium on the basis of competitively submitted bids in which the auctioned item(s) are allocated endogenously in an efficient way when the price of the auctioned item is (ex ante) unknown. Empirical evidence also suggests that, in the absence of distortive factors, auctions function efficiently. 22/ In addition, the auction mechanism can achieve this objective more effectively than alternative trade arrangements, such as price setting by the seller, or by negotiation between buyer and seller. Of the four auction formats, the English and second-price result in an efficient, or Pareto-optimal, allocation in the case of private-values auctions. If, in addition, we impose the assumption of symmetric information, all four formats result in an efficient, or Pareto-optimal, allocation. In the case of common values, efficiency also requires the assumption that all bidders base their strategies on information drawn from the same distribution, as opposed to asymmetric information. This is discussed further in Section IV, which focuses on privatization of public enterprises.

4. Collusion

The extent to which incentives to collude vary under different auction formats can be of great practical concern in deciding on the type of auction mechanism to use. Indeed, the indictment in the United States of a primary securities dealer in 1991 for fraudulent activities in the government securities market has focused attention on the collusive potential of standard auction formats. These are briefly dealt with below. It is important to keep in mind that all auctions are susceptible to collusive behavior in one way or another--what we review here is the comparative incentive for collusion under different auction formats.

A basic hypothesis, first formulated in the literature by Mead (1967), is that ascending-bid formats are more susceptible to collusion than sealed-bid auctions. This belief may explain the popularity of sealed bidding, even though (as discussed in Section III) the ascending bid format has superior revenue-generating potential. Intuitively, auction formats where covert “side-deals” are possible are more likely to support bidder manipulation. Thus, an open-bid English auction is particularly vulnerable to manipulation, since a subset of bidders (“ring”) simply has to agree not to outbid each other to effectively lower the winning bid. The item can then be re-auctioned among the ring members, and the profits shared. The open-format inculcates adherence to the agreement since any ring member attempting to exploit the ring by a side deal of his own would, effectively, negate the ring and restore the auction to a competitive footing. The format ensures that compliance is easily monitored. It should be noted, however, that problems with collusion under the English format should diminish as both the actual number of bidders or the potential number of bidders increase. Intuitively, for a ring to be successful it has to have a significant proportion of the total number of bidders under its control. To help achieve this result, it is advantageous to have no new bidders be able to enter the auction. Further, with a higher number of actual bidders, it becomes more difficult to control a significant proportion, and more than one ring can form and try to outbid each other.

Sealed-bid auctions, by comparison, are vulnerable to collusion involving the auctioneer--that is, collusion between the auctioneer and one or more bidders, and between the auctioneer and the seller. 23/ This format is, however, less prone to rings, since sealed bidding tempts the participants in any conspiracy to bid just above the agreed-on price, effectively dissolving the cartel. This result also holds true for the Dutch format, even though it is an open, instead of sealed-bid, auction since the first bidder to defect from the ring ends the auction. As Smith (1987) points out, the Dutch auction is perhaps most effective against collusion because “In this auction, since none of the losing bids is known to anyone, they cannot even be leaked let alone announced and conspiracy is therefore infeasible.” (page 52) Milgrom (1987) succinctly states that “collusion is hardest to support when secret price concessions are possible, and easiest to support when all price offers must be made publicly.” (page 27)

Theoretically at least, the four formats can be ranked from (potentially) most collusive to least collusive as follows:

  • (1) English auction

  • (2) Uniform second-price auction

  • (3) Discriminatory first-price auction

  • (4) Dutch auction

The English auction is potentially the most susceptible to collusive because there is no incentive to get out of the line--i.e., more aggressive bidding does not result in the award--and such attempts are highly visible to the other members of the ring. On the other hand, the Dutch, descending-price, auction is potentially the least susceptible to collusion because of the difficulties ring members would have in supporting and enforcing collusive behavior. Once a ring member bids more aggressively than agreed, his actions are not only obvious, but the auction is won before others can react.

IV. Applications

This section discusses four applications of the various mechanisms for auctioning different items, taking in turn the auction of government securities, refinance credit, foreign exchange, and state assets in the context of privatization.

1. Government securities

There is considerable controversy over the type of auctions that are most suitable for selling government securities. As we have seen, the theoretical analysis deals with bidders who demand only one indivisible unit of the commodity being auctioned. However, frequently in the case of auctioning government securities, bidders may submit bids for multiple units of the auctioned item, and they may also be permitted to submit multiple bids, thus, in effect, demanding differing quantities and prices at the same auction. In such circumstances, theoretical models can offer only limited insight, and care needs to be taken in applying theoretical results to real world settings.

Consider first the U.S. Government securities market. The weekly auction of Treasury securities by the U.S. Government is structured differently than the simpler theoretical formats discussed earlier, and offers an excellent example of the gap between stylized models and real world settings. In addition, this market has been subject to much recent analysis and proposed changes; accordingly, the details of the market are readily available.

The U.S. Treasury’s offering of some two and a half trillion dollars in new debt annually is auctioned off in a multiple-price, sealed-bid auction with active, open trading both preceding and following the event. Thirteen and twenty-four week maturities are auctioned weekly; longer maturities are offered several times a year. The Department of the Treasury publicly announces the amount of debt securities it is offering. These are traded in an active “when-issued” market, a forward market for the same, where the actual issue date is the delivery date for the forward contract. This forward market serves two important functions: one, allocative, and the second, evaluative, in that it provides insight into the common value beliefs of the participant regarding the securities’ marketability.

At present, there are 39 bidders--“primary dealers”--who can participate in the Treasury auction. They submit sealed bids specifying a price and the number of securities they are willing to purchase at that price. These are referred to as “competitive bids” and approved dealers can submit them in several price-quantity combinations. In addition, the proceedings are open to the general public/individual investors through the submission of “non-competitive” bids that specify a quantity sought, up to a (fairly conservative) maximum, determined by the Treasury. The price paid by these non-competitive bidders is a quantity-weighted average of the winning competitive bids. Starting with the highest competitive bid, the Treasury awards the amount specified (at the stated price); the next highest bidder is awarded the amount demanded at his stated price, and so on until the supply is allocated. Winning bidders thus pay their bid, and all of them may pay different prices. The securities are delivered within a few days and may be resold in active secondary markets. Recently, starting in September 1992, the Treasury began selling two- and five-year bonds using a uniform price auction on an experimental basis.

In addition to the forward market mentioned earlier, there is a “repurchase and reverse” market in Treasury securities where short-term borrowing and lending is collaterized by these instruments. One can borrow funds overnight by selling securities with an agreement to repurchase them the next day at a predetermined price, with the difference between the purchase and the selling price being the return earned.

The potential for profit in Treasury auctions lies at the intersection of the three trading forums--the auction itself, the forward markets, and the repurchase and reverse market. Sealed bidding combined with multiple prices create the potential for any determined bidder to corner the post-auction market. Well-informed and deep-pocketed groups can, by submitting deliberately high-valued bids, receive the bulk of awarded securities. Unsuccessful bidders who have taken a position in the “when-issued” or forward market are caught in a “short squeeze,” where they are forced to pay heavily to close their positions or purchase securities at a premium in the repurchase market to be able to honor their commitments. Under current treasury auction procedures, the winner’s curse places a premium on information regarding competitive bids, (an important outcome of the “when-issued” market), creating the basis for a bid that will corner the primary auction and squeeze the post-auction market.

Having described how the market works, we look at the arguments in favor of switching to a uniform second price auction. One main argument rests on the belief that it will probably increase the revenue to the Treasury because, following the theoretical section, the “winner’s curse” would be muted, leading to more aggressive bidding. The magnitude of this increase may, however, be small. In any case, it is not clear that revenue maximization is an appropriate goal for the Treasury. Economic efficiency seems to be more appropriate.

Another often-cited advantage of uniform auctions is that they will increase participation and, hence, competition, since the winner’s curse is muted. It can, however, be argued that the number of bidders (n) participating in a Treasury auction is endogenously determined. The potential number of competitive bidders includes the primary dealers (39) and all depository institutions (a couple of thousand). There is nothing to prevent the (n+1)th bidder from entering. Clearly the intra-marginal investor does not think it profitable to bid. It is possible that this is related to the costs of “certification” and establishing “creditworthiness.” These costs are unlikely to change if one moves from one auction format to another. Hence, we would not expect any significant increase in the number of competitive bidders if the Treasury moved to a uniform price auction. 24/25/ The recent experience in Mexico, when it moved to a uniform-price auction from a discriminatory auction, and in Italy, in moving from a uniform to a discriminatory auction, bears this out. The number of primary dealers in either country has not changed significantly.

A third advantage of a uniform-price auction is that it reduces socially suboptimal information gathering. The incentive to collect information diminishes in a uniform auction. Since gathering this information only redistributes wealth among bidders, it adds nothing to society as a whole. This, we believe, is a strong argument in favor of a uniform auction: it promotes economic efficiency. A final consideration is that it may be easier to implement. On the other hand (as discussed in Section III.4), a major disadvantage of uniform price auctions, is that they are more prone to collusion than discriminatory auctions.

Hence, any policy recommendation would be country specific. If a fairly active (competitive) market exists, a uniform-price auction would seem appropriate, since collusion is minimized and there could be some gain to society from less information acquisition. Revenues to the Government may also increase. 26/

If the market in a particular country is thin and subject to collusion, a discriminatory auction would seem more appropriate. The benefits would exceed the dead-weight loss implied by excessive information gathering. In an immature market, information collecting encouraged by the discriminatory format may be useful in the initial stages of market development; nevertheless, in the absence of concerns about collusion, a later shift to a uniform format would be desirable. Needless to say, we would recommend measures to increase participation thus making the market more competitive and safeguarding against monopoly positions. These measures might include lowering barriers to entry and, hence, increasing the number of participants.

In some situations, English auctions might also be considered, particularly in cases where it is possible to run a centralized open-outcry auction (such as might be the case in a small country where all direct auction participants could meet in the capital city in one location, as for example, with the foreign exchange auction in Romania). 27/ In such a situation, a Dutch auction may also constitute a feasible and desirable option.

2. Refinance credit

Another situation in which auction techniques can be usefully applied is in the allocation of refinance credit. In general terms, refinance credit represents direct lending by a central bank, usually to the financial sector but sometimes directly to ultimate users. Lending to the financial sector can be for the specific purpose of implementing monetary policy, for example, by providing liquidity to commercial banks to meet specified monetary targets. It can also represent “targeted” lending--for example, in some developing countries to support investment and economic development in key sectors--in which the Central Bank provides funds to the financial sector for on-lending directed to targeted activities. When collateral is required to obtain refinance credit, for example, government securities, such credit is more likely to be called a repurchase agreement. 28/ In this case, instead of extending a simple credit, the transaction entails an agreement that the borrower sell to the Central Bank a given security and later buy it back at the maturity date of the repurchase agreement. Alternatively, “refinance” facilities are also referred to as “rediscount” facilities when lending takes place against securities. 29/

A main issue that arises is how to allocate refinance credit. One non-price approach with less of a market orientation has been to set the price of refinance credit at a given interest rate and provide that credit on a first-come, first-served basis up to some quantity limit. Some countries allocate refinance credit entirely on an administrative basis, directing such credit and setting its price. Frequently, when such lending is at administered rates, a substantial subsidy is involved because the administered rate is low compared with market-based interest rates. Operationally, in these cases, commercial banks have recourse (sometimes automatically) to the refinance facility at the Central Bank at below market-related interest rates for loans to specified sectors of, or agents in, the economy. Direct controls are sometimes used, instead of the first-come-first-served approach, to achieve the desired distribution of credit and deal with the excess demand that would prevail. In any event, the types of transactions described above can bring with them significant distortions in the financial system.

Auction techniques may be viewed as a mechanism to allocate refinance credit. 30/ They have the advantage of tying the refinance rate to market conditions and improving efficiency. A potential added benefit is that auctions may improve transparency while lessening discretion in the allocation of credit. Importantly, although an issue outside the realm of this paper, it should be noted that to the extent refinance credit is directed toward development objectives--for example by providing (subsidized) credit to key sectors of the economy--such policy actions might more appropriately be handled as a fiscal matter, with subsidies being budgeted directly instead of implemented through interest rate policy. Otherwise, the Central Bank may be carrying quasi-fiscal operations on its balance sheet, potentially generating central bank losses and thereby complicating monetary policy and disguising the underlying fiscal position. In any event, whether the lending takes place through the fiscal authority or the Central Bank, auction techniques would be useful. 31/

The discussion above on auction techniques in the context of auctioning government securities would be applicable to refinance credit. Thus, much of the analysis presented earlier on government securities is also applicable in this context, although an important qualification deserves emphasis. Auctioning refinance credit may be different in the sense that payment to the seller of the auctioned item may not be required upfront as is the case with government securities. Instead, payment is effectively made when the (refinance) credit matures, thus subjecting the seller to the risk of nonpayment in the interim. Collateral requirements would reduce this risk, as would an appropriate evaluation of the creditworthiness of the auction participants and associated certification. A concern is to avoid problems of adverse selection in which allocating credit by price alone creates a situation where borrowers with the poorest credit risk always place the highest bids. Such a difficult situation might arise, for example, when demanders of refinance credit have strong incentives to seek credit at higher prices because they themselves hold “nonperforming” assets in their portfolios and are ready to go under. 32/

3. Foreign exchange

Countries adopting market-related arrangements for their exchange rate have been confronted with two basic choices: operating an interbank type of market within the private sector, which may, in addition to commercial banks, include other licensed foreign exchange dealers; or an auction system whereby foreign exchange is surrendered to the central bank for auction to the highest bidders. 33/

As noted earlier, under the auction system countries have used different techniques that are basically divided between discriminatory pricing (including a Dutch auction) and a uniform pricing approach. A possible difficulty with discriminatory pricing is that it may discourage potential participants from entering the market, or impede more aggressive bidding, because of the winner’s curse. 34/ Other difficulties, using this format, are also raised concerning the appropriate exchange rate to be used for transactions outside the auction (e.g., for government transactions and customs purposes). 35/ Uniform pricing would deal with some of these difficulties and more closely match how private foreign exchange markets work. In any event, based on the country experience, the interbank approach has gained comparative favor because it involves less government control over the availability of foreign exchange to the private sector than is implied by auctions, which rely on the government specifying the quantity available, at times meeting its own needs first.

The double auction is less restrictive in the sense of fixing the supply of foreign exchange, as it brings in the private sector on both the supply and demand sides. This technique is, in a way, implicit in an interbank market when brokers match the supply and demand orders that they receive. A main difference is that rather than being a continuous market, as would be the case with an interbank market, a double auction would be run at discrete points in time--like a fixing session. Such an approach may be appropriate where insufficient institutional capacity or experience is in place to operate an interbank market, but some of the flexibility of that type approach is desirable.

4. Privatization 36/

The rapid decline of socialism in Eastern Europe has focused theoretical and empirical interest on the process of privatization, and how best to reallocate resources freed from centralized control to maximize their productive potential. The crux of the problem is to determine the most efficient private uses of the existing state assets.

As mentioned earlier, in the absence of well-defined markets for resources, auctions can function as an effective and efficient mechanism, reallocating resources to those that value them most (and where they will be most productive), and ensuring the seller a return that is superior to alternative trade arrangements. Typically, centrally-planned economies lack institutions such as a well-functioning stock market and a market for corporate control that facilitate the resource allocation function. In such economic environments, auctions can provide a conduit for efficient resource allocation. Our focus here is on the essential thrust of privatization, that is, the achievement of productive and allocative efficiency. We therefore focus on various auction formats and evaluate them based on their potential for achieving allocative efficiency. This contrasts with the major thrust of the literature on auctions, which primarily emphasizes their revenue generating potential.

Earlier we distinguished between private-value and common-value auctions. Here, we discuss the efficiency of various auction formats under both these assumptions, paying particular attention to the nuances that more accurately reflect observed behavior. We motivate our discussion by assuming n potential buyers for one indivisible unit of capital. As we have already seen (Section III), under the private-value assumption the English auction and its strategic equivalent, the second-price, sealed-bid auction, are efficient. In both formats, bidders bid their reservation price, since this is, in both cases, the dominant strategy to pursue, and thus both formats are efficient. Complications arise in the case of the first-price, sealed-bid auction and the Dutch (descending price) auction. In the most commonly analyzed case of “symmetric” environments--where bidders are identical, draw their information from the same distribution, and cannot differentiate among their competitors--these auction formats are efficient. In general, however, with first-price, sealed-bid and Dutch formats, it is not optimal to bid one’s reservation price resulting in bid shading and consequently an inefficient allocation. 37/

In common-value auctions where all bidders value the unit of capital equally and have access to the same information set, the choice of auction formats is clearly not so crucial. However, under the more realistic assumption that different bidders have private information, the analysis is not so straightforward. In particular, Maskin (1992) distinguishes between two cases: (a) where private information can be modeled as a scalar, that is, a single item of information; and (b) where the bidders’ private information can be represented by a vector that is, multiple units of information. In the former case, under fairly general conditions, the English auction is efficient but the uniform price (second-price, sealed-bid) format is efficient only if there are two bidders. In this case, the first-price and the Dutch auctions will typically not be efficient except in highly restrictive cases. In the second case, when bidders have several items of private information, efficiency is unattainable in any auction format. It can be shown, however, that when the informational asymmetry among bidders is not too great, the English and second-price auctions function better than alternative formats. 38/

To summarize, in the context of privatization, if the focus is efficiency, the auction of choice would be the English auction followed closely by the second-price, sealed-bid auction. As we have seen, the two formats are identical only in the case where there are two bidders.

V. Summary and Conclusions

Auctions play a useful role in price discovery and resource allocation. They are routinely used in market economies and can be expected to play an important role in the emerging market economies of Eastern Europe and the former Soviet Union. At present, the institutional structure in these countries may not, to varying degrees, be fully conducive to free-market economic arrangements and incentives, and the advantages accruing from the use of auction techniques could productively be exploited. Indeed, auctions can play a pivotal role in acclimating economic agents to decision making in market-determined economic structures.

This paper has focused on four applications of auction technique, namely, auctioning government securities, refinance credit, foreign exchange, and state assets in the context of privatization programs. In assessing the pros and cons of different auction formats, our starting point was to survey the theoretical literature. Auctions offer the advantages of simplicity in determining market-based prices where markets may be thin or nonexistent, and in allocating the auctioned items efficiently. The appropriate choice of auction format is less clear-cut. This ambiguity stems in part from the difficulties in applying theoretical results to real world settings and in part from the importance of individual country circumstances. Based on our earlier discussion we conclude that there are no unambiguous answers to the question of what is the “best” auction technique to use.

We attempt to provide broad guidelines to appropriate auction formats in different circumstances. Our survey indicates that uniform, second-price auctions, due to their administrative simplicity, economic efficiency, and revenue enhancing potential are perhaps the most widely applicable format. The ascending-price, English auction may be preferred in the context of privatization, in auctioning government securities or refinance credit where revenue maximization is the prime objective. However, unless individual country circumstances provide for a bidding forum conducive to the open-outcry format, this mechanism is technically infeasible. In addition, the English auction is, potentially, the most prone to collusive agreements and should be avoided if prevailing institutional arrangements are conducive to “side-deals”. 39/ We emphasize that, independent of the chosen format, auctions should be conducted competitively, with stringent safeguards against monopoly positions.

Some of the arguments need to be qualified for the case of foreign exchange, in part because auctions may not be desirable in the first place. In using any of the four basic formats, the government retains a great deal of discretion in determining the amount of foreign exchange to be auctioned. This discretion can be seen as particularly disadvantageous at a time when the thrust of the reform effort is to develop the private sector. Double auctions offer an alternative, as the government participates on the same basis as the private sector. Nevertheless, in the end, the main goal is to encourage trading among participants of double auctions, and others, not only at the time of the auction, but on a more continual basis. Thus, while auctions, especially double auctions, may be a useful intermediate step, development of an interbank market should ultimately be pursued under a floating rate system.

In conclusion we reiterate the two-pronged thrust of this paper: (i) auctions play a central role in efficiently allocating resources in the absence of alternative market mechanisms and can be--at least theoretically---productively applied in the transition and restructuring of the countries of Eastern European and the former states of the Soviet Union; and (ii) the choice of an appropriate format depends crucially on the specific item being auctioned and on the institutional arrangements prevailing in the country adopting, and choosing between, different auction techniques.