Informs Annual Meeting Phoenix 2018

INFORMS Phoenix – 2018

TC23

2 - Customer Preference and Station Network in the London Bike Share System Pu He, Columbia University, Uris Hall, Cub 4H, New York, NY, 10027, United States, Fanyin Zheng, Elena Belavina, Karan Girotra We study customer preference for the bike share system in the city of London. We estimate a structural demand model on the station network to learn the preference parameters and use the estimated model to provide insights on the design and expansion of the bike share system. We highlight the importance of network effects in understanding customer demand and evaluating expansion strategies of transportation networks. We develop a new method to deal with the endogeneity problem of the choice set in estimating demand for network products. Our method can be applied to other settings, in which the available set of products or services depends on demand.

n TC21 North Bldg 129B

Joint Session RMP/Practice Curated: Data-driven Applications and Dynamic

Learning in Revenue Management Sponsored: Revenue Management & Pricing Sponsored Session Chair: N. Bora Keskin, Duke University, Durham, NC, 27708-0120, United States 1 - The Exponomial Choice Model: Assortment Optimization and Data-driven Applications Jacob Feldman, Olin Business School, 6 Portland Court, Saint Louis, MO, 63108-1291, United States, Mohammed Ali Aouad, Danny Segev We study the assortment optimization problem under the Exponomial choice model. In this problem, a retailer seeks the revenue maximizing set of products to offer to each arriving customer. Our main contribution comes in the form of the first polynomial time approximation scheme with a provable guarantee for the assortment problem under the Exponomial choice model. We follow up this result with a series of estimation studies using real data, which show that the predictive power of the Exponomial model is on par with the classic MNL model. 2 - Personalized Advertising and Learning Through High-dimensional Data with Limited Samples Mingcheng Wei, University at Buffalo, 326C Jacobs Management Center, Buffalo, NY, 14260, United States, Xue Wang, Tao Yao In this paper, we propose a Minimax Concave Penalized Multi-Armed Bandit algorithm for a decision-maker facing high-dimensional data with latent sparse structure in an online learning and decision-making process. This algorithm performs favorably compared to other algorithms, especially when there is a high level of data sparsity or when the sample size is not too large. 3 - Data-driven Pricing for a New Product Mengzhenyu Zhang, University of Michigan, Ross School of Business, Ann Arbor, MI, 48105, United States, Hyun-Soo Ahn, Joline Uichanco Decisions regarding new products are difficult to make, and mistakes can have grave consequences to a firm’s bottom line. Firms have little foresight on information about new product demand such as the potential market size, the rate of customers’ adoption. We study the interplay between pricing and learning to maximize the expected revenue of a new product over a finite time horizon. We consider a setting where a firm can learn by observing sales data. To capture the stochastic adoption process, we develop a continuous-time Markovian Bass model. We derive the optimal pricing policy with learning and propose two simple and computationally tractable pricing policies that are provably near- optimal. 4 - Dynamic Pricing-and-Learning Strategies in Service Operations Yuan-Mao Kao, Duke University, Durham, NC, 27708, United States, N. Bora Keskin, Kevin Shang We consider a firm providing a service to its customers under limited information about demand and service requirements. The firm can obtain more information by offering the service over multiple periods. In this setting, we formulate how pricing strategies affect the firm’s learning, and we study the performance of near- optimal dynamic pricing strategies. n TC22 North Bldg 130 Marketplace Innovation: Field and Empirical Studies Sponsored: Revenue Management & Pricing Sponsored Session Chair: Chiara Farronato, Harvard Business School, Soldiers Field Road, Morgan Hall 412, Boston, MA, 02163, United States

n TC23 North Bldg 131A Algorithmic Trading

Sponsored: Finance Sponsored Session Chair: Sebastian Jaimungal, University of Toronto, Toronto, ON, M5S3G3, Canada 1 - Mean Field Games with Differing Beliefs for Algorithmic Trading Sebastian Jaimungal, University of Toronto, Toronto, ON, Canada, Philippe Casgrain Even when confronted with the same data, agents often disagree on a model of the real-world. Here, we address the question of how interacting heterogenous agents, who disagree on what model the real-world follows, optimize their trading actions. The market has latent factors that drive prices, and agents account for the permanent impact they have on prices. This leads to a large stochastic game, where each agents’ performance criteria is computed under a different probability measure. We analyse the mean-field game (MFG) limit of the stochastic game and show that the Nash equilibria is given by the solution to a non-standard vector-valued forward-backward stochastic differential equation. Under some mild assumptions, we construct the solution in terms of expectations of the filtered states. We prove the MFG strategy forms an $\epsilon$-Nash equilibrium for the finite player game. Lastly, we present a least-squares Monte Carlo based algorithm for computing the optimal control and illustrate the results through simulation in market where agents disagree on the model. 2 - A Mathematical Analysis of Technical Analysis Bin Zou, University of Connecticut, Storrs, CT, United States, Mattew Lorig, Zhou Zhou In this paper, we study trading strategies based on exponential moving averages (ExpMA), an important indicator in technical analysis. We seek optimal ExpMA strategies when the drift of the underlying is modeled by either an Ornstein- Uhlenbeck process or a two-state continuous-time Markov chain. Closed-form solutions are obtained under the logarithm utility maximization and the long- term growth rate maximization. 3 - Hedging Non-Tradable Risks with Transaction Costs and Price Impact Ryan Donnelly, University of Washington, Seattle, WA, United States, Alvaro Cartea, Sebastian Jaimungal An agent hedges exposure to a non-tradable risk factor U using a correlated traded asset S and accounts for the impact of trades on both factors. We obtain in closed-form the optimal strategy when the agent holds a linear position in U. With non-linear exposure to U, we provide an approximation to the optimal strategy in closed-form, and prove that the value function is correctly approximated when cross-impact and risk-aversion are small. With non-linear exposure, the approximate optimal strategy can be written in terms of the optimal strategy for linear exposure with the size of the position changing dynamically according to the exposure’s “Delta” under a particular probability measure. 4 - The Shadow Price of Latency: Improving Intraday Fill Ratios in Foreign Exchange Markets Alvaro Cartea, University of Oxford, Oxford, United Kingdom, Leandro Sánchez-Betancourt, Philippe Casgrain Liquidity takers face a moving target problem because of their latency. They send marketable orders that aim at a price and quantity they observed in the LOB, and by the time their order is processed, prices could have worsened (improved), so the order may not be filled (be filled at a better price). We show how to choose the discretion of orders to walk the LOB to increase the chances of filling them. The strategy balances the tradeoff between the costs of walking the LOB and targeting a percentage of filled orders. We employ prop data of FX trades to analyze the performance of the strategy and compute the price of latency that a trader would be willing to pay for co-location and hardware to reduce their latency.

1 - Consumer Protection in an Online World: When Does Occupational Licensing Matter? Chiara Farronato, Andrey Fradkin, Bradley Larsen, Erik Brynjolfsson

This paper studies the effects of occupational licensing on outcomes in a large online platform for home services, where professional service providers bid on consumers’ projects. We exploit exogenous variation in the time at which licenses are displayed on the platform to identify the causal effects of licensing information and reviews on consumer choices. We find that platform verified licensing status is not valued by consumers but digital reputation is. Next, we use zip-code by job category variation in licensing stringency to measure the effects of licensing on aggregate market outcomes. Our results show that more stringent licensing leads to less competition and higher prices, but does not improve customer satisfaction.

315