Key Areas of Research
Learning from Earnings Calls: Graph-Based Conversational Modeling for Financial Prediction
Information Systems Research
Earnings conference calls are valuable venues for business communication. Empirical research has shown that the content of earnings calls contains predictive signals about future market risks, which has motivated a line of computational studies that utilize earnings transcripts for financial forecasting tasks. However, earnings call transcripts are typically very long, and the predictive signals within them are often sparsely distributed across different sections of the transcript. As a result, existing computational methods often fail to capture the essential information within the transcript that is relevant to market risks. In this work, we design a novel method to model earnings transcripts as a conversational graph where graph nodes represent discussed topics and graph edges indicate the similarity between topics. By doing so, we aim to explicitly model what is discussed (i.e., topical content), how it is discussed (e.g., cross-referencing or newly introduced topics), and in what manner it is discussed (e.g., sentiment and other linguistic features) within the transcript. We then leverage a graph neural network to learn transcript-level representations for financial risk forecasting. Experimental results show that the proposed method significantly reduces risk forecasting errors, demonstrating its capability of modeling earnings call transcripts. Moreover, this predictive power holds even after considering the firm’s fundamentals, and the performance gain over baseline models continues to grow as transcript length increases. The interpretability analysis shows that the proposed method identifies cross-referencing and newly introduced topics as influential for risk prediction. Moreover, the model tends to associate transcripts with a higher number of new topics in the Q&A section, more cross-referencing discussions, and more positive sentiment with lower predicted financial risks. This work contributes methodologically by proposing a novel predictive approach specifically tailored to the domain-specific challenge of transcript-based risk forecasting. We also discuss key design insights and implications.
Yi Yang (HKUST), Yixuan Tang (HKUST), Yangyang Fan (HK PolyU), and Kunpeng Zhang (UMD)
Market Formation, Pricing, and Revenue Sharing in Ride Hailing Services
Manufacturing & Service Operations Management, September 2025
Problem definition: We empirically study the market for ride-hailing services. In particular, we explore the following questions: (i) How do the two-sided market and prices jointly form in ride-hailing marketplaces? (ii) Does surge pricing create value and for whom? How can its efficiency be improved? (iii) Can platforms' strategy on revenue sharing with drivers be improved? (iv) What is the value generated by ride-hailing services, including hosting rival taxi services on ride-hailing apps? Methodology/Results: We develop a discrete choice model for the formation of mutually dependent demand (customer side) and supply (driver side) that jointly determine pricing. Using this model and a comprehensive data set obtained from the largest mobile ride platform in China, we estimate customer and driver price elasticities and other factors that affect market participation for the company's two main markets, namely basic ride-hailing and Taxi services. Based on these estimation results and counterfactual analysis, we demonstrate that surge pricing improves customer and driver welfare as well as platform revenues, while counterintuitively reducing Taxi revenues on the platform. However, surge pricing should be avoided during non-peak hours as it can hurt both customer and platform surplus. We show that platform revenues can be improved by increasing drivers' revenue share from the current levels. Finally, we estimate that the platform's basic ride-hailing services generated customer value equivalent to 13.25 Billion USD in China in 2024, and hosting rival Taxi services on the platform boosted customer surplus by 3.6 Billion USD. Managerial Implications: Our empirical framework provides ride-hailing companies a way to estimate demand and supply functions, which can help with optimization of multiple aspects of their operations. Our findings suggest that ride-hailing platforms can improve profits by containing surge-pricing to peak hours only and boosting supply by increasing driver compensation. Finally, our results demonstrate that restricting ride-hailing services create significant welfare losses while including taxi services on ride-hail platforms generate substantial economic value
Liu Ming, Tunay I. Tunca, Yi Xu, and Weiming Zhu
Unintended Consequences of Closing Pay Gaps Across Multiple Groups: A Formal Modeling and Simulation Analysis of Allocation Methods
Organization Science, October 2025
In recent years, many firms have prioritized both pay equity (i.e., closing pay gaps associated with target groups such as women and racial minorities) and equitable representation (i.e., ensuring these target groups are fairly represented across a firm’s hierarchy). We use formal modeling and simulations to show how efforts to close pay gaps across multiple groups can undermine equitable representation. Specifically, our analysis suggests that pressure for pay equity creates a cost-based financial incentive to enact a subtle form of tokenism: A firm may minimize the cost of closing pay gaps if it maintains a workforce with a small number of minority women whom it pays well in order to compensate for underpaying larger numbers of majority women and minority men who resemble each other in terms of job attributes and personal qualifications. A firm can avoid these outcomes if it focuses on ensuring that employees from target groups are equitably rewarded for job attributes and personal qualifications rather than minimizing cost. But an equitable-rewards approach can be substantially more expensive than a cost-minimization approach, especially if pay gaps are larger in high-wage jobs or if there are many target groups. We conclude by offering testable empirical predictions and recommending a practical solution, namely to include terms for intersectional categories (e.g., minority women) in the regressions used to estimate pay gaps.
David Anderson (Villanova University); Margret Bjarnadottir (University of Maryland);
David Ross (University of Florida)
Cost-Saving Synergy: Energy Stacking in Battery Energy Storage Systems
Management Science
Despite the great potential benefits of battery energy storage systems (BESSs) to electrical grids, most standalone uses of BESS are not economical due to batteries’ high upfront costs and limited lifespans. Energy stacking, a strategy of providing two or more services with a single BESS, has been of great interest to improve profitability. However, some key questions, for example, the underlying mechanism by which stacking works or why and how much it may improve profitability, remain unanswered in the literature. Using two popular battery services, we analytically show that there often exists cost-saving synergy -- the cost of performing both services at the same time (simultaneous stacking) is smaller than the sum of individual costs if we had performed each service alone -- which allows for bigger profits. Furthermore, we perform comparative statics on the optimal mix of the services to systemically characterize grid/market conditions that maximize/minimize this synergy. We also derive a theoretical upper bound on simultaneous stacking’s benefits, showing that it can approximately double the profit of the best standalone service. Several generalizations of the base model not only show that the main lessons continue to hold but also that stacking’s benefits may become even stronger.
Joonho Bae (Indiana University), Roman Kapuscinski (University of Michigan), John Silberholz (UMD)
Can Employees' Past Helping Behavior Be Used to Improve Shift Scheduling? Evidence from ICU Nurses
Management Science, November 2025
Employees routinely make valuable contributions at work that are not part of their formal job description, such as helping a struggling coworker. These contributions, termed organizational citizenship behavior, are studied from many angles in the organizational behavior literature. However, the degree to which the past helping behavior of employees scheduled to a shift impacts that shift’s operational outcomes remains an underexplored question. We define two measures of past helping behavior for members of a shift -- the total past helping of each employee and the past helping between each pair of employees -- and hypothesize that they are associated with shift performance. We empirically confirm our hypotheses with detailed scheduling and patient outcome data from six intensive care units (ICUs) at a large academic medical center, using the hospital’s electronic medical records to identify cases of one nurse helping another. Our empirical results indicate that both measures of past helping are predictive of patient length of stay (LOS), more so than the broadly studied notion of team familiarity. Counterfactual analysis shows that relatively small changes in shift composition can yield significant reduction in total LOS, indicating the managerial significance of the results. Overall, our study suggests the potential value of shift scheduling using data on past helping behaviors, and this may have promise far beyond the selected application to ICU nursing.
Zhaohui (Zoey) Jiang (CMU), John Silberholz (UMD), Yixin (Iris) Wang (UIUC), Deena Kelly Costa (Yale), Michael Sjoding (University of Michigan)
Celebrity messages reduce online hate and limit its spread
Online hate spreads rapidly, yet little is known about whether preventive and scalable strategies can curb it. We conducted the largest randomized controlled trial of hate speech prevention to date: a 20-week messaging campaign on X in Nigeria targeting ethnic hate. 73,136 users who had previously engaged with hate speech were randomly assigned to receive prosocial video messages from Nigerian celebrities. The campaign reduced hate content by 2.5% to 5.5% during treatment, with about 75% of the reduction persisting over the following four months. Reaching a larger share of a user's audience reduced amplification of that user's hate posts among both treated and untreated users, cutting hate reposts by over 50% for the most exposed accounts. Scalable messaging can limit online hate without removing content.
Eaman Jahani, Assistant Professor, UMD
Blas Kolic, Post-doc, Universidad Carlos III de Madrid
Manuel Tonneau, PhD Student, Oxford University
Hause Lin, Post-doc, MIT
Daniel Barkoczi, University of Southern Denmark
Edwin Ikhuoria, Middlesex University
Victor Orozco, World Bank
Samuel Fraiberger, World Bank and NYU
User Innovation and Product Stickiness: Evidence from Video Games
Journal of Economics & Management Strategy
Prior research on user innovation fails to explain its low adoption rate and neglects its impact on increased product stickiness. To bridge these gaps, we conducted an empirical investigation into user innovations within the video game sector. Our study reveals that embracing user innovation leads to an upsurge in the number of active players for a game. Furthermore, the marginal effect of user innovations varies depending on their recency and quality, with low-quality user innovations leading to user attrition. The effect is also contingent on the stage in the product life cycle in which user innovation is adopted.
Yunfei Wang, UMD and Peng Huang, UMD
Seed Accelerators, Information Asymmetry, and Corporate Venture Capital Investments
Management Science
Beyond financial incentives, investments by Corporate Venture Capitalists (CVCs) are often motivated by strategic objectives, such as gaining early exposure to emerging technologies. However, in the presence of information asymmetry, CVCs tend to invest in startups with a high degree of business relatedness—startups that are less risky but lacking in knowledge novelty—which are not ideal for achieving their strategic objectives. With startup accelerators showing promise in mitigating the information asymmetry problem, we examine how a CVC’s investment pattern in a region shifts following a startup accelerator’s entry, with a particular interest in the degree of business relatedness between the CVC’s parent corporation and its portfolio companies. Analyses reveal that CVCs increase investments in startups that are dissimilar to their parent’s business following the entry of startup accelerators. We show that the two pathways through which accelerators reduce information asymmetry—quality signals, and mentorship and training—likely contribute to this change. In addition, the change is most pronounced for CVCs whose parent firm operates in an IT-using—rather than an IT-producing—industry, suggesting that accelerators help IT-using firms gain a foothold in the technology space through CVC investments. These findings deepen the understanding of the role that startup accelerators play in the entrepreneurial ecosystem against the backdrop of digital transformation occurring in nearly every industry.
Raveesh Mayya, NYU and Peng Huang, UMD
Prompt Adaptation as a Dynamic Complement in Generative AI Systems
As generative AI systems rapidly improve, a key question emerges: How do users keep up—and what happens if they fail to do so. Drawing on theories of dynamic capabilities and IT complements, we examine prompt adaptation—the adjustments users make to their inputs in response to evolving model behavior—as a mechanism that helps determine whether technical advances translate into realized economic value. In a preregistered online experiment with 1,893 participants, who submitted over 18,000 prompts and generated more than 300,000 images, users attempted to replicate a target image in 10 tries using one of three randomly assigned models: DALL-E 2, DALL-E 3, or DALL-E 3 with automated prompt rewriting. We find that users
with access to DALL-E 3 achieved higher image similarity than those with DALL-E 2—but only about half of this gain (51%) came from the model itself. The other half (49%) resulted from users adapting their prompts in response to the model’s capabilities. This adaptation emerged across the skill distribution, was driven by trial-and-error, and could not be replicated by automated prompt rewriting, which erased 58% of the performance improvement associated with DALL-E 3. Our findings position prompt adaptation as a dynamic complement to generative AI—and suggest that without it, a substantial share of the economic value created when models advance may go unrealized.
Eaman Jahani, UMD
Benjamin Manning, MIT
Hong-Yi TuYe, MIT
Mohammed Alsobay, MIT
Christos Nicolaides, University of Cyprus
Siddharth Suri, Microsoft Research
David Holtz, Columbia
Improved LISA Analysis for Zero-Heavy Crack Cocaine Seizure Data
INFORMS Journal of Data Science
Local Indicators of Spatial Association (LISA) analysis is a useful tool for analyzing and extracting meaningful insights from geographic data. It provides informative statistical analysis that highlights areas of high and low activity. However, LISA analysis methods may not be appropriate for zero-heavy data, as without the correct mathematical context the meaning of the patterns identified by the analysis may be distorted. We demonstrate these issues through statistical analysis and provide the appropriate context for interpreting LISA results for zero-heavy data. We then propose an improved LISA analysis method for spatial data with a majority of zero values. This work constitutes a possible path to a more appropriate understanding of the underlying spatial relationships. Applying our proposed methodology to crack cocaine seizure data in the U.S., we show how our improved methods identify different spatial patterns, which in our context could lead to different real-world law enforcement strategies. As LISA analysis is a popular statistical approach that supports policy analysis and design, and as zero-heavy data is common in these scenarios, we provide a framework that is tailored to zero-heavy contexts, improving interpretations and providing finer categorization of observed data, ultimately leading to better decisions in multiple fields where spatial data is foundational.
Eunseong Jang, The Robert H. Smith School of Business, University of Maryland
Margret Bjarnadottir, The Robert H. Smith School of Business, University of Maryland
Marcus Boyd, National Consortium for the Study of Terrorism and Responses to Terrorism, University of Maryland
S. Raghavan, The Robert H. Smith School of Business & Institute for Systems Research, University of Maryland