CONTACT: Stanford University News Service (650) 723-2558
STANFORD -- A new type of government auction devised by economists at Stanford University and the University of Texas will be used by the Federal Communications Commission in selling radio spectrum licenses for wireless technologies later this year, the federal commission decided on Tuesday, March 8.
The auction, called a "simultaneous ascending-bid auction," differs from familiar public auctions in which the items are sold in a one-at-a- time sequence, said Stanford University economist Paul Milgrom, one of the Stanford economists who designed the new auction.
In contrast to auctions for wine or antiques, where the pieces are valued independently, Milgrom said, the value of a spectrum license to a bidder can depend on the other licenses that the bidder acquires - the reason a different auction design was needed.
For example, a license to use a portion of the radio spectrum in Northern California - say, for cellular telephone service - may be more valuable if the bidder can also acquire a license to use the same band in Southern California at low cost, Milgrom said. Unlike sequential auctions, the simultaneous auction allows bidders to learn something about the prices of all the licenses before having to make their final bids on any single license.
Milgrom worked with Robert Wilson, another Stanford economist, on the auction design model as consultants for Pacific Bell, and both said this case set a precedent for use of economic game theory in designing government auctions.
"The U.S. government makes extensive use of auctions to sell treasury bills, mineral, oil and timber rights, and emissions permits," Milgrom said. "But this is the first time that game theory has been used to forecast how the rules of the auction game are likely to affect outcomes, to test those predictions in laboratory experiments, and to use the results to devise and implement superior designs for federal government auctions."
Another economist who independently came up with a similar model for the FCC is Preston McAfee of the University of Texas, Milgrom said. Most of the major communication companies sought out economic experts who could analyze the FCC auction proposal, which led to an ongoing discussion of how to make the design as efficient as possible, Milgrom said.
At issue for the government was how to fairly and efficiently auction off 2,562 licenses for 10- and 30-megahertz communications bandwidths that were recently allocated to personal communication systems - the next generation of cellular phones and other wireless communication devices such as two-way pagers and hand-held computers. Overall, 120 megahertz of access are to be sold, which will more than triple existing capacity for this type of use. The White House Office of Management and Budget has estimated the access to the airwaves is worth about $12 billion to companies that would provide various wireless services on local, regional and national basis.
Previously, the FCC awarded licenses first by public hearings and later by a lottery; the latter provided windfalls to individual citizens who could sell their rights to the highest bidder. Last year, Congress decided an auction was the preferable method to provide opportunities for broad competition and protect the taxpayers' interests.
In September, the FCC proposed it hold a sequence of auctions to sell the licenses by region. A national company that wanted to acquire the same band throughout the nation would submit a sealed bid for nationwide rights to that band, and then regional public auctions would be held to give those who only wanted to buy regional access a chance to compete. If the sum of the regional bids exceeded the price offered by the highest national bidder, the licenses would be allocated to the highest regional bidders. If the highest national bid, however, exceeded the sum of the regional bids, the national bidder would get the band nationwide.
"It sounded sensible to a lot of people, but it isn't," said Milgrom, who determined that the rules both gave the national bidders an advantage and would tend to reduce the price the government would get for the licenses overall.
Sequencing the sales by region also would create bias in prices, he said.
"The price each license would attract would change depending upon whether you started the sales in New York and ended up in California or started in California and ended up in New York," Milgrom said.
The problem, he said, is that "the value of the licenses are interdependent, so that the whole is worth more than the sum of the parts."
Two reasons for this are based on the technology itself, he said.
"If you sell a cellular phone service to customers in Northern California, they will want the phone to continue to work when they drive into Southern California," Milgrom said. "If the service provider has acquired the same band in both regions, the phone will work in both, and the service provider should be able to charge the customer more for the service."
If the service provider acquires only the Northern California band, he or she not only loses this marketing advantage, Milgrom said, but will also find that some of the potential customers who live and work at the "geographical interface" of the two regions will experience some communications interference.
"That will reduce the value of the service to those potential customers," he said.
Finally, he said, the buyers of larger areas probably will enjoy economies of scale based on their ability to spread the fixed costs of designing a system over a larger customer base.
As an alternative, Milgrom proposed that the FCC sell all the licenses simultaneously in an ascending-bid auction that might take six weeks to two months to finish.
In this scenario, each bidder would offer a sealed bid for the licenses he or she was interested in, the bids would be opened at the end of the day, and the highest bid would be listed as the "tentative price" on the license. The tentative price becomes the lowest acceptable bid for another round of sealed bids the next day. The process continues for each license until no new bids are received for it.
At first, Milgrom said, critics said such an auction would be too complicated for the government to run and for bidders to bid. However, he said, "we provided them with software that could keep track of the auction, and showed them what the problem looked like from the standpoint of an individual bidder."
From the bidder's standpoint, he said, the problem is no different than deciding what to purchase from among thousands of listings on the New York Stock Exchange.
"You look only at those properties you are interested in," Milgrom said. "As long as you have ample time to reconsider your situation [before the next round of bidding], then once one license gets too expensive for you, you can switch from bidding on an expensive 30-MHz band to bidding on a less expensive 10- or 20-MHz band."
To compare the performance of the Milgrom-Wilson design to the government's proposal, Pacific Bell sponsored experimental tests of the two. Subjects in the experiments were Cal Tech students who bid for licenses and were later permitted to resell them to the experimenters for specific prices. The prices were determined to incorporate the kinds of value interdependencies found in the personal communication system licenses.
"The crucial advantage of experiments over actual experience," Milgrom said, "is that the experimenter can control the values of the licenses to the bidders. That allows us to tell whether the auction results in licenses being assigned to the bidders who value them most, or whether the auctions show some systematic biases."
The experiments did show systematic differences between auction designs. In particular, he said, "the government's originally proposed design had a bias toward national bidders."
"In the trials using the original design, the national bidders won licenses 92 percent of the time when they were the bidders with the highest values, but they also won 53 percent of the time when the regional bidders had a higher total value."
In the simultaneous auction design, the national bidders won 88 percent of the time when the value for them was higher than for the regional bidders, but never won when their value was lower than for the regional bidders.
"The results show that the new design is both less biased and more efficient," Milgrom said.
The simultaneous auction is likely to take longer, he said, "but so do sales of major companies take a long time. There are billions of dollars at stake here, and there is no reason to rush it when we are talking about permanently affecting the structure of a new industry."
While the model that was devised was specifically for the FCC's situation, Wilson said that "the principles used in the design have broader applicability to other government auctions."
"The FCC auction is likely to attract a lot of attention from economists seeking to analyze the outcomes," Milgrom said. "There is a breakthrough occurring in the quality of analysis underpinning auction design decisions, and the lessons are likely to be applicable to a wide range of government auctions."
This is an archived release.
This release is not available in any other form.
Images mentioned in this release are not available online.
Stanford News Service has an extensive library of images, some of which may be available to you online. Direct your request by EMail to email@example.com.