Informs Annual Meeting 2017

TC03A

INFORMS Houston – 2017

TC03A Grand Ballroom A Financial Engineering in Applied Probability II Sponsored: Applied Probability Sponsored Session Chair: Ning Cai, Hong Kong University of Science & Technology, Kowloon, Hong Kong, Hong Kong, ningcai@ust.hk Co-Chair: Xuewei Yang, Nanjing University, Nanjing, 210093, China, xwyang@nju.edu.cn 1 - Stochastic Game of Finite Fuel Problem Renyuan Xu, University of California-Berkeley, 4141 Etcheverry Hall, OR.Department, Berkeley, CA, 94720, United States, dorisxu1026@gmail.com, Xin Guo In general, finding a Nash equilibrium or proving Pareto optimality for continuous-time stochastic game with multiple players is hard, even when there are only two players in the game. Mean Field Game (MFG) provides a powerful tool and analytically feasible framework to approximate the Nash equilibrium of the N-player stochastic games. In this talk, we will first present the stochastic game for the classical finite fuel follower problem in the seminal paper by Benes, Shepp, and Witsenhausen (1980). We will then solve and compare with explicit solutions for the Nash equilibrium and the Pareto optimality for the two-player game, and the corresponding MFG when the number of players goes to infinity. 2 - An Optimal Stopping Problem under General Levy Model Wei Zhang, Hong Kong University of Science and Technology, Hong Kong, Hong Kong, wzhangas@ust.hk We consider a pricing problem of one financial instrument, which is essentially an optimal stopping problem, under the general L\’{e} model. We give both the probabilistic and the algebraic characterization of the continuation region. Finally, we give some examples under some specific L\’{e}vy models. 3 - Credit Ratings as Regimes: Evidence from the Sovereign Credit Default Swaps Market Xuewe I. Yang, Nanjing University, Hankou Road 22, Gulou District, Nanjing, 210093, China, xwyang@nju.edu.cn We study sovereign credit risk through a rating-based regime switching model. Each country’s default intensity consists of a rating-modulated systematic factor and an idiosyncratic factor. The rating-dependence enables the model to conduct cross-sectional predictions, generate model-implied ratings, and jointly capture CDS spreads of multiple countries. Consequently, a 17-parameter version of the model fits CDS spreads of 68 countries equally well with the classical country-by- country model with hundreds of parameters. The systematic factor along with the observed (model-implied) ratings can explain more than 60% (80%) of the variations of sovereign CDS spreads of all countries. 4 - Mean-variance-Skewness Portfolio Selection Jingnan Chen, Singapore University of Technology and Design, 20 Dover Drive, Singapore, 138682, Singapore, jingnan_chen@sutd.edu.sg, Fang Zhen We introduce a mean-variance-skewness portfolio selection model to account for the asymmetry in portfolio returns. Assuming a multivariate skew-normal distribution for asset returns, we obtain a mean-variance-skewness efficient frontier and derive a closed-form optimal portfolio strategy. In addition, we propose a three-moment capital asset pricing model incorporating skewness for asset valuation. The superiority of our mean-variance-skewness portfolio strategy over the classical mean-variance strategy is demonstrated through empirical tests. TC03B Grand Ballroom B Advances in E-commerce Sponsored: Revenue Management & Pricing Sponsored Session Chair: Van-Anh Truong, Columbia University, Mudd Hall, New York, NY, 10027, United States, vt2196@columbia.edu 1 - Online Resource Allocation under Arbitrary Arrivals: Optimal Algorithms and Tight Competitive Ratios Will Ma, MIT, willma353@gmail.com, David Simchi-Levi We consider the problem of allocating fixed resources to heterogeous customers arriving sequentially, under the framework of competitive analysis, which does not assume any predictability in the sequence of customer arrivals. We derive the first optimally-competitive algorithms when there are both multiple resources, and different rewards for allocating resources to different customers. Our algorithms are simple, intuitive, and robust, and broadly expand the applicability of competitive analysis, in areas such as online advertising, matching markets, and personalized e-commerce. We test our methodology extensively on the publicly-accessible hotel data set of Bodea et al.

2 - Approximation Algorithms for Dynamic Assortment Optimization Models Ali Aouad, Massachusetts Institute of Technology, Cambridge, MA, United States, aaouad@mit.edu, Retsef Levi, Danny Segev Motivated by retailing in highly-differentiated markets, we study joint assortment optimization and inventory management. Here, customers show a dynamic substitution behavior elicited by stock-out events. We devise the first provably- good algorithms by revealing hidden submodular-like structure. Our approximation factor is best-possible under a general rank-based model, while we obtain constant-factor guarantees under popular specialized structures, such as Multinomial Logit. Our approach is an order of magnitude faster than existing heuristics and increases revenue by 6% to 12% in the experiments. 3 - Peak-end Demand Models and their Impact on Promotion Planning Tamar Cohen, Massachusetts Institute of Technology, Cambridge, MA, United States, tcohen@mit.edu, Georgia Perakis We consider the problem of promotion planning for a single product. We propose a new demand model that predicts actual sales more accurately than current methods. By utilizing the structure of this proposed demand model, we suggest a tractable Dynamic Programming approach and provide conditions under which it solves the promotion planning problem exactly. In the cases where these conditions do not hold, we provide an analytical guarantee. Finally, we test our approach on sales data and demonstrate improvement in the precision of the demand prediction as well as an increase in revenue. 4 - Dynamic Optimization of Mobile Push Advertising Campaigns Xinshang Wang, Columbia University, New York, NY, United States, xw2230@columbia.edu, Van-Anh Truong, Shenghuo Zhu, Qiong Zhang We study a novel resource-allocation problem where mobile push messages must be sent to hundreds of millions of users. Each message yields a reward when it generates a clickthrough, subject to a budget constraint. The problem aims to maximize total reward without overwhelming each user with too many messages. We propose an algorithm that sends push messages in two cycles and makes use of information observed in the first cycle to adapt decisions in the second cycle. The algorithm has a provable regret that is asymptotically smaller than the regret of simple LP-base algorithms. Numerical experiments on real data sets show that the former improves the regret of the latter by at least 10%-50%. TC03C Grand Ballroom C Economic Models in OM Sponsored: Manufacturing & Service Oper Mgmt Sponsored Session Chair: Manu Goyal, University of Utah, Salt Lake City, UT, 84112, United States, manu.goyal@eccles.utah.edu 1 - Price and Customization of a Conspicuous Good in Monopolies and Duopolies Cheryl Druehl, Associate Professor, George Mason University, Fairfax, VA, 22030, United States, cdruehl@gmu.edu, Jesse Bockstedt Consumers increasingly desire uniqueness, but within recognizable status- oriented brands. The model investigates the customization and pricing decisions of conspicuous goods in monopoly and duopoly settings. The firm trades-off the cost vs the demand-enhancing impact of customization, which allows better matching of customers’ tastes. We find that the monopolist always benefits from offering horizontal product customization. In a duopoly market, however, a prisoner’s dilemma scenario results. The dominant strategy is for both firms to offer a customizable good, but this outcome is not Pareto optimal. 2 - RFQ, Sequencing, and the Most Favorable Bargaining Outcome Leon Zhu, University of Southern California, University Park Campus, Bridge Hall 401T, Los Angeles, CA, 90089, United States, leonyzhu@usc.edu, Ying Rong, Huan Zheng We study quantity-dependent pricing contracts with exclusion clauses in a dual- sourcing setting when the suppliers are imperfect substitutes. We analyze the negotiations between a buyer and two suppliers both with and without a request for quotation (RFQ) stage that precedes the negotiation. We show that the buyer can leverage the RFQ stage even under a full information setting when the negotiation sequence is endogenously determined by the final quotations of the RFQ process. Specifically, the buyer’s equilibrium payoff with RFQ dominates the most favorable equilibrium under bargaining without RFQ.

336