RESEARCH & Society ECONOMICS Playing Games

Ten years ago, it is more than likely that others in the profession would have shaken their heads in despair over the

research methods employed by DR. AXEL OCKENFELS of

the MAX PLANCK INSTITUTE FOR RESEARCH INTO ECONOMIC

SYSTEMS in Jena. Or simply dismissed him as an outsider on account of his penchant for using real people to put his theories to the test – for example, by exposing his students to the “prisoner’s dilemma”. What should one make of an economist who spends as much time in the laboratory as many a chemist, and who actually films his “guinea pigs” as they struggle with specific game-playing situations? For most economists, that would have been taking things just a little too far. Chess is not the only game of strategy – the moves that make up economic behavior owadays economic science is no stran- toral thesis that was hailed by the Economic patterns could Business School before returning to Germany if only to take account of human error and Nger to experimentation. Dr. Ockenfels is Science Association as the best of the year; at also be described to take up a post as research group leader on certain moral concerns. However, in terms of a man in great demand. The 33 year-old pro- 33 he gained his post-doctoral qualification as such. Who bids the Emmy Noether Program run by the theory, such deviations from the norm played ject manager simply cannot accept all of the at the University of Magdeburg, picking up what and when in Deutsche Forschungsgemeinschaft (DFG). no more than a subordinate role. an online auction? invitations he receives. From Harvard to another research prize on the way. He has Ockenfels attributes the quality of his train- But just how do people behave when mak- Berlin, Utrecht to Oxford, Ockenfels has al- added to his experience by working in the Innovative talent: ing above all to : “He was my ing day-to-day decisions in their business Axel Ockenfels, ready presented his theses in countless audi- Management and Information Systems de- 33, of the Max academic teacher, my mentor, and a stroke of lives? Are we really as forward-looking, toriums in Europe and the USA. Universities, partment at Penn State University in Penn- Planck Institute luck for my career.” rational and self-interested as economic companies and economic associations are sylvania, and in 1999 he spent a year as a for Research It was while Ockenfels was busy with his science would long have had us believe? starting to take notice of this young man who Postdoctoral Research Fellow at the Harvard into Economic master’s thesis that his professor was award- Reinhard Selten’s pioneering work in experi- works with the Strategic Interaction Group Systems in Jena. ed the highest of honors: in 1994 Selten, who mental economic research has forever shaken under the leadership of Werner Güth, a new has spent almost 50 years researching into the construct of earlier economic ideas, as research group that was only recently set up , became the first and so far only evidenced by the award of the Nobel Prize for in Jena, in July 2001. German to be awarded the Nobel Prize for Economics in 2002. In selecting the two Ockenfels is one of the most innovative tal- Economics. However, even while the Swedish Americans, Daniel Kahneman and Vernon ents in a sub-discipline, which is growing Academy was praising his achievements in Smith, the Swedish Academy of Sciences progressively more influential in the world of the theory of games, it was impossible to chose to honor two researchers who, like economics. So who is this unusual economist overlook the fact that Selten was also one of Selten, have radically questioned the nature

whose academic career took off in a flash and OCKMARKET the most vehement critics of many of the ba- of homo economicus through their work T

who has already been offered professorships -S sic tenets of classic economics. Selten recog- with experimental subjects in the laboratory.

both at home and abroad? To describe Ock- ORBIS nized at an early stage that most economic Kahneman, who teaches psychology at Prince- / C enfels as determined would be an understate- models are based on false premises: namely ton University, shared joint responsibility for ILSER ment: at the age of 25 he earned a prize-win- F on the assumptions that humans are rational developing the “Prospect Theory” with his

ning degree at the ; at 29 OLFGANG beings motivated only by their own self-in- late colleague Amos Tversky, who died in

he completed his doctorate under the super- : W terest. Deviations from these assumptions 1996. On the basis of experimental findings, HOTOS

vision of Joachim Weimann, writing a doc- P had been acknowledged in individual cases, this theory postulates that human beings

64 M AXP LANCKR ESEARCH 1/2003 1/2003 M AXP LANCKR ESEARCH 65 RESEARCH & Society ECONOMICS

often make decisions spontaneously – less on in the first place. It is an interesting fact that game” in which several competitors take part rational grounds than as a result of a given economists frequently leave the laboratory is aimed specifically at competitive situations situation. People have a tendency to overesti- empty-handed. “They assume themselves to in an anonymous market. And here it is evi- mate and attach too much weight to new in- be extremely cunning, and coldly offer the dent that considerations of fairness can have formation, as Kahneman demonstrated with absolute minimum,” says Ockenfels. no impact on market performance since the his “nervous frogs theory”, using dollar fluc- The “ultimatum game” is related to the participants are no longer able to influence tuations as an example. Vernon Smith, who “dictator game”: here Player A is at liberty to the distribution simply through their own be- carried out some of the first laboratory ex- split up any sum of money at will, while havior. Under extreme competitive condi- periments at around the same time as Selten, Player B, on the other hand, is unable to veto tions, so the study indicates, the classic self- is regarded as the pioneer of experimental the split. Many “dictators” give their partner interest hypothesis works well, even if the methodologies in the USA. what seems to them to be a fair share – on players would, in principle, prefer a fairer The work done by Axel Ockenfels is in- average, about a quarter of the profits. A distribution. spired by the theoretical and experimental similar phenomenon is to be observed with Game theory and experimental economics endeavors through which his teacher Rein- the “prisoner’s dilemma” that has likewise undoubtedly has its educational attraction, hard Selten revolutionized this field of re- been the subject of much research and is believes Ockenfels. He uses auction games on search. “Very simple, but enormously infor- studied by economists, sociologists, psychol- occasion in order to provide his students with mative,” says Ockenfels of the “ultimatum ogists, and political scientists alike. Here the an impressive demonstration of their restrict- game” that he and his colleagues at the Max profit accrues to the partner who fails to co- The eBay website, ed rationality. It is simple enough to experi- Planck Institute have been repeating in vary- operate. Therefore in theory, cooperation be- one of the best prospect of a quid pro quo, a fair exchange. ment for oneself: players are asked to esti- ing forms in their laboratory in Jena. Funda- tween two players should never take place. known online Thus, while behavior systematically deviates mate the contents of a glass full of euro coins mentally the “ultimatum” depicts a quite or- The reason: there is a temptation to refuse to auction houses: from standard economic predictions, it is still and then make an offer. “I regularly find that many participants dinary commercial relationship: for example, cooperate at the expense of one’s game part- make their bids clearly strategic in nature. the student who wins the bidding makes a Player A is given 100 euros which he must ner in order to pocket a proportionately high- in the final seconds Axel Ockenfels has published numerous share with an anonymous Player B. If Player er profit. This applies irrespective of whether of an auction. papers in which he investigates complex pat- The winner overestimates the value – B declines the offer, both parties are left emp- one’s partner cooperates or not. “In fact, terns of trading, negotiation and cooperation. and comes to grief ty-handed; if he accepts, each may keep their however, we notice a conditional behavior One of his most recent contributions has been share. According to the theory of homo eco- pattern. Many players are willing to cooper- widely acclaimed even beyond economic cir- bad bargain, there being nowhere near as nomicus, both participants will prefer to have ate, provided that the other does so too. cles. Together with his colleague Gary E. many coins in the glass as he or she expect- more money rather than less. Consequently, These people adopt a reciprocal behavior and Bolton of the Smeal College of Business at ed.” On average, students’ estimates are fairly Player A will keep 99 euros and give Player B in so doing improve their financial position Penn State University, Ockenfels published an realistic. However, since the winner of the far beyond anything homo economicus could essay in the American Economic Review en- auction tends to be the one whose estimate of People don’t just consider their own ever manage. Given such results, it must be titled “ERC – A Theory of Equity, Reciprocity the value of the coins in the glass was higher self-interest – they can also act cooperatively asked, who is the cleverer: our experimental and Competition”. At the core of this paper than all the others, it is all too often the case subjects in the laboratory, or homo economi- lies the mathematical formulation of a simple that the winner overestimates the true value. the remaining one euro. Player B accepts cus?” adds Ockenfels. behavioral principle that is responsible for According to Ockenfels, this statistical ef- even this minimal amount on the basis that it If the “prisoner’s dilemma” is repeated mul- organizing a host of apparently unconnected fect is also to be observed in the real econo- is better than getting nothing at all. Player A tiple times, however, by the final round the Human evolution – behavioral phenomena. “We believe that peo- my: “That is the curse of winning.” An exam- can accordingly claim almost the entire cake willingness to cooperate typically declines to culminating in ple are far more oriented towards the profits ple might be the sell-off of UMTS licenses for himself. a minimum. The players seem to know very homo ERC in the of others than has previously been assumed: which in Germany were bought in an extrav- theoretical field of The “ultimatum game” was innovated and well that cooperation can no longer be recip- conflict between they feel bad if they get less of the cake.” Put agant auction procedure by six mobile com- first investigated by Werner Güth and in its rocated in the future. In the moments before equity, reciprocity another way: in addition to pure self-interest, munications providers, some of whom are many variations it is today one of the most the game ends, each of them is pursuing their and competition. players are also motivated by their relative now indeed coming to grief as a result of the widely studied phenomena in the field of ex- own personal advantage – there is no further pay-off. huge investment. A situation which might perimental economic research. The reason Two questions are of importance in arriving have been avoided given better game theo- lies in the fact that forecasting the outcome at a decision. Firstly: how much money do I retical analyses that could be supported and proves dramatically ineffective in the experi- get? And secondly: how do I stand in relation illustrated by experimental testing. mental laboratory, as Ockenfels confirms: “In to my opponent? Ockenfels and Bolton do not In the end, believes Axel Ockenfels, the terms of human behavior, there is a wide gap believe that altruism plays a significant role success of game theory and experimental re- between theory and practice. Many people in such economic decision-making situations. search will be measured by the extent to

are willing to reject unfair procedures or out- EUTSCHLAND It is not a question of doing something to which the knowledge gained can be imple- D

comes, even if it comes at a non-trivial cost IMES benefit the other party, but of reciprocal be- mented in practice. As a scientist, he is opti- T to them.” With a split of 90 euros to 10, it is havior: as you treat me, so will I treat you mistic: methods have meanwhile become so

highly likely that Player B will turn down the INANCIAL (cf. MAXPLANCKRESEARCH 2/2002, p. 10 f.). sophisticated that it often proves possible to offer. If Player A rightly suspects this to be : F Even complex market relationships can be close the gap between theory and practice. ARTOON

the case, he will avoid making an unfair offer C simulated using the ERC model. The “auction Even the strategic interaction between nu-

66 M AXP LANCKR ESEARCH 1/2003 1/2003 M AXP LANCKR ESEARCH 67 RESEARCH & Society ECONOMICS

merous people with varying goals and de- sellers the opportunity of artificially inflating However, dangers lurk in the very anony- grees of experience in highly complex mar- the price of their item by making phony bids. mity of Internet market players. In tradition- kets could be investigated in detail by closely On the other hand, there is a greater risk that al markets, a variety of personal relation- combining both theory and experimentation. in the closing hustle, chance will decide the ships frequently exist which provide a basis The challenge is to use these methods in such outcome: the winner is the one who clicks for trust. Not so on the Internet. Here there a way as to exert control over the economi- their mouse at the right moment. is a lack of institutions to exclude fraud and cally relevant complexities of human behav- What do economists stand to learn from misuse. Just how do you prevent an eBay ior and market institutions. such studies? Axel Ockenfels is fascinated by seller from breaking the rules by supplying For Ockenfels, the Internet represents a how strongly the subtleties of market rules defective goods? How do you prevent the hugely promising area of application – for impact on market performance. Even small fraud which is becoming increasing preva- the first time, it is now possible to dissect and changes in structure can have significant lent in online markets in particular? Ocken- monitor real markets and real behavior at consequences in terms of market perfor- fels believes that economists can contribute will. Ockenfels in his research is therefore mance, because the behavior of the partici- paying particular attention to the “market de- pants in the market adapts immediately. Mar- Control is good, sign” of online Internet auctions. Together ket design opens up a whole new field of ex- but trust is better – even online with colleague Alvin E. Roth of Harvard Uni- perimental economics, Ockenfels believes – versity, Ockenfels has been comparing the the more so because economists can “invent” towards closing the virtual “trust gap”. He platforms used by eBay and Amazon – two Amazon, known new markets and test them at once: “Online himself is currently engaged in analyzing Internet companies with enormous growth ket design. Of 240 bids made at Amazon, primarily as a digital markets can be reproduced on a one to one and developing electronic reputation sys- potential and a gigantic array of products on only one was placed in the final minute. bookseller, prevents basis in the laboratory, or varied at will. But tems, which will make the previous conduct offer. In addition to game theory and labora- Things are quite different at eBay: cus- last minute bidding they can also serve to directly showcase eco- of market players transparent. “So far the by automatically ex- tory experiments, the researchers also have tomers have a tendency towards last minute tending the auction. nomic behavior patterns. The transition from work has been done by programmers, but bidding, even though the risk exists that with theory through laboratory experiment to re- slowly but surely we economists must also Specific market design prevents all the hustle and bustle just before an auc- ality is fluid. In the end, everything is con- contribute our expertise.” “sniping” – and cuts out the hustle tion ends, their bid might get lost somewhere trollable.” The difficulty lies in the fact that reputa- in the data network. To avoid this the Inter- On the other hand, it is comparatively sim- tion systems as already deployed by Ama- field data to fall back on: eBay allows public net auction houses offer their subscribers the ple to simulate interactions between anony- zon or eBay are always dependent on the access to the bids made in all auctions during assistance of an automatic “proxy” to safe- mous participants who meet but once at an willingness of participants to provide infor- the preceding four weeks, whilst Amazon ac- guard their interests. At the beginning of an Internet auction. “An interesting challenge mation. Purchasers can assess the quality tually stores the data for eight weeks. With auction they can make a “proxy bid” which is for market design is to explore relatively and delivery of products they have bought; this pool of data it is possible to precisely stored in the system. Even without actively complex structures – for example, energy or whereas sellers can grade buyers on how recreate the progress of an auction: in other participating, bidders then have a chance of telecommunications markets, the technologi- they pay and how cooperative they are. words, to see who bid what price at what acquiring the object they are bidding for pro- cal and economic constraints of which neces- However, such reputation systems may fail time for which specific product. vided no higher bid is made – the purchase In his games sitate particularly innovative rules. Initial in practice because participants can change Ockenfels and Roth have made some inter- price is then only slightly higher than the laboratory, Axel studies offer some impressive evidence that their Internet identities – no problem even esting observations. The decisive difference second-highest bid and not necessarily as Ockenfels tries out here too with the aid of game theory, rule for a layman. The willingness to give a vol- his theories on between the two systems lies in the closing high as the bid stored. real people – for constructs can be developed which ex ante untary, fair assessment is also lacking. In phase of the auctions, most of which last for Applying game theory and experimental example by exposing experimentally and ex post empirically prove the opinion of some scientists, without some several days. Whereas eBay puts a precise economics, Ockenfels and his colleague Roth students to the themselves to be successful,” says Ockenfels. financial incentive the system cannot func- time limit on its auctions (the “hard close”), have discovered a number of strategic rea- “prisoner’s dilemma”. tion effectively. at Amazon the cut-off is automatically ex- sons for why far fewer eBay customers use a Dr. Ockenfels in his laboratory in Jena is tended for a further ten minutes if a bidder “proxy”, but prefer to make a bid shortly be- aware that in this case too, economists have makes an even higher offer in the closing fore the auction ends. For example, experi- barely begun their task. “There is a huge phase (the “soft close”). enced bidders can use the chaos in the clos- amount of ground to be made up. What we Amazon’s intention is to avoid “sniping” ing seconds to exclude less cunning competi- need is to more closely link theoretical and (bidding at the very last minute). At eBay, on tors. “Sniping” often results in participants empirical research. We must learn how to de- the other hand, “sniping” is the order of the acquiring an item at a relatively low price, sign intelligent reputation systems.” Without day: the data indicates that many bidders do simply because the auction never progresses mutual trust between buyer and seller, with- not bid until right at the end. In 240 eBay as far as an exchange of blows between bid- out clearly defined rules and sanctions, the auctions, there were 89 bids made during the ders and thus leaves no time for bidders not New Economy won’t work. A realization that final minute and 29 in the last ten seconds. using eBay’s proxy to respond. Furthermore, may have come a little late but that is steadi- At Amazon every bid made in the closing sniping is a strategy which enables experi- ly gaining acceptance: there is a whole lot of phase automatically leads to the auction be- enced customers to avert two risks: on the work in store for experimental economists ing extended. The strategic attraction of one hand, they avoid price wars with other and market architects like Axel Ockenfels. “sniping” is thus countered by a specific mar- bidders; while on the other they deny shrewd CHRISTIAN MAYER

68 M AXP LANCKR ESEARCH 1/2003 1/2003 M AXP LANCKR ESEARCH 69