Applied Mathematics, 2011, 2, 12791291 doi:10.4236/am.2011.210178 Published Online October 2011 (http://www.SciRP.org/journal/am) Copyright © 2011 SciRes. AM Joint Bidding under Capacity Constraints Beatriz De OttoLópez University of Oviedo, Oviedo, Spain Email: bdeotto@uniovi.es Received July 14, 201 1; revised August 30, 2011; accepted September 7, 201 1 Abstract In this paper we analyze the anticompetitive effects of concentration of ownership in auction markets. We compare two different auction formats with uniform price. In the first, the price equals the highest accepted bid, whereas in the second the price equals the lowest rejected bid. For the former, and for a twounit, twoplants, twofirms model, we find an equilibrium where all plants (all firms) bid according to a common bidding function. The concentration of the ownership has the same effect on the bidding behavior as elimi nating one plant. However, the expected price is lower than the one expected in such three independent plant scenario. More surprisingly (and special to this 2 × 2 × 2 case), the equilibrium is efficient. In the latter, al ternative auction format, firms bids asymmetrically for its two plants. Hence, the equilibrium is inefficient. Also, with this format, we show that the market price may be arbitrarily large. Thus, and contrary to some plausible expectation base in received auction theory, a (sealedbid) auction format in which the price for a bidder is unrelated to his bid becomes less efficient than one in which the price may coincide with that bid der’s bid, when one admits that several bidders may coordinate (through ownership) their bids. The results add to a literature that favors more winner’sbid pricing rules. Keywords: UniformPrice Auction Formats, Capacity Constraints, Ownership Structure, Collusion 1. Introduction The concentration of ownership in an industry—a smaller number of firms, each of which owns a larger size of capacity increases market power, and this results in price increases compared to a situation in which own ership is disperse. This has been well studied and docu mented in decentralized markets. The purpose of this paper is to investigate whether the concentration has the same anticompetitive effect in auction markets. This paper is motivated by some experiences in the reform of regulation of the electricity industry aiming at introducing competition at the generation level, usually characterized by a high degree of concentration. As an example of such experiences, the Spanish electricity in dustry is dominated by two major generators, Iberdrola and Endesa, which own most of the generation plants and set the market price at the pool near 90 per cent of the times. However, the bidding units in the pool are the generation plants, which must submit their price asks simultaneously. Obviously, the bids of the plants be longing to the same firm are not independent. In other words, firms have the ability to strategically choose the bids of their plants in a manner that can be different to the one we could expect if each plant were owned by a different firm. This ability allows the firms to increase the market price relative to a situation in which the own ership is disperse. How large is this price increase de pends on the price formation rules of the mechanism. In this paper we study a multiunit auction mechan ism for two units (the market demand), with four production plants—the bidderswith one unit production capacities, whose cost are independent draws of the same random variable. The two pants that submit the lowest bids are called to produce one unit each. We consider two differ ent uniformprice auction formats, one with price equal to the bid of the last seller called to produce, and the oth er with price equal to the lowest unsuccessful bid. We describe a concentrated ownership structure in which two firms own two plants each. Thus, a given firm decides on the bids of the two plants it owns. In principle, a bidding strategy of a firm could consist on a pair of bidding functions (one for each plant), each one depend ing on the costs of the two plants this firm owns. How ever, we show that at any symmetric eq uilibriu m, and for either of the two auction formats, the bidding functions
1280 B. DE OTTOLÓPEZ are independent in the sense that its bid depend on its cost (apart from its ranking among the plants of the same firm), and not on the cost of other plants of the same firm. For the auction format with price equal to the highest successful bid, we show that there exists a monotone symmetric equilibrium in which the two types of pants (the more and the less efficient ones belonging to the same firm) use the same bidding function. Moreover, such function is the symmetric equilibrium bidd ing func tion of an auction for two units with the same price for mation rules buy only three plants owned by independent firms. It is easy to check that this function is everywh ere above the symmetric equilibrium one for the case of four independent plants. Thus, the concentration of the own ership induces the plants to bid higher compared to a situation with disperse ownership. More precisely, they bid as if there were just three plants in competition. Hence, the concentration has the same effect on the bid ding behavior as eliminating one plant. As the plants bid higher with concentrated ownership, the expected price is greater than with four independent plants, but it is lower than with three independent plants. Indeed, with three plants, the price is set by the plant with the second lowest of three realizations of the cost, whereas with concentrated ownership, the price is set by the one with the second lowest of four realizations. In other words, the concentration of the ownership increases the price with respect to a situation in which the plants belong to inde pendent firms, but not as much as eliminating one plant. More surprisingly, this equilibrium is efficient, as all plants bid according to a monotone increasing function, and hence, by scheduling the lowest bidders, the mecha nism calls to produce the plants with the lowest costs. For the auction format with price equal to the lowest unsuccessful bid, Vickrey [1] has shown that bidding the cost is a weakly dominant strategy when bidders can at most be awarded with one unit (in our model, this is as to say that ownership is disperse). Ausubel and Cramton [2] and EngelbrechtWiggans and Kahn [3] analyze a me chanism in which (exchanging the roles of buyers and sellers) bidder can win more than one unit of an indivisi ble good1. For a bidder that desires more than one unit there is a positive probability that the bid for the second or latter units determines the price paid for the other units that he wins. Therefo re, there is an incentive to bid truthfully on the first unit, but to shade the true valuation of the second and subsequent ones, in order to decrease the price of the unit it wins. Thus, there is a positive probability that the mechanism result in ex post ineffi cient allocations. There is also a recently increasing re search analyzing alternative auction designs and pricing rules for wholesale electricity markets (see, for instance, Cramton et al. [5], Federico and Rahman [6], Fabra [7], Cramton and Stoft [8], and Tierney et al. [9]). Our model of concentrated ownership, where the plants bid for the right to supply the market demand, is equivalent to auction models where the bidder wants more than one unit. Thus, for the auction format where the price equals the lowest rejected bid our results are in line with those in EngelbrechtWiggans and Kahns’; bidding the true cost of the first plant is a weakly domi nant strategy, but the bid of the second plant must be above the cost. As in their model, inefficiencies arise as there is a tendency towards disseminating the units across firms more than what the relative costs would indicate. Moreover, also in line with their results, we show that there exists a continuum of monotone sym metric equilibria in which the market price is arbitrarily large. Certainly, the efficiency result for the alternative auc tion format is special to the 2 × 2 × 2 model we analyze (although it is also true with some other special market configurations). Nevertheless, it points to better effi ciency and revenues properties of an auction format that is more similar to a “pay your bid” auction, as compared to one more similar to second price auction2. Our claims is that the auction format with price equal to the best unsuccessful bid gives larger opportunities to tacit collu sion among the bidders than the format with price equal to the worse successful bid. Indeed, in the former, all equilibria are inefficient, and the b idders have the ability to coordinate on “split award” equilibria at which they can increase the price with no bound. The rest of the paper is organized as follows. In Sec tion 2 we describe the model and we prove the inde pendence of the bidding functions. In Section 3 we ex amine the auction format with price equal to the highest successful bid, and we show that there exists an equilib rium in which all the plants bid according to the same bidding function. In Section 4 we analyze the auction with price equal to the lowest unsuccessful bid, and we show that there exists a collection of equ ilibria with price arbitrarily large. Section 5 contains some of the con cluding remarks. The appendix contains some of the proof. 2. The Model There are two firms, which own two production plants each. Each plant has production capacity for one unit. 1Also Brusco and Lopomo [4] analyze a multiobject version of the English oral auction with heterogeneous objects that can be sold at different prices. They find that the possibility of signaling trough the bids allows the buyers to split the objects among them at low prices. 2This is yet another example of how intuitions based on single objects auctions may be inadequate for multiunit auctions; see for instance the discussion in Ausubel and Cramt o n ( 1 9 9 8 ). Copyright © 2011 SciRes. AM
B. DE OTTOLÓPEZ 1281 The unit costs of the plants are constant. They are inde pendent draws of a random variable with cumulative density function F and distribution function f. The sup port of the distribution is ,cc . This is common knowledge. The cost of a plant is private information to the firm that owns the plant. The market demand of the good is equal to two units for any price. The sellers compete in a pool mechanism for the right to supply one unit of output. Each plant submits a bid that represents the price at which this plant offers this unit. The pool ranks the bids in ascending order and calls the two plants which submit the two lowest bids to pro duce. The auction is a uniform price one. That means that the two plants in the schedule are paid the same price. However, we consider two auction formats which differ in the manner this uniform price is determined. In one case, the price is equal to the bid of the last plant called to produce, that is, the highest successful bid. In the second case, the market price is equal to the lowest unsuccessful bid. The Bidding Strategies Each firm observes the costs of the plants it owns, that we denote by 1 and 2, with 12 . Then, the firms simultaneously submits two bids each, 1 and 2, with 12 . A bidding strategy for a firm is a pair of bidding functions 2 and c 11 , cccb b bb bcc ,c 212 , which determines the bids of the two plants that the firm owns. It is straightforward that at any equilibrium the lowest bid must correspond to the plant with the lowest cost . bc 1 b 1 cWe look for symmetric equilibria of these games. Our first result, which applies to both au ction formats, greatly simplifies this question. Proposition 1. In any symmetric equilibrium, the bid ding functions are independent in the following sense: conditional on that the cost of a given plant is greater that (or less than) the cost of the other, its bid depends only on its cost, and not on the cost of the other plant owned by the firm. This holds for the two auction formats; with price equal to the highest successful bid and with price equal to the lowest unsuccessful bid. Proof. See the appendix. Thus, at any symmetric equilibrium any firm must behave as follows. First, it must observe the costs of its two plants in order to identify the one with the low cost 1. Then, the firm assigns a bidding function to each plant, so that the plant with low cost bids according to the function 1 (depending only on the cost of the first plant), and the plant with the high cost bids according to the function 2 (depending only on its cost), where for any c b b 1 c 1 bc b ,ccc. The intuition behind this result is the following. Let us consider first a mechanism in which the uniform price is equal to the bid of the last plant called to produce. The bid 1 can affect the profits of the firm as much as it affects whether the first plant producer or not or if it af fects the price this plant obtains. In particular, it cannot affect whether or not plant 2 produces or the price in the market when it happens. Hence, the bid must depend solely on . b 1 b b 1 Suppose now that the firm decreases 2 by c (with 2 b still above1 b). The profits of the firm change only if plant 2 is at the second position before and after lowering 2 (then the profits are reduced by b2 , that is, the reduction of the price times the number of units the firm produces), or if plant 2 moves from the third to the second position, in which case the profits increase by 22 bc . As before, the changes in the profit caused by a change in 2 do not depend on the cost 1 of the other plant. Therefore, the op timum bidding function does not depend o n . b c 2 b 1 Let us consider now a mechanism with price equal to the highest unsuccessful bid. Again, a change in 1 cannot affect the revenues or costs accruing from plant 2. Suppose now that the firm reduces 2 by c b b . This change affects the profits only if plant 2 is at the third position before and after (then the profits are reduced by , as much as the market price), of if the change in 2 moves plant 2 from the third to the second position, in which case plant 2 enters the production schedule and the profits increase by 22 b bc . Again, these changes do not depend on 1, and therefore, the optimum bidding func tion for plant 2 must depend only on . c 2 Next we describe the conditions which define any strictly monotone symmetric equilibrium strategies for each of the two different price mechanisms we have con sidered. c 3. When the Highest Successful Bid Sets the Price Let us consider first the case in which the price is equal to the bid of the last plant called to produce. Think of a firm with plants 1 and 2, whose costs are 1 and 2 respectively, which bids 1 and 2 b. Suppose that the rival firm bids according to some strictly increasing (and hence invertible) and differentiable functions and with c c b 1 b 2 b 12 bc bc for any c. If both plant 1 and plant 2 are called to produce, the price is set by the second one and hence equal to 2. Then, the profits of the firm are 212 . This oc curs when the lowest rival bid is higher than b 2bcc 12 b 1 b , and hence, with probability Copyright © 2011 SciRes. AM
B. DE OTTOLÓPEZ Copyright © 2011 SciRes. AM 1282 2 1 12 1.Fb b If plant 1 is the first one in the ranking and plant 2 is off the production schedule, then the price is equal to the lowest rival bid. The expected profits conditional on 1 being the first on the ranking and being strictly above the second position are b 2 b 1 12 1 11 1 12 1 11 11 21 d. 21 d bb bb bb bb bzcFzfz z Fzfz z The denominator in this expression is the probability that the lowest rival bid is between and . 1 2 Finally, if plant 1 is the second lowest bidder, that is, the one that sets the price, then the profits of the firm are 11 bc . This occurs when the lowest cost of the rival firm is less than 1 11 bb , and the highest is greater than 1 2 bb 1 (notice that 11 21 11 bb bb ), and hence with probability 1 21 1 11 1 21 1 21 21 d 21 d, bb c bb bb Fbbfz z Fzfz z or equivalently, 22 11 1 11 1121 2.Fb bFb bFb b Summarizing, the expected profits of a firm with costs and bidding and are 1 c2 c1 b2 b b b 1 12 1 11 2 1 12122 1 21211 2 111 11 111121 ,,, 2121d 2 bb bb bbccbccFbbbzcFzf zz bcFb bFb bFbb (1) Setting the partial derivative with respect to equal to zero for 1 b 111 bbc and we obtain 222 bbc 2 111 1121111 1211211211 22FcFcFbbcbccFbbcfbbcDbbc 1 0, (2) where 1 represents the deri vat ive of t he f unction 1 2 2 Dbb . Second, if plant 1 was the second one in the ranking before raising the bid and moves to the third position (off the schedule) by increasing 1. Then, the profits de crease by b 11 1 bcc . This occurs with probability The intuition behind this condition is the following. Suppose that the firm slightly increases 1. This affects the profits of the firm only in two cases. First, if plant 1 is the second in the ranking before and after raising the bid, in which case the profits increase as much as the bid (the market price). This occurs with probability b 111 211211 211 2 bbcfbbc Dbbc Similarly, we obtain the F.O.C. with respect to 2 by setting the partial derivative of the profit function with respect to this variable equal to zero. That is b 2 1 11211 2FcFcFb bc 2 111 12222 2122122122 21 210FbbcbccFbbcfbbcDbbc 1 (3) To understand this expression, again, suppose that the firm increases 2 by an infinitesimal amount (say b ). Then, its profits change only in two cases. First, if plant 2 is at the second position in the ranking before and after raising the bid. In this case, the market price increases by , and hence, the profits of the firm increase by 2 (notice that in this situation the two plants of the firm are called to produce). This occurs with probability These two effect must balance at the optimum bid 22 bc. This is what the second F.O.C. represents. The initial conditions that complete the differential system which defines the symmetric equilibrium bidding strategies are 12 12 bc bcc bc bc (4) 2 1 122 1Fbb c Second, if the plant moves from the second to the third position by increasing 2. Then, the profits of the firm fall by . This occurs with probability b 22 2 bc c 111 122122122 21 bbcfbbc Dbbc The later condition is a usual one in asymmetric auc tions when the support of the distribution of the cost is the same for the two type of bidders. Indeed, the differ ential system above and the boundary conditions define a problem which is very similar to that of an asymmetric auction with two type of bidders. In our case, each firm owns one plant of each type; the plant with the low cost
B. DE OTTOLÓPEZ 1283 for a given firm is of one type, say type 1, and the other is of type 2. In our case, we know that 1 bc must be less or equal to 2 bc, so that a plant of type 1 with cost c is called to produce with probability one. Suppose that 1 bc is strictly less than 2 bc. Then, the price paid to plant 1 is less than 2 bc with some positive prob ability (the probability that the cost of the rival plant of type 1 is between c and 1 12 bbc . Plant 1 could, instead, bid exactly 2 bc. By doing so the probability that the plant enters into operation remains unchanged, since the other firm never bids below 2 bc for its second plant, but the market price may raise at least to 2 bc with probability one. So this would be a profitable deviation. Hence, 1 bc must be no less than 2 bc . Let us now explain the first initial condition above. We need to show first that 1 bc must be no less than c. Otherwise, a plant of type 1 and cost c would be at the second position with some positive probability (recall that for any c and, in particular, 12 bc bc 1 bc c 2 and, hence, there could be some plant of type 2 whose bid is greater or equal to b 1 bc). In this case, plant 1 with cost c would make negative profits, as price would be less than the cost c. If, instead, plant 1 bids exactly c, it would make zero profits with prob ability one. Now, we need to show that at any symmetric mono tone equilibrium 1 bc must be equal to 2 bc. Sup pose not, that is, 12 bc bc, and think of a plant of type 2 with cost slightly below c, say c , bidding something between 1 bc and 2 bc (by the continu ity of the function 2 b, there must exist a cost c such that 2 bc is between 1 bc and 2 bc). As the two plans of type 1 bid less than 1 bc, the plant bidding 2 bc is called to produce with probability zero. If, instead, this plant submits a bid between c and 1 bc (this is possible as long as 1 bc c), it would be the marginal plant with positive probability, making positive profits. Finally, 1 bc must be equal to c. Suppose that 1 bc c. Then plant 1 with cost c is off the schedule with probability one (recall that 12 bc bc). If, in stead, this plant submits a bid between c and c, it would be called to produce with positive probability at some price above c, and its expected profits would be positive. Proposition 2. At any symmetric, strictly monotone equilibrium of the auction with price equal to the highest successful bid, the bidding functions 1 and 2 b must sat isfy the boundary conditions (4) and also it must hold that b 1 1 21 11 1 d. 1 c cbbzfzz bc Fc Proof. See the appendix. As is everywhere above 2, it must hold that 1 b b 1 21 bbz z for any ,zcc. Thus, at any sym metric equilibrium, it must hold that 1 1 1 21 11 1 1 1 d 1 d 1 c c c c bbzfzz bc Fc zf zzbc Fc (5) where the function b is a symmetric equilibrium bidding function in an auction for two objects (two units of de mand) and three independent bidders, with price equal to the highest successful bid. The following proposition shows that there is a symmetric equilibrium with con centrated ownership in which the four plants bids ac cording to the same bidding function and they all bid as if there were just three plants owned by independent firms. Proposition 3. Bidding , where 12 bc bcbc d 1 c czfzz bc c constitutes an equilibrium when the price equals the highest successful bid. That is, with this auction format there exists an efficient equilibrium for the two firms, two units, two plants case. Proof. See the appendix. Notice that bc coincides with the bidding strategy of three independent bidders competing for two units when the price is the highest successful bid. The intuition behind this result is simple. The bid 1 affects the profits of the firm only in case that plant 1 bidding 1 is the marginal plant. And if so, plant 2 is off the schedule with probability o ne. That is, at the mar ginal position, the competitors of plant 1 are the two rival plants which behave as independent bidders using a common bidding strategy b. This is the same situation as if there were just three plants in competition for the first and second positions in the ranking. b b Now consider the bid 2. As before, 2 affects the profits of the firm only in case that plant 2 is the mar ginal one. If so, plant 1 is th e first in the ranking and the firm produces two units. The gains from a higher price following an increase in 2 is now twice as much as the ones corresponding to higher . However, the “compe tition” faced when increasing 2 is also twice (two ex tramarginal rival plants, instead of one), and hence the probability of incurring in loses is also twice higher. Thus, the incentive of higher are exactly the same as b b b 1 b b 1 b Copyright © 2011 SciRes. AM
B. DE OTTOLÓPEZ Copyright © 2011 SciRes. AM 1284 the incentives for higher . 2 Remark. Notice that this result hinges on the fact that the number of competing plants for the less efficient plant of a firm equals the number of plants of that firm. b th A similar result holds for some other special market configurations. Suppose that there are N firms that own m plants each, which bid in an auction for k units with unit price equal to the lowest bid. Then , it is easy to check that if k 1mN k , then there exists a symmet ric equilibrium with all the plants bidding according to the highest successful bid and bidders. In other words, the plants bid as if the other plants of the same firm where not real competitors. For a brief outlin e of the proof, consider the first order condition for th e bid of the plant with the lowest cost 1, given that all the plants bid according to the same function b, that is, 11mN c 1 1 111 11 11 1 111 11 11 1 11 1 mN k k mN k k mN cmNFcFcf c k mN cF cF c k bc Db (6) Here, the right hand side represents the gains from increasing the bid if the plant is at the marginal position before and after the change, and the left hand side is the reduction of the profits if by increasing the bid the plant moves from the marginal to the position. This condition above coincides with the first order condition of an auction for k and independent bid ders and price equal to the lowest bid. 1th k 11mN th k th h Analogously, the first order cond itio n for the bid of the plant with cost h, the lowest cost of the firm, given that they all bid according to the same function is cb 11 1 11 11 11 mNkh kh hhh hh mNk h kh hhh mN cmNFcFcfc kh mN cF cFc kh bc hDb th k where the expression after in the right hand side is the probability that the plant with cost h bid ding is the lowest bidder. Notice that this implies that the plants with costs 1 , that bid less than , are among the first positions in the ranking, and hence, they produce one unit each. Thus, if the plant is the bidder after and before in creasing its bid, the profits of the firm change by h times the price increase. h hDb c 12 cc 1k c h bc th h th k,,, h c h bc Corolla ry . The expected market price in the symmet ric equilibrium defined by proposition 3 is below that of a symmetric equilibrium in the auction with just three independent bidders and above the price when all four plants are independent. Summarizing, in an auction with uniform price equal to the highest successful bid, the concentration of the ownership affects the bidding behaviour of the plants in the same manner as eliminating one plant. In other words, a single plant that belongs to a larger firm does not con sider the other plant of the same firm as a real competitor. Hence, it is true that the concentration increases the ex pected market price relative to a situation with disperse ownership, but not as much as eliminating all but one of the plants that are merged. With four plants bidding as they were only three, the price is set by the plant whose cost is the second lowest of four independent draws of the same random variable, whereas when there are only three plants in competition, the cost of the marginal plant is the second lowest of three realizations of that random variable. To illustrate this point, when the random proc ess is uniform in the interval 0,1 , the expected price with four independent plants is 0.6, whereas with con centrated ownership is 0.7 and with three independent The condition above coincides with (6) when 1mNk, that is, when the number of units to be sold is equal to the production capacity of 1N firms. Notice that when there are just two firms this condition stipulates that the demand must coincide with the capac ity of a single firm. The fact that at the equilibrium defined by proposition 3 all the plants bid according to a common and monotone bidding function greatly simplifies the comparisons be tween the expected prices with concentrated and with disperse ownership. In addition, the fact that there is a unique and monotone function from costs to bids has a desirable consequence in terms of efficiency; by sched uling the lowest bidders, the mechanism calls to produce the plants with the lowest costs.
B. DE OTTOLÓPEZ 1285 plants is 0.75. 4. When the Lowest Unsuccessful Bid Sets the Price Consider now a price mechanism in which the market price is equal to the bid of the plant at the third position in the ranking, that is, the lowest unsuccessful bid. Con sider a firm with two plants and costs 1 and 2 (12 c), that bids 1 for plant 1 and 2 for plant 2. Suppose that the rival firm bids according to some strictly increasing and differentiable functions 1 (for the plant of type 1) and 2 (for the plant of type 2), where is everywhere below . c c c B B B B B 1 2 If plants 1 and 2 are called to produce, the price is equal to the lowest rival bid. Hence, the expected profits of the firm conditional on its two plants operating are B 1 12 1 12 112 221 , 21 d c BB c BB BzccFzfz z Fzfz z d where the denominator is the probability of this event. If only plant 1 is operative (at the first or second posi tion in the ranking) and plant is at the fourth position, then the price is set by the plant with the highest cost of the rival firm, which bids according to . Hence, the conditional expected profits are 2 B 1 22 1 21 1 22 1 21 21 2d . 2d BB BB BB BB Bz cFzfzz Fzfz z Again, the denominator is the probability of this event. Finally, if plant 1 is called to run and plant 2 sets the price, the profits of the firm are 21 . This occurs whenever the lowest cost of the rival is below Bc 1 12 BB and the highest is above , and hence with probability 1 2 B 1 B 1 22 1 12 1 22 1 22 21 d 21 d BB c BB BB BB fzz zfzz or equivalently, 2 111 1212 22 2FB BFB BFB B Summarizing, the expected profits of a firm with costs 1 and 2 bidding and , given that the rival firm bids according to and are c c1 B 1 B2 B 2 B 1 22 1 1 12 21 121211221 2 111 21121222 ,,,221d2 d 2 cBB BB BB BBccBzccFzfz zBzcFzfzz BcFB BFBBFBB As before, the equilibrium bidding functions must sat isfy the F.O.C. of the problem. Setting the partial derivative of with respect to equal to zero at 1 B 11 Bc we have 111 1112112 11211 20Bc cFBBcfBBcDBBc (7) where represents the derivative of the function . 1 2 DB 1 2 B From the F.O.C. above it is clear that either for any or 11 1 Bc c1 c 1 211 0FB Bc c 21 10Bc . On the one hand, 11 1 for any 1 means that the plants of type 1 bid their true costs. This bidding function would be a dominant strategy if the four plants were in dependent (if each one were owned by a different firm). On the other hand, holds if, for any 1, FB Bc c 1 c 1 211 BBc does not belong to the support ,cc , as it occurs if the plants of type 2 always bid above the maximum bid of the plants of type 1. That is, if the bidding function is everywhere above 2 B 1 Bc. Consider first the case that for any 1 c. Then, by the second F.O.C. (setting equal to zero the partial derivative with respect to 2 at 11 1 Bc c B 22 Bc) and taking into account that is the identity function, we have 1 B 2 22 2222222222 212 0Bc cFBcfBcFBcFBcFc (8) The intuition behind this condition is the following. By changing 2 the firm may reduce its profits by (if plant 2 was the second plant in the rank ing and becomes the third one after the change) or in crease the profits as much as the market price (the bid 2) in case that plant 2 is at the third position before and after raising its bid. The optimum bidding function must balance this trade off for any cost . B 22 2 Bc cB 2 B 2 c Copyright © 2011 SciRes. AM
1286 B. DE OTTOLÓPEZ Proposition 4. For the auction with concentrated ownership and price equal to the lowest unsuccessful bid, the following conditions 11 11 2 2222 2 22 2 22 22 2 2 ,, 2 21 ,, Bccc cc FB cFB cFc Bc cFB cfB c ccc Bc c (9) define a strictly monotone symmetric equilibrium at which for any c in 222 Bc c2 ,cc . Proof. The first order condition for is an imme diate consequence of (8). 2 B First, we need to prove that for any in 22 2 Bc c2 c ,cc . Think of a plan t with cost 2 bidding 2 cBc . By bidding 2 c instead of B, the profits of the firm change only in two cases. First, if the plant was the sec ond in the ranking and moves to the third position. If this was the case, the price before the change was below 2 and plant of type 2 was produ cing one unit at some price below its cost. By increasing the bid, the plant increases the market price for the plant of type 1 (which was and still is operative) and, moreover, stops making losses with its plant of type 2. And second, if the plant of type 2 was setting the price before the change. Then, by in creasing the bid, the plant makes the market price in crease (no matter the position of the plant after the change) and, hence, the profits accruing from the plant of type 1 increase too. Then, c 22 2 c Bc c Bc for any . 2 Now, we have to show that for any 2 c in c 222 ,cc 0 . Let 2 be such that . Then, by condition (8), we have c 22 Bc2 c 2 22222 221FcFc FcFcFc Thus, for this co st , either and 2 c 20Fc 2 cc , or 21Fc and 2 cc. It only remains to prove that 2. Suppose not, and let Bc c 2.Bc c Then, by the continuity of 2, there is some cost at the left of B c (say c ) for which 2.Bc c By condition (8), and taking into ac count that 20fBc fc and 21FB cFc , we have 1 20,Fc that is, 21Fc and ,cc what is not pos sible. Hence, 2 Bc must be exactly equal to c. Corollary. At any symmetric equilibrium defined by, (9) the market price is higher than in the unique domi nant strategy equilibrium with four independent plants with probability one. Also, the equilibrium is inefficient. The equilibrium is inefficient in the sense that there is a positive probability that the plants that are called to produce are not the ones with the lowest cost; if the plant that sets the price is of type 2, it can occur that its cost is less than the most efficient rival plant which is scheduled. Moreover, in this case the market price is, with certainty, higher than when the four plants are independent–in which case the price is the third lowest cost, whereas with concentrated ownership, when a plant of type 2 sets the price it is because either it is the third most efficient one, bidding now above its cost, or either it is the second most efficient one, but bidding now above the three low est costs. When a plant of type 1 sets the price, the allo cation is efficient—the cost of the plant of type 2 that does produce is lower than the cost of that of type 1 th at sets the price, as otherwise it would not have bid below that quantity, and the price is the same that would have prevailed with disperse ownership, as it is set by the third most efficient one which is of type 1 and, hence, bids its cost. Summarizing, it the two plants of one firm are called to produce, the result of the auction process is the same that would have appeared with four independent plants. But is the mechanism calls to produce to one plant of each firm, then the price is, with certainty, higher than when ownership is disperse and, moreover, there is a positive probability that the allocation is inefficient. Exchanging the roles of buyers and sellers, Vickrey (1962) showed that when a single bidder can obtain at most one unit (in our case this is as to say that the own ership is disperse), bidding the true valuation (the cost) is a weakly dominant strategy in this multiunit auction where the price is determined by the best rejected bid. When a single bidder can obtain up to two units, Engel brechtW iggans and Khan (1998) find an incen tive to b id truthfully for the first unit (the first plant) but to shade the bid of the second one. The reason is that, with some positive probability, the second bid determines the price for the units he obtains. Our findings are in the same direction; the bid for the first plant coincides with its cost, but the second plant bid s above its cost. Thus, if th e plan t that sets the price is of type 2, the price is greater than with disperse ownership. Moreover, there is a positive probability that the cost of this plant setting the pricethat is, off the schedule is less than the cost of the last plant called to run. Hence, at this equilibrium inefficient allo cations arise with positive probability. Let us go back to the first F.O.C. (7). As we have seen before, this condition holds if 1 Bis the identity function or, else, if the function 2 B is bounded below by some upper bound of 1 B. In fact, we will show that there is a collection of symmetric equilibria in which the plants of type 1 bid according to some bounded function 1 B and plants of type 2 bid some upper bound M of 1 B, atever the cost of the plant. the wh Copyright © 2011 SciRes. AM
B. DE OTTOLÓPEZ 1287 Think of a firm with plants 1 and 2 with costs 1 and 2 whose rival firm bids according to a function 1 bounded above by 1 for its plant of type 1, and sub mits a bid c c B B 1 B for its plant of type 2. Then, by bid ding anyth ing less th an M for plant 1 the firm makes sure that its plant will be called to run, and the market price will be M or the bid this firm submits for plant 2 if it is below M. The firm has no incentive to submit a bid greater than M for its first plant unless its cost 1 is greater than M. So let us suppose that c c. It is clear that any bid greater or equal to M for plant 2 is equally profitable for the firm. Moreover, if M is large enough, the firm should bid exaclty M for plant 2. On the other hand, the firm should not bid anything between 1 and M, since this would reduce the market price below M without making plant 2 enter the production Schedule. By bidding less than 1, say B B1 B , plant 2 is called to run with some positive probability, and in this case the market price p is between 1 B and 1. Suppose that this is the case. Then the profits corresponding to plant 2 increase by 2, and the ones accruing from decrease by the price reduction B pc p. If M is large enough, bidding less than1 B (and hence than M) for the plant of type 2 would reduce the profits of the firm. Proposition 5. For the auction with concentrated ownership and price equal to the lowest unsuccessful bid, the following conditions 11 11 22 2 1 ,, ,, and 2 BcBc cc BcMccc cMB c (10) define a symmetric equilibrium strategy. At this equilib rium the market price is M. Proof. Any function 1 bounded above by B c is equally profitable for the plants of type 1. If plant 2 with cost 2 bids c1 B instead of M, there is some positive probability that this plant is scheduled. This oc curs when the lowest rival bid is some value p between 1 B and . In this case the profits of the firm change by 1 B pc p 22 Hence, for M to be more profitable than any bid less than for any cost , it must hold that 2.Mpc M 1 B2 c2 2 pc 2,.ccc As 2 is less than 2pc12 2Bc, the above condi tion holds for any whenever M is greater or equal to 2 c 1 Notice that any equilibrium of this type is equally in efficient; with probability 1/3 the cost of one of the plants of type 2, which are never scheduled, is below the cost of one of the plants called to produce. This is the probability that the two plants with the lowest costs of the industry belong to the same firm. The expected effi ciency losses are, then, 1/3 times the expected value of the difference between the third and the second lowest of four realizations of the random variable c. When c is uniformly distributed on 2Bc. 0,1 , the expected efficiency loses are 1/15. At the most favourable equilibrium of this type, the market price M is exactly c. Suppose that 11 0Bc for any cost 1 c, and 22 Bc c for any 2. A firm has not any incentive to bid more than zero for its first plant. By bidding anything less than c c this plant is scheduled with probability one, and the price is c or the bid of its second plant if less than c. And there are not incentives to bid less than c for the second plant, as this will only reduce the price for the first plant below c and the second plant is unable to enter into operation unless it submits a bid equal to zero. To sum up, the concentration of the ownership is more harmful under this auction format with price equal to the lowest unsuccessful bid than with price equal to the low est successful bid, both in terms of efficiency and price. Indeed, the price is, at any equilibrium of the type de scribed by (10) in the former, with certainty, no less than the upper bound for the price in the later. To illustrate the different effects of the concentration on the price across the auction formats, when the random process is uniform in 0,1 , the price in the most favourable equilibria de scribed by (10) is 1, whereas the expected price when it is set equal to the lowest successful bid it is 0.7. This means that in the first case, the expected price is a 66 per cent higher than with disperse ownership, whereas in the later the increment is of 16 per cent. 5. Conclusions The concentration of the ownership in auction markets implies that a single bidder submits bids for the different units offered, and it may win more than one unit. It is already known (see, for instance, Ausubel and Cramton (1998)) that when a bidder can be awarded with more than one unit, uniformprice auctions for multiple units do not inherit the desirable efficiency and revenue prop erties of the auctions for a single object, except in very particular settings (as, for instance, with pure common values). The reason is that in these multiunit auctions the bidders have an incentive to shade their true cost (or valuation), as their bid for one unit affects with positive probability the price of the other units they win. Inefficiency is not a result of this shading per se, but rather a consequence of differential bid shading; for a bidder, the incentives to shade are different for the dif ferent units. As there is not a monotone mapping from costs (or valuations) to bids, inefficient outcomes arise with positive probability. Copyright © 2011 SciRes. AM
B. DE OTTOLÓPEZ Copyright © 2011 SciRes. AM 1288 The auction format with price equal to the best unsuc cessful bid has been well studied by Ausubel and Cram ton (1998) and EngelbrechtWiggans and Khan (1998), among others. In line with their results, and exchanging the roles of buyers and sellers, differential bid shading appears in our model as bidders have not an incentive to shade their first bid, since it cannot affect the price that this bidder gets. But there is a positive probability that the bid for the second unit determines the price of the first. Hence, the bidders increase this second bid in an attempt to increases the price they receive for the first unit. Indeed, in equilibrium, they can increase it with no bound. In the auction format with price equal to the worst suc cessful bid, we find that the incentives for bid shading are stronger when the ownership is concentrated than when each plant is an independent firm. This causes that the expected price is higher when ownership is concen trated. More surprisingly, there are some special markets con figurations for which we find bid shading, but not dif ferential bid shading. In our 2 × 2 × 2 case (two units, two firms and two plants each) there exists a symmetric equilibrium in which all the plants bid according to the same bidding function. More precisely, they bid as in a symmetric equilibrium for this auction format with three bidders that can win up to one unit. Of course, this func tion lies everywhere above the symmetric equilibrium bidding function for the case with disperse ownership, and this implies that the expected price is higher, but not as much as it would with three independent plants. Sym metry and monotonicity guarantee efficient outcomes. Summarizing, the two auction formats we analyze create incentives to strategic bid shading when a single bidder can win several units, but, at least for this 2 × 2 × 2 and some other special market configurations, the auc tion format with price equal to the highest successful bid dominates any equilibrium of the former, alternative uniformprice auction format both in terms of price and in terms of efficiency. 6. References [1] W. Vickrey, “Auctions and Bidding Games,” Recent Advances in Game Theory, Princeton University Con ference, 1962, pp. 1529. [2] L. Ausubel and P. Cramton, “Demand Reduction and Inefficiency in MultiUnit Auctions,” Mimeo, University of Maryland, Baltimore, 1998. [3] R. EngelbrechtWiggans and C. M. Khan, “MultiUnit Auctions with Uniform Prices,” Economic Theory, Vol. 12, No. 2, 1998, pp. 227258. doi:10.1007/s001990050220 [4] S. Brusco and G. Lopomo, “Collusion via Signaling in Simultaneous Ascending Bid Auctions with Multiple Objects and Complementarities,” The Review of Eco nomic Studies, Vol. 69, No. 2, 2002, pp. 407436. doi:10.1111/1467937X.00211 [5] P. Cramton, A. E. Kahn, R. H. Porter and R. D. Tabors, “Uniform Pricing or PayasBid Pricing: A Dilemma for California and Beyond,” Electricity Journal, Vol. 14, No. 6, 2001, pp. 7079. [6] G. Federico and D. Rahman, “Bidding in an Electricity PayasBid Auction,” Journal of Regulatory Economics, Vol. 24, No. 2, 2003, pp. 175211. doi:10.1023/A:1024738128115 [7] N. Fabra, “Tacit Collusion in Repeated Auctions: Uni form versus Discriminatory,” Journal of Industrial Eco nomics, Vol. 51, No. 3, 2003, pp. 271293. doi:10.1111/14676451.00201 [8] P. Cramton and S. Stoft, “Why We Need to Stick with UniformPrice Auctions in Electricity Markets,” Electric ity Journal, Vol. 20, No. 1, 2007, pp. 2637. doi:10.1016/j.tej.2006.11.011 [9] S. Tierney, “PayasBid vs. Uniform Pricing: Discrimi natory Auctions Promote Strategic Bidding and Market Manipulation,” Public Utilities Fortnightly, Vol. 146, No. 3, 2008, pp. 4048.
B. DE OTTOLÓPEZ 1289 Appendix Proof of Proposition 1 Consider first an auction with price equal to the highest successful bids. Think of a firm bidding 1 and 2 b for its plants 1 and 2 with costs 1 and 2 c respectively. Suppose that the rival firm bids according to some bid ding functions 1 b and 2 which depend both on the costs of the plants this firm owns. b c b When the two plants of the firm bidding 1 and 2 b are called to run, then the market price is set by the plant bidding 2. Hence, the expected profits of the firm con ditional on that the two plants are called into operation are 212 . This occurs whenever the two rival plants bid above 2. That is, when the costs of the rival, 1 and , are such that . Or equivalently, if where b b bc 2 t 2c 121 2 ,tt ub b t 112 2 ,btt b 1212112 2 ,,, ,ubttcc ccbttb This is the upper set corresponding to the value of the function . In general, 2 b 1 b 12 12 ,,, , ii ukttcc ccbttk Thus, the firm bidding and will produce two units with probability 1 b 2 b 12 12 1 dd, ub 2 zf zz z where 12 ub represents the double integral over the set . 12 ub If the plant with cost 1 is the first in the ranking and the one of type 2 is at the third of fourth position, the price is set by the rival plant of type 1, that bids accord ing to 1. This occurs when the lowest rival bid is below 2 and above 1. That is, if c b b b 12121 1 ,ttlb ub, where 2 is the lower set corresponding to the value of the function . In general, lb 1 2 b1 b 12 12 ,,, , i lkttcc ccbttk i Hence, the expected profits of the firm conditional on that the price is set by the lowest rival bid are 12 11 12 11 112112 12 1212 ,d dd lb ub lb ub bzzc fzfzzz fz fzzz d 1 The denominator in this expression is the probability that the lowest rival bid is between and . 1 2 Finally, if the plant that bids 1 is the one that sets the price, this is the only plant of the firm which is called to run. The profits of the firm are with probabil ity b b b 1 bc 11 21 121 dd lbu b2 zfz zz This is the probability that the bid 1 is between the two rival bids, and hence, at the second position in the ranking. b Summarizing, the expected profits of a firm with cost 1 and 2 bidding 1 and 2, given that the rival firm bids according to the functions and are ccb b 1 b2 b 12 12 11 11 21 12122 1 21212 112112 12 111212 ,,, 2dd ,dd dd ub lb ub lb ub bbccbccf zf zzz bzzc fzfzzz bcfzfz zz Differentiating with respect to 1 and setting this de rivative equal to zero we get the first F.O.C. of the prob lem, which his b 12 11 11 21 11 21 112112 12 1 1212 1212 11 1 ,d dd dd 0 lb ub lbu b lbu b bzzc fzfzzz b fz fzzz fz fzzz bc b d Clearly, this condition does not depend on 2 neither on 2 (notice that, although the integral in the second term of the expected profits depends on 2, its deriva tive with respect to 1 does not, as the only frontier that changes when 1 changes is that of the set b cb b b 11 ub). Thus, the bidding function 1 for the plant of type 1 depends only on the co st of this p lan t, and not o n th e cost of the second plant of the firm. b Taking this into account, and setting the derivative of the expected profits with respect to equal to zero we have the second F.O.C., which is 2 b 12 12 12 11 12 11 1212 121 212 2 1112 12 2 1212 12 2dd dd 2 dd dd 0 ub ub lb ub lb ub fz fzzz 2 zfz zz bccb bz fzfzzz b fz fzzz cb Copyright © 2011 SciRes. AM
B. DE OTTOLÓPEZ Copyright © 2011 SciRes. AM 1290 As and the interior of are complementary sets, it holds that 12 ub 12 lb 1212 11 1212 121 22 dd dd ublb ub2 zfz zzfzfz zz bb Also, as the frontier of the set is 12 lb 1 ,cccbcb2 , we have that 12 1112 11 12 11 111212 121 112 2 2 1212 22 dd dd , dd lb ublb ub lb ub bz fzfzzzfz fzzz bccc bcb bb fz fzzz bb 2 12 21 22 12 22 1212 1121 21212 212112 12 211212 ,,, 2, dd ,d dd UB UB LB LBU B BBcc Bzzccfz fzzz Bzzc fzfzzz Bcfzfz zz Thus, we can rewrite the second F.O.C. as 12 12 1212 121 22 2 2dd dd 0 ub ub fz fzzz 2 zfz zz bc b d where U and L represent the upper and lower sets of the functions 1 and 2. Setting the derivative with re spect to equal to zero, we have that B 1 BB Now, and similarly to before, this condition defines the bidding function as depending solely on . 2 2 The proof for the auction format with price equal to the lowest unsuccessful bid is analogous. In this case, the expected profits of a firm with costs 1 and 2 bid ding 1 and 2, given that the rival firm bid according to some functions 1 and 2 (which, in principle, depend on the two costs of the firm) are b c 21 22 212112 12 1 ,d 0 UB LB Bzzc fzfzzz B d c c BB B BSimilarly to before, as the frontier of 21 UB is the set 122 121 ,,ttB ttB, this is equivalent to 21222122 1212 1212 212 2121111 11 ddd d ,, 0 UB LBUB LB fz fzzzfz fzzd BttBttBcBc BB Again, the function only depends on the cost of the plant of type 1. 1 BSetting now the derivative of the expected profits with respect to equal to zero, we have 2 B 12 21 22 12 22 12 22 112 121212212 11212 2 2 1212 1212212 2,dd, dd dd 0 UBUB LB LBU B LBU B Bzz ccfzfzzzBzz cfzfzzz BB fz fzzz fz fzdzdzBcB That is, 12 12 22 1212 221 212 2 dd dd 0 UB LBU B fz fzzz Bcfzfz zz B and depends on and not on the cost of the plant of type 1. 2 B2 c
B. DE OTTOLÓPEZ 1291 1 Proof of Proposition 2 Consider the F.O.C (3) 2 111 12222 2122122122 21 210FbbcbccFbbcfbbcDbbc As the bidding functions 1 and 2 cross at the upper and lower bound of the support of the distribution (see the boundary conditions of this problem), for any given cost c there is a cost 2 such that b b c 221 bc bc 1 bc . The condition above at this cost is 1 22 cb 1 1121 1DbcFcbcfcbbcfc Integrating this expression in the interval ,cc we have 1 21 1 d 1 c cbbzfzz bc Fc as we wanted to prove. Proof of Proposition 3 Suppose that one of the firms uses the same bidding function b for its two plants, and this function b is de fined by d. 1 c czf zz bc c By setting in (1), the expected profits of a firm with costs and bidding and , we have 12 bc bcbc 1 c2 c1 b 2 b 12 11 2 1 12122 1 22 1 11 11 11 ,,,21 21 21 bb bb bbccbc cFbb bzcFzf zdz bc FbbFbb The first F.O.C. of this problem (setting equal to zero the partial derivative with respect to ) is 1 b 11 11 111 11111 21 20 Fb bFbb bcFbb fbbDbb Or equivalently, the optimum for cost must satisfy 1 b1 c 11 111 1 11 1 bb bcfbbDbb (12) By the expression which defines the function b, we know that 1d c c cbczfzz Differentiating this expression, 1 cDbcbc f ccfc At 11 cb b 1 1 , and taking into account that 11 Db bDbb b 1 , we have 111 11 11 11 1 bbbbb fbbDbb Substituting this expression in the first F.O.C. above, it must hold that 1 111 bbb bc 1 , that is, 11 .bbc c (11) Hence, the optimum bid 1 for the cost 1 is given by the bidding function b. Or, in other words, if a given firm bids according to b for its two plants, then, the other firm must bid likewise for its low cost plant. b Consider now the second first order condition which we obtain by setting equal to zero the differential of the expected profits (11). That is, 2 111 22222 2 21 210FbbbcFbb fbbDbb 1 1 2 or equivalently, 11 2222 1 bb bcfbbDbb This expression is equivalent to (12) for the bid . Hence, the rest of the proof is analogous. 2 b Copyright © 2011 SciRes. AM
