II. Alternative Auction FormatsIII. A Preliminary Experiment: Threshold versus ExposureIV. Main Experimental DesignFigure 2: Eight-Bidder Design with Three RegionsSubjects and Sessions: Before conducting the sessions that Revenues: Figure 4 shows the revenues by auction format andProfits: Figure 5 shows bidders’ profits by auction format Table 4: Summary Statistics by Auction FormatCombinatorial Auctioneering Christoph Brunner, Jacob K. Goeree, Charles A. Holt, and John O. Ledyard∗ July 26, 2006 Abstract The simultaneous multi round (SMR) auction, introduced by the FCC in 1994, has been successfully applied in the sales of high-valued market licenses around the world. The FCC now contemplates setting a new standard that incorporates the possibility of package, or combinatorial, bids. This paper provides the first comprehensive laboratory test of several combinatorial auction formats that have been proposed in the recent literature. We find stark differences in terms of efficiencies and revenues, sometimes caused by seemingly minor design details. In general, however, the interest of policy makers in combinatorial auctions is justified by the laboratory data; there are simple package bidding formats that yield improved performance, especially in terms of seller revenue. We perform “stress tests” by considering environments where collusion is sustained in equilibrium in any of the (single-stage) multi-round formats. We introduce the two-stage Anglo-Dutch combinatorial auction, consisting of a multi-round clock stage followed by a single round of sealed bids, and find it effectively breaks collusion. ∗ Brunner, Goeree, and Ledyard: Division of the Humanities and Social Sciences, California Institute of Technology, Mail code 228-77, Pasadena, CA 91125, USA. Holt: Department of Economics, PO Box 400182, University of Virginia, Charlottesville, VA 22904-4182. We acknowledge partial financial support from the Federal Communications Commission (FCC contract 05000012), the Alfred P. Sloan Foundation, the Bankard Fund, the National Science Foundation (SBR 0094800), and the Dutch National Science Foundation (VICI 453.03.606). The conclusions and recommendations of this paper are those of the authors alone and should not be attributed to the funding organizations. We would like to thank Martha Stancill, Bill Sharkey, and Mark Bykowski for useful discussions. We are grateful to Raj Advani, Charlie Hornberger, Jou McDonnell, Anton Shuster, Walter Yuan, Arzyx, and Eastek Design for programming support, to Lauren Feiler, Joel Grus, Guido Maretto, Alan Miller, Laura Panattoni, Brian Rogers, and David Young for helping run the experiments, and to Dash Optimization for the free academic use of their Xpress-MP software.I. Introduction Simultaneous auctions for multiple items are often used when the values of the items are interrelated. An example of such a situation is the sale of spectrum rights by the Federal Communications Commission (FCC). If a telecommunications company is already operating in a certain area, the cost of operating in adjacent areas tends to be lower. In addition, consumers may value larger networks that reduce the cost and inconvenience of “roaming.” As a consequence, the value of a collection of spectrum licenses for adjacent areas can be higher than the sum of the values for separate licenses. There can also be important synergies in the spectrum frequency dimension, where adjacent bands may improve capacity and reduce interference. For instance, in the FCC auction for air-to-ground communications frequencies in May 2006, a package of three bandwidth units sold for about 4.5 times as much as a single unit, and similar synergies were implied by unsuccessful bids. Value complementarities arise naturally in many other contexts, e.g. aircraft takeoff and landing slots, pollution emissions allowances for consecutive years, and coordinated advertising time slots. This paper reports a series of laboratory experiments to evaluate alternative methods of running multi-unit auctions, in both high and low-complementarities environments. Various auction formats have been suggested for selling multiple items with interrelated values. The most widely discussed format is the simultaneous multiple round (SMR) auction, first used by the FCC in 1994. In the SMR auction, bidders are only allowed to bid on single licenses in a series of “rounds,” and the auction stops when no new bids are submitted. To win a valuable package of licenses in this type of auction, bidders with value complementarities may have to bid more for some licenses than they are worth individually, which may result in losses when only a subset is won. Avoidance of this “exposure problem” may lead to conservative bidding, lower revenue, and inefficient allocations. The obvious solution to the exposure problem is to allow bidding for packages of items. In such combinatorial auctions, bidders can make sure they either win the entire package or nothing at all. As a result, bids can reflect value complementarities, which should raise efficiency and seller revenue. Combinatorial bidding, however, may introduce new problems. Consider a situation in which a large bidder submits a package bid for several licenses. If other bidders are only interested in buying one of the licenses contained in that package, they might find it hard to coordinate their actions, even if the sum of their values is higher than the value of the package to the large bidder (the threshold problem). Thus, there is no clear presumption that package bidding will improve auction performance, and laboratory experiments may be used to evaluate alternative auction formats. Experiments can also be used to evaluate the extent to which auction procedures facilitate or deter bidder collusion. The next section summarizes the main features of the auction formats to be considered, and laboratory experiments to compare these formats are reported in the sections that follow. 2II. Alternative Auction Formats The various combinatorial auctions to be considered are best understood in terms of how they differ from the incumbent standard, the FCC’s simultaneous multi-round auction procedure. Therefore, we will begin by explaining how the SMR auction was implemented in the experiments. Each auction consists of multiple rounds. Bidders have 40 seconds to submit their bids in each round. After time runs out or all bidders