Informs Annual Meeting Phoenix 2018

INFORMS Phoenix – 2018

MA24

4 - Multi-stage Assortment Problems under the Multinomial Logit Model Yuhang Ma, Cornell Tech, 2 West Loop Road, New York, NY, 10044, United States, Nan Liu, Huseyin Topaloglu We consider an assortment problem where we offer sets of products in multiple stages and the choice process at each stage is driven by the multinomial logit model. In particular, we have K stages. At each stage, we offer a distinct set of products. If the customer makes a purchase at a certain stage, then her choice process terminates. If the customer does not make a purchase at a certain stage, then she observes the set of products offered at the next stage. The goal is to find a set of products to offer at each stage to maximize the expected revenue obtained from a customer. We show that the problem is NP-hard and develop an FPTAS.

n MA24 North Bldg 131B User Behavior in Online Marketplaces and Platforms Sponsored: EBusiness Sponsored Session Chair: Yingjie Zhang, Pittsburgh, PA, 15206, United States 1 - Machine Prediction for Better Investment Decision on P2P Platform Runshan Fu, Carnegie Mellon University, 4800 Forbes Ave., Pittsburgh, PA, 15213, United States, Yan Huang, Param Vir Singh Two major claimed advantages of P2P lending over traditional banks are 1) funding opportunity for borrowers unqualified for traditional bank loans due to their unattractive credit profiles, and 2) steady and attractive return on investment for lenders. In practice, however, P2P lending platforms often find it difficult to deliver both promises. In this paper, we show that we can achieve both higher return on investment for lenders and more opportunities for borrowers at the same time, by leveraging the predictive power of machine learning techniques. We also demonstrate that while machine learning can be powerful in prediction, realizing its value requires carefully designed economic framework. 2 - Privacy Concerns and Social Recognition Trade Off- Two Natural Experiments on the Role of Online Identity in Crowdfunding Alvin Zuyin Zheng, 5555 Wissahickon Ave, Philadelphia, PA, 19144, United States, Ohad Barzilay, Paul Pavlou While online identify has been shown to have significant economic outcomes in online marketplace, disclosure of online identity may cause privacy concerns since it tends to persist across transactions and interactions and thus can carry their own reputation. In this study, we explore the effect of online identity on backers’ contribution behavior in crowdfunding by leveraging two natural experiments happened on Kickstarter. Specifically, we are interested two questions: 1) What’s the effect of online identity on backer’s contribution in crowdfunding? 2) How does this effect differ across different backers (large vs. small)? 3 - Home Bias in Online Employment Chen Liang, Yili Hong, Bin Gu We study the nature of home bias in online employment, wherein employers prefer workers hailing from the same countries. Using a large-scale dataset from a major online labor market containing employers’ consideration sets and selection of workers, we first estimate home bias in online employment. We disentangle two types of home bias, i.e., statistical and taste-based home bias, using a quasi- natural experiment wherein the platform introduces a monitoring system for employers to observe workers’ effort in time-based projects. After matching comparable fixed-price projects as a control group, our DID estimations indicate that home bias is primarily driven by statistical discrimination. 4 - Does Interaction with Promoted Content Help or Hurt Publishers’ Readership Xiaoli Yang, Boston University Questrom School of Business, Boston, MA, 02215, United States, Nachiketa Sahoo We investigate how interactions with promoted content affect users’ tendency to revisit the publisher afterwards. Using a public dataset on users’ interactions with promoted content during their visits to multiple publishers, we found that after engaging with promoted content users were 2% more likely to return in a future session to the publisher hosting the links to promoted content. This suggests that interactions with promoted content might have a positive effect on user experience on the publisher’s site. 5 - When an Event Comes: A Empirical Analysis on Taxi and Sharing-economy Drivers’ Responding Behavior Yingjie Zhang, CMU, Pittsburgh, PA, 15206, United States This paper aims to explore the links between users’ physical and digital behavior within the social-cyber-physical systems. Specifically, I propose to compare the effects of different types of city events on taxi and sharing-economy drivers. I conduct the study on New York City, using a large-scale trip data of both taxis and for-hire vehicles. From the perspective of city planners, the understanding of responding behavior of drivers, as well as the traffic/demand flows, will allow us to better design countermeasures when an emergency event happens.

n MA23 North Bldg 131A Asset Pricing and Portfolio Theory

Sponsored: Finance Sponsored Session Chair: Chanaka Edirisinghe, Rensselaer Polytechnic Institute, Troy, NY, 12180-3590, United States 1 - On New Approaches to Modelling and Trading at the Transaction-level Time Scale James Primbs, California State University Fullerton, 800 N. State College Blvd., Fullerton, CA, 92831, United States, B. Ross Barmish, Sean Warnick This talk describes our recent research on modelling and trading at a transaction- level time scale. In particular, we utilize ITCH data, which allows for use of order message data to reconstruct the so-called NASDAQ Limit Order Book (LOB). Within this context, we describe numerical experiments aimed at validating our models and evaluating the efficacy of our new high-frequency trading algorithms under development. This talk also includes suggestions for future research motivated by the results of our simulations. 2 - Robo-advising as a Human-Machine Interaction System Agostino Capponi, Columbia University, 500 W. 120th Street, New York, NY, 10027, United States The advent of robo-advising has led to an increased interest in applying artificially intelligent agents to the field of portfolio management. Central to all human- machine interactions is the value alignment problem: ensure that an autonomous agent acts according to the human clients that it serves. We introduce a framework to quantity the value of human-machine interaction in a dynamic mean-variance framework. We show that the human-machine tandem exhibits superior performance over traditional robo-advisors. 3 - Portfolio Theory of Three Tales: Risk-adjusted Returns, Liquidity, and Leverage Chanaka Edirisinghe, Rensselaer Polytechnic Institute, Lally School of Management, Pittsburgh 2118, Troy, NY, 12180-3590, United States, Jingnan Chen, Jaehwan Jeong Under liquidity costs, we show analytically the Sharpe-maximizing unlevered portfolio is no longer a tangency portfolio, and proportionate-leveraging is not an optimal strategy. As return targets increase, the required minimum portfolio- leverage increases at an increasing-rate, while the Sharpe-Leverage frontiers are progressively-dominated. Empirical analysis verifies our analytical findings, which also shows that ignoring liquidity impact can lead to severe portfolio under- performance. 4 - Optimal Portfolio Deleveraging under Liquidity Costs and Margin Restrictions Jaehwan Jeong, Assistant Professor, Radford University, Department of Management, P.O. Box 6954, Radford, VA, 24142, United States, Chanaka Edirisinghe We develop a portfolio deleveraging model under margin restrictions, where trading impacts asset prices. The model objective and constraints are non-convex separable quadratic functions; hence, it is extremely difficult to solve. We develop a new dual-cutting plane technique for solution and test it with leveraged portfolios of ETF assets. Solution efficacy and sensitivity results are reported on leverage and margin limits.

127