«The Effect of Incumbent Bidding in Set-Aside Auctions: An Analysis of Prices in the Closed and Open Segments of FCC Auction 35 Peter Cramton ...»
The effect of a shift in demand by value-seeking firms such as Leap and Northcoast toward second-tier properties was an increase in the price of second-tier markets. Hence, even legitimate entrepreneurs that did not bid directly against Alaska Native felt its presence.
3.2. RESULTS OF THE OPEN AUCTION The open auction was conducted at the same time as the closed auction. Bidders that qualified as small firms received a 25 percent bidding credit on licenses in the open auction—that is, a bid of $100 was interpreted as commitment by a small firm to pay only $75 conditional on winning the license. The effect of such a bidding credit was to permit qualifying firms to switch between the open and closed auction whenever the ratio of prices in the closed auction to the prices in the open auction exceeded 0.75.
The average price per MHzPop in the open auction was $4.96, compared to the average price of $3.12 in the closed auction. Verizon Wireless, which did not use a front to gain access to the closed auction, won roughly half (113 of 252) of the open licenses. On a population-weighted basis, Verizon won 65.2 percent of the open licenses. On a value-weighted basis, Verizon won
76.2 percent of the open licenses (Verizon submitted high bids of $8.78 billion).
The most intense competition in the open auction occurred in the New York Basic Trading Area (BTA). Verizon won each of the two 10 MHz licenses covering the New York BTA for a combined price of $4.1 billion. Had Verizon allowed the next-highest bidder to acquire one of the two open licenses, the final price per license would have been significantly lower.18
4. ESTIMATING THE BUT-FOR PRICES IN THE CLOSED AUCTIONAs stated above, Alaska Native paid $2.6 billion to win 29 closed licenses that represented 36 percent of the population-weighted spectrum made available in the entire closed auction, and 50 percent of that spectrum on a dollar-weighted basis. Below, the process of bidding in an FCC spectrum auction is explained, and then the concept of supply and demand within the concept of the FCC’s spectrum auction formation is reviewed. This paper then introduces a supply and demand model that defined the market equilibrium in the closed auction, and estimates that model to determine the effect of AT&T’s bidding on the closed licenses on the average prices paid the closed licenses.
4.1. BIDDING IN AN FCC SPECTRUM AUCTION The FCC’s spectrum auctions are multiple round ascending price auctions.19 In any given auction, the FCC lets a specific number of spectrum licenses for sale and establishes minimum bid prices (reserve prices) for each license. Bidding occurs during rounds. At the start of an auction, rounds may last as long as two hours. Toward the end of the auction, when very few bids are submitted, rounds may be as short as 5 or 10 minutes.
At the start of a round, the FCC announces the current price of a license and the minimum price at which it will accept a new bid for that license in the subsequent round. A valid bid must be at least as high as that minimum price. In most auctions, bidders may exceed that minimum price, but only by bidding increment amounts stipulated by the FCC. For example if the current price of a 20 MHz license in New York is $50 million and the bid increment is $5 million, then the minimum acceptable bid in the next round is $55 million. However, a bidder could, if it desired, submit a bid of $50 million + I x $5 million, where I is an integer typically between 1 and 9.
Currently, the FCC specifies bid increments as a percentage of the current price of a license.
Increments typically range between 20 percent and 5 percent of the current price of the license.
The percent increment depends on largely on the number of new bids submitted for the license in the prior round. The more bids there were in the prior round, the higher the increment will be.
Therefore, increments are typically 20 percent near the start of the auction, when bidding is intense, and gradually decrease to 5 percent as the auction ends. With the exception of the very last round of the auction, in which no new bids are submitted, the average price of spectrum is, by construct, higher than that of the previous round. Hence, the name ascending price auction accurately describes an FCC spectrum auction.
To be eligible to bid in an FCC auction, a bidder must make an initial down payment. The size of a bidder’s down payment determines the maximum quantity of spectrum, expressed in MHz-pop, it can bid on at auction. Small regional operators who are interested only in a few licenses typically make small initial payments to bid only on those few licenses. National carriers typically make large payments that allow them to bid simultaneously on all licenses at auction, so long as those licenses are not set-aside for small carriers.
Each bidder has a specific demand for spectrum. The auction ends when prices rise high enough so that no new bids are submitted. When this occurs, the sum of all bidders’ individual quantities demanded for spectrum will equal the fixed quantity of spectrum that the FCC has let at auction. Put simply, the market clears when supply equals demand.
This paper focuses on the effect of AT&T’s participation through Native Alaska even though other large firms used fronts because Native Alaska dominated the closed auction for spectrum.
Cingular Wireless bid through Salmon, which enjoyed a 25 percent discount in the open auction as well as access to the closed auction. Voicestream bid through Cook Inlet VS, which had no open auction discount, but still did have access to the closed auction. Verizon did not bid through any smaller front. This may have been due to the unclear legal status of the front setups at the time, or more likely the falling marginal returns on fronting after another large market player has done so. If Verizon entered through a front, they would still compete with other large firms in the closed auction; meanwhile, Verizon could take advantage of lower prices in the open auction with AT&T’s absence.
4.2. CALCULATING THE MARKET DEMAND FOR CLOSED SPECTRUMLet Qrc denote the quantity demanded of all bidders in the closed auction in round r. Quantity demanded in a given round is a function of the price of closed spectrum in that round, prc. As round numbers rise, and license prices increase, the model can trace out the market demand curve
for closed spectrum:
To estimate the market demand for the closed licenses, the model first retrieves the market quantity demanded for closed spectrum, measured in MHzPop,20 in each round of the auction.
The model adds the MHzPop from all new bids submitted in round r to MHzPop from all standing high bids from round r–1, and then subtracts out the MHzPop of any bidder who raised its own high bid in round r. The model then calculates the price in dollars per MHzPop on all standing high bids in round r to get the associated price. This procedure yields a market demand curve for spectrum in the closed licenses. Specifically, the model performs the following
calculation for each auction round:
winning bidder” increased its own high bid from round r – 1. N rc is obviously a valid component of round r demand, because those bids were submitted in that round. High bids from the prior round are also part of this round’s demand, because a bidders new bids and standing high bids are not allowed to exceed its total eligibility in any given round. Finally, if a bidder increased its own high bid, the spectrum from that license would be included in both new bids, N rc, and standing
bid in round r, and adds its standing high bids in MHzPop from round r–1. The model then subtracts off any intersection that occurs when a high bid is raised,21 and matches these quantities to the price variable discussed above for each of the 100 auction rounds.22 Matching AT&T’s quantity demanded for closed spectrum against prices for that spectrum yields AT&T’s demand
for closed spectrum:
4.4. MARKET EQUILIBRIUM AND AT&T’S IMPACT IN THE CLOSED AUCTIONThere was a fixed supply curve of spectrum in both the open and closed auctions. In the closed auction, 1.7 billion MHzPop of spectrum was let at auction. Thus, the market supply of
closed spectrum was given by the following, perfectly inelastic, supply curve:
With AT&T bidding in the closed, the market-clearing price was roughly $3.12 per MHzPop, which was found by setting Q c equal to 1.7 billion. Specifically, supply in Equation  was equated with demand from Equation . The two curves intersect at a price of $3.12 per MhzPop.
The market demand curve is the sum of the individual bidder’s demands. Therefore, to calculate AT&T’s impact on prices in the closed auction, AT&T’s quantity demanded for closed spectrum is subtracted in each auction round from the market demand in the closed auction. In particular,
which is the graphing of the residual quantity demanded for closed licenses in round r against the price for the closed spectrum in round r. The model then looks along this residual market demand curve to find the price at which total market demand was roughly equal to the market supply of
1.7 billion MHzPop. Formally, the model solves:
Solving for price yields $1.97 per MHzPop. Therefore, AT&T’s presence in the closed auctions caused an increase in price of roughly $1.15 per MHzPop, or 58 percent.
5. ESTIMATING THE BUT-FOR PRICES IN THE OPEN AUCTION
5.1. METHODOLOGY What AT&T would have paid in the open if it was not allowed to create a bidding front in the closed is now estimated. First, AT&T’s total demand for spectrum—that is, what Alaska Native bid in MHzPop in both the open and the closed for a given round is determined. Above, AT&T’s demand for spectrum in the closed was found. To determine AT&T’s total demand for spectrum, AT&T’s demand for open spectrum is horizontally added to AT&T’s demand for closed
r. That quantity was bid at a price of pro. Matching those price-quantity combinations yields
AT&T’s demand for spectrum in the open segment of the auction:
Next, AT&T’s total demand for spectrum as the horizontal summation of quantity in the open and the closed segment of the auction is determined, and matched against the average price of spectrum in both auctions. Thus, AT&T’s total demand for spectrum in Auction 35 is
where q ATT is AT&T’s total quantity demanded, and p is the average price of spectrum.
To transfer AT&T’s total demand into the open auction, first AT&T’s demand curve for spectrum is estimated. Then what AT&T’s quantity demanded of spectrum would have been at the open prices is predicted. Formally, econometric techniques are used to estimate q ATT ( p ). The
ˆ where q ATT is AT&T’s total demand for spectrum had it only been allowed to bid in the open segment of Auction 35.
Next, Alaska Native’s actual bids (in MHzPop) in the open are subtracted from the market
yields an estimate of the demand for spectrum in the open that would have been realized if AT&T had been precluded from bidding for the closed licenses. Formally, let Qro denote total quantity demanded in the open segment of Auction 35 in round r. As in the closed segment, demand is a
function of price:
With this equation, the price that would have occurred at the fixed supply of 2.32 billion MHzPop in the open portion of the auction is estimated. Further, the price obtained in this last step is used to predict the quantity of spectrum that AT&T would have won in the open, given their demand for total spectrum, and its savings from having bid in the closed is calculated.
5.2. ESTIMATING AT&T’S DEMAND FOR SPECTRUM First, AT&T’s demand for spectrum in each round of the auction, at the average price per MHzPop in the entire auction, is determined.23 The price for spectrum ranged from $0.10 per MHzPop in round 1, and increased steadily to over $4.18 per MHzPop by round 100. AT&T’s total quantity demanded for spectrum ranged from a maximum of 34.3 billion MHzPop in round 4 to 648 million MHzPop in round 98. At auction’s end, AT&T won a total of 649 million MHzPop of spectrum. Note that this demand curve is not perfectly downward sloping—the FCC’s activity rules required a bidder to be active on 80 percent to 99 percent of its total demand during different stages of the auction. Thus, a bidder can, to a limited extent, expand the MHzPop on which it bids as prices rise. For this reason, the demand curves are smoothed, and in so doing, the fact that if bidders were active on 8 billion MHzPop at a price of $2.50 per MHzPop is acknowledged at higher price points in the market demand curve. So, the market would demand no more than 8 billion MHzPop at a price greater than $2.50.24 Alternatively, AT&T’s demand could be smoothed by filling back smaller quantities demanded with any larger quantities that occurred in later rounds, thus creating a demand curve that is a true step function. This would serve as an upper bound of AT&T’s demand in any round of the auction, and might lead the model to overstate the savings that AT&T garnered from bidding in the closed. For this reason analysis is conducted using the smoothed demand curves as well. Footnotes will reference these results in later sections of the paper.
An ordinary least squares is now run on AT&T’s total demand in Auction 35. The model uses a log-linear model—namely, prices are regressed on the log of quantity, for the sample of 100 auction rounds. Thus, the estimates of α and β are sought in the regression equation