09. – 14. Juni 2002, Dagstuhl Seminar 02241
Electronic Market Design
Daniel Lehmann (The Hebrew University of Jerusalem, IL)
Rudolf Müller (Maastricht University, NL)
Tuomas Sandholm (Carnegie Mellon University – Pittsburgh, US)
Rakesh V. Vohra (Northwestern University – Evanston, US)
Auskunft zu diesem Dagstuhl Seminar erteilt
About the Seminar
During the week of June 9-14 the Dagstuhl Seminar on Electronic Market Design was held. The aim of the seminar was to provide researchers and practitioners in economics, mathematics and computer science working on topics in electronic trading, auction design, mechanism design and artificial intelligence with a platform where they could interchange results and ideas and profit from the achievements of each other's field of expertise. The seminar attracted leading researchers from the fields, in particular from the United States and Israel. This included for example leading figures on auction design, like Sushil Bikhchandani, UCLA, Peter Cramton, University of Maryland, Paul Milgrom, Stanford University, Benny Moldovanu, University of Mannheim, leading researchers from mechanism design, like Mark Satterthwaite, Northwestern University, Eric Maskin, Princeton, and Dove Monderer, Technion, as well as from Computer Science, like Noam Nisan, Hebrew University, Christos Papadimitriou, Berkeley, and Yoav Shoham, Stanford. The working atmosphere was excellent, in particular thanks to the perfect local organization in Dagstuhl. In the following we will give a short overview of the main topics addressed during the seminar and the impact the workshop has on the development of the field.
The seminar included 37 lectures as well as a rump session on diverse problems such as winner determination for combinatorial auctions, the trade-off between the informational and economic efficiency of markets, implementation of incentive compatible mechanisms as well as analyses of the strategic consequences of the design of real-life auctions on and off the Internet as they currently exist, in particular the UMTS spectrum auctions, eBay, and the electricity auctions. Because of the diverse background of the participants the central issues in electronic market design were approached from many different angles. Specifically the following four different aspects of market design were discussed extensively.
- economic efficiency From an economic perspective auctions and markets are instruments that can be used to allocate scarce goods in an efficient way, meaning that the goods are to be divided among the agents participating in the auction or the economy in such a way that the overall welfare of the agents is maximized. In for example combinatorial auctions or markets for perishable goods (with severe time constraints) enforcing economic efficiency may be a complicated or sometimes even impossible task.
- strategic behavior Market design relies on the assumption that the agents engaged in trade within the market behave according to the ideas envisaged by the designer. Thus it is of crucial importance that the design is immune to manipulation by the participants in the market. Especially in electronic markets, where buyers and sellers engage in anonymity, shill bidding, sniping, and false-name bidding do often occur. It is an important task for designers to develop trading mechanisms that discourage manipulation of this type.
- computational complexity Trade that takes place in a complex environment, such as a combinatorial auction or market for heterogeneous goods, puts a heavy computational burden on the buyers and sellers in such an environment. One of the problems market designers face is to develop trading mechanisms for these complex situations that are still transparent from the trader's perspective and nevertheless guarantee outcomes that also are sufficiently close to efficiency in the economic sense of the word.
- evidence from the field Case studies from the day-to-day ongoing practice of electronic trade such as the sales on eBay, the FCC and UMTS auctions and the auctions for surplus electricity indicate that electronic market design is an area of research that is and still very much needs to be developed further. The call for faster, simpler and more robust designs is heard everyday and everywhere throughout the internet!
Related to each of these topics, most recent research results were presented. In comparison to a previous seminar, organized by the same organizers at the International Institute of Infonomics in the Netherlands (see www.etrade.infonomics.nl/workshop), several scientific breakthroughs were presented. For example, the development of fast algorithms for combinatorial auctions has reached a strength such that these auctions can be used by the FCC in future spectrum auctions in the US. Experimental and empirical results let us better understand the strategic behavior in online auctions. Several presentations illustrated new iterative auctions with bundle bids. Furthermore, new insights in the interplay of complexity of communication, computation, and bidders decision making were presented, for example by Daniel Lehmann and Noam Nisan and their students from Hebrew University, researchers from Universiteit Maastricht, and from Tuomas Sandholm and his team from Carnegie Melon University. Notably, most of these results strongly benefit from the interplay between economics, game theory and computer science. Without the fruitful exchange of ideas between disciplines, like it has been facilitated by this Dagstuhl seminar, many of them would not have been possible. Finally, this seminar contributed to lay out a research agenda on electronic market design for the following years. For example, it remains still a big puzzle how integer programming theory can be successfully applied to understand auction markets with budget constraints.
Training and European Impact
Very remarkable for the seminar was the intensity of communication between very different fields, and between junior and senior researchers. This communication is even more remarkable given the heterogeneous scientific background of the participants. Experienced colleagues, like the economist Eric Maskin, told us after the seminar that this quality of inter-disciplinary communication is extremely rare, and how greatly he enjoyed this experience. Almost half of the lectures were given by young researchers, many of them still PhD students. For many young colleagues from Israel and the US, this was likely the first time that they got the opportunity to meet the senior colleagues from the field. It has to be said that the field is still not very well developed inside Europe, at least when it comes to the theoretical foundation of electronic markets. European research seems to focus more on the adoption of technologies in business settings. Nevertheless several research groups were present (e.g., Maastricht, Karlsruhe, Kiel, München, Cambridge), including PhD students from these groups. For these groups, and for others in Europe hopefully too, the seminar provided certainly a large stimulation to catch-up with the international research agenda. All presentations enjoyed a large audience, and presenters got plenty of feedback, despite a dense schedule. At the same time it was observable that the tutorial speakers as well as other senior speakers had put a big effort into their presentations, such that PhD students could get a maximum out of it. It might for some of the senior researchers have been a rare occasion to have time for a whole week to listen all day to presentations. Remarkably, and to the enjoyment of the young presenters, they made full use of it and missed almost none of the presentations. Partly inspired by the seminar, a European consortium on the field of market design is currently emerging, and preparing project proposals for the sixth framework program.
Electronic market design is a new field of research that builds on theories from various established fields. The challenges of market design are to create rules for trading interaction, in particular for auctions, that lead to economically desired allocations of items and payments, and that are immune against manipulation by strategic behavior of the participants. When market design is implemented electronically, in principle more complex market designs are realizable because of computerized transactions. However, computational complexity increases rapidly and excludes therefore certain designs. These problems have led to plenty of research in computational issues of market design, which was widely presented at the seminar. At the same time, most electronic markets will still have human participants interacting with them. The impact of design on their behavior cannot completely be captured by theoretical models, but requires an empirical or experimental investigation. Finally, every (electronic) market design has to be embedded into a broader set of issues, for example the industrial environment in which it takes place, asking for careful economic considerations of the impact of design on market outcome in the short and the long term. The Dagstuhl seminar on electronic market design has successfully provided a forum, in which world-wide leading researchers from these fields could exchange their ideas. European research projects on the topic are currently in preparation.