Nobel Laureate Paul Milgrom at International ZEW Workshop

Workshop

Interview with Market Design Experts

The participants of the ZEW Workshop on Market Design with Paul Milgrom in Paris

The fact that US economist Paul Milgrom received the Nobel Prize in Economics drew worldwide attention to the economic research field of market design about two years ago. As a comparatively young discipline, market design deals with the question of how market rules can be designed to improve the performance of existing markets. In order to promote academic exchange in this field, ZEW Mannheim, together with other institutes, organised the 2nd KIT-Paris-ZEW Workshop on Market Design at the end of June. The two-day workshop welcomed Paul Milgrom (Stanford University) as guest of honour. In his lecture, the economist, who was awarded for his contributions to the theory and practice of auction design, presented research into new auction designs. Their algorithm properties can lead to better investment incentives in the run-up to auctions for the allocation of scarce resources, such as in the allocation of take-off and landing rights in air traffic. Milgrom then explained the goals and possibilities of market design in an interview with ZEW researcher Marion Ott. ZEW also conducted short interviews with Gary Bolton (University of Texas), Gary Charness (University of California), Ksenia Shakhgildyan (Bocconi University) and Sander Onderstal (University of Amsterdam), who all participated in the workshop.

Paul Milgrom in an interview with Marion Ott

Podigee

Click the button below to reload the content. (I agree to external content being displayed to me. Read more in our privacy policy).

At the workshop at the Panthéon-Assas University in Paris, a total of twelve distinguished researchers and outstanding young talents from leading universities presented their findings to an expert audience and discussed the results. Basic research met applied theory and experimental research. The topics ranged from rating systems on marketplaces and the influence of algorithms in auctions to the best way to motivate children to eat healthier.

1) Experiments

Gary Charness (University of California Santa Barbara, with Ramón Cobo-Reyes, Erik Eyster, Gabriel Katz, Ángela Sánchez and Matthias Sutter): Improving children’s food choices: Experimental evidence from the field

Nutrition is an important factor in maintaining our health. In a field experiment at Spanish schools, the researchers investigated how to encourage children to eat healthier. The result is clear: Lectures on healthy food are of little help. In contrast, introducing “grading scales” for food are more successful. However, the most effective way to persuade children to eat healthily is to tell their parents the “grades” of the food.

Sander Onderstal (University of Amsterdam, with Andrej Woerner and Arthur Schram): Comparing Crowdfunding Mechanisms: Introducing the Generalized Moulin-Shenker Mechanism

Crowdfunding has become very important for the financing of many projects. There are already several crowdfunding providers (e.g. Kickstarter, Indiegogo). Such platforms often use an all-or-nothing mechanism. Here, the company sets a fundraising target and gets nothing if the target is not met, shifting the risk to the entrepreneur. The study examines the introduction of a new crowdfunding mechanism.

Gary Bolton (University of Texas at Dallas): Rate this Transaction: Coordinating Mappings in Market Feedback Systems

Rating systems are a common feature in online markets. Customers share their reviews about providers. For interested new customers, this feedback puts the reputation of individual companies on display. However, a large part of the rating depends on the system used. How must rating systems be designed to be informative? What questions should be asked? What scales should be used (e.g. 3 stars or 5 stars?). In this study, the question of optimal design was investigated by testing different systems in the laboratory.

2) Applied theory

Philippe Jehiel (Paris School of Economics, with Konrad Mierendorff): Auction Design with Data-Driven Misspecifications

Auctions serve to sell the auctioned item to the person for whom it has the highest value. However, this does not always succeed. The study shows that a certain mix of bidders hinders the efficiency of the auction. This is the case when some bidders act economically rational and others – so-called novices – have to rely mainly on data from previous auctions due to a lack of experience. The source of inefficiency here is not the quality of the available information, but the limited intellectual capacity of some bidders.

Helene Mass (University of Bonn, with Claudia Herresthal): Optimal Transparency in Task Design

This study seeks to find the optimal balance between transparency and difficulty in tasks. How should tasks be designed if you want people to acquire as much knowledge as possible? When is it optimal to have lengthy or short tasks? When should the evaluation scale be detailed or approximate?

Paul Milgrom (Stanford University, with Mohammad Akbarpour, Scott Kominers, Kevin Li and Shengwu Li): Investment Incentives in Truthful Approximation Mechanisms

The study examines auctions for optimal resource allocation in very complex problems in terms of their investment incentives for bidders in the run-up to the auction. It is shown that standard algorithms that use auctions to solve such problems may provide less appropriate investment incentives. In contrast, Milgrom demonstrates that a novel property of algorithms can provide better incentives. Their application becomes relevant in practice when it comes to the distribution of scarce goods, such as the allocation of take-off and landing rights at airports.

Julien Combe (Ecole Polytechnique, with Umut Dur, Olivier Tercieux, Camille Terrier and M. Utku Ünver): Market Design for Distributional Objectives in (Re)assignment: An Application to Improve the Distribution of Teachers in Schools

It is not only in France that there is a great imbalance in the distribution of teachers. In so-called problem regions, experienced teachers are needed above all, but it is precisely these experienced teachers who do not want to move to regions that are unattractive for teachers. Conversely, there are many experienced teachers and too few young teachers in the less problematic “good” regions. The study therefore investigates how to improve the matching mechanism for teachers.

Ksenia Shakhgildyan (Bocconi University, with Francesco Decarolis, Gabriele Rovigatti and Michele Rovigatti): Artificial Intelligence, Algorithmic Bidding und Collusion in Online Advertising

In online advertising (e.g. on Google), auctions are used millions of times to sell advertising space. Bidders often use artificial intelligence and algorithms. What impact does this have on these markets? What is the optimal market design for such markets? To investigate this, a number of simulations were examined. One result is that the more detailed the information of the algorithms, the better the efficiency of the allocation and the profit of the advertiser.

3) Basic research (theory)

Anne-Katrin Roesler (University of Toronto, with Rahul Deb): Multi-Dimensional Screening: Buyer-Optimal Learning and Informational Robustness

How can a monopolistic seller design a mechanism to sell multiple goods when neither buyer nor seller know ex ante the buyer’s valuation? The study investigates the problem and introduces a novel feature, i.e. buyer learning. With this feature, the optimal design for the buyer involves bundling his goods rather than selling them individually.

Songzi Du (University of California San Diego, with Benjamin Brooks): On the Structure of Informationally Robust Optimal Auctions

Often the information structure in auctions is less clear than assumed. In particular, it is unclear to the seller what expectations bidders have about the value estimates of other bidders. How should auctions be designed in such cases? The study shows that in order to achieve the optimal profit, it is not necessary for sellers to get bidders to explicitly communicate their bid expectations. The authors hope that this approach will lead to the discovery of new auction designs that are convincing in terms of both their high degree of security and their simplicity.

Ina Taneva (University of Edinburgh, with Thomas Wiseman): Strategic Ignorance and Information Design

How does strategic ignorance change the interaction between a designer and different agents? An unspoken assumption of information design is that participants agree to be informed according to the designer’s ideas. However, this is not necessarily beneficial for the agents in every case. There are situations where it is better for them to remain uninformed. Strategic ignorance means that agents deliberately refuse to view information in order to benefit from it in equilibrium. Ignoring agents’ strategic ignorance when designing can lead to divergence between the designer’s intent and the actual outcomes.

Harry Pei (Northwestern University): Building Reputations via Summary Statistics

In economic interaction, trading partners often have limited information about each other. They can only draw conclusions from observing statistics or from the behaviour of their counterparts. In addition, there are markets where consumers inform others about their experiences. This is limited, however, because while the collected evaluations make a statement, the underlying circumstances are usually not conveyed. So what are the equilibria in the interaction between a seller who is not necessarily honest and buyers who only have access to aggregated reviews of that seller? The paper shows that a “long memory” is bad for buyers, i.e. it is better not to include old reviews in the aggregation for too long.