Abstract.

When multiple items are sold through the use of a simultaneous ascending-bid auction, bidders can find it in their mutual interests to reduce their aggregate demand for the items while prices are still low relative to the bidders' valuations. The FCC's first broadband PCS auction provides examples of how such mutual reductions might be "arranged" even when the bidders are not allowed to communicate with one another outside of the auction arena.

Introduction.

On December 5, 1994, the Federal Communications Commission began to accept bids for 99 licenses that would permit licensees to offer broadband personal communications services (PCS) in regions covering the United States and its territories. The auction continued for more than three months, and at its conclusion the total of the winning bids stood in excess of $7 billion. Thirty bidders were involved in the early rounds of the auction, and eighteen of them eventually acquired licenses.

For each of 48 regions, two licenses designated A and B, covering adjacent 30 MHz blocks of spectrum, were offered; for three additional regions (including New York City, Los Angeles, and Washington, DC), only single B licenses were offered, the A licenses having been previously awarded by the FCC through a non-auction-based process. Qualified bidders would eventually be allowed to hold only one of any pair of licenses. (Each region was referred to as an MTA, or "major trading area".)

The auction format.

Within the FCC, Evan Kwerel had suggested years earlier that licenses be allocated by auction, and in 1993 Congress had finally given the FCC a mandate to do so. However, the Congressional mandate was written in broad terms, and left the FCC to decide what particular type of auction process to use. On September 22, 1993, the FCC issued a Request For Comments (RFC) inviting outside suggestions. A common theme to a number of the resulting recommendations (and to a paper this author submitted to the FCC prior to issuance of the RFC) was that licenses should be sold via some type of simultaneous ascending-bid process, within which more than one license would be on the auction block at the same time.

The simultaneous auction procedure implemented by the FCC proceeded in discrete rounds. Following a suggestion made by Paul Milgrom and Robert Wilson, the FCC required bidders to remain active in every round in order to protect their eligibility to acquire licenses at the end of the auction. A bidder's "activity" in a round equaled the total population coverage of licenses for which the bidder was the high bidder in the previous round, added to the total population coverage of other licenses for which the bidder submitted new bids in the current round. The auction began in "Phase I", where a bidder's activity, a, in each round placed a permanent upper bound of 3a on the bidder's activity in all subsequent rounds. The FCC would later move the auction into "Phase II", where the permanent upper bound was 3a/2 , and finally into "Phase III", where the upper bound on all future activity equaled the activity in the current round. These activity rules, in combination with an FCC-controlled minimum bid increment, served to keep bidders from laying in wait (not bidding until they saw the intentions of others). Each bidder was granted five waivers, which could be exercised to keep the bidder's activity in a specified round from imposing any upper bound on future activity. (The primary intent behind the creation of waivers was apparently to protect bidders from potential errors in a day's bid-submission process.) Bids would be submitted as integral numbers of U.S. dollars.

After every round of bid submissions, the bidders would be provided with a listing of all submitted bids on each license; the listing would identify the submitter of each bid. Then a bid withdrawal round would occur, during which a bidder could withdraw any bid that was currently the standing high bid for any license. (The primary rationale behind allowing bid withdrawals was to enable a bidder to adjust the package of licenses it was attempting to acquire as some licenses became, relative to the bidder's valuations, prohibitively expensive.) If a bid was withdrawn, the FCC would list itself as the current high bidder for the license, and would announce a minimum acceptable bid typically equal to the highest of all other bids the license had ever drawn (although the FCC reserved the right to set the re-offering price, and adjust it in future rounds, in whatever way it chose). To discourage frivolous bids, the bid withdrawer would be obliged, at the end of the auction, to pay a penalty equal to the difference between the withdrawn bid and the final winning bid, if the final bid were less than the withdrawn bid. In essence, this meant that the FCC was guaranteed to receive at the end of the auction at least the amount of the withdrawn bid for the license.

The auction would end when a round passed with no new bid submissions. Until the end, any license could be bid upon, even if several rounds had passed during which that license received no new bids.

Strategic demand reduction.

In the sale of a single object through an ascending-bid auction, each active bidder has an expressed demand of one unit. As long as the total expressed demand is greater than the single unit of supply, the auction price continues to rise: The auction cannot end until the aggregate expressed demand drops to one unit. However, an individual bidder has only one way to reduce its expressed demand, and that is to drop out of the auction, i.e., reduced its expressed demand from one to zero. As long as the auction price is below the bidder's current valuation of the object being sold, this act of "demand reduction" is strategically dominated by the alternative act of continuing to bid.

When more than one item is being sold, and some of the bidders have the desire to acquire more than a single item, the situation changes dramatically. It becomes possible for a bidder to gain from strategic demand reduction, even if all of the items are currently priced below the bidder's valuation.

As an extreme example, consider two bidders, each of whom desires either one or two of three items being sold. Let each bidder assign a valuation of 100 to each single item, and 200 to any pair of items. If one of the bidders is prepared to continue bidding on the two cheapest items until their prices cross the 100 threshold, the other does better by immediately reducing its demand to a single item (and letting the auction end immediately, while the prices of all three items are near zero) than by following any other bidding strategy. Demand reduction when prices are below a bidder's valuations is no longer strategically dominated by other bidding behavior.

Indeed, it is simple to construct examples where strategic demand reduction occurs in equilibrium. Consider the sale of two items, in a simultaneous ascending-bid auction involving two bidders. Assume that each of the bidders values either item at 100, and the pair at 200. And finally, consider the following strategy for a bidder:

In any round, if the bidder is not the current high bidder on either item, bid at the minimum increment for the lower-priced of the items (as long as that minimum increment leaves the price below 100). If both are priced the same, select a single one at random and bid on it. If the bidder is currently the high bidder on at least one item, don't submit any bids for the round.

If both bidders follow this strategy, the items will sell within two rounds, at prices equal to the minimum increment over the opening price. Obviously, no change in strategy by one of the bidders can improve upon this outcome. (The only prospective change that could be advantageous would have to involve the deviating bidder having some chance of winning both licenses. But this will only happen when both licenses are priced within a single bid increment of 100.) As long as the minimum bid increment is less than  100/3, the selection of this strategy by both bidders stands in equilibrium.

Of course, with multiple bidders and many more items, there is a vast number of equilibria to the bidding game. With laws against prior agreements amongst the firms, the preselection of a specific equilibrium outcome is impossible. Rather, in order to bring the auction to a conclusion at relatively low prices, the bidders must "negotiate" with one another, and reach agreements that are inherently stable, inside the auction arena. In order to negotiate, it must be possible for the bidders to communicate with one another.

Methods of communication.

The most obvious way of communicating one's interest is by bidding on a desired item. The format of the FCC auction provided several other ways for bidders to communicate with one another.

Perhaps the most blatant means of communications would be to encode messages using the lower-order digits of one's bids. Most of the submitted bids were in excess of $1 million. To use the lowest three or four digits of a bid to send a signal would be of relatively negligible cost. For example, assume that one bidder is competing with another on a particular license, and the second bidder is standing as the high bidder, at an attractive price, on some other license. Then a bid by the first bidder on the license in contention, with trailing digits equal to the identification number of the other license, could serve as a threat: "Top my bid here, and I'll raise the price on your license next round." Alternatively, an immediate bid on the second license, with digits pointing to the first, could be construed to mean: "Get off the first license, or I'll continue to bid here." As long as the bidders have some idea of one another's general interests, either message would be easy to read.

Properly-timed bid withdrawals could play a similar signaling role. A bidder who tops another's bid on one license, and during the bid withdrawal period withdraws a bid on another license, could be interpreted as offering the second license to the displaced bidder in return for being allowed to remain high on the bid-upon license.

Withdrawing one's own bid on several licenses could signal an intent to use the freed-up eligibility to strike elsewhere. By taking a waiver in the following round, the withdrawer could hold the eligibility in hand long enough to see if the threatened strike elicits a response from others.

Alternatively, if bidders are all bidding near the minimum acceptable bid in each round (i.e., the current high bid plus the minimum bid increment), one could submit a bid slightly higher than the minimum acceptable bid for a license, and if high, immediately withdraw the bid. Under the FCC's rules, the license would be reoffered at a somewhat lower price in the next round, rather than at a price equal to the high bid plus the minimum bid increment. This would retard the rate of price increase by a round, and signal to others a willingness to slow the general rate of price increases while other forms of communication take place.

Enforcement.

What can sustain a tacit agreement among bidders concerning an allocation of licenses, when no binding agreements are legal? The force of threats can serve to stabilize an agreement. If two bidders have ceded licenses to one another, a subsequent attempt by one to violate the agreement can be immediately met with a response by the other, raising the prices of licenses held by the violator.

Of course, attempted violations would be most common in the final phase of the auction, where, under the FCC's rules, a bidder's activity level in each round bounds its activity level in all future rounds. Differing packages of licenses covering exactly the same total population were, in the actual auction, nonexistent for all practical purposes. Hence, any attack, or response, would necessitate a permanent reduction in a firm's eligibility: It could never return to the original package it held before the attack or response.

The substantial reduction of eligibility - a clear, public, and binding restriction of demand - by one bidder before others do the same, places that bidder in a very vulnerable position, since it sacrifices the ability to respond to an attack, and then return to its original holdings. In order for agreements to remain stable, the bidders must engage in a process of detente, gradually and mutually reducing eligibility as the auction approaches its conclusion. To protect their positions, bidders must hold onto a small amount of excess eligibility until the very end. This can be accomplished by continuing to bid for both licenses on some MTAs, or by bidding for licenses on small, inexpensive MTAs in which a bidder holds no real interest, until very near the end of the auction.

A tale of two bidders: The interaction between WirelessCo and American Portable Telecommunications during the auction.

The auction potentially might have progressed in a vast number of very different ways. No bidder could anticipate all possible happenings, and preplan a response to each. At best, a general approach to the strategic issues could be planned, and that approach adapted as the situation required.

APT stood in a particularly advantageous strategic position in the auction. It did not seek to acquire expansive national coverage; indeed, APT's upfront payment in the auction qualified it to acquire licenses covering a population of at most 34,000,000, although APT was eligible to bid on licenses covering a much larger population. One of APT's siblings in the Telephone and Data Systems (TDS) family, U.S. Cellular, held cellular licenses covering approximately 10% of the U.S. population. Another sibling, American Paging Corporation, had obtained paging licenses covering the entire country in an earlier narrowband auction. Many different packages of broadband PCS licenses would complement TDS's aggregate holdings, and therefore APT was in a flexible position, where it could adjust its objectives in order to seek licenses offering substantial value relative to their auction prices. [Disclaimer: I consulted with the American Portable Telecommunications (APT) bidding team before every round through the final phase of the auction (Rounds 65-112). It therefore is important to note that the comments offered here are purely speculative interpretations, based on the bid information released each round by the FCC, and do not necessarily reflect the considerations of the bidding team at the time of the auction, or the intent of any action taken by APT.]

The first two phases of the auction (Phase I, Rounds 1-11, and Phase II, Rounds 12-64) were somewhat uneventful. Most of the bidders engaged in just enough activity in each round to preserve the total eligibility for which they had originally qualified, and prices generally stayed well below what were expected to be the final selling prices.

On February 3, 1995, the FCC announced that the auction would move into Phase III, beginning with Round 65. From that point forward, each bidder would be walking down a one-way street. A bidder's total activity in any round would bound its total activity in any future round. Any two packages of licenses covering different sets of MTAs would differ at least slightly in total population coverage. Therefore, any shift in the package on which a bidder was active would preclude the bidder from ever returning to bid on the previous package.

Maintenance of eligibility was a critical issue for the first time in the auction. Indeed, most of the bidders carried into Round 65 an eligibility level not exactly equal to the population covered by the set of licenses on which they were currently active, and stood to lose eligibility immediately. So it wasn't particularly surprising that 11 of the 24 remaining bidders took waivers in Round 65. Major exceptions were AT&T, which took several striking actions, including bidding on both Chicago and Washington, DC for the first time in 20 rounds, and PCS PrimeCo, which preserved most of its eligibility both by bidding on licenses where it had been inactive, and by bidding for both licenses covering Miami. Of particular interest to APT was that WirelessCo, while taking its waiver, hit the B license in Tampa, where APT had been sitting high since Round 55.

After this one-round pause to look at developments, most bidders chose not to exercise waivers in Round 66. APT sacrificed less than 11,000 units of population eligibility by combining a Minneapolis-Houston-Miami-Milwaukee-Pittsburgh-Denver-Kansas City-Columbus package with a bid for the 46,000 units of population in American Samoa. This met the joint goals of retaining flexibility by not sacrificing eligibility in future rounds, and "marking" the MTAs of primary interest to APT. APT maintained its activity on this package in Round 67.

Many of the licenses of interest to APT were also of apparent interest to WirelessCo, which was seeking to acquire a broad national collection of licenses. For APT doggedly to bid head-to-head against WirelessCo in these markets would accomplish nothing except to rapidly raise the prices that APT would eventually have to pay. Clearly, both firms would gain if some accommodation could be reached. WirelessCo had to be made to see what would happen if it didn't make room for APT.

Conveniently, AT&T's bid for the single license covering Washington, DC had not been topped in Rounds 66 and 67. It was generally perceived that this license was one on which AT&T was prepared to outbid - within reason - any other bidder, and the license was still priced quite low. In Round 68, APT restructured its activity package, losing less than 2,000 units of population eligibility while bidding atop AT&T in Washington, DC.

[In this and the following tables, both the bid amount, and the bid per unit of population coverage, are given. Round numbers in parentheses indicate the last bid on a license prior to the rounds listed in the table.]

Washington-Baltimore				population : 7,777,875
A license				B license
(65)					130,000,000	16.71	AT&T Wireless
68					136,500,333	17.55	American Portable
69					143,325,000	18.43	AT&T Wireless
70					150,501,222	19.35	American Portable
71					158,026,000	20.32	AT&T Wireless
72					165,927,111	21.33	American Portable
73					174,223,000	22.40	AT&T Wireless
74					182,934,999	23.52	American Portable
75					192,082,000	24.70	AT&T Wireless
76					201,686,333	25.93	American Portable
77					211,771,000	27.23	AT&T Wireless

For the next ten rounds, APT ping-ponged AT&T, as the price of the Washington, DC license eventually rose above $200 million. That APT could do the same thing to other bidders, in other markets, was clear.

In Round 78, APT moved into the seventh-largest market, Dallas, hitting WirelessCo while restructuring its activity package to sacrifice less than 10,000 units of population eligibility. WirelessCo struck back in the next round, APT then moved to the cheaper license, topping PCS PrimeCo, which in turn struck back. At the same time, APT continued to bid for Minneapolis, Houston and Miami, all markets contested by WirelessCo. In Miami, APT went so far as to hit WirelessCo on the more expensive side of the market.

Dallas-Fort Worth				population : 9,694,157
A license				B license
(74)	68,489,000	7.06	WirelessCo	(69)	68,601,000	7.08	PCS PrimeCo
78	71,913,789	7.42	American Portable
79	75,509,000	7.79	WirelessCo
80					72,031,789	7.43	American Portable
81					84,232,323	8.69	PCS PrimeCo

In Round 81, WirelessCo submitted bids for licenses covering Houston and Miami, then withdrew both bids later in the day. Both of the bids were on the more expensive licenses in those markets, hitting APT. This appeared to be an offer directed squarely at APT: "Leave us alone elsewhere, and we'll let you have these two markets."

Houston				population : 5,190,849
A license				B license
74	72,316,483	13.93	American Portable	(73)	71,420,412	13.76	PCS PrimeCo
75					74,991,000	14.45	WirelessCo
76	75,932,536	14.63	PCS PrimeCo
77					78,741,333	15.17	American Portable
78	79,729,000	15.36	WirelessCo
79					82,680,425	15.93	PCS PrimeCo
80	83,715,789	16.13	American Portable
81	87,902,000	16.93	WirelessCo
			[bid withdrawn]
82	83,716,837	16.13	American Portable
83	87,903,000	16.93	WirelessCo

Miami-Fort Lauderdale				population : 5,136,581
A license				B license
(71)	103,201,135	20.09	PCS PrimeCo	(71)	108,255,000	21.08	WirelessCo
78					113,668,789	22.13	American Portable
79
80
81					119,352,000	23.24	WirelessCo
82							[bid withdrawn]
82	108,361,837	21.10	American Portable		113,669,000	22.13	WirelessCo
83	113,785,432	22.15	PCS PrimeCo
84					119,352,837	23.24	American Portable
85	119,475,000	23.26	WirelessCo
86					126,020,126	24.53	PCS PrimeCo
87	125,450,001	24.42	GTE Macro Com.
88	131,723,000	25.64	WirelessCo

In Round 82, APT rejected WirelessCo's offer, apparently finding it insufficient. APT moved into the cheaper side of the Houston and Miami markets, and also attacked the fifth-largest market, Detroit, again with a bid topping WirelessCo. At the same time, APT bid into Tampa, another market that WirelessCo was contesting.

Detroit				population : 10,001,009
A license				B license
(36)	81,177,000	8.12	AT&T Wireless	(44)	78,101,277	7.81	WirelessCo
82					82,006,837	8.20	American Portable
83					86,107,000	8.61	WirelessCo

The next natural step up the ladder was to the fourth-largest market, San Francisco. This was a prime target for WirelessCo, since a member of the WirelessCo consortium already had been awarded a Los Angeles license by the FCC. WirelessCo had sat undisturbed on one of the San Francisco licenses from Round 32 to Round 75. However, WirelessCo and ALAACR Communications had contested the New York license sporadically from Round 7 until Round 74. Finally, in Round 75 ALAACR had left WirelessCo high in New York, and had hit them instead in San Francisco while attacking Pacific Telesis (PacTel) in Los Angeles, maintaining enough eligibility to move back into New York later. It appeared that WirelessCo might be afraid to touch off a return to New York by hitting ALAACR in San Francisco.

Rather than move into San Francisco immediately, APT waited for two rounds, preserving eligibility by bidding in the still-hotly-contested Atlanta market in Rounds 84 and 85. APT also continued to contest the Tampa and Miami markets with WirelessCo.

In the meantime, ALAACR had stopped bidding for Los Angeles after being topped by PacTel in Round 82, and had preserved its eligibility in Rounds 83 and 84 by taking its last two waivers. Out of waivers, ALAACR became active on both sides of the San Francisco market in Round 85. APT moved into San Francisco in Round 86.

San Francisco-Oakland				population : 11,891,177
A license				B license
(32)	123,395,983	10.38	WirelessCo	(33)	132,000,943	11.10	Pacific Telesis
(75)	130,000,000	10.93	ALAACR
85					138,601,000	11.66	ALAACR
86	136,500,837	11.48	American Portable
87	143,326,000	12.05	Pacific Telesis
88					145,541,004	12.24	American Portable
89	150,492,238	12.66	WirelessCo		152,818,000	12.85	ALAACR
90	158,021,837	13.29	American Portable		160,460,000	13.49	Pacific Telesis
91	165,925,000	13.95	ALAACR
92	174,221,367	14.65	WirelessCo		168,494,837	14.17	American Portable
93					177,001,000	14.89	Pacific Telesis
94	182,947,837	15.39	American Portable
95	192,095,838	16.15	WirelessCo
96					192,099,000	16.15	ALAACR
97					202,150,000	17.00	Pacific Telesis
98	206,500,000	17.37	WirelessCo

In Round 87, WirelessCo withdrew its standing high bid on Minneapolis. It wasn't clear whether this was an offer to APT, to GTE (which had been contesting Minneapolis and Miami), or merely an act to free up enough eligibility to move into San Francisco (for which WirelessCo submitted a bid just below APT's in Round 88). APT was certainly willing to move into Minneapolis, and did so the very next round.

Minneapolis-St. Paul				population : 5,986,039
A license				B license
81	35,981,000	6.01	WirelessCo		34,790,000	5.81	Continental Cable
82					36,590,001	6.11	GTE Macro Com.
83
84
85	37,785,000	6.31	Continental Cable
86					38,420,000	6.42	WirelessCo
87							[bid withdrawn]
88					36,600,012	6.11	American Portable

The contest for San Francisco continued, with ALAACR, APT, PacTel and WirelessCo competing for the two licenses. In Round 95, WirelessCo topped APT on the high side of the market. The trailing digits of WirelessCo's bid are of interest. APT had been trailing its bids with the digits "837", which spell out "TDS" on a telephone touchpad. WirelessCo chose to end its bid with an "838". This might be read as a message to APT: "San Francisco is a priority for us. Please go away." In Rounds 96 and 97, APT submitted bids on the low side of San Francisco, and WirelessCo sat untouched.

After Round 97, WirelessCo's standing bid was again the lower of the bids for the two San Francisco licenses. Presumably, APT was poised to bid on that side of the market, were it not given a reason to bid elsewhere. WirelessCo provided a reason: It withdrew its standing high bids on both Tampa and Houston. In the next round, APT moved into both of those markets, and stopped bidding on San Francisco.

Tampa-St. Petersburg				population : 5,417,788
A license				B license
(84)	94,255,827	17.40	WirelessCo	(85)	99,327,723	18.33	PCS PrimeCo
97			[bid withdrawn]
98	89,786,837	16.57	American Portable

Houston				population : 5,190,849
A license				B license
(84)	87,903,000	16.93	WirelessCo	(79)	82,680,425	15.93	PCS PrimeCo
97			[bid withdrawn]
98	83,737,837	16.13	American Portable

At the same time, there was a difficult-to-explain flurry of activity elsewhere. WirelessCo raised its own bid on San Francisco by more than $14 million. Perhaps this was an attempt to drive ALAACR out of the San Francisco competition. Yet ALAACR had already decided to move elsewhere. It bid over WirelessCo's standing bid on Dallas in Round 98, WirelessCo hit back in Round 99, and ALAACR dropped out of the auction in Round 100. Whether WirelessCo's increase of its own bid in San Francisco in Round 98 helped to drive ALAACR out of the bidding is a question that might go forever unanswered.

WirelessCo also topped APT's standing bid on Columbus, which had been quiet for ten rounds. It then withdrew the bid immediately. APT rebid for the license at essentially the same price as its previous standing bid in Round 99, only to find that WirelessCo had resubmitted the bid just withdrawn. It took three more rounds before Columbus was back in the hands of AT&T and APT again, at slightly higher prices than before. Again, it's not clear precisely what WirelessCo's objective was.

Columbus				population : 2,145,561
A license				B license
(87)	20,217,000	9.42	AT&T Wireless	(86)	20,113,837	9.37	American Portable
98					21,120,000	9.84	WirelessCo
							[bid withdrawn]
99					21,120,000	9.84	WirelessCo
100	21,228,837	9.89	American Portable
101	22,290,000	10.39	AT&T Wireless
102					22,176,837	10.34	American Portable

By Round 100, the auction was nearing its conclusion. From this point on, no more than ten new bids were submitted in any round. APT was settled into Minneapolis, Tampa, Houston, Kansas City, and was resettling into Columbus. It continued to compete for Denver and Seattle, and to maintain eligibility through bids at various times on Honolulu, Alaska, Guam, and American Samoa.

In Round 101, WirelessCo re-entered Minneapolis, carefully avoiding APT by hitting Continental Cable on the high side. Continental chose not to respond, and after using its last waiver, left the auction. APT, apparently reading WirelessCo's action as a signal that a similar such action would not jeopardize APT's positions in Tampa and Houston, withdrew its standing bid on Alaska at the end of Round 102, combined the freed eligibility with that it was using to compete for Seattle, and hit CCI Data on the high side of Pittsburgh in Round 103, avoiding WirelessCo's lower standing bid. Several rounds later, after exhausting its remaining waivers, CCI left the auction.

As the auction wound down, there was only one way that APT's holdings could be at risk: If WirelessCo were to attempt to reclaim some of the licenses held by APT. And WirelessCo continued to cling to eligibility, through actions such as bidding repeatedly for both licenses covering Oklahoma City.

On Friday morning, March 10, Round 108 drew only five new bids. During the bid withdrawal round, however, WirelessCo withdrew a long-standing bid on Milwaukee and its recently-uncontested bid for a second license covering Oklahoma City, bringing into hand enough eligibility to hit APT in one of its largest markets. And that afternoon in Round 109, WirelessCo did exactly that, bidding over APT's standing bid on the high side of Houston.

APT's response was immediate. In the withdrawal period following Round 109, it withdrew its standing high bids on both Pittsburgh and Kansas City. Combined with the eligibility freed by WirelessCo's hit in Houston, APT carried into the weekend just enough free eligibility (12,206,919 in population) to bid for San Francisco on Monday morning.

Curiously, however, during the same withdrawal period, and therefore with no knowledge of APT's response, WirelessCo withdrew the bid it had just submitted for Houston. What was WirelessCo up to? One possible interpretation is that WirelessCo was seeking to be relieved of the bid penalties it was facing in Houston, and perhaps in Tampa as well, by asking APT to raise its bids for those licenses. Another is that this was a swap offer - Milwaukee for Houston - although that seems unlikely, since Houston was both larger in population and more attractively priced.

Whatever the explanation, WirelessCo had the weekend to consider APT's threatened response. On Monday morning, the auction entered its first day in which three bidding rounds were scheduled. In Round 110, APT and WirelessCo put their withdrawn bids back into place, APT reclaiming Houston and reoccupying Pittsburgh and Kansas City while WirelessCo returned to Milwaukee and its second Oklahoma City license. At midday, in Round 111, only four bids were submitted: GTE raised its standing bid on Cincinnati to the level it was obliged to pay due to an earlier withdrawal, WirelessCo did the same on Milwaukee, Western PCS took one of the two Oklahoma City licenses from WirelessCo, and APT replaced its bid on an Alaskan license from which it had withdrawn 9 rounds earlier, and which would otherwise have reverted into the hands of the FCC. Finally, in Round 112, no bids were submitted, and the auction ended.

Assessing the results.

It is difficult to compare directly the results obtained by the winning bidders in the A- and B-block auction, due to the demographic and physical differences between the regions covered by the various licenses. However, a year later the FCC began a similar auction of C-block licenses. These licenses covered the same amount of bandwidth at roughly the same frequency, and in aggregate covered the same territory. Each individual license covered a smaller area (a BTA, or "Basic Trading Area"), and a total of 493 licenses were put up for bid. Participation in the auction was restricted to small businesses, businesses controlled by women or minorities, and qualifying rural telephone companies.

On May 6, 1996, after 184 rounds, the BTA auction ended. The resulting prices established by that auction are not directly comparable to the prices from the MTA auction, since installment payments are available to the BTA bidders. Still, it is of interest to note the relative prices of population coverage acquired by bidders in the MTA auction, and the prices of licenses covering the same regions in the BTA auction. Prices in the MTA auction were, on population-weighted average, only 40.5% of the net (after adjusting for bidder credits) BTA prices. The prices of licenses acquired by each of the winning bidders in the earlier auction, as a percentage of the net prices of the BTA licenses covering the same territories, is displayed in the percentage column of the table below. APT acquired licenses covering more than 10% of the total population, at a cost of less than one-third the cost of C-block licenses covering the same population.

After adjusting for differences in the payment arrangements for the BTA licenses, it still appears that the C-block prices exceed those of the MTA auction. Why would this be? A primary reason would seem to be the increased level of competition for the licenses. The BTA auction drew 255 bidders, of which 89 acquired licenses. However, the increase in prices is more than one would expect if both auctions yielded prices near the market-clearing prices of individual licenses. It seems that at least part of the increase must be attributed to the licenses in the first auction selling at prices well below their market valuations. and the primary reason this could happen is because, with fewer bidders, the ability of the bidders to coordinate their strategic demand reductions is substantially increased.

Final observations.

WirelessCo was a consortium of Sprint and three cable television companies, and PCS PrimeCo was a consortium of four regional Bell operating companies. By permitting such consortia to form prior to the auction, the FCC facilitated strategic demand reduction in two ways: by permitting some of it to occur before the auction had even begun, and by lessening the number of other bidders with which any one bidder needed to establish within-auction communications.

The release after each round of detailed information concerning all submitted bids, and the identities of all the bidders, similarly facilitated the strategic communication process. However, this policy of full information revelation also likely served to increase the economic efficiency of the final outcome.

The presence of bidders with narrow interests in only a few licenses in the BTA auction probably also served to bring prices in the BTA auction closer to full market value. In particular, bidders demanding only a single specific license have no incentive for strategic demand reduction. If every license offered has such bidders seeking it, one would expect final prices to be comparable to the prices obtainable by selling licenses separately and individually. A structuring of licenses into smaller coverage areas would probably draw more such bidders into the auction.

Alternatively, the sale of licenses through a sequence of smaller simultaneous ascending-bid auctions would reduce the opportunities for mutually-advantageous strategic demand reduction among the bidders. Of course, if the early auctions in such a sequence sold only relatively large licenses, the ability of firms to preserve spare eligibility in order to be able to adjust their bid packages in the final phases of the auctions would be reduced. The FCC's decision to impose a weaker Phase III bound on eligibility in the BTA auction (a bidder need be active only on licenses covering 95% of the total population for which it wishes to remain eligible) fits well with this alternative.

Addendum.

The chart below indicates the licenses on which APT was active in Rounds 63-111. The beginning of any vertical interval indicates the round in which APT submitted a high bid, and the end indicates the round in which some other firm topped APT's bid. The symbol "x" indicates a submitted bid that was not high. A lowercase "w" indicates a bid withdrawal by APT, and an uppercase "W" a bid withdrawal by WirelessCo. In rounds 66 and 67, APT was active on both licenses for American Samoa.

The table below lists the identification numbers associated with each MTA upon which APT bid.

Robert J. Weber
J.L. Kellogg Graduate School of Management
Northwestern University
May 6, 1996