Informs Annual Meeting 2017

MC26

INFORMS Houston – 2017

4 - Robustness Analysis of Power Grid under False Data Attacks Against AC State Estimation Ming Jin, jinming@berkeley.edu, Javad Lavaei The integration of sensing and information technology renders the power grid vulnerable to cyber-attacks. To understand how robust state estimators are, we study their behavior under worst attacks possible. Three types of data injection attacks are formulated in an optimization with “unobservability” and “sparse attack” constraints. A novel treatment of the continuous nonconvexity and discrete nonlinearity of the problem is proposed based on semidefinite programming (SDP) relaxation with a sparsity penalty, where a rank-1 and sparse solution exists under mild conditions. In light of the analysis, measures to improve the robustness of power grid are discussed. 350B Innovative Methods in Productivity and Efficiency Measurement Invited: Data Envelopment Analysis Invited Session Chair: Abolfazl Keshvari, Aalto University School of Business, Helsinki, 00100, Finland, abolfazl.keshvari@aalto.fi 1 - A Comparative Study of Alternative Approaches to Estimate Productivity Joseph Atwood, Montana State University, Bozeman, MT, United States, jatwood@montana.edu, Saleem Shaik The main objective of the paper is to estimate & compare productivity measures between & within three approaches for single output-input technology. The following models are considered - index (Tornqvist and Fisher), programming (Malmquist Input, Output & Graph; Malmquist total factor productivity) and parametric (Input & Output; total factor productivity). Results support differences in productivity growth rate across approaches due to the composition of the frontier and the underlying assumptions. However, within nonparametric (parametric) approach, differences are due to the use of distance functions (residuals) or ratio of conditional distance functions (conditional residuals). 2 - Estimating Parameter Uncertainty in the Stoned Model Endre Bjørndal, Norwegian School of Economics, Bergen, Norway, Endre.Bjorndal@nhh.no, Ørjan Mydland, Jonas Andersson We investigate the performance of estimators for parameter uncertainty in the StoNED model (Kuosmanen and Kortelainen, 2012). More specifically, we study the performance of the bootstrap for making inference on the estimation of model coefficients. The estimators, based on different versions of the bootstrap, are investigated by means of a Monte Carlo study. We apply the method to a dataset for Finnish electricity distribution companies, and we discuss the relevance and importance of measuring the uncertainty in cost function coefficient estimates in a regulatory context. 3 - A Data Mining Approach to Hedonic Pricing Model via MC26

conversations and web traffic for the brand, with audiovisual placements yielding the largest increases. We also find that television advertising can enhance the impact of product placement on social media activity. 2 - Exit, Voice, and Response in Digital Platforms: An Empirical Investigation of Online Management Response Strategies Naveen Kumar, University of Memphis, Memhis, TN, United States, nkumar7@memphis.edu, Liangfei Qiu, Subodha Kumar In the past decade, we have witnessed the growing importance of management responses to online reviews in digital platforms. In this study, we examine the impact of such responses on business performance. We also uncover the spillover effect of such responses on the nearby businesses and identify conditions under which business owners are more likely to respond to consumer comments. Our findings have direct implications for both the business owners and digital platforms that can help businesses develop the right engagement strategies by taking the online response strategies into account. 3 - Influence of Flavors on the Propagation of Electronic Cigarettes Related Information Jiaqi Zhou, City University of Hong Kong, 83 Tat Chee Avenue, 6/F, Academic 1, Kowloon, 12180, Hong Kong, jiaqi.zhou@my.cityu.edu.hk, Qingpeng Zhang, Daniel Dajun Zeng, Kwok-Leung Tsui Depicting the influence of e-cigarette flavors on information propagation is critical for understanding the use of e-cigarettes and the reception of flavors. Modeling such influence could provide quantitative policy decision support concerning smoking initiation and contagion, and e-cigarettes regulations. This study aims to characterize the influence of flavors on the propagation of e-cigarette related information on social media. The occurrence of three flavors (coffee, fruits, and herbs&spices) did not enhance the popularity of information, indicating users’ lack of interests in flavors. Invited: Auctions Invited Session Chair: Thayer Morrill, North Carolina State University, Raleigh, NC, 27695, United States, thayermorrill@gmail.com 1 - The Simple Structure of Top Trading Cycles in School Choice Irene Yuan Lo, Columbia University, New York, NY, 10025, United States, iyl2104@columbia.edu, Jacob Leshno Top Trading Cycles (TTC) is attractive for school choice, as it is strategy-proof, Pareto efficient, and allows school boards to guide the allocation via priorities. However, the combinatorial description of TTC makes it non-transparent to parents and designers. We give a simple characterization of TTC for school choice: the TTC assignment is given by n2 admission thresholds, where n is the number of schools, and these thresholds can be easily observed after running the mechanism. To calculate these thresholds, we define TTC through trade balance equations, which correspond in a continuum model to a differential equation. The model allows us to compute comparative statics and evaluate welfare. 2 - Communication Requirements and Informative Signaling in Matching Markets Peng Shi, Assistant Professor, USC Marshall School of Business, 3670 Trousdale Pkwy, Los Angeles, CA, 90089, United States, pengshi46@gmail.com, Itai Ashlagi, Mark Braverman, Yash Kanoria We study how much communication is needed to find a stable matching. Previous results show that any protocol must incur a high communication cost per agent, which scales linearly in the number of agents. On the other hand, we show that when the distribution of preferences satisfy natural restrictions, the amount of communication needed is much less. We give efficient protocols based on giving agents an estimate of whom they have realistic chances of matching, and recommending that agents reach out to easier-to-match partners while waiting for harder-to-match partners to reach out to them. Our protocols have good incentive properties and give insights on how to reduce congestion in matching markets. MC28 350D Matching and Assignment

Nonparametric Piecewise Linear Regression Abolfazl Keshvari, Abolfazl, Helsinki, Finland, abolfazl.keshvari@aalto.fi

This paper first reviews the main developments in nonparametric piecewise linear regressions. The statistical properties and applications of a family of such models are discussed. Numerical techniques to solve the related problems are also explained. Next, nonparametric concave regression is used to build a hedonic pricing model, which determines similar pricing strategies and categorizes observations accordingly. Advantages of this model are demonstrated and it is compared with similar approaches.

MC27

350C Marketing Insights from Social Media Invited: Social Media Analytics Invited Session Chair: Ali Hojjat, University of New Hampshire, 10 Garrison Ave., Durham, NH, 03824, United States, ali.hojjat@unh.edu 1 - The Impact of Product Placement on Brand-related Social Media Conversations and Website Traffic Beth Fossen, Indiana University, 3013 E.Daniel St, Bloomington, IN, 47401, United States, bfossen@indiana.edu, David Schweidel Despite the popularity of product placement, there is limited research that empirically explores the impact of placement activities in television programs. We examine the effect of television product placements on the volume of social media activity and web traffic for the placed brand. Using data on 2,800 product placements, we find that placements contribute to both more online

3 - Competitive Equilibria in School Assignment Thayer Morrill, North Carolina State University, Campus Box 8110, Raleigh, NC, 27695, United States, thayermorrill@gmail.com, Umut Dur

Top Trading Cycles was originally developed as an elegant method for finding a competitive equilibrium of Shapley and Scarf’s housing market. We extend the definition of a competitive equilibrium to the school assignment problem and show that there remains a profound relationship between Top Trading Cycles and a competitive equilibrium. Specifically, in every competitive equilibrium with weakly decreasing prices, the equilibrium assignment is unique and exactly corresponds to the Top Trading Cycles assignment. This provides a new way of interpreting the worth of a student’s priority at a given school. It also provides a new way of explaining Top Trading Cycles to a school board.

212