Informs Annual Meeting 2017

MA03C

INFORMS Houston – 2017

substitutive choices. This approach enables us to recover proven, near global optimal parameter values with respect to the chosen loss-minimization objective function. We test the method on simulated and real data, and present results for a variety of single- and multi-item demand prediction scenarios, and for learning the unobserved market shares of competitors. 2 - Demand Estimation under the Multinomial Logit Model from Sales Transaction Data Tarek Abdallah, New York University, Stern School of Business, New York, NY, United States, tabdalla@stern.nyu.edu, Gustavo Jose Vulcano We study a multinomial logit (MNL) model of demand when customers arrive over time in accordance to a non-homogeneous Poisson process. The model is suitable for retail settings where only sales and product availability data are recorded, not all products are displayed in all periods, and the seller has information about her own aggregate market share. We characterize conditions under which the model is identifiable and the maximum likelihood estimates are consistent. We propose a Minorization-Maximization (MM) algorithm that is guaranteed to converge to the global optimal solution when the model is identifiable. 3 - Salesforce Incentives for Managing Product Returns Rashmi Sharma, Pennsylvania State University, University Park, PA, United States, rashmi.sharma@psu.edu, Aydin Alptekinoglu We study a setting where product sales and returns are effort-dependent and the selling activity is conducted by a salesforce. We investigate the effect of salesforce behavior on net sales and compare different incentive schemes to identify optimal incentive strategies. 4 - An Expectation-maximization Algorithm to Estimate the Parameters of the Markov Chain Choice Model A. Serdar Simsek, University of Texas-Dallas, Naveen Jindal School of Management, 800 W. Campbell Road, Richardson, TX, 75080, United States, serdar.simsek@utdallas.edu, Huseyin Topaloglu We develop an expectation-maximization algorithm to estimate the parameters of the Markov chain choice model (MCCM). We prove theoretical convergence guarantees and our computational experiments show that the MCCM, coupled with our expectation-maximization algorithm, can yield better predictions of customer choice behavior when compared with other commonly used alternatives. MA03C Grand Ballroom C Behavioral Modeling in Operations Sponsored: Manufacturing & Service Oper Mgmt Sponsored Session Chair: Andrew M. Davis, Cornell University, Ithaca, NY, 14853, United States, adavis@cornell.edu 1 - Heterogeneity of Reference Effects in the Competitive Newsvendor Problem Anton Ovchinnikov, Queen’s School of Business, 143 Union Street, Kingston, ON, K7L.3N6, Canada, ao37@queensu.ca, Sam Kirshner This paper examines two recently-proposed reference effect formulations for the newsvendor problem and extends them to a competitive setting. The analysis of the resultant game shows that the heterogeneity of newsvendors’ reference effects can explain multiple regularities observed in recent experimental studies of newsvendor competition. Specifically, the observations that a behavioral newsvendor may effectively ignore the orders of the competitor, receive a significantly smaller profit, and over-order when there is no expected demand overow can all be attributed to the heterogeneous reference effects in our model’s equilibrium. 2 - Omnichannel Service Operations with Online and Offline Self-order Technologies Fei Gao, Indiana University, Bloomington, IN, United States, fg1@iu.edu, Xuanming Su Many restaurants have recently implemented self-order technologies across both online and offline channels. We study the impact of both online and offline self- order technologies on consumer demand, employment levels, and restaurant profits. 3 - Predicting Order Variability in Inventory Decisions: A Model of Forecast Anchoring Dayoung Kim, Cornell University, 201D Sage Hall, Ithaca, NY, 14853, United States, kdy928@gmail.com, Andrew M. Davis, Li Chen We develop a forecast anchoring model that explains and predicts order variability behavior in a multi-period newsvendor problem. Our model assumes that people anchor on random point forecast and insufficiently adjust toward the

profit-maximizing quantity. We apply GMM to fit our model across different experimental data sets, and further conduct our own experiment for a robustness check. Our results show that the order variability at the individual level and the variability at the aggregate level is caused by different sources of behavior. By providing better predictions, our model can help firms better anticipate the order variability from the downstream buyers. 4 - How Bounded Rationality Generates Bias in Demand Forecasting and Product Choice Jordan Tong, University of Wisconsin-Madison, Madison, WI, United States, jordan.tong@wisc.edu, Daniel Feiler Humans are boundedly rational and make random error. Can initial random error, even if it is unbiased, generate future systematic decision bias? We uncover and model an important operations context for which the answer is yes: when managers both choose which product to sell and make a forecast for their chosen product, random error (and failure to account for it statistically) causes them to systematically forecast too high. Two experiments, one with managers in a single- period setting and the other with students in a periodic review setting, test the behavioral model’s predictions - illustrating how and why human random error causes bias and illuminating mitigation techniques. 320A Operations in Electricity Supply Chains Sponsored: Manufacturing & Service Oper Mgmt, Sustainable Operations Sponsored Session Chair: Ozge Islegen, Kellogg School of Management, Kellogg School of Management, Evanston, IL, 60208, United States, o-islegen@kellogg.northwestern.edu Chair: Asligul Serasu Duran, Kellogg School of Management, Kellogg School of Management, Chicago, IL, 60657, United States, a-duran@kellogg.northwestern.edu 1 - An Analysis of Demand Response Programs in the Wholesale Electricity Market Asligul Serasu Duran, Kellogg School of Management, 825 W. Cornelia Avenue, Apt GDN, Chicago, IL, 60657, United States, a-duran@kellogg.northwestern.edu, Baris Ata, Ozge Islegen This project explores the impact of the participation and compensation of demand response (DR) providers in the wholesale electricity market on the electricity supply, electricity prices, generators and DR providers’ profits, and consumer welfare. Specifically, we model a supply function equilibrium for generators and DR providers where DR resources offer “negawatts” in the wholesale market. Then, we analyze the change in the electricity supply, and in the welfare of the market participants due to the varying compensation rates of DR providers. We show that generators’ profits decrease, and consumers’ energy bills may increase with the entry of DR providers to the wholesale electricity market. 2 - Government Financing for Clean Technology Development MA04 We study government financing for a firm’s clean technology development under a financial constraint. The main interest of this paper is to investigate the impact of government financing on environment and the firm’s bankruptcy risk when market uncertainty exists. Our analysis shows that government financing improves environment, compared to bank financing. Surprisingly, the government financing, however, exposes the firm to higher bankruptcy risk. However, our analysis further shows that the government financing benefits both the firm and consumers. Hence, our results also shed light on the benefit of the government financing. 3 - Mind the Gap: Coordinating Energy Efficiency and Demand Response Eric Webb, PhD Student, Indiana University, 2915 W. Winterberry Ct., Bloomington, IN, 47404, United States, eric.michael.webb@gmail.com, Owen Wu, Kyle D.Cattani Traditionally, energy demand-side management programs are each designed in isolation. Breaking with this tradition, we examine the interactions between long- term energy efficiency upgrades and daily demand response participation at an industrial firm. Upon re-examining the long-studied energy efficiency gap, we find that the gap is typically smaller when demand response is considered. Policies aiming to close or reduce the energy efficiency gap, such as investment subsidies and carbon taxes, may fail to achieve their desired outcomes when firms participate in demand response. Seung Hwan Jung, Washington University in St. Louis, 1552 E. Swan Cir., Brentwood, MO, 63144, United States, seunghwan.jung@wustl.edu, Lingxiu Dong

132