Latest Seminars

When Popularity Meets Position: Disentangling Popularity and Position Using an Experimental Approach
Ms. Qianran JIN, Jenny, McGill University

Date 06.02.2020
Time 10:30 am - 12:00 pm
Venue ISOM Conference Room, LSK 4047

To Compete or Contract? Assessing the Effectiveness of Contests in Online Labor Market
Dr. Jiahui MO, Assistant Professor, Nanyang Technological University

Date 31.01.2020
Time 10:30 am - 12:00 pm
Venue ISOM Conference Room, LSK 4047

Two-sided Adverse Selection and Bilateral Reviews in Sharing Economy
Mr. Murat M. TUNC, University of Texas at Dallas

Date 21.01.2020
Time 10:30 am - 12:00 pm
Venue ISOM Conference Room, LSK 4047

Optimal Commissions and Subscriptions in Networked Markets
Mr Kevin Hongfan Chen, Department of Operations Management, University of Chicago Booth

We consider a platform that charges commission rates and subscription fees to sellers and buyers for facilitating transactions but does not directly control the transaction prices, which are determined by the traders. Buyers and sellers are divided into types, and we represent the compatibility between different types using a bipartite network. Traders are heterogeneous in terms of their valuations, and different types have possibly different value distributions. Buyers may have additional value for trading with some seller types. The platform chooses commissions-subscriptions to maximize its revenues. Two salient features of most online platforms are that they do not dictate the transaction prices, and use commissions/subscriptions for extracting revenues. We shed light on how these commissions/subscriptions should be set in networked markets.

Using tools from convex optimization and combinatorics, we obtain tractable methods for computing the optimal commissions/subscriptions and provide insights on revenues and welfare. We provide a tractable convex optimization formulation to calculate the revenue-maximizing commissions/subscriptions, and establish that, typically, different types should be charged different commissions/subscriptions depending on their network positions. We establish that the latter result holds even when the traders on each side have identical value distributions, and in this setting we provide lower and upper bounds on the platform’s revenues in terms of the supply-demand imbalance across the network. Motivated by simpler schemes used in practice, we show that the revenue loss can be unbounded when all traders on the same side are charged the same commissions/subscriptions, and bound the revenue loss in terms of the supply-demand imbalance across the network. Charging only buyers or only sellers leads to at least half of the optimal revenues, when different types on the same side can be charged differently. Our results highlight the suboptimality of commonly used payment schemes, and showcase the importance of accounting for the compatibility between different user types. Under mild assumptions, we establish that a revenue-maximizing platform achieves at least 2/3 of the maximum achievable social welfare.

Date 21.01.2020
Time 3:00 pm - 4:15 pm
Venue Room G012, LSK Business Building

Dynamic Pricing of Relocating Resources in Large Networks
Mr Chen Chen, Fuqua School of Business, Duke University

Motivated by applications in shared vehicle systems, we study dynamic pricing of resources that relocate over a network of locations. Customers with private willingness-to-pay sequentially request to relocate a resource from one location to another, and a revenue-maximizing service provider sets a price for each request. This problem can be formulated as an infinite horizon stochastic dynamic program, but is quite difficult to solve, as optimal pricing policies may depend on the locations of all resources in the network. We first focus on networks with a hub-and-spoke structure, and we develop a dynamic pricing policy and a performance bound based on a Lagrangian relaxation. This relaxation decomposes the problem over spokes and is thus far easier to solve than the original problem. We analyze the performance of the Lagrangian-based policy and focus on a supply-constrained large network regime in which the number of spokes (n) and the number of resources grow at the same rate. We show that the Lagrangian policy loses no more than O(sqrt{ln n/n}) in performance compared to an optimal policy, thus implying asymptotic optimality as n grows large. We also show that no static policy is asymptotically optimal in the large network regime. Finally, we extend the approach to general networks with multiple, interconnected hubs and spoke-to-spoke connections, and to incorporate relocation times. We also examine the performance of the Lagrangian policy and the Lagrangian relaxation bound on some numerical examples, including examples based on data from RideAustin.

Date 20.01.2020
Time 3:00 pm - 4:15 pm
Venue Room 3005, LSK Business Building

Promotional Design for Small Businesses: The Operational Value of Online Deals
Ms Simin Li, Kellogg School of Management, Northwestern University

Among the limited ways for small service providers to balance demand and supply, launching temporary consumer offer may be attractive. However, relatively little work has empirically examined whether and how such offers pay off service providers. In this paper, using a comprehensive dataset from two leading deal platforms in China, we empirically study a new business model: the online deal. Service providers, who face predictable demand swings and capacity constraints, launch online deals for customers to prepay online and redeem later in store. Using a structural model, we show that online deals effectively facilitate demand-supply coordination through two levers, the discount and, more interestingly, the advance sales period. To our knowledge, using the advance sales period as revenue management tool has not been studied in the literature. Tailored to demand fluctuations and the service provider's operating margin, the advance sales period and the discount serve two separable operational roles to achieve profit maximization – adjusting demand mean and reducing variance-related costs. Furthermore, our model estimates enable us to quantify the operational value of the online deal. Via counterfactual analyses, we show that by using these two levers instead of solely a discount, 82.1% of the service providers see a mean profit improvement of 23.6%. The additional lever, advance sales period, helping to mitigate the extreme discounts is likely where the profit boosts come from.

Date 17.01.2020
Time 3:00 pm - 4:15 pm
Venue Room 1001, LSK Business Building

Learning Customer Preferences from Personalized Assortments
Mr Yifan Feng, School of Business, University of Chicago Booth

A company wishes to commercialize a single version of a product from a menu of alternative options. Unaware of true customer preferences, the company relies on a system that allows potential buyers to provide feedback on their preferred versions. Under a general ranking-based choice model framework, we study how to dynamically individualize the set of versions shown to each customer to collect feedback efficiently. We prove an instance-specific lower bound on the sample complexity of any policy that identifies the top-ranked version with a given (probabilistic) confidence. We also propose a robust formulation of the company's problem and derive a sampling policy (Myopic Tracking Policy), which is both asymptotically sample optimal and intuitive to implement. We conduct computational studies on both synthetic and real-life data to assess the performance of our proposed method and compare it to alternative methods proposed in the literature.

Joint work with René Caldentey (Chicago) and Chris T. Ryan (UBC).

Date 15.01.2020
Time 3:00 pm - 4:15 pm
Venue Room G012, LSK Business Building

Innovative Public Policies to Improve Social Welfare – from an OM/OR Perspective
Dr Jiayi Joey Yu, Department of Industrial Engineering & Operations Research, University of California, Berkeley

1. Improving Consumer Welfare and Manufacturer Profit via Government Subsidy Programs: Subsidizing Consumers or Manufacturers?

Most consumers in rural areas of many developing countries cannot afford to purchase certain livelihood improvement products such as home appliances. To improve consumer welfare and manufacturer profit, many governments launch different types of subsidy programs that offer subsidies to consumers, manufacturers, or both. Motivated by a subsidy program developed by the Chinese government in 2007, we present a parsimonious model to determine the optimal subsidy program in different settings.

2. A Balancing Act of Regulating On-demand Ride Services

Regulating on-demand ride-hailing services (e.g., Uber and DiDi) requires a balance of multiple competing objectives: encouraging innovative business models (e.g., DiDi), sustaining traditional industries (e.g., taxi), creating new jobs, and reducing traffic congestion. This study is motivated by a regulatory policy implemented by the Chinese government in 2017 and a similar policy approved by the New York City Council in 2018 that regulate the “maximum” number of registered Uber/DiDi drivers. We examine the impact of these policies on the welfare of different stakeholders.

Date 10.01.2020
Time 3:00 pm - 4:15 pm
Venue Room 1003, LSK Business Building

The Economics of Cyber Crime
Mr. Xiaofan LI, University of Texas at Austin

Date 09.01.2020
Time 10:30 am - 12:00 pm
Venue ISOM Conference Room, LSK 4047

Nudging Drivers to Safety: Evidence from a Field Experiment
Mr Vivek Choudhary, Department of Technology & Operations Management, INSEAD, Singapore

Driving is an integral component of many operational systems and any small improvement in driving quality can have significant effect on accidents, traffic, pollution, and the economy in general. However, the complexity and multidimensionality of driving as a task make it challenging to improve. With motor vehicles at the core of their business processes, many industries employ financial and non-financial incentives for their drivers to promote better driving. Such interventions are often expensive and not very effective as evident from accident statistics. Instead, we devise and test a behavioral intervention called a nudge in a field experiment. We use telematics technology (i.e., real-time sensor data in a mobile device such as accelerometer and gyroscope) to measure driving performance as well as to deliver nudges to the drivers via notifications. Leveraging a smartphone application launched by our industry partners, we sent three types of performance nudges to drivers, indicating how they performed with respect to their personal best, personal average, and latest driving performance. We are the first to study effectiveness of different types of nudges and compare their relative strength in the context of driving in a field experiment. We find that personal best and personal average nudges improve driving performance, on average, by 18.17% and 18.71% standard deviations of the performance scores calculated by the application respectively, translating into an increase in the inter-accident time by nearly 1.8 years, while also improving driving performance consistency (as measured by the coefficient of variation of the performance score). We further study heterogeneity of this effect using generalized random forest. We show that high-performing drivers who are not frequent feedback seekers benefit the most from personal best nudges, while low-performing drivers who are also frequent feedback seekers benefit the most from the personal average nudges. Using these findings, we construct personalized nudges that outperform either of these nudges.

Date 03.01.2020
Time 3:00 pm - 4:15 pm
Venue Room G003, LSK Business Building

Joint Statistics Seminar - Sufficient Dimension Reduction for Classification
Prof. Xin CHEN, Southern University of Science and Technology

We propose a new sufficient dimension reduction approach designed deliberately for high-dimensional classification.  This novel method is named maximal mean variance (MMV) stimulated by the mean variance index first proposed by Cui, Li and Zhong (2015) which measures the dependence between a categorical random variable with multiple classes and a continuous random variable.  Our method requires quite mild restrictions on the predicting variables and keeps the model-free advantage without the need to estimate the link function. Consistency of the MMV estimator is established under regularity conditions for both fixed and diverging dimension (p) cases and the number of the response classes can also be allowed to diverge with the sample size n.  We also construct the asymptotic normality for the estimator when the dimension of the predicting vector is fixed.  Furthermore, although without any definite theoretical proof, our method works pretty well when p << n.  Surprising classification efficiency gain of the proposed method is verified by numerical studies.

Date 02.01.2020
Time 11:00 am – 12:00 noon
Venue Room 5047 (LSK Business Building)

“Un”Fair Machine Learning Algorithms
Dr. Manmohan ASERI, Visiting Assistant Professor, Tepper School of Business, Carnegie Mellon University, USA

Date 15.11.2019
Time 10:30 am - 12:00 pm
Venue Case Room G001, G/F, LSK Business Building

Blockchain Adoption for Combating Deceptive Counterfeits
Dr Hubert Pun, Professor and PhD Program Coordinator, Ivey Business School, Western University

In this paper, we study combating deceptive counterfeiting using blockchain technology. When a product is tagged with a non-duplicable unique identifier customers know whether a product is authentic or fake. However, customers have privacy concerns while using blockchain. We consider a market with a manufacturer and a deceptive counterfeiter. The manufacturer can either use blockchain or signal through pricing to validate product authenticity. The government can incentivize blockchain adoption by providing subsidy to the manufacturer while optimizing social welfare. We find that without government subsidy, blockchain should be used only when the counterfeit quality is intermediate or when customers have intermediate distrust about products in the market. When customers have serious distrust about products, differential pricing strategy is more effective than blockchain. With subsidy, however, we show that differential pricing strategy should never be used. Blockchain can be more effective than differential pricing strategy in eliminating post-purchase regret, and our result advocates that government should participate in the manufacturer’s blockchain adoption decision because this benefits customers and society.

Key words: blockchain, customer privacy, deceptive counterfeit, government subsidy

Date 11.11.2019
Time 11:00 am - 12:15 pm
Venue Room G001, LSK Business Building

Bug Bounty Programs, Security Investment and Law Enforcement: A Security Game Perspective
Mr. Jiali Zhou, PhD student, ISOM

Date 01.11.2019
Time 10:30 am - 12:00 pm
Venue ISOM Conference Room 4047, LSK Business Building

Assigning Priorities (or not) in Service Systems with Nonlinear Waiting Costs
Dr Huiyin Ouyang, Assistant Professor, School of Business and Economics, The University of Hong Kong

For a queueing system with multiple types of customers differing in service time distributions and costs for waiting, it is well known that the rule is optimal if costs for waiting are incurred linearly with time. It is also known that when costs are convex increasing functions, the generalized rule, which is dependent on queued customers' already experienced waiting times, is asymptotically optimal under heavy traffic. In this paper, we seek to identify policies that minimize the long-run average cost under possibly nonlinear waiting cost functions and arbitrary traffic conditions but within the set of static policies that require information only on the type identities of jobs and their order of arrival. For a single-server queueing system with two types of customers, our main result gives conditions under which type-based priority policies and the first-come-first-serve (FCFS) policy can be ordered for non-decreasing cost functions that are first-order differentiable. We then apply this result to polynomial cost functions and obtain useful insights into when prioritization should be preferred over FCFS and when it should be avoided. For example, unlike in the linear-cost case, it could be better not to give priority to a certain type at all and employ FCFS under quadratic cost functions, especially when the traffic intensity is high. Finally, by means of a numerical study, we test how the best static policy compares with the generalized rule, which requires information on the current waiting times of customers and precise structure for the cost functions. In these experiments with quadratic waiting costs, we find that using the best static policy performs comparably with (sometimes even better than) the generalized rule except when the traffic intensity is high and there is not a clearly more “important” type in regards to dominance with respect to rates of cost and service.

Date 01.11.2019
Time 11:00 am - 12:15 pm
Venue Room 1001, LSK Business Building