INFORMS Workshop on Market Design 2023
(in conjunction with the ACM EC 2023 conference)
July 11-13, 2023
Sunday, July 9, 8:30-9:00 BST:
Employees versus Contractors: An Operational Perspective
Ilan Lobel, New York University
We consider a platform’s problem of how to staff its operations given the possibilities of hiring employees and setting up a contractor marketplace. We aim to understand the operational difference between these two work arrangement models. We consider a model where demand is not only stochastic, but also evolving over time, which we capture via a state of the world that determines the demand distribution. In the case of employees, the platform controls the number of employee hours it uses for serving demand, while in the case of contractors it sets the wage paid to them per utilized hour. We show that while the employee problem is equivalent to a standard newsvendor, the contractor one corresponds to an unusual version of the newsvendor model where utilization is the control variable. This distinction makes the contractor model more flexible, allowing us to prove that it performs significantly better, especially if the order of magnitude of demand is unknown. Meanwhile, hybrid solutions that combine both employees and contractors have complex optimal solutions and offer relatively limited benefits relative to a contractor marketplace.
(Joint work with Sebastien Martin and Hoatian Song)
Sunday, July 9, 9:00-9:30 BST:
Strategic Decentralized Matching: The Effects of Information Frictions
Leeat Yariv, Princeton University
(Joint work with Andrew Ferdowsian and Muriel Niederle)
Sunday, July 9, 10:00-10:30 BST:
Quality and Externalities on Platforms
Peter Coles, Airbnb
We propose an intuitive measure of platform quality externality, applicable across a range of platforms. Guest Return Propensity (GRP) is the aggregate propensity of a seller’s customers to return to the platform after a transaction (controlling for customer characteristics). This metric differs from more traditional quality metrics, like star ratings, in that it is entirely based on revealed preference – future actions buyers take.
We validate this metric using data from Airbnb, a peer-to-peer accommodation platform. Using an instrumental variable analysis to account for unobservable guest characteristics, we find that matching customers to listings with a one standard deviation higher GRP causes them to take 17% more subsequent trips. By directing buyers to higher-GRP sellers, platforms may adjust for this quality externality and increase overall platform surplus.
Sunday, July 9, 10:30-11:00 BST:
Learning Bayesian Nash Equilibria in Auction Games
Martin Bichler, Technical University of Munich
Equilibrium problems in Bayesian auction games can be described as systems of differential equations. Depending on the model assumptions these equations might be such that we do not have a mathematical solution theory. The lack of analytical or numerical techniques for the equilibrium problem has plagued the field and limited equilibrium analysis to rather simple auction models such as single-object auctions.
Recent empirical results in equilibrium learning led to algorithms that find equilibrium under a wide variety of model assumptions. We analyze first- and second-price auctions where simple learning algorithms converge to an equilibrium of the discretized game. The equilibrium problem in auctions is equivalent to solving an infinite-dimensional variational inequality (VI). Monotonicity is the central condition for learning algorithms to converge in such VIs. We show that neither monotonicity nor pseudo- or quasi-monotonicity hold for the respective VIs.
Monday, July 10, 8:30-9:30 BST:
Auctions Between Regret-Minimizing Agents
Noam Nisan, Hebrew University of Jerusalem
We analyze a scenario in which software agents implemented as regret-minimizing algorithms engage in a repeated auction on behalf of their users. We study first price and second price auctions, as well as their generalized versions (e.g., as those used for ad auctions). Using both theoretical analysis and simulations, we show that, surprisingly, in second price auctions the players have incentives to mis-report their true valuations to their own learning agents, while in the first price auction it is a dominant strategy for all players to truthfully report their valuations to their agents.
(Joint work with Yoav Kolumbus)
Tuesday, July 11, 8:30-9:00 BST:
Redesigning Volunteer Match’s Ranking Algorithm: Toward More Equitable Access to Volunteers
Vahideh Manshadi, Yale University
In collaboration with VolunteerMatch—the world’s largest online platform for connecting volunteers with nonprofits—we designed and implemented a novel ranking algorithm called SmartSort to improve equity in access to volunteers. Based on promising experimental results in three large metro areas, each showing a statistically significant 8-9% increase in our metric for equity, VolunteerMatch has deployed SmartSort nationwide. We expect it to provide an additional 30,000 volunteer sign-ups annually to opportunities with limited access to volunteers.
Tuesday, July 11, 9:00-9:30 BST:
Redesigning Framework Agreement Auction in Chile Reduces Government Spending
Daniela Saban, Stanford University
Tuesday, July 11, 10:00-11:00 BST:
Differentiable economics: Using deep learning to discover new market designs
David C. Parkes, Harvard University and DeepMind
Since Duetting et al. (ICML’19), we have been exploring the use of deep learning for economic design, and in particular end-to-end, unsupervised deep learning as a way to discover the design of auctions, matching mechanisms, data markets, and optimal contracts. This talk will illustrate current techniques and results, and give a series of open problems that are important to address in building up a solid framework for ML-based discovery in economics.
(Joint work with many collaborators, including Paul Duetting, Zhe Feng, Dima Ivanov, Jeff Jiang, Scott Kominers, Sai Srivatsa Ravindranath, Inbal Talgam-Cohen, and Tonghan Wang)