Speaker: Yaroslav Rosokha, Ph.D. Candidate, University of Texas - Austin
Title: Learning Under Ambiguity: An Experiment
Abstract: We design and conduct an economic experiment to investigate the learning process of the agents under risk and under ambiguity. We gather data for subjects choosing between lotteries involving risky and ambiguous urns. Decisions are made in conjunction with a sequence of random draws with replacement, allowing us to estimate the beliefs of the agents at different moments in time. For each of the urn types we estimate a model of Bayesian updating allowing for base rate fallacies. Our findings suggest an important difference in updating behavior between risky and ambiguous environments. Specifically, when updating under ambiguity subjects significantly underweight the new signal, while when updating under risk subjects are essentially Bayesian.
Speaker: Yang Zhang, Ph.D. Candidate, Pennsylvania State University
Title: Network Effects on Decentralized Operations: A Laboratory Investigation
Abstract: How does the network structure affect decentralized decision making of firms or individuals being networked? We examine networks in which the connected agents (firms/individuals) engage in a coordination game where the player actions are complementary (e.g. technology adoption, consumer herding, firm collaboration). Agents know the number of local links they possess but otherwise have incomplete information about the global network structure. All the networks we study have the same pooling equilibria. Each also exhibits separating equilibria in which agent actions differ by her connectivity in the network. The behavior in our experiment exhibits a clear pattern of equilibrium selection. When network density is low, agents with more connections exhibit higher levels of coordination, with overall coordination levels consistent with separating equilibria. When network density is high, we observe nearly full coordination consistent with pooling equilibrium. Our findings suggest that both global and local networks impact strategic behavior and simple topological measures (e.g. network density and one’s number of connections) may be sufficient for predicting the level of successful coordination
Speaker: Yufei Ren, Visiting Assistant Professor of Economics, Department of Economics, Union College, NY
Title: Overconfidence in Newsvendor Orders: An Experimental Study
Abstract: Previous studies have shown that individuals make suboptimal decisions in a variety of supply chain and inventory settings. We hypothesize that one cause is that individuals are overconfident (in particular, overprecise) in their estimation of order variation. Previous work has shown theoretically that underestimating the variance of demand causes orders to deviate from optimal in predictable ways. We provide two experiments supporting this theoretical link. In the first, we elicit the precision of each individual's beliefs, and demonstrate that overprecision significantly correlates with order bias. We find that overprecision explains almost 1/3 of the observed ordering mistakes, and that the effect of overprecision is robust to learning and other dynamic considerations. In the second, we introduce a new technique to exogenously reduce overprecision. We find that participants randomly assigned to this treatment demonstrate less overprecision and less biased orders than those in a control group.
Speaker: Pengyi Shi, Ph.D. Candidate, School of Industrial and Systems Engineering, Georgia Institute of Technology
Title: Data-driven Modeling and Decisions for Hospital Inpatient Flow Management
Abstract: Emergency department (ED) overcrowding negatively impacts patient safety and public health, and hence, has become one of the most challenging problems facing healthcare delivery systems worldwide. It is known that prolonged waiting time for admitted patients to be transferred from ED to inpatient beds (i.e., ED boarding) is a key contributor to ED overcrowding. Our research focuses on gaining insights into effective inpatient flow management to reduce this waiting time, and eventually, to reduce ED overcrowding.
Based on an extensive empirical study of a Singaporean hospital, we build a new stochastic network model of inpatient flow. The model contains several novel features including the service times being endogenous, and these features are critical for the model to predict the time-dependent empirical performance measures such as the hourly average waiting time and the fraction of patients waiting more than 6 hours. By simulating the stochastic model, we identify certain operational policies that can reduce ED boarding and eliminate the excessively long waiting times for patients requesting beds in the morning. These policies focus on discharging patients at an earlier time of the day. The model also allows one to study the impact of other operational policies including staffing and expanding step-down-care facilities on ED boarding. To obtain structural insights, we further develop a novel “two-time-scale” analytical framework to analyze the model. This framework overcomes many challenges, including the service times being extremely long compared to the time-variations of the arrival rate, faced by existing methods for large-scale queuing systems. In addition to exact analysis, we employ a heavy-traffic approximation. Finally, we discuss future directions for research and practice.
Speaker: Agnieszka Tymula, AXA Post-Doctoral Fellow,Center for Neuroeconomics, New York University
Title: Separating Risk and Ambiguity Preferences Across the Life Span: Novel Findings and Implications for Policy
Abstract: We experimentally examine how attitudes toward risk and ambiguity, as well as other properties of the decision-making process such as choice consistency and frequency of dominance violations, change over the life span. We assess all of these characteristics of the decision process, both in the gain and in the loss domain, in a sample of 135 subjects that range in age from 12 to 90 years old. We also collected extensive demographic information and psychological measures that we use as covariates in our analysis. Our results indicate that there is substantial and domain-specific variability across subjects in all parameters with indications of age-related differences in both risk and ambiguity preferences as well as choice consistency and frequency of dominance violations. Several of our findings challenge widely held assumptions about the preferences of representative agents.