Market Design: A Selective Review

Market Design: A Selective Review Thomas Kittsteiner Axel Ockenfels Abstract We argue that the design of market and negotiation rules significantly affects outcomes, and that small design changes can have substantial consequences. We focus on three particular environments online markets, electricity markets and partnership dissolution mechanisms and show how a combined theoretical and empirical approach to market design research may help to improve the governance of markets and negotiations. Dr. Thomas Kittsteiner: London School of Economics, Department of Management, Houghton Street, London WC2A 2AE, email: t.m.kittsteiner@lse.ac.uk. Prof. Dr. Axel Ockenfels: University of Cologne, Department of Economics, Albertus Magnus Platz, D-50923 Cologne, Germany, email: ockenfels@uni-koeln.de (http://ockenfels.unikoeln.de/ao).

A. Introduction The design of market rules that govern the interaction of traders and other market participants dates back to the ancient world. Auction markets were common in ancient Rome and even the Roman empire was once sold in an auction. 1 Goethe designed new market rules when he sold one of his manuscripts to his publisher Vieweg. Around 200 years later it turned out that Goethe's rules are a special case of Vickrey's famous auction design, and that Goethe was probably aware of the economic advantages of his design (Moldovanu and Tietzel (1998), and Vickrey (1961)). Despite the long history of auctions and other exchanges, market design has not gained momentum as a scientific discipline until recently, when the theoretical and empirical market design literature started to grow rapidly and at accelerating rates. This development has many explanations, including a surge in privatization activities, availability of reliable and fast computer technologies, and progress in game theory and experimental economics. These developments made it possible, for instance, to design smart auction markets that can deal with various complexities, such as complementarities among objects (e.g., in spectrum licenses 2 ), or to design various stable matching markets that suffered from congestion, unravelling and incentive problems. 3 Moreover, e-commerce and online trading pose new design challenges, partly because trade is among anonymous, geographically dispersed traders in C2C markets, and products are complex and multi-dimensional in most B2B markets. In general a market-designer's task is to find market rules that eventually will lead to a desired market outcome. 4 A market outcome is given by a (final) allocation of traded objects and by monetary payments between market participants. The task of the market-designer is often complicated by allocational complexities and asymmetric information (about preferences). To illustrate the first point, consider a local government who wants to privatize a bus network by selling the individual routes separately. Due to complementarities, a buyer's preference over bus-routes specifies his willingness to pay for any possible package of routes. Even if the seller is completely informed about the (potential) buyers' preferences, it is still a non-trivial task to split up and allocate the network in the profit-maximising way. 5 In fact, a considerable literature in Operations Research is concerned with allocational complexities in auctions and markets (for a recent survey see Anandalingam et al. (2005) and de Vries and Vohra (2003). Another difficult, but less understood class of allocational complexities arises 1

in the presence of allocative externalities; e.g., the value of a spectrum license may depend on the entire allocation of licenses to (possibly different) bidders (see Jehiel and Moldovanu (2001, forthcoming)). In this article, however, we will mainly focus on the second issue: governance structures in case of asymmetric information. From the economist's perspective, a market where participants are ignorant about each others preferences is a game with incomplete information. The literature in mechanism- and auction design, a sub-discipline of game theory, analyses how equilibrium behaviour in these market games depends on their rules, and, by the same token, how rules that implement desired market outcomes look like. That is, mechanism design and auction theory examine market incentives, reveal market failures, and thus provide useful guidance in market design. 6 However, although important, theoretical considerations alone typically do not guarantee a sensible market design. This is true in particular in those cases, in which reality is too complex to be solved analytically. Here, the role of theory is in developing intuition and identifying trade-offs in design choices by isolating a particular effect rather than capturing as much reality as possible. But typically, theory needs to be complemented with other tools such as computational, empirical and experimental methods (see Roth (2002), Ariely et al. (2005)). In particular, laboratory experiments can be used to test the validity of various economic theories, to test hypotheses that cannot be investigated with the help of field data, and to test new market mechanisms that do not yet exist. In section B.I.3., we will demonstrate how a mixture of theoretical, empirical and experimental research can help to investigate the issues involved in designing online-auctions. One of the important lessons that can be learned from theoretical and empirical research is that there is generally no one-fits-it-all design. Every market environment is different and has its own particularities. Indeed, one aim of this article is to demonstrate that details can be very important; that is, even small design flaws can have a substantial impact on traders' behaviour and the performance of the market. For this purpose, we chose three examples from very different kinds of markets: 7 online auctions, electricity markets and partnership dissolution. B. The design of online auctions EBay is an attractive area to study market design. Part of the reason is ebay's enormous success. In 2005, more than 180 million users were registered on ebay (of which 72 million are 2

actively bidding), who offered about 1.9 billion items, leading to consolidated net revenues of $4.6 billion. Furthermore, online auctions (as those conducted by ebay) have precise and simple rules. This greatly facilitates the empirical and theoretical analysis as the complexity of strategic decisions of market participants is limited, and ambiguity with respect to implicit rules is eliminated or at least low. The huge amount of transactions and the huge amount of data readily available in electronic form further facilitates the empirical analysis. As a result, the literature on online auctions (unlike the literature in many other areas of empirical Industrial Organization) has quickly led to robust insights based on a mixture of empirical, experimental and theoretical research, and (single-unit) online auctions have become one of the best understood subjects in market design research. B.I. Controlling the pace of electronic auctions 8 B.I.1. Basic issues Unlike offline auctions, which typically last only a few minutes, Internet auctions such as those on ebay, Yahoo and Amazon run many days. This reflects that the Internet allows bidders to bid from basically everywhere at any time, and that a short auction would give less bidders a chance to spot the object and to enter the auction. Lucking-Reiley et al. (1999) and Hasker et al. (2004) observed that longer auction durations on ebay tend to attract more bidders and lead to higher prices. E.g., Lucking-Reiley et al. reported that prices in 7-day auctions are approximately 24 percent higher than in shorter auctions, and in 10-day auctions they are 42 percent higher, on average. However, long durations also create challenges because bidders cannot be expected to always keep an eye on the auctions as they unfold. Many auction houses including ebay respond to this by providing bidders with artificial proxy agents. These agents bid on the bidders' behalf, automatically responding as other bids come in, and thus free bidders from the necessity to follow the auctions and the price discovery process themselves. A related design question is whether the auction should be conducted sealed-bid. That is, bidders may submit their (proxy) bids over an extended period of time - but without the opportunity to react to the bidding activity of other human or proxy bidders. Or should bidding be open, so that bidders can see how others' bidding activities evolve during the course of auction? This way, bidders would retain the right to change (proxy) bids in response to the bid 3

history. Many (but not all) online auction houses, such as ebay, chose an open format. From a theoretical point of view, open ascending-price auctions tend to generate more revenue for the seller. Milgrom and Weber (1981) show that competitors' bidding activities may convey relevant information that the bidders can then use in revising their estimates of value. This translates into more aggressive bidding behaviour and higher prices. 9 Another argument for open auction formats comes from behavioral economics. It has been shown in laboratory experiments that the observability of bidding activity in open secondprice auctions substantially accelerates the speed of learning the optimal bidding strategy compared to second-price sealed-bid auctions (Ariely et al. (2005)). This improves the price discovery process and increases competition among bidders so that efficiency and revenue can be enhanced, even in purely private-value environments. In line with this finding, Ivanova-Stenzel and Salmon (2004) reported that, when having the choice between sealed-bid and open, ascending-bid auctions, laboratory subjects in a private-value environment have a strong preference for the open format. However, there are also disadvantages from open bidding: open auctions can lead to lower revenues when bidders are risk-averse, and when ex ante asymmetries among bidders are strong or competition is weak (e.g., Cramton (1998)). Furthermore, open auctions facilitate collusion among bidders which can result in lower revenue (see Klemperer (2004)). This might be part of the reason why ebay in the U.S. recently introduced a sealed-bid format as an option for sellers; in the best offer format, bidders can make sealed-bids, and sellers can accept any bid at any time they wish. B.I.2. The timing of bids Open auctions raise the question of the optimal timing of the bid. Many researchers found that bids on ebay, where auctions run typically for a week, often arrive very near to the closing time -- a practice called sniping. For instance, in the sample of computer and antiques auctions with at least two bidders, Roth and Ockenfels (2002) found that about 50 percent of all auctions still have bids in the last five minutes, 37 percent in the last one minute, and still 12 percent in the last 10 seconds. 10 At first glance, last-minute bidding of this sort cannot easily be reconciled with economic theory, because, as explained, ebay makes available a software bidding agent, called proxy bidding, to make bidding simple for bidders without having to 4

be constantly vigilant or to be online at the close of the auction. A bidder can submit a proxy bid and the bidding agent then bids on behalf of the potential buyer up to the proxy bid. As a consequence, not the last bid (as in ascending-price auctions) but the highest bid wins, regardless of submission time. (And the winner pays a small increment above the second highest submitted proxy bid, which is why this is a second-price auction.) Furthermore, there is a risk involved in late bidding in online auctions. Because the time it takes to place a bid may vary considerably due to erratic internet traffic or connection times, last-minute bids have a positive probability of coming in too late, after the close of the auction. EBay explains the simple economics of second-price auctions and the risks involved in late-bidding and comes to the conclusion: ebay always recommends bidding the absolute maximum that one is willing to pay for an item early in the auction. ( ) If someone does outbid you toward the last minutes of an auction, it may feel unfair, but if you had bid your maximum amount up front and let the Proxy Bidding system work for you, the outcome would not be based on time. However, Ockenfels and Roth (2006) demonstrated within an auction theoretic model that sniping on ebay can be a best response to a variety of strategies. In particular, inexperienced, naïve bidders might make an analogy with (open) ascending-price auctions, and be prepared to continually raise their bids to maintain their status as high bidder. In an ebay style auction that closes at a predetermined deadline ( hard close ), bidding very late might be a best response to incremental bidding (or multiple bidding) of this sort. That is, bidding very near the end of the auction would not give the incremental bidder sufficient time to respond, and so a sniper competing with an incremental bidder might win the auction at the incremental bidder's initial, low bid. In contrast, bidding one's value early in the auction, when an incremental bidder is present, would win the auction only if one's value were higher than the incremental bidder's, and in that case would have to pay the incremental bidder's value. 11 The evidence in the laboratory and the field indicates that incremental bidding is common, and that sniping is likely to arise in part as a response to incremental bidding. Wilcox (2000) indicated that 1.5 to 2 bids are submitted by the average bidder. Ockenfels and Roth (2006) reported that 38 percent of the bidders bid at least twice. Furthermore, the number of bids per bidder is increasing in the number of other active bidders who bid multiple times, suggesting that incremental bidding may induce bidding wars with like-minded incremental bidders. 5

There are other ways to explain late and incremental bidding without positing inexperience on the part of the bidders, including implicit collusion (Roth and Ockenfels (2002)) and common-value aspects of the item being sold. 12 Some observers of ebay believe that the amount of sniping will decrease over time because it is mainly due to inexperience and unfamiliarity with ebay's proxy bidding system. This section showed, however, that there are a variety of rational, strategic reasons for sniping. In fact Wilcox (2000), Roth and Ockenfels (2002), Wintr (2004), and Ariely et al. (2005) observed both in laboratory and field studies that more experienced bidders snipe more often than less experienced bidders. Thus, it seems likely that late bidding will remain a persistent phenomenon on ebay as long as the auction rules remain unchanged. B.I.3. Rule for ending electronic auctions On ebay, auctions end at a predetermined time: a hard close. Amazon, on the other hand, emulates the Going, Going, Gone feature of traditional auction houses and employs a soft close. That is, Amazon automatically extends an auction if a bid comes in late, so that all bidders always have the opportunity to respond to the opponents' bids. Ockenfels and Roth (2006) show that, although the risks of last-minute bidding remain, the strategic advantages of last-minute bidding are eliminated or severely attenuated in Amazon-style auctions. That is, a bidder who waits to bid until the last seconds of the auction still runs the risk that his bid will not successfully be transmitted in time. However, if his bid is successfully transmitted, the auction will be extended for ten minutes, so that, no matter how late the bid was placed, other bidders will have time to respond. Thus on Amazon, an attentive incremental bidder, for example, can respond whenever a bid is placed. The differences in the strategic environment are reflected in the data: there is significantly more late bidding on ebay than on Amazon. For instance, in the sample of Roth and Ockenfels (2002), 40 percent of ebay-computers auctions and 59 percent of ebay-antiques auctions have last bids in the last 5 minutes, compared to about 3 percent of both Amazon computer and Amazon antiques auctions that have last bids in the last five minutes before the initially scheduled deadline or later. Further analyses reveal that while the impact of the bidders' feedback numbers on late bidding is significantly positive in ebay, it is negative in Amazon, suggesting that more experienced bidders on ebay bid later than less experienced bidders, but experience in Amazon has the opposite effect. 6

Experiments by Ariely et al. (2005) replicate these findings in a controlled laboratory privatevalue setting in which the only difference between auctions is the ending rule. The experiment thus controls for differences other than the closing rule that might affect behaviour on Amazon and ebay, such as the number of auctions being conducted at a time and the number of potential bidders. Controlled field experiments, on the other hand, seem to have more difficulties to find evidence for the impact of the ending rule. Brown and Morgan (2005) and Houser and Wooders (forthcoming) took advantage of the fact that Yahoo sellers are allowed to choose whether to end the auction with a hard or a soft close. In both studies, identical items were sold using both ending rules. None of these studies found a significant effect of the ending rule on the amount of late bidding. However, Houser and Wooders (forthcoming) observed that, ceteris paribus, hard-close auctions tend to raise somewhat less revenue than soft-close auctions (as it was found in Ariely et al. (2005)). Online market design includes the design of artificial agents, such as ebay's proxy bidder (Ockenfels and Roth (2002)). Because late-bidding involves a good deal of planning and effort, artificial agents can also help executing late-bidding strategies. In fact, there is a market for artificial sniping agents (like esnipe.com) that will allow a bidder not only to submit a proxy bid, but to do so at the last moment. Note that to the extent sniping becomes more widespread on ebay, ebay will be gradually transformed into a sealed-bid second-price auction. If a large part of the late bidding activity takes place on third party sites like esnipe.com, ebay faces a number of design and rule choices. One is to ban sniping services. In fact, ebay.de (Germany) banned third party sniping services in its general terms and conditions (which is, of course, difficult to enforce). A second choice would be just the opposite: recapturing the sniping market by offering a sniping option on ebay itself. Under this option, last minute bids submitted in advance directly to ebay could all be counted at the same time, immediately after the auction close. This would give bidders certainty both that their bids would be successfully transmitted and that there would be no time for other bidders to react. Of course, if all bidders used this option, the auction would be precisely a sealed bid auction. As we have argued above, ebay might prefer not to encourage this development towards sealed-bid bidding. Yet, ebay supports its own sniping service, which enables last-minute bidding via phone. While bidding by phone will still 7

involve risks that the bid fails to be successfully submitted, it probably further increases the number of snipes. Finally, ebay could consider changing the ending rule of the auction to a soft close. Such a change, however, may also cause adverse effects. Sniping cannot only help bidders to get better prices on ebay, but sellers too can profit from sniping, because the possibility of sniping may attract more bidders. For instance, sniping can lead to more bidding from experts in auction environments with information externalities, because by bidding late, experts can avoid giving information to others through their own early bids. Sniping can also increase the excitement and entertainment value of bidding, which also can attract more bidders. B.II. Multi-unit auctions Most of the auctions on ebay and other C2C online market places are single-unit auctions, i.e., a seller sells only one item per auction. Obviously, a seller with a lot of (maybe identical) objects for sale can use such auctions and sell the items in separate auctions (either simultaneously or sequentially). But this is inconvenient as the seller needs to initiate and monitor many auctions. 13 Due to the increase of B2C sales and the associated demand for market designs that allow selling multiple units in a single auction, auction platforms started to offer multi-unit auctions. 14 The analysis of multi-unit auctions is considerably more difficult and the theoretical as well as empirical literature is less developed. Part of the reason is that, when objects are heterogeneous or bidders demand multiple items, new difficulties arise such as market power and strategic and computational complexities. In addition there are many more design-alternatives, and general insights are difficult to come by. 15 In particular, the sale of complementary objects typically poses non-trivial challenges, because efficient auctions have to allow bids on bundles of objects. For example, the celebrated Vickrey auction requires that bidders specify their preferences on any possible package of the N objects, i.e., each bidder has to submit 2 N -1 numbers. Especially for a large number of objects such an auction is infeasible. 16 Fortunately, multi-unit auction design is considerably simpler if each bidder just demands one object (or, more generally, if objects are substitutes). Under this unit-demand assumption, ebay's single-unit auction format has a natural extension: every bidder submits a (proxy) bid and the N highest bidders win one object each and pay a price that is an increment above the N+1'th highest (proxy) bid, i.e., the price is (almost) equal to the highest losing bid. In ebay's multi-unit auction 17, however, the final price is equal to the lowest winning bid. Consequently, one of the winners (the bidder with the N th highest bid) will determine the 8

price. If the N th highest bid is different from the N+1 th highest bid (e.g., because bidders bid valuations and not incrementally), each of the winning bidders can reduce the price by lowering his bid to an amount just above the highest losing bid. That is, the auction creates incentive for bid shading. But because the highest losing bid is usually not known before the auction is over, bidding is difficult. These kinds of arguments convinced ebay Germany to change their multi-unit format in summer 2005 to the more natural format discussed above. However, in a more general auction environment, neither ebay's old nor the new multi-unit format share the desirable properties of the second-price auction in the single-object case. The reason is that if a bidder can demand more than one unit in ebay's multi-unit format, there is a positive probability that his bid on a second or later unit will be pivotal, thus determining the price for the first and possibly other inframarginal units. With discrete goods, the bidder will thus in equilibrium bid his true value on the first unit demanded, but strictly less on all subsequent units. As a consequence, in equilibrium, bidders understate their values, or (equivalently) reduce demanded quantities, hampering revenue and efficiency. 18 Furthermore, efficiency cannot be expected in case of complementarities; that is, when the value of a bundle of objects is larger than the sum of values of each object separately. This may not only be the case on ebay but also in many applications including spectrum auctions, electricity auctions, airport landing slots auctions, supply chains, and transportation services auctions. In such cases, a bidder may end up stuck with objects that are worth little because he failed to win complementary objects (exposure problem), and may quit early because of fear of this. As a result, inefficiencies are likely to arise in all auction formats, in which bidders cannot make sure to purchase desired packages of objects. The way ebay addresses these problems is to allow bidders, who receive less objects than demanded, to withdraw the bid for all units. Allowing bid withdrawal, however, raises further design questions. For instance, what is the price to be paid by the winners when, after the auction, a bid is withdrawn? If the withdrawn bid counts (which is the case on ebay), the winner may rightly ask why he has to pay the higher price even though there is, after the withdrawal, no competition that justifies the higher price. We believe that designing a robust multi-unit auction that can account for even complex preferences of the bidders into account is an important challenge to online auctions. The changes on ebay s German platform go into the right direction. More radical changes could involve, 9

for instance, a declining buy-it-now price: the price decreases from a high initial price and then declines by a predetermined decrement at predetermined times so that a bidder can indicate how many items he is willing to buy at the current price. The auction ends when all objects are gone (or when the auction runs out of time). This auction format would combine ebay's successful buy-it-now format, which is often used to sell multiple units, with the merits of an auction. As bidders who demand a certain amount of objects know at any time whether they will be able to win that bundle, one may expect bids (and revenue) to be higher than in ebay's multi-unit auction, where, due to the exposure problem, bidders might be reluctant to bid. B.III. Reputation systems Moral hazard problems seriously challenge the performance of online markets. Traders are usually anonymous and geographically dispersed, exchange is done sequentially, and the quality of objects is difficult to verify. 19 In addition, the huge number of transactions, the lack of repeated interaction with the same trading partner 20 and the possibility to create new identities without cost aggravates the problem of fraud in online markets. In fact, a report by the research group GartnerG2 (2002) concludes that Internet transaction fraud is 12 times higher than in-store fraud. 21 In markets with few, well known traders reliance on legal enforcement might be sufficient (this is probably the case in energy markets, see section C.), whereas almost all online market platforms incorporate additional measures that encourage bidders to sanction fraudulent traders. In particular, ebay responds to this challenge with a mixture of policies and mechanisms. 22 In this section, we describe ebay's most visible mechanism, the reputation system, from a market design perspective. On ebay, after each encounter, buyers and sellers evaluate each other by giving the trading partner either a positive (+1), neutral (0) or negative (-1) feedback (and maybe an additional verbal commentary). These feedbacks are publicly available and easy to access such that each buyer can look at a seller s feedback history before he engages in bidding. This way, ebay's feedback mechanism can weaken the incentives for moral hazard. If ebay traders punish sellers with negative feedback information by refusing to buy from them or reducing the price they are willing to pay, then the threat of leaving negative feedback may discipline the 10

seller. 23 However, two points are central for the functioning of online reputation systems. First, they must generate reliable information about market participants' trading patterns. Second, they must evoke the desired reactions of traders to this information. In online reputation systems, feedback information must come from voluntary self-reporting of personal experiences. But feedback constitutes a public good; the benefit from giving feedback goes to other traders, and nobody can be excluded from using this feedback information. Furthermore, there is a kind of information externality, as observed by Bolton et al. (2004) in an experiment, and theoretically discussed in Bolton and Ockenfels (2006). As an illustration, observe that the risk of fraud is highest for trades with newbies, i.e. for trades with a seller, who has not been evaluated yet. So, buyers may only be willing to trust sellers, who have proven to be reputable, but then nobody would be willing to generate this information. However, despite these various free-riding problems, about 50 percent of the transactions on ebay receive feedback (Resnick and Zeckhauser (2002)). One of the main motives for providing the public good and giving feedback appears to be the expectation that a positive feedback is reciprocated by the trading partner. 24 It also rationalizes ebay's decision to allow bilateral evaluations (buyers can leave feedback on sellers and vice versa). But, by the same token, bilateral feedback might also discourage negative feedback if negative feedback is repaid by negative feedback. 25 Thus, reciprocal feedback may yield more feedback giving at the cost of lower information value. To avoid retaliatory feedback, some observers recommend changes in the rules of ebay's reputation system. Güth et al. (2006) recommend that buyers only evaluate sellers, and not vice versa, because there is typically no adverse selection or moral hazard issue on the buyer side. Alternatively, they as well as Klein et al. (2005) recommend a blind period, in which both trading partners can leave feedback on each other, which however is not revealed before the end of the blind period. Other more sophisticated proposed rule changes include an idea by Jurca and Faltings (2004), who propose that traders simultaneously rate the transaction. If the reports disagree, one of the traders must be lying, and both parties have to pay fines. Obviously, this simple game has a truth-telling equilibrium. Alternatively, Miller et al. (forthcoming) propose a mechanism based on scoring rules. The underlying idea is that a feedback giver is paid according to the extent the feedback predicts future seller performance. Under certain assumptions, a proper scoring rule also induces truth-telling behaviour. 11

Another source of noisy and manipulative feedback information is the possibility of identity change. The costs of changing an online trader identity is often close to zero, implying that fraudulent sellers can exploit their buyers and then reappear with a clean record. Ockenfels (2003) shows in a theoretical model, however, that a feedback mechanism can work, even if sellers with a bad history can change their identity and start again as a newbie. If buyers are only willing to buy from a newbie if he offers a lower price (as compared to a seller with a good history) then starting again as a newbie can be more costly than avoiding to get a bad reputation in the beginning. The empirical evidence in Ockenfels (2003) and in the studies cited in Bajari and Hortaçsu (2004) show that indeed newbies offer goods at a price discount. But market design can also play a decisive role, because the change of online identities can be made more costly by design. One technical way would be the use of modern authentication technologies. A more economic design approach would be to implement an entrance fee ; every trader who wants to sell on the market platform has to pay a fee that needs to be sufficiently high to deter fraudulent strategies of identity changes (Friedman and Resnick (2001)). To see how market participants react to provided feedback information, Bolton et al. (2004), conducted an experiment and compare trading in a market with online feedback to a market without feedback. 26 They find that the feedback mechanism induces quite a substantial improvement in transaction efficiency and buyers respond to feedback information in a strategic way. If a seller shipped the last order, a buyer is twice as likely to trust him. 27 They also find that a trader's behaviour is influenced by own experience: a buyer who has been treated badly in the past is more likely to mistrust other sellers - regardless of their reputation. Furthermore, buyers put more weight on negative than positive feedback, and more on recent than old feedback (see also Lucking-Reiley et al. (1999), and Resnick and Zeckhauser (2002), and Keser (2002)). These findings suggest, among other things, that market performance can be increased if market participants are provided with summary statistics about overall feedback. This might mitigate the negative effect of own bad experience and hides information that affects buyers' decisions (in non-desirable way). But they also note that a problem with relying solely on cumulative feedback measures comes from the fact that the seller's actions have diminishing impact on the overall assessment. This may lead to increasing incentives to exploit one's good reputation. 28 12

C. The design of electricity markets The liberalization of energy markets created a new interest in the design of electricity markets, though the theoretical and empirical literature is still in its infancy. One reason is the complexity of electricity markets, which involve many layers of intertwined trading institutions, and require sound knowledge of the technical constraints related to energy trading. Another reason is that electricity trading rules widely differ across the nations, as do the political, environmental and economic conditions. Because of space restrictions, we will suppress many complicating aspects here (see Müsgens and Ockenfels (2006) and the references cited therein for a more detailed description of market design issues). Our goal is to demonstrate that the microstructure of the market rules can be decisive for the market performance. Because electricity markets are organised as auctions 29 most of the design-issues discussed in this section have relevance beyond electricity markets. In particular questions concerning the timing of (interlinked) auctions and problems that arise because of the multi-object nature of electricity auctions are also relevant for online market design (and market design in general). Electricity markets are, in some sense, inherently fragile. Energy cannot be stored at reasonable costs so that the market, which suffers from stochastic supply and demand, has to be balanced in real time. As a result, equilibrium prices are extremely volatile. Furthermore, shortrun demand is inelastic, as is supply near the capacity limits, and transmission of electricity is often constrained. Therefore the market can be prone to the exercise of market power (which may be amplified by high entry costs and asymmetric information). And finally, electricity markets consists of several (strategically connected) sub-markets, which interact in various ways. 30 On forward markets, future contracts (and other derivatives) are traded (long) before the physical exchange of electricity takes place. This gives suppliers and buyers the possibility to lock in prices and reduce risks. On the day-ahead market electricity is traded for each of the 24 hours of the next day (in Germany, this is done on the Electricity Exchange, EEX, in Leipzig). Capacity and reserve markets (which typically take place immediately after and sometimes immediately before the day-ahead market) should help to improve reliability by mitigating the effects of unanticipated supply and demand shocks. Suppliers offer capacity as reserves which, if purchased and called for, are activated to balance supply and demand in real time to avoid outages. And finally, the markets for transmission (in particular interconnector) capacity needs to be cleared. 31 13

All four markets are obviously closely intertwined and an isolated consideration can easily be misleading. Indeed, already the establishment of the markets, as well as the timing and the detailed rules of each of the markets are the result of conscious market design considerations. Regarding the establishment, observe that California decided in the nineties not to allow forward markets in order to trigger a liquid and thick day-ahead market. This design policy - as has been thought - would weaken problems of market power, because all capacity is bid in the day-ahead market, and thus eventually yield low prices. However, in fact, the opposite is true; forward markets not only reduce price risks for both demand and supply, but also tend to make the day-ahead market less vulnerable to manipulation. For instance, a utility that has sold 100 percent of the capacity over the forward market has no incentive to influence dayahead prices. Also, it is profitable for each seller to commit himself to larger quantity in forward markets (not much unlike in the common Stackelberg model, see Allaz and Vila (1993)), which however, in the aggregate, leads to larger quantities and thus smaller profits for all sellers (e.g., Bolle (1993)). Finally, if buyers cannot lock-in prices long before consumption, generators can better exploit the fact that close to real-time demand is very inelastic. Empirical work strongly supports the view that forward markets tend to reduce electricity prices. For instance, it has been forcefully argued that banning forward markets in California was essential for the energy crisis in 2000 and 2001 to emerge (e.g., Borenstein et al. (2002)). The empirical study conducted by Wolak (2000) provides strong evidence that forward markets may mitigate market power. Experiments under controlled conditions, without the complicating aspects that prevent unambiguous interpretation of the field data, come to the same conclusion (see Brandts et al. (2003), see also Rassenti et al. (2002) for a discussion about how experiments may help shaping the deregulation efforts in energy markets). Once the existence of markets is established, the timing becomes the next crucial market design issue. Because different markets may trade complements or substitutes, an inappropriate timing of the markets imposes serious strategic complexities and coordination problems for bidders. They also limit the information available to bidders, and how bidders can respond to this information, in particular in the presence of complementarities. Bidders must guess future prices, and incorrect guesses hamper revenue and efficiency. For example, Dutch-German interconnector (transmission) capacities are often auctioned off before the corresponding electricity markets clear. However, electricity and transmission capacity are obviously complements, and bidding in one market depends on the outcomes in the other market. As a result of this uncertainty, the exposure problem yields coordination problems and result in large ineffi- 14

ciencies. 32 Other European interconnectors (e.g., NordPool) solved the problem more efficiently by implementing what is sometimes called an implicit auction, where transmission capacity is managed implicitly by two or more neighbouring spot markets. Here, network users submit bids for electricity in the geographical zone where they wish to generate or consume, and the market clearing procedure determines the most efficient amount and direction of physical power exchange between the market zones. Hence, cross border capacity and energy are traded together. An even more comprehensive market would also include reserve markets, because the timing of electricity reserve auctions, which are partly cleared before and after the day-ahead market, pose similar problems. From a generator's point of view, selling electricity to the day-ahead market or the reserve market are substitutes; capacity that is sold in the day-ahead market cannot be sold anymore in the reserve market. An obvious market design solution to such timing problem is the simultaneous clearing of all these markets. A timing problem also occurred in capacity markets. 33 Take the capacity market used in New England in the nineties (Cramton (2003)). This market was intended to assure that there is sufficient capacity in the system to cover the peak load plus the reserve margin. Generators could submit bids expressing their willingness to supply capacity, and a market clearing price was calculated. Eventually, generators with bids below the price provided the capacity at the clearing price. However, the price for capacity is determined day-ahead - at a time where the cost of offering capacity is sunk (as investments in increasing capacity are long-term). So, absence of market power, the only rational bid is zero. With market power, the price is arbitrarily high. Obviously, the market design failed to promote efficient capacity investments. Given the timing of markets, the market rules need to be designed. Let us, for example, take a look at the details of the bidding formats and rules in the day-ahead market. Should bids be binding, should there be a bid cap, and what costs are bids allowed to reflect? In day-ahead markets participants submit their bids one day ahead, for every hour of that day. Such a bid describes a bidders supply or demand function, and is then used to derive a supply schedule and a clearing price for every hour of the following day. In some markets, however, bidders could change their bids without any penalty until the electricity is physically exchanged, because suppliers sometimes have to deal with unpredictable outtakes. But this can lead to gaming in a way that suppliers initially submit bids to win contracts and then exploit the market 15

power they have close to real time when demand becomes price-inelastic. One solution to this dilemma is to allow bid changes, but to impose costs when initial bids are changed (see Milgrom (2004) who describes a similar problem and solution in spectrum auction design). A similar problem to that of non-binding bids occurred in the day-ahead markets of California. To prevent exaggerated prices the Californian day-ahead market did not allow for prices above a certain, pre-determined level. But this rule turned out to be ineffective as bidders could avoid the day-ahead market and offer their electricity on the reserve market, which did not restrict prices by such a bid-cap. As the bid cap could not be enforced (in the sense that it could easily be bypassed), it was ineffective and eventually given up. Day-ahead markets also suffer from the exposure problem, if bidders cannot adequately express their preferences. The per-hour cost of electricity crucially depends on the total running time of the generator as suppliers face substantial start-up and no-load costs (Müsgens and Neuhoff (2005)). As in the day-ahead auction bidders submit demand and supply functions for every hour of a day but cannot express their preferences for sequences of hours, they might be reluctant to bid. Similar to a buyer who demands several units on an ebay multi-unit auction and who fears winning too few objects (making it impossible to realize synergies), a supplier in the day-ahead market is afraid of not winning subsequent hours. Therefore, electricity may not be generated at lowest costs. The exposure problem can be avoided if suppliers can express their complex cost structure in their bids. Usually this implies that they have to be able to submit supply functions for all possible combinations of hours of a day. As this is hardly feasible, Cramton (2003) suggests that in addition to a supply function suppliers submit their start-up and no-load costs, as this sufficiently describes their production costs. 34 D. The design of partnership dissolution mechanisms The market discussed in this section differs from those in previous sections. First, the number of participants is low and second, seller(s) and buyer(s) are determined endogenously by the market rules. The following discussion of partnership dissolution mechanisms does not merely show (once more) that details matter but also provides an example for how marketdesign techniques can be applied to a less conventional market. When a business partnership or joint venture comes to an end, partners have to decide what to do with their jointly owned business. 35 Under the assumption that the partnership is profitable 16

it is more sensible for one partner to buy out the other or to sell the entire business to a third party and realize the profits rather than to liquidate. In many instances assets may be less valuable to a third party than to either partner and thus a buy-out should be preferred to a sale. 36 In a buy-out partners essentially have to determine who should become sole owner of the partnership and how leaving partners should be compensated. To obtain an efficient allocation, 37 the shares of the partnership should be traded on a market where the partners are the market participants. A particularity of this market is that buyers and sellers are endogenously determined by market interactions, as it is possible that a certain partner will either end up selling his shares or buying-out the other partners (but does not know whether he will be buying or selling at the time the market opens). In this section, we want to demonstrate how market design matters for such a dissolution market. In particular we once more want to point out the importance of seemingly small design details and how rules affect individual behaviour. Rules that govern the division of a commonly owned object are already mentioned in the Old Testament [Genesis 13: 8-9] and Hesiod's Theogony. The rules described there are known as the cake-cutting mechanism or the divide-and-choose mechanism, and work as follows. One of two partners has to divide a commonly owned asset (e.g., a cake), and the other partner then decides which share he wants to possess. 38 The divide-and-choose mechanism is also applicable to partnership dissolution, though it needs to be modified as a partnership (usually) is an indivisible business entity. Such a modification of the divide-and-choose mechanism is the so called buy-sell clause or shoot-out mechanism. Assume that each of two partners owns 50% in a partnership. 39 The buy-sell clause states that one partner (the proposer) makes a price-offer and the other partner (the chooser) decides on whether to sell her share or buy the other partner's share at this price. This buy-sell clause is the prevalent dissolution procedure in Anglo-Saxon corporate law and is recommended by corporate lawyers and included in most partnership agreements (see Brooks and Spier (2004), Stedman and Jones (1990) and Cadman (2004)). It is important to stress that the buy-sell clause becomes effective when the partnership needs to be dissolved because of a contingency like a dispute. The performance of the clause can be measured in terms of efficiency, i.e. it should allocate the partnership to the partner who values it most (e.g., because he is most capable of running the business) at the time of the dissolution. 40 Its broad application suggests that the buy-sell clause performs well. Nevertheless, de Frutos and Kittsteiner (2004) cite cases in which the clause seems to have failed. The origin for these 17

failures can be found in the fact that partners prefer to be chooser rather than proposer. In fact, in most theoretical models with asymmetric information (about the value of the partnership), a partner is better off choosing than proposing in the buy-sell clause. 41 This is because a proposer does not know whether eventually he will buy or sell at the proposed price, which makes it difficult for him to exploit his private information on his valuation for the partnership. A chooser does not have this disadvantage. He can decide on whether to buy or sell at the proposed price. This shows that the determination of the proposer is an essential part of the dissolution process and the buy-sell clause has to be very specific on this matter and needs to be carefully designed to avoid disputes and inefficiencies. Examples cited in de Frutos and Kittsteiner (2004) indicate that the dissolution has been problematic in cases where it was specified that the partner who initiates the dissolution has to propose the price. 42 This is because both partners (who are in dispute) want to dissolve but no one wants to make the (initial) price offer (as this is disadvantageous). Consequently, the partners engage in a war of attrition and wait for the other to move first. This can lead to an inefficiently long dispute on who has to propose which may even end up before court. 43 De Frutos and Kittsteiner show that a different rule that governs the determination of the chooser can circumvent these problems. They show that if partners negotiate on who has to propose, the buy-sell clause can perform better than in the case where partner's just wait for the other to come up with a price. In their theoretical model they demonstrate that the buy-sell clause where the proposer is determined endogenously by negotiation is indeed the best possible dissolution rule (as it is efficient). They also provide a case that suggests that a partnership dissolution where the proposer has been determined by negotiations worked flawlessly. 44 Here again we find that a seemingly small difference in the rules (how the proposer is determined) can make all the difference with respect to the performance of an entire market (which here governs the dissolution). But also other designs for partnership dissolution have been proposed. Cramton et al. (1987) suggest a simple auction where both partners 45 simultaneously submit a bid. The entire partnership is awarded to the higher bidder who pays an amount proportional to a convexcombination of the two bids. 46 Cramton et al. show that (under their assumptions) this auction, like the buy-sell clause with negotiations, is an optimal dissolution mechanism. 47 Even though the auction might seem a more natural design for a dissolution mechanism it is hardly recommended by lawyers and therefore absent in many partnership agreements. 48 Whereas de Frutos and Kittsteiner can rationalize the use of the buy-sell clause (if used with the correct 18

rule for selecting the proposer), an explanation of its pre-dominance with respect to auctions still requires further theoretical and empirical/experimental research (see Kittsteiner et al. (2006)). It should be noted that both mechanisms, the buy-sell clause and the auction, can fail to deliver an efficient dissolution if partners do not know the exact value of the partnership, as this depends on information only available to the other partner, and in addition partners only want to dissolve if the gains from trade that are associated with dissolution are large. Kittsteiner (2003) shows that the auction might result in an outcome where partners stay together even though a dissolution is desirable (i.e., it generates higher total profits). Fieseler et al. (2003) show that any market design for environments where the firms value depends on all partner's private information, exhibits this problem that partners stay together even though they should not. This implies that the buy-sell clause cannot be efficient in such environments either. The optimal market design for such a situation is still unknown. 49 We believe that the design of negotiation procedures, such as in partnership dissolution but also in online bargaining etc., is one of the most promising research areas in economic design. However, for significant progress to be made, more experimental and empirical work is needed. E. Conclusion Institutions matter. All our examples demonstrate that the design of rules and procedures significantly affects market outcomes. Even small design changes can have substantial consequences. But behaviour matters too. Experimental and field research shows that decision makers and traders respond to incentives. They do not always do so in a rational way, as predicted by standard theory, but they typically do so in a systematic and predictable way. This opens the door for market design as an engineering science (Roth (2002)), in which various theoretical and empirical tools interact to improve the performance of markets and negotiations. We think that the recent examples of market design, some of which are discussed and cited in this article, are only the beginning of a fruitful and exciting engineering literature in business and economics. 19

Literature Anandaligam, G., Day, R. and S. Raghavan (2005): The Landscape of Electronic Market Design, Management Science 51, 316-327. Allaz, B. and J.-L. Vila (1993): Cournot Competition, Forward Markets and Efficiency, Journal of Economic Theory, 59, 1-16. Anwar,S., McMillan, R. and M. Zheng (2004): Bidding Behavior in Competing Auctions: Evidence from ebay, Working Paper, Economics Working Paper Archive. Ariely, D., Ockenfels, A., and A.E. Roth (2005): An Experimental Analysis of Ending Rules in Internet Auctions, The RAND Journal of Economics, 36, 790-809 Barbaro, S. and B. Bracht (2005):. Shilling, Squeezing, Sniping: Explaining Late Bidding in Online Second-price Auctions, Working paper, University of Mainz, Germany. Bajari, P. and A. Hortaçsu (2003): Winner's Curse, Reserve Prices and Endogenous Entry: Empirical Insights from ebay Auctions, Rand Journal of Economics, 2, 329-355. Bajari, P. and A. Hortaçsu (2004): Economic Insights from Internet Auctions, Journal of Economic Literature XLII, 457-486. Bolle, F.(1993): Who Profits from Futures Markets?, Ifo-Studien, 3-4, 239-256. Bolton, G. E., Katok, A.E. and A. Ockenfels (2004): How Effective are Online Reputation Mechanisms? An Experimental Investigation, Management Science, 50, 1587-1602. Bolton, G., Katok, E. and A. Ockenfels (forthcoming): Bridging the Trust Gap in Electronic Markets. A Strategic Framework for Empirical Study, In: E Akcali, J. Geunes, P.M. Pardalos, H.E. Romeijn and Z.J. Shen (eds.), Applicatons of Supply Chain Management and E-Commerce Research in Industry, Kluwer Academic Publishers. Bolton, G., C. Loebbecke, and A. Ockenfels (2006): How Social Reputation Networks Interact with Competition in Anonymous Online Trading: An Experimental Study, Working paper, University of Cologne. Bolton, G. and A. Ockenfels (2006): The Influence of Information Externalities on the Value of Reputation Building. An Experiment, Working Paper, University of Cologne. Borenstein, S., J. Bushnell and F. Wolak (2002): Measuring Market Inefficiencies in California's Wholesale Electricity Industry., American Economic Review, 92, 1376-1405. Brams, S.J. and A. Taylor (1006): Fair Division: From Cake-Cutting to Dispute Resolution, Cambridge University Press. Brandts, J., Pezanis-Christou, P. and A. Schram (2003): Competition with Forward Contracts: A Laboratory Analysis Motivated by Electricity Market Design, Working paper, Institute for Economic Analysis, Barcelona. 20

Brooks, R, and K.E. Spier (2004): Trigger happy or gun shy? Dissolving common-value partnerships with texas shootouts, Kellogg School of Management, Working Paper. Brosig, J. (2006): Communication Channels and Induced Behavior, Zeitschrift für Betriebswirtschaft, this issue. Brosig, J., A. Ockenfels, and J. Weimann (2002): The Effect of Communication Media on Cooperation. German Economic Review, 4(2), 217-241. Brown, J. and J. Morgan (2005): Reputation in Online Markets: Some Negative Feedback, Working Paper, University of California, Berkeley. Cadman, J. (2004): Shareholders' Agreements, 4th edition, Thomson. Cassady, R. Jr. (1967): Auctions and Auctioneering, Berkely and Los Angeles, CA: University of California Press. Cramton, P. (1995): Money Out of Thin Air: The Nationwide Narrowband PCS Auction, Journal of Economics and Management Strategy, 4, 267-343. Cramton, P. (1998): Ascending Auctions, European Economic Review, 42, 745-756. Cramton, P. (2003): Electricity Market Design: The Good, the Bad, and the Ugly, Proceedings of the Hawaii International Conference on System Sciences, January 2003. Cramton P., Gibbons, R. and P. Klemperer (1987): Dissolving a Partnership Efficiently, Econometrica 55, 615-632. de Frutos, M., and T. Kittsteiner (2004): Efficient partnership dissolution under buy-sell clauses, Working Paper, University of Bonn. Dellarocas, C. (2003): The Digitization of Word-of-Mouth: Promise and Challenges of Online Reputation Mechanisms, Management Science, 49,1407--1424. Dellarocas, C. (forthcoming): Reputation Mechanisms, Handbook of Information Systems and Economics, forthcoming. Dellarocas, C., Fan, M. and C. A. Wood (2004): Self-Interest, Reciprocity, and Participation in Online Reputation Systems, Working Paper. de Vries, S. and R. Vohra (2003): Combinatorial Auctions: A Survey, INFORMS Journal on Computing 15, 284-309. Duffy, J. and U. Unver (2005): Internet Auctions with Artificial Adaptive Agents: A Study on Market Design Working Paper, Univeristy of Pittsburgh. Engelmann, D. and V. Grimm (2004): Bidding Behavior in Multi-Unit Auctions - An Experimental Investigation and some Theoretical Insights, Working Paper, Charles University, Center for Economic Research and Graduate Education. 21

Fieseler, K., Kittsteiner, T. and B. Moldovanu (2003): Partnerships, Lemons and Efficient Trade, Journal of Economic Theory 113, 223-234. Friedman, E. and P. Resnick (2001): The Social Cost of Cheap Pseudonyms, Journal of Economics and Management Strategy, 10, 173-199. GartnerG2 (2002): Online transaction fraud and prevention get more sophisticated, www.gartnerg2.com/rpt/rpt-0102--0013.asp. Gatzen, C., A. Ockenfels and M. Peek (2005): Sind die Gesetze des Wettbewerbs auf dem Strommarkt außer Kraft gesetzt? Analyse der Strompreisentwicklung auf dem Großhandelsmarkt in Deutschland, Energiewirtschaftliche Tagesfragen, 11, 4-11. Grimm, V., Riedel, F. and E. Wolfstetter (2003): Low Price Equilibrium in Multi-Unit Auctions:The GSM Spectrum Auction in Germany, International Journal of Industrial Organization 21, 2003, pp. 1557 1569 Grimm, V., and G. Zoettl (2005): Capacity Choice under Uncertainty: The Impact of Market Structure, Working paper, University of Magdeburg. Güth, W., Mengel, F. and A. Ockenfels (2006): The Dynamics of Trust and Trustworthiness on EBay. An Evolutionary Analysis of Buyer Insurance and Seller Reputation, Working paper, University of Cologne. Güth, W. and A. Ockenfels (2000): Evolutionary Norm Enforcement, Journal of Institutional and Theoretical Economics, 156, 335-347. Güth, W. and A. Ockenfels (2003): The Coevolution of Trust and Institutions in Anonymous and Non-anonymous Communities, In: M.J. Holler, H. Kliemt, D. Schmidtchen and M. Streit (Eds.), Jahrbuch für Neue Politische Ökonomie, 20, 157-174, Tübingen: Mohr Siebeck. Hasker, K., Gonzales, R. and R. C. Sickles (2004): An Analysis of Strategic Behavior and Consumer Surplus in ebay Auctions, Working Paper, Rice University. Hauswald, R and U. Hege (2003): Ownership and Control in Joint Ventures: Theory and Evidence, Discussion paper No. 4056, C.E.P.R. Hayne, S., Smith, C.A.P. and Vijayasarathy, L. (2002): Predicting Sniping in ebay Auctions, Working Paper, Colorado State University. Heyman, J., Orhun,Y. and D. Ariely (2004): Auction Fever: The Effect of Opponents and Quasi-Endowment on Product Valuations, Journal of Interactive Marketing, 18(4),7-21. Houser, D. and J. Wooders (forthcoming): Reputation in auctions: Theory, and evidence from ebay, Journal of Economics and Management Strategy. 22

Ivanova-Stenzel, R, and T. Salmon (2004): Bidder Preferences among Auction Institutions, Economic Inquiry, 42, 223-236. Jehiel, P. and B. Moldovanu (2001): Efficient Design with Interdependent Valuations, (with Philippe Jehiel), Econometrica 69(5), 1237-1259. Jehiel, P. and B. Moldovanu (forthcoming): Allocative and Informational Externalities in Auctions and Related Mechanisms, The Proceedings of the 9th World Congress of the Econometric Society, edited by Richard Blundell, Whitney Newey, and Torsten Persson, Cambridge University Press. Jin, G. Z. and A. Kato (2005): Quality and Reputation: Evidence from An Online Field Experiment, Working Paper, University of Maryland. Jurca and B. Faltings (2004): Eliciting Truthful Feedback for Binary Reputation Mechanisms, The International Conference on Web Intelligence (WI 2004), 214-220, Beijing, China. Kagel, J. H., and D. Levin (2001): The Winner's Curse and Public Information in Common Value Auctions, American Economic Review, 76, 849-920. Kagel, J. H. and A.E. Roth (2000): The dynamics of reorganization in matching markets: A laboratory experiment motivated by a natural experiment, Quarterly Journal of Economics, February, 2000, 201-235. Keser, C. (2002): Trust and reputation building in e-commerce, Working Paper, IBM, New York. Kittsteiner, T. (2003): Partnerships and Double Auctions with Interdependent Valuations, Games and Economic Behavior 44, 54-76. Kittsteiner, T., Nikutta, J. and E. Winter (2004): Declining valuations in sequential auctions, International Journal of Game Theory 33, 89-106. Kittsteiner, T., Ockenfels, A. and N. Trhal (2006): An Experimental Investigation of Partnership Dissolution Mechanisms, Work in progress, University of Cologne. Klein, T., Lambertz, C., Spagnolo, G., and K. Stahl (2005): Last Minute Feedback, Consip Working Paper VI/2005. Klemperer, P. (2004): Auctions: Theory and Practice, Princeton University Press. Krishna, V. (2002): Auction Theory, Academic Press. Li, J. and E. Wolfstetter (2005): Partnership Dissolution, Complementarity, and Investment Incentives, Working Paper, Humboldt Universitat Berlin. List, J., and D. Lucking-Reiley (2000): Demand Reduction in Multi-Unit Auctions: Evidence from a Sportscard Field Experiment, American Economic Review, 90, 961-972. 23

Lucking-Reiley, D., Bryan, D., Prasad, N. and D. Reeves (1999): Pennies from ebay: the Determinants of Price in Online Auctions. Working paper, Vanderbilt University. McAfee, R.P. (1992): Amicable Divorce: Dissolving a Partnership with Simple Mechanisms, Journal of Economic Theory 56, 266-293. McMillan, J. (1994): Selling Spectrum Rights, Journal of Economic Perspectives 8, 145-162 Milgrom, P. (2004): Putting Auction Theory to Work, Cambridge University Press. Milgrom, P. and R. Weber (1981): A theory of auctions and competitive bidding, Econometrica 50, 1081-1122. Miller, N., Resnick, P., and R. Zeckhauser (forthcoming): Eliciting Honest Feedback: The Peer Prediction Method, Management Science. Minehart, D. and Z. Neeman (1999): Termination and Coordination in Partnerships, Journal of Economics & Management Strategy 8, 191-221. Moldovanu, B, and M. Tietzel (1998): Goethe's Second-Price Auction, Journal of Political Economy 106, 854-859 Mongell, S. and A. E. Roth (1991): Sorority Rush as a Two-Sided Matching Mechanism, American Economic Review, vol. 81, 441-464. Müsgens, F. and K. Neuhoff (2005): Modelling Dynamic Constraints in Electricity Markets and the Costs of Uncertain Wind Output, EPRG Working Paper 05/14, Cambridge, UK. Müsgens, F. and A. Ockenfels (2006): Marktdesign in der Elektrizitätswirtschaft, working paper, University of Cologne. Nöldeke, G., and K.M. Schmidt (1998): Sequential Investments and options to own, RAND Journal of Economics 29, 633-653. Ockenfels, A (2003): Reputationsmechanismen auf Internet-Marktplattformen: Theorie und Empirie, Zeitschrift für Betriebswirtschaft 73, 295-315. Ockenfels, A., and A.E. Roth (2002): The Timing of Bids in Internet Auctions: Market Design, Bidder Behavior, and Artificial Agents. Artificial Intelligence Magazine, 23 (3), 79-87. Ockenfels, A. and A.E. Roth (2006): Late and Multiple Bidding in Second Price Internet Auctions: Theory and Evidence Concerning Different Rules for Ending an Auction, Games and Economic Behavior, 55, 297 320. Ockenfels, A., K. Sadrieh, and D. Reiley (forthcoming): Online Auctions, Handbook of Information Systems and Economics. Ockenfels, A., and R. Selten (2000): An Experiment on the Hypothesis of Involuntary Truth- Signalling in Bargaining, Games and Economic Behavior, 33(1), 90-116. 24

Ockenfels, A., and R. Selten (2005): Impulse Balance Equilibrium and Feedback in First Price Auctions, Games and Economic Behavior, 51, 155--170. Ockenfels, A. and J. Weimann (1999): Types and Patterns - An Experimental East-West- German Comparison of Cooperation and Solidarity, Journal of Public Economics, 71, 275-287. Peters, M. and S. Severinov (2004): Internet Trading Mechanisms and Rational Expectations, Working Paper, University of British Columbia. Rassenti, S., Smith, V. and B. Wilson (2002): Using experiments to Inform the Privatization/Deregulation Movement in Electricity, The Cato Journal, 22, 3. Rasmusen, E. (2003): Strategic Implications of Uncertainty Over One's Own Private Value in Auctions, Working paper, Indiana University Resnick, P., and R. Zeckhauser (2002): Trust Among Strangers in Internet Transactions: Empirical Analysis of ebay's Reputation System, in M. R. Baye (editor), The Economics of the Internet and E-Commerce, Advances in Applied Microeconomics, 11, Amsterdam, Elsevier Science. Roth, A.E. (1984): The Evolution of the Labor Market for Medical Interns and Residents: A Case Study in Game Theory, Journal of Political Economy, 92, 991-1016. Roth, A.E.(1991): A Natural Experiment in the Organization of Entry Level Labor Markets: Regional Markets for New Physicians and Surgeons in the U.K., American Economic Review, vol. 81, 415-440. Roth, A.E. (2002) The Economist as Engineer: Game Theory, Experimental Economics and Computation as Tools of Design Economics, Econometrica, 70, 1341-1378. Roth, A.E.. and A. Ockenfels (2002), Last-Minute Bidding and the Rules for Ending Second- Price Auctions: Evidence from ebay and Amazon Auctions on the Internet, American Economic Review, 92, 1093-1103. Roth, A. E. and E. Peranson (1994): The Redesign of the Matching Market for American Physicians: Some Engineering Aspects of Economic Design, American Economic Review, 89, 748-780. Roth, A. E., Sönmez, T. and U. Ünver (2005): A Kidney Exchange Clearinghouse in New England, American Economic Review, Papers and Proceedings, 95, 376-380. Roth, A.E. and M. Sotomayor (1990): Two-Sided Matching: A Study on Game-Theoretic Modelling, Cambridge University Press. Roth, A.E. and X. Xing (1994): Jumping the Gun: Imperfections and Institutions Related to the Timing of Market Transactions, American Economic Review, 84, 992-1044. 25

Rothkopf, M.H., Teisberg, T.J. and E.P. Kahn (1990): Why are Vickrey Auctions Rare?, Journal of Political Economy. 98, 94-109. Simonsohn, U. (2005): ebay Evenings are a Buyers' Market: Excess Entry During Peak- Demand Hours for Online Auctions, Working paper, The Wharton School, University of Pennsylvania. Stedman, G. and J. Jones (1990): Shareholders' Agreements, 2nd edition, Longman. Stryszowska, M. (2005): On the Ending Rule in Sequential Internet Auctions, Working Paper, Tilburg University. Utz, S. (2005): What's in an ID card? Trustworthiness judgments in ebay, Working paper, Free University, Amsterdam. Veugelers, R. and K. Kesteloot (1996): Bargained Shares in Joint Ventures among Asymmetric Partners: Is the Matthew Effect Catalyzing?, Journal of Economics 64 23-51. Vickrey, W. (1961): Counterspeculation, Auctions and Competitive Sealed Tenders, Journal of Finance 16, 8-37. Wilcox, R. (2000): Experts and Amateurs: The Role of Experience in Internet Auctions, Marketing Letters, 11, 363-374. Wintr, L. (2004): Some Evidence on Late Bidding in ebay Auctions, Working paper, Clark University, Worcester MA. Wolak F. (2000): An Empirical Analysis of the Impact of Hedge Contracts on Bidding Behavior in a Competitive Electricity Market, Working Paper, Stanford University. Wolf, J., Arkes, H. and W. Muhanna (2005): Is Overbidding in Online Auctions the Result of a Pseudo-Endowment Effect?, Working paper, Ohio State University. Wolfram, C. D. (1998): Strategic Bidding in a Multiunit Auction: an empirical Analysis of Bids to supply Electricity in England and Wales, RAND Journal of Economics, 703-725. Notes We are grateful to Felix Müsgens for valuable comments. Ockenfels gratefully acknowledges the support of the Deutsche Forschungsgemeinschaft. We advised governments and firms on market design; the views expressed are our own. 1 For an overview of the history of auction markets and more historic examples of market design see Cassady (1967). 26

2 See Milgrom (2004), McMillan (1994) and Klemperer (2004) for a detailed overview on spectrum privatization auctions. The next large auction in Germany is planned at the end of 2006 to allocate spectrum for Broadband Wireless Access. 3 See Roth (1984, 1991, 2002), Roth and Xing (1994), Kagel and Roth (2000), Mongell and Roth (1991), Roth and Sotomayor (1990), Roth and Peranson (1994) and Roth et al. (2005). 4 Obviously market-design is a directed activity but very different objectives are possible. Whereas a seller might simply want to maximize profit or revenue, a government or a marketplatform provider might want to consider the interest of all market participants (and maybe even the interest of non-market participants) and therefore would like to create an efficient market. In this article we will assume the designer's objective as being given and discuss how it is served by a particular design or design element. 5 No algorithm with a computation time that is a polynomial function of the number of objects and bidders is known. For practical purposes this means that for a large number of objects and bidders the allocation cannot be calculated in a reasonable amount of time (see de Vries and Vorah (2003)). 6 For surveys on mechanism- and auction design that go beyond the issues discussed in this article, see Milgrom (2004), Krishna (2002) and Klemperer (2004). 7 Most work covered here is taken from our own recent or ongoing research projects, but readers are also referred to the books of Krishna (2002), Milgrom (2004) or Klemperer (2004), and to the webpage maintained by Alvin Roth (http://kuznets.fas.harvard.edu/~aroth/alroth.html) for much more material. 8 This section partly draws from the survey by Ockenfels et al. (forthcoming), who provide a comprehensive review of economic online auction research. 9 The economic explanation for this result relies on the so called linkage principle (see Milgrom and Weber (1981)): the more information linked to the final price is revealed the higher this price will be (see Klemperer (2004)) for the economic intuition behind the effect). 27

10 Similar data has been reported by Bajari and Hortaçsu (2003), Anwar et al. (2004), Simonsohn (2005), Hayne et al. (2002), among others. 11 Late bidding may also be a best response to other incremental bidding strategies like shill bidding by confederates of the seller in order to push up the price beyond the second highest maximum bid (Barbaro and Bracht (2005)). Other reasons for incremental bidding include protecting information (see Rothkopf et al. (1990)), ignorance of the own valuation and reluctance to think about it (see Rasmusen (2003)) and emotional, non-rational factors (Heyman et al. (2004), Wolf et al. (2005)). 12 Roth and Ockenfels (2002) provided survey evidence and Ockenfels and Roth (2006) field evidence supporting the common value explanation. However, the fact that Ariely et al. (2005) observed substantial sniping in the laboratory even in a pure private-value context strongly suggests that different causes contribute to sniping behavior. For another explanation, based on the multiplicity of listings of identical objects, see Peters and Severinov (2004) and Stryszowska (2005). 13 From a theoretical point of view as long as bidders just demand one unit each and objects for sale are similar and auctioned in the right order, it should (under certain assumptions) not matter whether items are sold sequentially or simultaneously (see Kittsteiner et al. (2004) for a short overview and a model with this feature). Note that the problems associated with multiunit auctions discussed below also apply to sequential auctions and cannot be avoided by selling objects sequentially. 14 Multi-unit auctions have been used, e.g., by governments to sell spectrum licences and bus routes (see Klemperer (2004)). Actually, Vickrey's famous article (which probably marks the birth of auction theory) was already concerned with multi-unit auctions. 15 For an excellent overview of common multi-unit auction designs and their features see Milgrom (2004), Klemperer (2004) and Krishna (2002). 16 It is not only infeasible because of the huge amount of information that needs to be transmitted but also because of the computational complexity involved in the determination of the allocation of the objects (see de Vries and Vohra (2003)). For a comprehensive overview on the most recent developments in combinatorial auction design refer to Milgrom (2004). 28

17 The term ebay's multi unit auction refers to the rules of the multi-unit auctions as they are conducted by ebay in the US. EBay Germany used a very similar format until recently but changed to a new format in 2005 (see below). 18 Several field studies provide direct evidence of strategic demand reduction in electronic auction markets, such as in the German auction of GSM spectrum (Grimm et al. (2003)), in the Austrian auction of third generation mobile wireless licenses (Klemperer (2004)), in the FCC's Nationwide Narrowband Auction (Cramton (1995)), in the UK electricity market (Wolfram (1998)), and in the California electricity market (Borenstein et al. (2002)). This field evidence is strongly supported by laboratory evidence (e.g., Kagel and Levin (2001), Engelmann and Grimm (2004)) and controlled field experiments (List and Lucking-Reiley (2000)). 19 Reasons why trade law might not be sufficient to induce trust and trustworthy behaviour are ambiguities in legislation, differences in legislation (for international transactions) nonverifiability of contracts, danger of bankruptcy of one of the trading partners, costs of lawenforcement etc. 20 Resnick and Zeckhauser (2002) report that 89% of all buyer-seller matches occurred only once in their sample of online-auction transactions. 21 For more evidence on the magnitude of fraud in online markets, see Jin and Kato (2005) and the report of the Federal Trade Commission on http://www.ftc.gov/bcp/conline/edcams/dotcon /auction.htm (July 2005). 22 See Dellarocas (2003, forthcoming). 23 Mathematically precise arguments using folk theorems and sequential game theory in the context of reputation systems can be found in Ockenfels (2003) and Bolton and Ockenfels (2006). For arguments using evolutionary game theory, see Güth and Ockenfels (2000, 2003). 24 This observation is remarkably in line with an enormous literature in experimental economics on voluntary provision of public goods (see, e.g., Ockenfels and Weimann (1999), and the references cited therein). 25 In fact, there is various evidence suggesting that negative feedback is retaliated by negative feedback. First, virtually all feedback given is positive; Resnick and Zeckhauser (2002) found 29

that only 0.4 percent of the feedbacks are negative. Second, differently to positive feedbacks, negative feedbacks are given late, in the last minute, or not given at all to prevent retaliatory feedback (Klein et al. (2005) and Dellarocas et al. (2004)). 26 They also analyse the effectiveness of the feedback mechanism in a market in which the same people interact with one another repeatedly (partners market). 27 That buyers' react to feedback ratings strategically (rather than having some intrinsic trust) is further confirmed by the fact that in the last trading rounds, trust and also trustworthy behaviour is significantly lower, demonstrating that buyers and sellers understand the end-game effect. 28 There are other studies investigating what kind of reputation information promotes trust and trustworthiness in online exchange; see e.g., Bolton et al. (2006), Bolton et al. (forthcoming), and Utz (2005). Similar studies investigate from a market design perspective what kind of feedback on auction behavior should be given to promote efficiency and auction revenue; see Ockenfels and Selten (2005) and the references cited therein. Finally, it is well known that the communication channel has a strong impact on trust and trustworthiness in anonymous groups (e.g., Brosig et al. (2002), Ockenfels and Selten (2000), and Brosig (2006)). This may partly explain ebay s recent move to integrate Skype, an Internet-based phone provider, into its platform. 29 To be more precise, electricity markets, as discussed here, are organised as two-sided multiunit auctions, where buyers and suppliers trade units of electricity. 30 In addition to the markets listed below, there may be also a market for CO 2 -certificates, which suppliers require to generate electricity. For more information on the role and impact of these markets, which is currently the focus of many debates in Germany, see Müsgens and Ockenfels (2006) and Gatzen et al. (2005). 31 Electricity traded on the day-ahead market and electricity traded on the reserve market are, from a supplier's point of view, substitutes while, from the demand perspective, they are complements. 32 In a recent report, the European Commission estimated that the financial loss resulting from inefficient (wrong sign nominations) and underutilisation of interconnector capacity in the 30

Dutch - German border in 2004 was almost 50 million Euro, which is 46 percent of the total value of this interconnector capacity. 33 There is also a dispute over whether the day-ahead market provides sufficient incentives for investments in generation, or whether a capacity market is generally needed in electricity trading. We believe that the answer is, at this point, open (see Grimm and Zoettl (2005) for a thorough theoretical treatment, and Cramton (2003) and Müsgens and Ockenfels (2006) for some discussion). 34 A different approach has been taken by the EEX, which allows for bids on certain blocks of hours. 35 Reasons for the termination of a partnership abound. We here consider a situation where the split-up is due to disagreement about the future strategy of the commonly owned firm. 36 Empirical evidence in Hauswald and Hege (2003) suggests that buy-outs are indeed the most common way of partnership dissolution. 37 An allocation is efficient whenever the partnership ends up in the hands of the partner whose willingness to pay for it is highest (and monetary transfers add up to zero). Partners might have different valuations for the business due to different capabilities of running the firm or because of individual synergy effects with their other businesses. 38 This mechanism has attracted a good deal of economic research, see Brams and Taylor (1996) for an overview and a discussion of properties of the divide-and-choose rules. 39 The mechanism can easily be modified to accommodate for a situation where both partners own different shares in the partnership. A generalization to three or more partners is, however, not straightforward (see Brams and Taylor (1996)). Empirical research suggests that most partnerships are indeed two-parent 50-50 partnerships (see Hauswald and Hege (2003) and Veugelers and Kesteloot (1996)). 40 Its performance can also be assessed in terms of its ability to provide incentives that encourage investment (see Minehart and Niemann (1999) and Li and Wolfstetter (2005)) even though other dissolution mechanisms like options to buy might perform better with respect to this objective (see Nöldeke and Schmidt (1998)). 31

41 This result is due to McAfee (1992). But note that in models with complete information (or if asymmetries are small) this result is reversed (see McAfee (1992)). 42 This is also the rule promoted by some corporate lawyers and it is consistent with what is advised in handbooks on partnership agreements (see Cadman (2004) and Stedman and Jones (1990)). 43 At least this happened in the case Jansezian vs. Hotoyan, [1999] O.J. No. 4486 (S.C.J.), in which partners disputed on who has to propose for more than three years (see de Frutos and Kittsteiner (2004)). 44 It should be noted that general empirical results are extremely difficult to come by as enterprises are very reluctant to provide the necessary information. 45 The auction proposed in Cramton et al. (1987) works for an arbitrary number of bidders. We here concentrate on the case of two partners to facilitate the exposition. 46 To be more precise, in a k-price auction (with 0 k 1), the higher bidder with a bid of b 1 wins and pays k b 1 + (1-k) b 2, where b 2 < b 1 denotes the lower bid, to the other partner. 47 De Frutos and Kittsteiner (2004) show that in a standard model of independent privatevalues both dissolution designs lead to the same outcome and therefore cannot be ranked (both are optimal). 48 For example books devoted to the drafting of shareholders' agreements (see e.g. Cadman (2004) and Stedman and Jones (1990)) do not even mention auctions as an alternative to the buy-sell clause. 49 But see Kittsteiner (2003) for a step towards a more efficient design. 32

Zusammenfassung Wir zeigen beispielhaft auf, wie individuelles Verhalten von Marktteilnehmern durch das Design der Marktregeln gelenkt wird. Insbesondere zeigen wir, dass bereits kleine Regeländerungen in Onlineauktionen, auf Elektrizitätsmärkten und bei Partnerschaftsauflösungen die Effizienz und Verteilung der Marktergebnisse signifikant beeinflussen können. Durch die enge Verknüpfung theoretischer und empirischer Methoden können Marktregeln einer systematischen Untersuchung zugänglich gemacht, und die Marktperformance gesteigert werden. Summary We demonstrate how the design of market rules influence and govern individual behaviour of market participants. In particular, we show that even small design changes can have a significant impact on market performance in online auctions, electricity markets and partnership dissolution mechanisms. A combined theoretical and empirical approach can be used to analyse market rules and institutions, and to improve market performance. 33