Informs Annual Meeting Phoenix 2018

INFORMS Phoenix – 2018

SC53

3 - Impact of Fake News and Ambiguity on the Equity Value of Social Media Platforms - Evidence from Twitter Srikar Velichety, University of Memphis, Memphis, TN, 38111, United States, Utkarsh Shrivastava We investigate the impact of Fake News and the ambiguity in its detection on the equity value of Social Media platforms. Using prior research, we develop hypotheses about the negative impact of fake news and ambiguity on the medium where it is incubated and propagated. We test these hypotheses using a large scale annotated dataset of real-world events posted to Twitter over a 140-day period. Using extended vector auto-regression, we estimate the impact of the presence of fake news and the ambiguity it causes on the short and long-term equity value of social media. Our results show that the presence of tweets about a single fake event is linked to a decrease in market return of the social media platform by 0.035% and increase the risk by 0.005%. 4 - Using DOE and Social Media to Spread Policy Information Theodore T. Allen, Ohio State University, 210 Baker Systems, 1971 Neil Ave, Columbus, OH, 43210-1271, United States We illustrate how experimental design can aid in directing social media campaigns to target voter types most receptive to policy information attitudinally and informationally. Both voter survey and google analytics experiments are studied together with resolution V designs and optimal alternatives. Issues relate to the environment, taxes, marijuana, and guns. n SC53 North Bldg 232A Pricing and Equilibrium Finding in Combinatorial Markets Sponsored: Auction and Marketing Design Sponsored Session Chair: Benjamin Lubin, Boston University, Boston, MA, 02215, United States Co-Chair: Sven Seuken, University of Zurich, Zurich, 8050, Switzerland 1 - Revealed Preference and Activity Rules in Dynamic Auctions We provide a general treatment of activity rules in auctions: constraints that limit bidding in future rounds based on past bids. Traditional point-based activity rules are effective for homogeneous goods, but are simultaneously too strong and too weak for general environments. Rules operationalizing the generalized axiom of revealed preference (GARP) enable straightforward bidding. We prove they are the weakest rules that prevent weak axiom (WARP) violations while never producing ``dead ends’’. In addition, GARP rules are robust to limited amounts of learning. We also provide empirical examples where bidders generally comply with GARP and violations suggest strategic manipulation. 2 - Adaptive Price Combinatorial Auctions Benjamin Lubin, Boston University, School of Management, 595 Commonwealth Avenue, Boston, MA, 02215, United States, Sebastien Lahaie We introduce a novel iterative combinatorial auction based on adaptive polynomial pricing. The mechanism starts with linear prices; then upon provably detecting that the current price structure cannot clear the market, expands the structure to guarantee progress. We provide theoretical and experimental evidence for the effectiveness of the design. 3 - Spectrum Repacking in the Incentive Auction Neil Newman, University of British Columbia, 1590 West 15th Avenue, Unit 5, Vancouver, BC, V6J 2K6, Canada, Kevin Leyton-Brown, Paul Milgrom, Ilya Segal In 2016-17 the FCC conducted an “incentive auction” to repurpose radio spectrum from broadcast television to wireless internet. The auction yielded $19.8 billion, $10 billion of which was paid to broadcasters for voluntarily relinquishing their licenses. A crucial element of the auction design was the construction of a solver, dubbed SATFC, that determined whether sets of stations could be “repacked” in this way; it needed to run every time a station was given a price quote. To evaluate the impact of our solver, we built an open-source reverse auction simulator. We found that SATFC substantially outperformed other alternatives at national scale. Oleg V. Baranov, University of Colorado-Boulder, 256 UCB, Boulder, CO, 80309, United States, Lawrence M. Ausubel

4 - Computing Bayes Nash Equilibria in Combinatorial Auctions Sven Seuken, University of Zurich, Binzmuhlestrasse 14, Zurich, 8050, Switzerland, Vitor Bosshard, Benedikt Buenz, Benjamin Lubin We present two new algorithms for computing symmetric pure epsilon-BNEs in CAs with continuous values and actions. We evaluate our algorithms in the well- studied LLG domain, against a benchmark of 16 CAs for which analytical BNEs are known. Furthermore, for CAs with quasi-linear utility functions and independently distributed valuations, we derive a theoretical bound on epsilon. Finally, we introduce the new Multi-Minded LLLLGG domain with eight goods and six bidders and apply our algorithms to finding an equilibrium in this domain. Our algorithms are the first to find an accurate BNE in a CA of this size. BOM Best Working Paper Competition Sponsored: Behavioral Operations Management Sponsored Session Chair: Julie Niederhoff, Syracuse University, Syracuse University, Syracuse, NY, 13244, United States 1 - Increased Transparency in Procurement: The Role of Peer-Effects Ruth Beer, Indiana University, Bloomington, IN, USA, Ignacio Rios, Ruth Beer. We study the effects of increased transparency in a setting where purchasing decisions are delegated to individual employees as opposed to being centrally managed by an organization. We develop a theoretical model which incorporates peer effects resulting from transparency and we show that there exists a spillover region where an employee is more likely to choose the expensive supplier when he observes that his peer did so. We confirm this finding with a laboratory experiment. We also find that employees whose decisions are observed are less likely to choose the expensive supplier, in line with what the social norm of appropriate behavior prescribes. 2 - Private Information and Endogenous Matching in Supply Chains Kyle Hyndman. University Texas-Dallas, Richardson, TX, USA, Andrew M. Davis We investigate a supply chain setting where a supplier’s cost may be private information (but they may disclose it) and buyers and suppliers may endogenously match into pairs. After forming pairs, the two parties engage in a dynamic bargaining setting. Suppliers always make less than theory predicts, whether their cost is known or private information. This effect is especially pronounced under private information for high cost suppliers, because buyers make more aggressive bargaining offers in such a setting. Thus, contrary to theory, a second result is that higher cost suppliers actually benefit from disclosing their private costs, in an effort to achieve a more favorable outcome while bargaining. Researchers in Behavioral Operations Management have made considerable progress on documenting the random error inherent to human decision behavior and incorporating it into Operations Management models. However, the field is also fundamentally interested in managing processes, and comparably little work has studied how noisy behavior plays out in sequences of unique linked decisions. In this paper, we study the fundamental choose-forecast-invest task combination: a decision maker must (a) choose which product to sell, (b) make a demand forecast for the chosen product, and (c) make an investment decision for the chosen product. Through behavioral models and laboratory experiments, we show how adding unbiased random noise to the process leads to a downstream systematic bias of overinvestment. This bias arises due to the way that random noise is filtered through the sequence of linked decisions and statistical naivety on the part of the decision-maker. Random noise in the demand forecast increases the overinvestment bias, but random noise in the product choice actually decreases the bias. Consequently, under certain parameters, being less rational (i.e., more random) in product choice can actually yield higher profits. Tasks in which information requires greater human interpretation are shown to increase the bias. Finally, employing a task-decomposition approach, we examine when separating product choice and investment decision-making across different people can reduce the bias and improve performance. 3 - From Noise to Bias in Linked Supply Chain Decisions Daniel Feller, Dartmouth College, Hanover, NH, USA, Jordan D. Tong n SC54 North Bldg 232B

80