2016 INFORMS Annual Meeting Program

WE43

INFORMS Nashville – 2016

2 - The Impact Of Dependence On Unobservable Queues Jamol Pender, Cornell University, jamol.pender@gmail.com In unobservable queueing systems, customers may choose to leave by balking, which is based on the queue length or they may leave by reneging from the queue, which is based on the virtual waiting time. The current literature assumes that the sequences of balking and reneging random variables are independent and a customer’s decision to join the line is independent of the their willingness to wait. Thus, we relax the independence assumption and assess the impact of dependence. We show that the joint density near the origin and not the correlation of the balking and reneging random variables describes the impact of the dependence for the queue length and workload processes. 3 - Service Systems With Correlated Service And Patience Times In most of the literature on service systems, a customer’s patience time is typically assumed to be independent of her service time. However, in many settings one expects a customer’s service and patience times to be related, as is empirically reported in call centers and intensive care units. In our work we capture the correlation in a many-server queuing model and explore the implications of the correlation on capacity decisions to maximize profit. The profit in our case is the difference between revenue generated by serving customers and personnel cost associated with capacity. We demonstrate a nontrivial relation between capacity and profit in a system with correlations. 4 - Service Systems With Slowdowns Jing Dong, Northwestern University, jing.dong@northwestern.edu, Pnina Feldman, Galit Yom-Tov Many service systems exhibit service slowdowns when the system is congested. In this project we investigate this phenomenon and its effect on system performance. We modify the Erlang-A model to account for service slowdowns and carry out the performance analysis in the heavy-traffic asymptotic regime. We find that when the load sensitivity is high, the system can alternate randomly between two performance regimes, a phenomenon which we refer to as bi- stability. We analyze how the system parameters affect the bi-stability phenomenon and propose an admission control policy to avoid the bad performance regime. Chenguang Wu, Northwestern University, ChenguangWu2013@u.northwestern.edu Chair: Brian Clark, Rensselaer Polytechnic Institute, Lally School of Management, Troy, NY, 12180, United States, clarkb2@rpi.edu 1 - The Implicit Value Of Tracking The Market Majeed Simaan, Rensselaer Polytechnic Institute, Troy, NY, 12180, United States, simaam@rpi.edu, Chanaka Edirisinghe, Brian Clark We find that the bias of the tracking error portfolio is mainly due to the mean- variance portfolio rather than the tracking error component. We find that shifting the weights of the portfolio toward the tracking error direction mitigates the higher estimation error that originates in the mean vector. Using bootstrap approach, we find that the committed estimation risk in the tracking error portfolio is significantly lower than that committed in the mean-variance portfolio. Additionally, this difference is amplified for cases that are associated with greater estimation risk. 2 - Optimal Portfolio Rebalancing Under Mean-risk Tradeoff, Market Impact, And Leverage Jaehwan Jeong, Radford University, Department of Management, P.O. Box 6954, Radford, VA, 24142, United States, jjeong5@radford.edu, Chanaka Edirisinghe A portfolio optimization problem is considered with frequent rebalancing, market impact costs, and random returns, to determine the efficient frontier between the Sharpe ratio and leverage ratio. Using an upper bound for portfolio variance, a nonconvex separable quadratic optimization model with two quadratic constraints is formulated. We develop an algorithm to solve this NP-hard problem and provide a strategy analysis. 3 - Disentangle Signals And Noises In Portfolio Optimization Long Zhao, University of Texas at Austin, Austin, TX, United States, zhaolong.soul@gmail.com, Deepayan Chakrabarti, Kumar Muthuraman Mean-variance portfolios constructed using the sample mean and covariance have a poor out-of-sample performance. There are two groups are better. The first group shrinks the covariance. The second imposes norm-constraints. However, the shrinkage targets and the norm-constraints are not validated. Our method WE41 207C-MCC Topics in Portfolio Optimization Sponsored: Financial Services Sponsored Session

disentangles signals and noises in sample covariance matrix. By using signals and bounding noises, this new method is parameter-free but can achieve significant lower out-of-sample variance than the naive portfolios across ten datasets. Finally, we show that the higher out-of-sample Sharpe ratio can be obtained by cautiously using the information in sample mean. WE42 207D-MCC Strategic Behavior and Dynamic Choice in Revenue Management Sponsored: Revenue Management & Pricing Sponsored Session Chair: Jue Wang, Queen’s School of Business, 143 Union St. West, Kingston, ON, K7L 3N6, Canada, jw171@queensu.ca 1 - To Ration Or Not To Ration? Selling To Strategic Customers Under Shortage Effect. We consider the dynamic pricing and rationing policy of a firm facing strategic customers under the influence of shortage effect. We provide conditions under which it is optimal for the firm to ration. We also identify the necessary and sufficient conditions for the existence of steady state. We also characterize the firm’s pricing and rationing policy under this steady state. 2 - Dynamic Pricing Of Vertically Differentiated Products For Consumers With Sequential Search Chi-Guhn Lee, University of Toronto, cglee@mie.utoronto.ca, Sajjad Najafi, Sami Najafi-Asadolahi, Steven Nahmias We consider a firm that wishes to maximize the expected revenue from a line of vertically differentiated products with fixed inventory over a finite horizon. Consumers are utility maximizers and sequentially check the products for one maximizing the utility. We analytically derive the optimal prices of products as well as the optimal order of products’ presentation. We show that if the reservation utility is stationary or increasing, it is optimal for the seller to present the products in the descending order of quality and to increase price over time under a certain condition. 3 - Competing With Responsive Follower: Imitator And Strategic Consumers Mike Mingcheng Wei, University at Buffalo, mcwei@buffalo.edu In this work, we study the production and pricing decisions of a market leader facing an imitator under strategic consumer behavior with possible network externalities. 4 - Dynamic Pricing Of The Fixed-term Subscription Contract Offered To The Strategic Customers Subscriptions are contracts that a company makes with its customers for regular service delivery or for providing access to the service. Service access limits can be stipulated in the subscription contract. We present a continuous time dynamic pricing model for a monopolist offering a fixed term subscription contract without per-use charges to strategic customers. We formulate the monopolist’s problem in terms of optimal control, derive its optimality conditions, and study the structure of the optimal solution. We also examine the stationary optimal pricing regime and evaluate it in numerical experiments. WE43 208A-MCC Decision Analysis Arcade II Sponsored: Decision Analysis Sponsored Session Chair: Alba Rojas, Virginia Tech, Progress Street, Blacksburg, VA, 24060, United States, albarc@vt.edu 1 - Adaptive Clinical Trials And The Hot Stove Effect Alba Rojas-Cordova, Virginia Tech, Blacksburg, VA, United States, albarc@vt.edu, Niyousha Hosseinichimeh Adaptive clinical trials promise cost savings to the pharmaceutical industry and have triggered significant regulatory changes. We model the dynamics of the drug development decision within the context of adaptive clinical trials. We consider the randomness surrounding the drug’s probability of technical success and the bias inherent to its estimation. We show a bias against candidates that appear to be worse than they actually are. Stephen Shum, City University of Hong Kong, swhshum@cityu.edu.hk, Peng Hu, Hanqing Liu Roozbeh Yousefi, Smith School of Business, 143 Union St., Kingston, ON, K7L 3N6, Canada, r.yousefi@queensu.ca, Yuri Levin, Mikhail Nediak, Jue Wang

497