Informs Annual Meeting Phoenix 2018

INFORMS Phoenix – 2018

SB25

3 - Marketplace Experimental Design at Airbnb Ruben Lobel, Airbnb, 888 Brannan St, 3rd Floor - Airbnb, San Francisco, CA, 94103, United States, David Holtz, Bar Ifrach, Sinan Aral Online A/B tests rely on the assumption that responses on treatment and control units are independent. However, experiments in online marketplaces such as Airbnb can be biased when listings in the treatment group affect the listings in the control group. We propose a method to reduce this bias by identifying a network of marketplace substitution and applying a clustering algorithm on this network. By delivering the experiment on clusters of listings, we hope to detect the true impact of our models, without sacrificing too much in the experiment’s sample size. We demonstrate the use of this method with a pricing experiment, measuring the trade-off between bias reduction and statistical power. 4 - An Overview of Uber’s Surge Pricing Mechanism Hamid Nazerzadeh, USC Marshall School of Business, Bridge Memorial Hall - BRI 401B, 3670 Trousdale Parkway, Los Angeles, CA, 90089, United States Uber is running one of the largest marketplaces in the world that matches millions of riders and drivers together. One of the key features of Uber’s marketplace is dynamic pricing, so called surge pricing, that aims to balance the demand for trips and the supply of available cars. The surge pricing also motivates drivers to relocate to parts of the city with higher demand. In this talk, I’ll discuss the design of the new generation of surge mechanisms that aim to increase the marketplace efficiency. Chair: Daniel Mitchell, University of Minnesota, Minneapolis, MN Co-Chair: Stathis Tompaidis, University of Texas-Austin, Austin, TX, 78712-1277, United States 1 - Regulation of Financial Networks Daniel Mitchell, University of Minnesota, Minneapolis, MN, United States, Stathis Tompaidis We investigate the problem faced by a regulator to impose capital constraints on financial institutions in a network. By imposing capital constraints on one bank, the riskiness of others may be affected. We show how the regulator can set cVaR capital requirements for every institution in the network simultaneously. We analyze this regulation and look at its implications in an example calibrated to match publicly available data from bank holding companies in the United States. 2 - Bank Risk Culture Identification by Machine Learning Abena Owusu, Rensselaer Polytechnic Institute, Troy, NY, United States, Aparna Gupta We develop a framework for evaluating the risk culture of banks and how it relates to banks’ performance, governance and risk-taking incentives. Textual analytics of banks’ annual 10-K reports using Loughran and McDonald’s (2011) sentiment dictionary and a risk culture framework develops a two-dimensional features space. Legal expense as a proxy allows a feature reduction, followed by k-means clustering to group banks into three risk culture groups. The good and moderate risk culture groups show better performance, more institutional investors, lower total executive pay and equity fraction of pay. Boards of better risk culture groups have more independent directors. 3 - Can Swing Pricing Prevent Mutual Fund Runs and Fire Sales? Agostino Capponi, Columbia University, 500 W. 120th street, New York, NY, 10027, United States We develop a model of the feedback between mutual fund outflows and asset illiquidity. Alert investors anticipate the impact on the fund’s NAV of other investors’ redemptions and exit first at favorable prices. Our analysis shows that (i) the first-mover advantage introduces a nonlinear dependence between a market shock and the aggregate impact of redemptions on the fund’s NAV; (ii) there exists a critical magnitude of the shock beyond which a run brings down the fund; (iii) properly designed {\it swing pricing\/} not only transfers liquidation costs from the fund to redeeming investors, but importantly reduces them and prevents fund failure. 4 - Managing Default Contagion in Inhomogeneous Financial Networks Nils Detering, University of California, Santa Barbara, Santa Barbara, CA, 93106, United States The aim of this paper is to quantify and manage systemic risk caused by default contagion in the interbank market. Our results allow us to determine the impact of local shocks to the entire system and the wider economy. As a central application, we characterize resilient and non-resilient cases. In particular, for the prominent case where the network has a degree sequence without second moment, we show that a small number of initially defaulted banks can trigger a n SB23 North Bldg 131A Systemic Risk Sponsored: Finance Sponsored Session

substantial default cascade. Paralleling regulatory discussions, we determine minimal capital requirements for financial institutions sufficient to make the network resilient to small shocks.

n SB24 North Bldg 131B Health IT Innovations: Opportunities and Risks Sponsored: EBusiness Sponsored Session Chair: Idris Adjerid, University of Notre Dame, Notre Dame, IN, 46556, United States 1 - When IT Creates Legal Vulnerability: Not Just Overutilization but Underprovisioning of Healthcare Could be a Consequence YeongIn Kim, University of Texas-Dallas, Richardson, TX, 75080, United States, Mehmet Ayvaci, Srinivasan Raghunathan, Turgay Ayer There is a growing interest in better understanding the new IT vulnerabilities. We examine one such vulnerability, viz., the heightened litigation risk due to information sharing enabled by IT, in a typical health care context. Using a game- theoretic model, we find that socially optimal policies may require underprovisioning of health care in the presence of health IT. Limits on malpractice damage can alleviate the underprovisioning of health care. 2 - Detecting Anomalous Patterns of Care Using Health Insurance Claims Sriram Somanchi, University of Notre Dame, 344 Mendoza College of Business, University of Notre Dame, Notre Dame, IN, 46556, United States, Edward McFowland, Daniel Neill This work provides a novel machine learning methodology to identify subpopulations for whom certain patterns of medical care have led to significantly anomalous health outcomes. We detect interventions in patient care (currently in terms of medications) that have significantly affected health outcomes either negatively (in which case they may represent suboptimal care that should be identified and corrected) or positively (in which case they may represent new, previously undiscovered best care practices). 3 - Effect of Gamification on Healthful Activity: The Case of Fitbit Leaderboards Idris Adjerid, Virginia Tech, Blacksburgh, PA, 46556, United States, Zia Hydari, Aaron Striegel Lack of physical activity represents a growing global health crisis. We evaluate whether gamification of health behavior using wearable devices is one intervention that has the potential to increase physical activity. Using a unique data set, our initial analysis suggests that gamification (in the form of leaderboards) leads to a significant increase in steps walked daily. 4 - Analyzing Healthcare Information Exchanges’ Strategies in a Competitive Environment Can Sun, University of Alberta, 03A Unit, 8909-112, Edmonton, AB, T6G 2C5, Canada, Yonghua Ji, Subodha Kumar Demand for health information exchange (HIE) among healthcare providers is growing very fast. In this paper, we investigate a setting where heterogeneous healthcare providers can join one of two competing HIEs. The utility for a healthcare provider is determined by the software quality offered by an HIE and also the network effect, i.e., the number of healthcare providers adopting the same HIE. We use a game theoretical model to investigate how HIEs should price the basic and premium services to maximize their profits.

n SB25 North Bldg 131C

Best Cluster Paper – II Sponsored: Service Science Sponsored Session Chair: Paul R. Messinger, University of Alberta, University of Alberta, Edmonton, AB, T6G 2R6, Canada 1 - Service Science Best Cluster Paper Paul R. Messinger, University of Alberta, University of Alberta, Edmonton, AB, T6G 2R6, Canada This session consists of finalists presentations (judged by an expert panel) to determine the Service Science Section Best Cluster Paper.

39