Knowledge that Transforms

Problem definition: Qualcomm, the largest cellphone chipmaker in the world, had adopted a cross-licensing agreement with its clients, downstream cellphone manufacturers. It requires cellphone manufacturers to allow each other to use their patents for free. This cross-licensing practice has received considerable scrutiny and attention around the world. We study the impacts of cross-licensing in a supply chain in which an upstream supplier requires its downstream competing manufacturers to cross-license, where they are asymmetric in their innovation capabilities. Methodology/results: We build a stylized model of a supply chain consisting of one supplier and two competing manufacturers and conduct game-theoretic analysis. We find that the supplier always prefers adopting cross-licensing ex post after manufacturers’ investments are sunk, but it may prefer committing to no cross-licensing ex ante. Specifically, the supplier should commit to not using cross-licensing if the inferior manufacturer’s cost of innovation is high or the effectiveness of cross-licensing is high. Furthermore, cross-licensing may increase innovation, the superior manufacturer’s profit, and social welfare under certain conditions. Interestingly, when the superior manufacturer’s cost advantage is intermediate, the inferior manufacturer’s innovation level first increases and then decreases in the effectiveness of cross-licensing. In addition, the inferior manufacturer’s profit also first increases and then decreases as the effectiveness level of cross-licensing increases. The cross-licensing policy benefits consumers when its effectiveness level is low and the superior manufacturer’s innovation cost is either high or low. Managerial implications: Our results provide guidance on when a supplier should adopt the cross-licensing strategy. For policy makers, our findings show that cross-licensing can be beneficial for consumers and the society. In particular, to increase social welfare, policy makers may consider encouraging cross-licensing with low effectiveness level when the superior manufacturer’s innovation cost is either low or high.

Problem definition: Assortment selection is one of the most important decisions faced by retailers. Most existing papers in the literature assume that customers select at most one item out of the offered assortment. Although this is valid in some cases, it contradicts practical observations in many shopping experiences, both in online and brick-and-mortar retail, where customers may buy a basket of products instead of a single item. In this paper, we incorporate customers’ multi-item purchase behavior into the assortment optimization problem. We consider both the uncapacitated and capacitated assortment problems under the so-called Multivariate MNL (MVMNL) model, which is one of the most popular multivariate choice models used in the marketing and empirical literature. Methodology/results: We first show that the traditional revenue-ordered assortment may not be optimal. Nonetheless, we show that under some mild conditions, a certain variant of this property holds (in the uncapacitated assortment problem) under the MVMNL model; that is, the optimal assortment consists of revenue-ordered local assortments in each product category. Finding the optimal assortment even when there is no interaction among product categories is still computationally expensive because the revenue thresholds for different categories cannot be computed separately. To tackle the computational complexity, we develop FPTAS for several variants of (capacitated and uncapacitated) assortment problems under MVMNL. Managerial implications: Our analysis reveals that disregarding customers’ multi-item purchase behavior in assortment decisions can indeed have a significant negative impact on profitability, demonstrating its practical importance in retail. We numerically show that our proposed algorithm can improve a retailer’s expected total revenues (compared with a benchmark policy that does not properly take into account the impact of customers’ multi-item choice behavior in assortment decision) by up to 14%.

Problem definition: Whereas firms are increasingly relying on sourcing globally as a key constituent of their supply chain strategy, there is no empirical evidence on whether investors of these firms adequately reflect firms’ global sourcing strategy (GSS) in their stock-valuation process. In this paper, we empirically test whether stock market participants are efficient in doing so. Methodology/results: Using the empirical asset-pricing framework, we find that information concerning firms’ GSS strongly predicts their future stock returns. We compile a transaction-level imports database for U.S.-listed firms and construct measures for five widely studied GSS aspects in the operations management literature: the extent of global sourcing, supplier relationship strength, supplier concentration, sourcing lead time, and sourcing countries’ logistical efficiency. For each measure, we examine returns of a zero-cost investment strategy of buying from the highest and selling from the lowest quintile of that measure. Collectively, these investment strategies yield an average annual four-factor alpha of 6%–9.6% (6%–13.9%) with value (equal)-weighted portfolios. Their return predictability is incremental over other operations- and cost arbitrage–motivated predictors, such as inventory turnover, cash conversion cycle, and gross profitability; is persistent across different supply chain positions; and is robust to alternate risk models, subsamples, and empirical specifications. Together, our results indicate that the GSS measures embody independent information about firms’ future profitability, and this information is mispriced by market participants, leading to predictable returns. In accordance with this mechanism, we find that the GSS measures strongly predict both firms’ future earnings and the surprise in market reactions around the earnings announcement days. Managerial implications: The robust return predictability of our GSS measures suggests that investors are not fully incorporating GSS-related information in their stock valuation frameworks. Therefore, our results call for greater investor education on global sourcing and better dissemination of global-sourcing information so as to mitigate valuation inefficiency.

Problem definition: Personalized medicine (PM) seeks the best treatment for each patient among a set of available treatment methods. Because a specific treatment does not work well on all patients, traditionally, the best treatment was selected based on the doctor’s personal experience and expertise, which is subject to human errors. In the meantime, stochastic models have been well developed in the literature for a lot of major diseases. This gives rise to a simulation-based solution for PM, which uses the simulation tool to evaluate the performance for pairs of treatment and patient biometric characteristics and, based on that, selects the best treatment for each patient characteristic. Methodology/results: In this research, we extend the ranking and selection (R&S) model in simulation-based decision making to solving PM. The biometric characteristics of a patient are treated as a context for R&S, and we call it contextual ranking and selection (CR&S). We consider two formulations of CR&S with small and large context spaces, respectively, and develop new techniques for solving them and identifying the rate-optimal budget allocation rules. Based on them, two selection algorithms are proposed, which can be shown to be numerically superior via a set of tests on abstract and real-world examples. Managerial implications: This research provides a systematic way of conducting simulation-based decision-making for PM. To improve the overall decision quality for the possible contexts, more simulation efforts should be devoted to contexts in which it is difficult to distinguish between the best treatment and non-best treatments, and our results quantify the optimal trade-off of the simulation efforts between the pairs of contexts and treatments.

Problem definition: We consider a customer-to-customer (C2C) platform that provides an inspection service. Uncertain about product authenticity, a seller sells a product through the platform. Before purchasing, a buyer obtains a signal of the product authenticity from the product’s price set by the seller. The platform’s inspection service can detect a counterfeit with a probability. If the product passes the inspection, the platform sends it to the buyer and charges the seller a commission fee. Otherwise, the platform returns it to the seller and charges the seller a penalty fee. Methodology/results: We develop a two-stage game-theoretical model. In the first stage, the platform designs a contract specifying the commission and penalty fees. In the second stage, the seller signals the product authenticity by setting a price and the buyer decides whether to purchase it. This results in a contract design problem that governs a signaling game. We find that the effect of inspection is beyond merely detecting counterfeits. The inspection, even an imperfect one, changes the signaling game’s structure and incentivizes the seller whose product is likely authentic to sell through the platform. This can only be achieved by carefully choosing the commission and penalty fees. Moreover, a larger platform’s expected profit does not imply a larger commission fee or price in equilibrium. Under some mild conditions, the optimal commission increases but the optimal penalty decreases as the platform’s inspection capability improves. Managerial implications: The inspection service is not widely available among leading C2C platforms as it is considered imperfect and costly. Our study suggests that its benefit may be underestimated in practice. Moreover, the inspection can eliminate the seller’s information rent and generate more revenue for the platform. This paper provides guidance on how to set commission and penalty fees when the inspection service is provided. Funding: L. Li is supported by the National Natural Science Foundation of China [Grant 72071198] and the Hong Kong Polytechnic University Distinguished Postdoctoral Fellowship Scheme [Grant 1-YWC7]. X. Fang and Y. F. Lim are grateful for the support from the Lee Kong Chian School of Business, Singapore Management University [Maritime and Port Authority Research Fellowship and Retail Centre of Excellence Research Grant]. Y. F. Lim is supported by the Association of South-East Asian Nations Business Research Initiative Grant [Grant G17C20421], the Research Grants Council of Hong Kong [Grants 15501920 and 15501221], and the Key Program of National Natural Science Foundation of China [Grant 71931009].

Problem definition: We seek to provide an interpretable framework for segmenting users in a population for personalized decision making. Methodology/results: We propose a general methodology, market segmentation trees (MSTs), for learning market segmentations explicitly driven by identifying differences in user response patterns. To demonstrate the versatility of our methodology, we design two new specialized MST algorithms: (i) choice model trees (CMTs), which can be used to predict a user’s choice amongst multiple options, and (ii) isotonic regression trees (IRTs), which can be used to solve the bid landscape forecasting problem. We provide a theoretical analysis of the asymptotic running times of our algorithmic methods, which validates their computational tractability on large data sets. We also provide a customizable, open-source code base for training MSTs in Python that uses several strategies for scalability, including parallel processing and warm starts. Finally, we assess the practical performance of MSTs on several synthetic and real-world data sets, showing that our method reliably finds market segmentations that accurately model response behavior. Managerial implications: The standard approach to conduct market segmentation for personalized decision making is to first perform market segmentation by clustering users according to similarities in their contextual features and then fit a “response model” to each segment to model how users respond to decisions. However, this approach may not be ideal if the contextual features prominent in distinguishing clusters are not key drivers of response behavior. Our approach addresses this issue by integrating market segmentation and response modeling, which consistently leads to improvements in response prediction accuracy, thereby aiding personalization. We find that such an integrated approach can be computationally tractable and effective even on large-scale data sets. Moreover, MSTs are interpretable because the market segments can easily be described by a decision tree and often require only a fraction of the number of market segments generated by traditional approaches. Disclaimer: This work was done prior to Ryan McNellis joining Amazon.

Problem definition: We model the development of effective agricultural supply chains (agri-chains) in emerging economies for better utilization of land and intermediate processing resources for harvested export-oriented goods. We study decisions made by farmers, intermediate processors, and government officials in agri-chains. The structure and management of supply chains and government minimum guaranteed prices to farmers affect the performance of these chains and are in the domain of our study. Methodology/results: We develop models of agricultural supply chains in which yields are correlated across regions, and farmers sell to competing capacitated processors. The models have two types of fundamental decisions: determining how much land to allocate for planting before the start of a growing season and, determining the prices offered by competing processors that purchase the harvest. We develop analytical results and algorithmic approaches for finding resulting equilibria that depend on the nature of decision making and on the structure of yield uncertainty. In particular, for all-or-nothing yields, we characterize the ranges of minimum guaranteed prices that lead to farmers’ no-production, under-production, full-production and over-production equilibria. Analytical results supported by numerical experiments allow us to conclude that appropriately setting minimum price guarantees, with the exact definition of such ranges dependent on agri-chain characteristics, can lead to first-best supply chain solutions. Managerial implications: The analysis also suggests that some farmer co-operation in land allocation or regional integration of farmer-processing assets, together with moderate minimum guaranteed prices, might be implementable pathways for achieving agri-chain efficiency in emerging economies. In an interesting result, farmlands with yields of positive correlation tend to inhibit over-production, whereas those with negative correlations tend to induce over-production.

Problem definition: We study the estimation of the probability distribution of individual patient waiting times in an emergency department (ED). Whereas it is known that waiting-time estimates can help improve patients’ overall satisfaction and prevent abandonment, existing methods focus on point forecasts, thereby completely ignoring the underlying uncertainty. Communicating only a point forecast to patients can be uninformative and potentially misleading. Methodology/results: We use the machine learning approach of quantile regression forest to produce probabilistic forecasts. Using a large patient-level data set, we extract the following categories of predictor variables: (1) calendar effects, (2) demographics, (3) staff count, (4) ED workload resulting from patient volumes, and (5) the severity of the patient condition. Our feature-rich modeling allows for dynamic updating and refinement of waiting-time estimates as patient- and ED-specific information (e.g., patient condition, ED congestion levels) is revealed during the waiting process. The proposed approach generates more accurate probabilistic and point forecasts when compared with methods proposed in the literature for modeling waiting times and rolling average benchmarks typically used in practice. Managerial implications: By providing personalized probabilistic forecasts, our approach gives low-acuity patients and first responders a more comprehensive picture of the possible waiting trajectory and provides more reliable inputs to inform prescriptive modeling of ED operations. We demonstrate that publishing probabilistic waiting-time estimates can inform patients and ambulance staff in selecting an ED from a network of EDs, which can lead to a more uniform spread of patient load across the network. Aspects relating to communicating forecast uncertainty to patients and implementing this methodology in practice are also discussed. For emergency healthcare service providers, probabilistic waiting-time estimates could assist in ambulance routing, staff allocation, and managing patient flow, which could facilitate efficient operations and cost savings and aid in better patient care and outcomes.

Problem definition: Employers across many sectors of the economy have been fast to adopt variable work scheduling policies. The cost of this flexibility for employers is usually borne by employees, for whom unstable work schedules create several disruptions. In the context of home healthcare, we examine how employer-driven volatility in nurses’ schedules impacts their decision to voluntarily leave their job. Methodology/results: Using an instrumental variables approach, we causally identify the effect of schedule volatility on nurses’ voluntary turnover. We begin by constructing an operational measure of schedule volatility using time-stamped work log data from one of the largest home health agencies in the United States. Because this measure may be endogenous to the worker’s decision to quit, we instrument for schedule volatility using paid days off taken by other nurses in the same branch. We find that higher levels of schedule volatility substantially increase a worker’s likelihood of quitting. Specifically, a one-standard-deviation increase in schedule volatility increases the average worker’s propensity to quit on a given day by more than threefold. Translated into yearly terms, 30 days of high schedule volatility over the course of the year increases the average worker’s probability of quitting that year by 20%. Our policy simulations of counterfactual scheduling policies suggest that excess schedule volatility can explain a significant portion of voluntary turnover, and some interventions have the potential to substantially reduce workers’ daily propensity to quit. Managerial implications: This work contributes to the understanding of the extent to which employees value control over their own work schedules and are averse to volatile work schedules that are dictated by employers. Especially in the current environment where there is a growing emphasis on work-life balance and employee-driven flexibility, finding a way to support stable schedules could be important for employers to attract and retain workers.

Problem definition: We study the economic rationale for certified pre-owned (CPO) programs and characterize the mechanism through which they influence secondary markets to create value for durable goods firms and consumers. Academic/practical relevance: Our study on understanding the market impact of CPO programs is highly relevant for industry given their prevalence in practice. Our work also contributes to the adverse selection literature in the context of the “lemons problem” by uncovering the mechanism underpinning CPO programs and how they improve market efficiency. Methodology: We use a game-theoretic durable goods model that captures CPO program features to identify how CPO programs improve market efficiency. We further investigate how CPO program size, as determined by program qualification criteria, plays a role in improving market efficiency. Results: We show that the mechanism underpinning CPO programs consists of two mutually reinforcing effects: a more effective segmentation of used product buyers through high- and low-quality product differentiation and a higher trade-in discount to used product owners. The former effect generates a higher revenue, which, in turn, allows for a higher trade-in discount. This translates to more high-quality products being traded in, more new products being purchased, and more used products being sold in the secondary market (i.e., increased market efficiency). We also show that these two effects are strengthened as the CPO program size increases. Finally, we test hypotheses deriving from our theoretical results in the context of the U.S. automotive industry. Our empirical findings are consistent with the conclusions of our theoretical work. Managerial implications: Durable goods firms can proactively manage their CPO programs by appropriately setting trade-in discounts and managing price discrimination between CPO and non-CPO products. CPO program qualification criteria can be used to alter CPO program size and, hence, the level of market efficiency improvement induced by these choices.