Simultaneous Ascending Auctions
We have learnt that the dominant strategy for bidders in an auction is to always bid at their ‘true value’. However, what if a bidder’s true value of an item depends on whether they are able to acquire another related item? An example given in one of Stanford Professor Jonathan Levin’s economics lectures is as follows: “Suppose a bidder has value 10 for either A or B alone, or 30 for A and B together. If both license prices reach 12, straightforward bidding means bidding on both A and B. If bidding on A stops and B’s price subsequently climbs to 20, the bidder will regret purchasing A.” This example illustrates the exposure problem. Due to the strategic uncertainty resulting from bidders being unsure of the value of items, the exposure problem often results in participants under bidding, hence reduced auction revenue. A notable example of a solution to this problem, accredited to Paul Milgrom and Robert Wilson, is the simultaneous ascending auction.
The first simultaneous ascending auction took place in 1994, as a result of the Federal Communication Commission’s need to auction licenses for particular frequencies of radio waves. Milgrom and Wilson developed a system whereby all licenses could be bid on at the same time, with information on bids released after every round of bidding. From the information, bidders can gauge demand for licenses and how much other bidders are willing to pay and adjust their strategies accordingly. In each new round, the FCC sets a minimum acceptable bid for each license. When no new bids are submitted, the auction ends and licenses are awarded to highest bidders. In this spectrum auction, the FCC made $617 million from selling 10 licenses over 47 rounds.
This first auction was a success but at later auctions several flaws of the simultaneous ascending auction became apparent. At a spectrum auction in 1996, it was discovered that bidders were using information released by the FCC between each round to communicate with competitors. One cellular carrier used the last three digits of its bids to indicate strong interest in particular area codes. Another flaw was identified in 2006 when cable consortium SpectrumCo was able to alter the relative pace of price increases of the large licenses and forecast total auction revenue. This informed their strategies enabling them to acquire licenses covering 91.2% of the US population at a billion dollar discount.
The effects of these flaws have since been minimized and the simultaneous ascending auction has become fundamental in the distribution of spectrum. It has also been modified for use in other areas, such as electricity and gas.
The FCC’s spectrum auctions have brought in over $120 billion in revenue to the U.S. Treasury, demonstrating huge potential of game theory in creating value.
Sources:
http://web.stanford.edu/~jdlevin/Papers/AWS.pdf
https://priceonomics.com/the-spectrum-auction-how-economists-saved-the-day/
http://web.stanford.edu/~jdlevin/Econ%20285/SAA%20and%20Spectrum%20Sales.pdf
http://www.cs.cornell.edu/home/kleinber/networks-book/networks-book-ch09.pdf