Informs Annual Meeting Phoenix 2018

INFORMS Phoenix – 2018

MB22

firm’s R&D investment, it may still trigger the exertion of a higher total hash rate. 2 - Propagation of Credit Freezes in Financial Lending Networks James Siderius, Massachusettes Institute of Technology, Boston, MA, United States We consider a network model of financial intermediation where banks decide whether to extend credit to other banks, which may then default on those loans. In contrast to previous literature on financial networks, the focus is on how “fear of future default” can lead to “credit freezes” before the realization of these uncertainties. Specifically, we show that increases in the riskiness of one or a few banks can lead to systemic credit freeze throughout the financial network. This occurs because the consequences from uncertainty travel throughout the network as well as decrease the profitability of loans. We use this framework to analyze the effects of policy interventions on systemic credit freezes. 3 - DGM: A Deep Learning Algorithm for Solving Partial Differential Equations Konstantinos Spiliopoulos, Boston University, 111 Cummington Mall, Boston, MA, 02215, United States We propose to solve high-dimensional PDEs by approximating the solution with a deep neural network which is trained to satisfy the differential operator, initial condition, and boundary conditions. We prove that the neural network converges to the solution of the partial differential equation as the number of hidden units increases. Our algorithm is meshfree, which is key since meshes become infeasible in higher dimensions. We implement the approach for American options in up to 200 dimensions. We call the algorithm a “Deep Galerkin Method (DGM)” since it is similar in spirit to Galerkin methods, with the solution approximated by a neural network instead of a linear combination of basis functions. 4 - Welfare Analysis of Central Bank Digital Currency: Privacy versus Efficiency Christoph Frei, University of Alberta, Mathematical and Statistical Sciences, CAB 621, Edmonton, AB, T6G 2G1, Canada, Agostino Capponi A Central Bank Digital Currency (CBDC) is electronically transacted and stored money, similar to existing cryptocurrencies like Bitcoin, but guaranteed by a central bank. If a central bank introduced a CBDC, it could provide the public with a cheaper payment system, leading to efficiency gains compared to a traditional currency. However, a CBDC may compromise users’ privacy because transactions could be monitored by the state’s authority. In this talk, we present a model featuring the trade-off between efficiency and privacy faced by people who decide between different payment methods, and analyze the welfare implications of a CBDC. n MB24 North Bldg 131B Emerging Business Models in Information Systems Sponsored: EBusiness Sponsored Session Chair: Manmohan Aseri, The University of Texas at Dallas, Richardson, TX 1 - Are You Paying Too Much for Financial Advice? Measuring the Value of Information on a Social Trading Platform Mingwen Yang, The University of Texas at Dallas, JSOM, UTD, 800 W. Campbell, SM 33, Richardson, TX, United States, Vijay S. Mookerjee, Zhiqiang Eric Zheng This study investigates the information value on a social trading platform. The platform releases the on-going transactions adding some time delays for public. Investors face the tradeoff between receiving real-time trading information paying higher commission and receiving delayed information with lower commission. Using individual trading data from a social trading platform and Forex spot price data, we simulate the profits with some time delays and measure the information value of following a trader in real time. We investigate how different factors (e.g., open time delay, close time delay, amount following by investors, market volatility, etc.) influence the information value. 2 - Selecting a Portfolio of Mobile Ad Exchanges Leila Hosseini, The University of Texas at DallasRichardson, TX, Vijay Mookerjee The trend of the last several years has been that people spend most of their mobile time in Apps. Following this trend, mobile advertisers and publishers have been focusing on running advertising campaigns inside mobile Apps to maximize their revenue. Mobile Ad delivery platforms participate in some auctions to buy supply from mobile Ad exchanges who help App owners to sell their Ad places. In this study, we analyze a planning model for a platform which needs to cover advertisers’ demand by procuring supplies from multiple Ad exchanges. The planning problem amounts to selecting which of the given set of Ad exchanges to retain, so as to minimize total costs while ensuring that the advertisers’ demand is met.

n MB22 North Bldg 130 Topics in Market Design and Multi-agent Systems Sponsored: Revenue Management & Pricing Sponsored Session Chair: Itai Ashlagi, Stanford University, Stanford, CA, 94305, United States 1 - Information and Pricing Mechanisms for On-demand Platforms Guido Martirena I consider the design of information and pricing mechanisms in an unobservable queue with strategic and heterogenous customers that are privately informed. I show that 1) if the platform releases all the information, then ùamong all possible pricing mechanismsù the platform wishes to set a sequence of posted prices, 2) for any information policy, the steady-state can be replicated by full-information and prices while achieving a higher revenue, 3) optimizing among all possible pricing and information mechanisms, the optimum is to disclose all the information and set a sequence of posted prices, and 4) I identify conditions under which the platform profits from releasing information strategically. 2 - Exploration vs. Exploitation in Team Formation for Collaborative Work Hannah Li, Stanford University, Palo Alto, CA, 94306, United States, Ramesh Johari, Vijay Kamble, Anilesh Krishnaswamy Service platforms face an online learning problem in matching workers with jobs and using the performance on these jobs to create better future matches, a problem complicated by the presence of complex tasks that require a team of workers to complete. The success of a complex job depends on the skills of all workers, which may be unknown to the platform. We analyze two settings where the performance of a team is dictated by its strongest and weakest member, respectively, and find that both problems pose an exploration-exploitation tradeoff. We establish regret bounds and show how certain myopic strategies are optimal depending on the setting as well as the proportion of high quality to low quality workers. 3 - Maximum Weight Online Matching with Deadlines Maximilien Burq, PhD Student, Massachusetts Institute of Technology, 77 Massachusetts Avenue, Cambridge, MA, 02139, United States, Itai Ashlagi, Amin Saberi We study the finite horizon problem of matching agents who arrive at a marketplace over time and leave after d time periods. Each pair of agents can yield a different match value, and we study matching algorithms that perform well over adversarial graphs, when the arrival order of agents is random. We show that the wildly used ``batching’’ algorithm, which periodically computes a maximum-weight matching, is 0.279-competitive when the period is d. Our analysis relies on a reduction to a graph-theoretic covering problem. We show that the competitive ratio is the limit solution of a family of linear programs, and we provide a computer aided bound on this limit. 4 - Markets for Public Decision-making Nikhil Garg, Stanford University, Ashish Goel, Ben Plaut A public decision-making problem consists of a set of issues, each with multiple possible alternatives, and a set of competing agents, each with a preferred alternative for each issue. We study adaptations of market economies to this setting. We first show that when each issue has a single price that is common to all agents, market equilibria can be arbitrarily bad. We then transforms the problem into an equivalent Fisher market, the simplest type of private goodsmarket. We show that the equilibrium prices in the constructed Fisher market yield a pairwise pricing equilibrium in the original public decision-making problem which maximizes Nash welfare. n MB23 North Bldg 131A Modern Trends in FinTech and Financial Networks Sponsored: Finance Sponsored Session Chair: Agostino Capponi, Columbia University, New York, NY, 10027, United States 1 - R&D in Bitcoin Mining Humoud Waleed Alsabah, Columbia University, New York, NY, United States We study a two-stage game where firms first invest in research and development (R&D) to subsequently compete in a Bitcoin mining game at reduced costs. We show that whether or not a firm is active in a mining game depends on the firm’s hash rate costs and not on the rewards that can be obtained from mining. Increasing rewards, on the other hand, induce firms to exert higher hash rates and R&D efforts. We show that while an increase in R&D spillover reduces the

156