An important debated topic in strategic management concerns the so-called “chief executive officer (CEO) effect,” which quantifies the impact that CEOs have on the performance of the firms that they lead. Prior literature has empirically investigated the CEO effect and found support for both theses: a significant effect and no effect at all. We note, however, that virtually all prior studies have relied on an empirical specification that leverages in-sample data, which could be unreliable in certain circumstances. In this paper, we utilize machine learning models and predictive analytics based on out-of-sample data to revisit the CEO effect. In particular, we operationalize the CEO effect as the gain in the out-of-sample predictive accuracy by adding the CEO information to the model input in addition to the firm information. By analyzing 1,245 firms and 1,779 CEOs over 20 years, we demonstrate that the results of the approach from the literature have limited external validity. More specifically, we convey that the analyses are purely based on in-sample data and that the predictive effects of CEOs are not substantive when out-of-sample test data sets are used. Although our main analysis relies on optimized distributed gradient boosting, we also conduct extensive robustness tests spanning close to 100 models with alternative algorithms and specifications, all of which yield consistent results. This paper was accepted by Joshua Gans, business strategy. Supplemental Material: The online appendix and data files are available at https://doi.org/10.1287/mnsc.2023.03625 .

This paper leverages new measurement of neighborhood consumption amenities to demonstrate that housing prices and rents in U.S. cities are likely determined nearly as much by access to amenities as by access to employment. We extend the Alonso-Muth-Mills model, allowing residents to derive utility from within-city trips to amenities. The model delivers standard estimable log-linear pricing equations as well as new measures of local amenities-based on a destination's popularity during leisure hours-and of access to consumption amenities city wide. We find our amenity measures add substantial explanatory power, have large effects in magnitude, and reduce naive estimates of commute costs by 30%. Elasticities of rents with respect to employment access are 20%-50% larger than those with respect to amenity access. The findings hold using a variety of alternative measures and are neither driven by density nor fully explained by the locations of business establishments. These results suggest the potential resilience of cities to changes in employment locations.

Following the increasing popularity of personalized pricing, there is a growing concern from customers and policymakers regarding fairness considerations. This paper studies the problem of dynamic pricing with unknown demand under two types of fairness constraints: price fairness and demand fairness. For price fairness, the retailer is required to (i) set similar prices for different customer groups (called group fairness) and (ii) ensure that the prices over time for each customer group are relatively stable (called time fairness). We propose an algorithm based on an infrequently changed upper confidence bound (UCB) method, which is proven to yield a near-optimal regret performance. We then leverage this method to address the extension of nonstationary demand, which is particularly relevant for time fairness, to prevent price gouging practices. For demand fairness, the retailer is required to satisfy the condition that the resulting demand from different customer groups is relatively similar (e.g., the retailer offers a lower price to students to increase their demand to a level similar to that of nonstudents). In this case, we design an algorithm adapted from a primal-dual learning framework and prove that our algorithm also achieves a near-optimal regret performance.

Discrimination in machine learning (ML) has become prominent as ML is increasingly used for decision-making. Although many "fair-ML" algorithms have been designed to address such discrimination issues, virtually all of them focus on alleviating disparity in the prediction results by imposing additional constraints. Naturally, in response, prediction subjects alter their behaviors. However, the algorithms never consider those behavioral responses. So, even if the disparity in prediction results may be removed, the disparity in behaviors may persist across different subpopulations of prediction subjects. When these biased behavioral outcomes are used for training ML algorithms, they can perpetuate discrimination in the long run. To study this issue, we define a new notion called "strategic best-response fairness" (SBR-fairness). It is defined in a context involving subpopulations that are ex ante identical and also have identical conditional payoffs. Even if an algorithm is trained on biased data, will it lead to identical equilibrium behaviors of subpopulations? If yes, we define the ML as SBR-fair. We then use this SBR-fairness framework to analyze the property of existing fair ML algorithms. We also discuss how the SBR-fairness framework can inform the design of fair ML algorithms and the practical and policy implications of SBR-fairness.

Predicting the demand of products in the retail industry is a complex task, especially when there are changes in the market. This paper examines three such market shifts in the retail industry: the COVID-19 pandemic, opening a new store, and introducing a new product. Our study found that the accuracy of demand prediction models decreases after these market shifts. To address this problem, we propose the use of domain adaptation methods, such as Frustratingly Easy and Kernel Mean Matching, to improve the accuracy of predictions by utilizing data from before the changes and adapting to the data after the changes. We show that using a pairing technique can further enhance prediction accuracy. Two retail demand forecasting models, XGBoost and Transformers, were assessed and XGBoost was found to be more effective. We demonstrate the effectiveness of the domain adaptation methods on a real-world case using point-of-sale data from 89 locations of Alimentation Couche-Tard convenience stores in Montreal between 2019-07 and 2021-02 focusing on the two best-selling product categories of coffee and energy drinks.

Free AccessAboutSectionsView PDF ToolsAdd to favoritesDownload CitationsTrack CitationsPermissionsReprints ShareShare onFacebookTwitterLinked InEmail Go to SectionFree Access HomeService ScienceAhead of Print Call for Papers: Service Science Special Issue on the Impact of AI on Service Design and DeliveryMaxime Cohen , Tinglong Dai , Beibei Li Maxime Cohen , Tinglong Dai , Beibei Li Published Online:8 Apr 2024https://doi.org/10.1287/serv.2024.cfp.v16.n2The integration of artificial intelligence (AI), including both predictive AI and generative AI, into the design and delivery of services represents a fundamental shift in how services are conceived, implemented, and experienced. This special issue of Service Science, an INFORMS journal, aims to explore the broad and diverse impact of AI on service sectors ranging from healthcare to finance, education, and beyond. We invite papers that explore how AI technologies and algorithms are transforming service landscapes, enhancing customer experiences, and reshaping operational processes.We are particularly interested in papers that address, but are not limited to, the following topics:The role of AI in innovating service design and improving service deliveryThe role of management science and principles in shaping AI development and deploymentAI-driven personalization and customization of servicesEthical considerations and challenges in AI-augmented servicesThe impact of AI on service agent roles and skill requirementsThe role of AI in improving service accessibility and inclusivityCustomer perceptions and trust in AI-augmented servicesModels and frameworks for understanding the impact of AI on servicesMethodological advances for studying AI in service contextsInnovative applications of generative AI to the service sectorWe welcome submissions using analytical, empirical, experimental, and qualitative methods. Papers should not only demonstrate the role of AI in service transformation but also discuss the implications for service theory, practice, and policy.Authors who are considering whether their research project fits the scope of the special issue are encouraged to email a brief description (no more than one page) of their project to the special issue editors. This initial interaction is intended to provide feedback on the relevance of the project to the goals of the special issue. Although this step does not evaluate the quality of the research, it does serve to ensure alignment with the themes of the special issue. The quality and appropriateness of full submissions will be determined through a peer review process involving both the existing Service Science editorial board and additional experts as needed.There is no obligation to submit a project description before submitting a full paper, but it is an available option for authors seeking preliminary feedback.Submission Process and TimelineAll submissions should be submitted via the Service Science online submission system: https://mc.manuscriptcentral.com/serv. All submissions will be subject to the journal's standard peer review process. Criteria for acceptance include originality, contribution, and scientific merit. For submission guidelines, please visit the journal's home page to learn more: https://pubsonline.informs.org/page/serv/submission-guidelines.Deadline for submission: October 1, 2024First-round decision and feedback: December 1, 2024Second-round submission (for those papers invited to revise): June 1, 2025Final decisions (subject to minor revisions): September 1, 2025We look forward to receiving your submissions and advancing the discourse on the transformative power of AI in service design and delivery. Back to Top Next FiguresReferencesRelatedInformation Articles In Advance Article Information Metrics Information Published Online:April 08, 2024 Copyright © 2024, INFORMSCite asMaxime Cohen, Tinglong Dai, Beibei Li (2024) Call for Papers: Service Science Special Issue on the Impact of AI on Service Design and Delivery. Service Science 0(0). https://doi.org/10.1287/serv.2024.cfp.v16.n2 PDF download

This paper introduces RISE, an innovative framework for robust individualized decision learning framework with sensitive variables. Sensitive variables are crucial data for intervention decisions, but often need to be excluded during decision-making due to factors like delayed availability or fairness concerns. Conventional practices neglect these variables in learning decision rules, resulting in significant uncertainty and bias. To rectify this, we propose a decision learning framework that integrates sensitive variables during offline training while excluding them from the input of the learned decision rule during model deployment. From a causal perspective, our framework aims to enhance worst-case outcomes for individuals influenced by sensitive variables unavailable at decision time. Diverging from prevalent mean-optimal objectives, we present a robust learning framework by identifying a newly defined quantile- or infimum-optimal decision rule. We introduce methods for both single-stage and multistage decision-making scenarios and extend the framework to accommodate multiple sensitive variables, continuous treatments, and enhance the robustness of learned decisions beyond their robustness in the value function. The efficacy of our proposed method is substantiated through synthetic experiments and real-data applications spanning economics, healthcare, and personalized pricing. Our approach represents a significant departure from existing literature, emphasizing robustness in decision learning amidst sensitive variables and showcasing tangible improvements across various domains.

Problem definition: How can retailers incentivize customers to make healthier food choices? Price, convenience, and taste are known to be among the main drivers behind such choices. Unfortunately, healthier food options are often expensive and not adequately promoted. However, we are observing recent efforts to nudge customers toward healthier food. Methodology/results: In this paper, we conducted a field experiment with a global convenience store chain to better understand how different add-on bundle promotions influence healthy food choices. We considered three types of add-on bundles sequentially: (i) an unhealthy bundle (when customers purchased a coffee, they could add a pastry for $1), (ii) a healthy bundle (offering a healthy snack, such as fruit, vegetable, or protein, as a coffee add-on for $1), and (iii) a choice bundle (the option of either a pastry or a healthy snack as an add-on to coffee for $1). In addition to our field experiment, we conducted an online laboratory study to strengthen the validity of our results. Managerial implications: We found that offering healthy snacks as part of an add-on bundle significantly increased healthy purchases (and decreased unhealthy purchases). Surprisingly, this finding continued to hold for the choice bundle, that is, even when unhealthy snacks were concurrently on promotion. However, we did not observe a long-term stickiness effect, meaning that customers returned to their original (unhealthy) purchase patterns once the healthy or choice bundle was discontinued. Finally, we show that offering an add-on choice bundle is also beneficial for retailers, who can earn higher revenue and profit. Funding: This research was supported by the James McGill Scholar Award Fund, the Scale AI Chair Program, IIVADO (Institut de valorisation des données) Fundamental Research Project Grant, and two Discovery Grants from the Natural Sciences and Engineering Research Council of Canada. Supplemental Material: The online appendix is available at https://doi.org/10.1287/msom.2023.0336 .

Problem definition: Restaurant review platforms, such as Yelp and TripAdvisor, routinely receive large numbers of photos in their review submissions. These photos provide significant value for users who seek to compare restaurants. In this context, the choice of cover images (i.e., representative photos of the restaurants) can greatly influence the level of user engagement on the platform. Unfortunately, selecting these images can be time consuming and often requires human intervention. At the same time, it is challenging to develop a systematic approach to assess the effectiveness of the selected images. Methodology/results: In this paper, we collaborate with a large review platform in Asia to investigate this problem. We discuss two image selection approaches, namely crowd-based and artificial intelligence (AI)-based systems. The AI-based system we use learns complex latent image features, which are further enhanced by transfer learning to overcome the scarcity of labeled data. We collaborate with the platform to deploy our AI-based system through a randomized field experiment to carefully compare both systems. We find that the AI-based system outperforms the crowd-based counterpart and boosts user engagement by 12.43%–16.05% on average. We then conduct empirical analyses on observational data to identify the underlying mechanisms that drive the superior performance of the AI-based system. Managerial implications: Finally, we infer from our findings that the AI-based system outperforms the crowd-based system for restaurants with (i) a longer tenure on the platform, (ii) a limited number of user-generated photos, (iii) a lower star rating, and (iv) lower user engagement during the crowd-based system. Funding: The authors acknowledge financial support from the Social Sciences and Humanities Research Council [Grant 430-2020-00106]. Supplemental Material: The online appendix is available at https://doi.org/10.1287/msom.2021.0531 .

Despite the extensive research on healthcare operations, a need remains for a deeper understanding of how to design and deploy new technologies to achieve efficiency, particularly concerning health-promoting behavior. Health-promoting behavior, driven by intrinsic motivations, often exhibits reduced efficacy as motivational impetus wanes over time. Despite the prevalence of various health interventions, there has been limited exploration of their efficacy within real-time motivation schemes. This study posits that real-time interventions grounded in social influence significantly impact users’ physical performance. In collaboration with a running app service platform, we examine how real-time voice-based interventions based on social facilitation influence users’ physical performance. Our key findings indicate that the recognition of communal exercise and the receipt of vocal encouragement in real time enhance users’ running performance. Notably, we observe a synergistic effect of these factors, particularly when a female supporter is involved. Additionally, we find heterogeneous effects based on the recipient’s gender, user performance, intervention timing, and running state. We also highlight that our intervention elicited a spillover effect on social interactions within the app. These findings have significant implications for the development of real-time interventions, accentuating the importance of social influence in enhancing health-promoting behavior.

Problem definition: : Opportunity zones (OZs) are designated census tracts in which real estate investments can gain tax benefits. Introduced by the U.S. Tax Cuts and Jobs Act of 2017, the goal of the OZ program is to foster economic development in distressed neighborhoods. In this paper, we investigate and optimize the OZ selection process and examine the impact of OZs by exploiting two data sets: a proprietary real estate data set that includes 36.1 million residential transactions spanning all 50 U.S. states and censustract demographics data between 2010 and 2019. Methodology/results: : We show that census tracts with higher poverty and unemployment rates were more likely to be selected. Counterintuitively, however, tracts with a higher average real estate price were also more likely to be selected. We then apply difference-in-differences, synthetic control, and matching techniques to rigorously assess the impact of the OZ program on two key real estate metrics: price and transaction volume. We find that the OZ program increased real estate prices by 4.03%-6.13% but do not observe a significant effect on the transaction volume. We also find that investors primarily targeted the high-end real estate market, namely, exhibiting a cherry-picking behavior. To better fulfill its intended societal and economic goals, we propose an optimization framework with fairness considerations for OZ assignment decisions. We show that the OZs assigned from our fairness-aware optimization formulation can better serve distressed communities and mitigate investors' cherry-picking behavior. Managerial implications: : Our paper underscores the importance of incorporating fairness in OZ designation to achieve a desirable real estate market reaction. Our largescale empirical analysis provides a comprehensive assessment of the current government OZ assignment, and our fairness-aware optimization framework provides concrete recommendations for policy makers.

Housing discrimination has been recognized as an important societal issue for decades. While this issue can manifest in multiple manners, one of the most common ways that housing discrimination is observed is through price discrimination, where houses in white-dominant neighborhoods are worth more than houses in black-dominant neighborhoods that are otherwise similar. Prior studies have empirically documented such pricing discrimination and attributed the source of discrimination to human biases. In addition, recent studies have shown that such an issue is unlikely to be addressed by traditional AI models, even for those specifically designed to address discrimination in AI. In this paper, we first compare AI-generated versus human-generated housing selling prices using a sample of 285,853 U.S. properties. We then study the impact of generative AI in the context of price discrimination in the housing market. Surprisingly, we find that generative AI can help alleviate this issue. Practical and policy implications are also discussed.

The principle that “eye-level is buy level” underpins planogram optimization and retailer-manufacturer agreements. A key assumption underlying this principle is that moving a product to an eye-level shelf consistently increases sales, regardless of how the overall planogram is reorganized. Our study challenges this assumption by examining how different planogram reorganizations moderate the eye-level effect. We conducted a field experiment in collaboration with a North American convenience store chain, testing eye-level placement under two planogram reorganizations: 1-swap (swapping the vertical locations of two product sets) and 2-swap (swapping three product sets). We find that planogram reorganization significantly moderates the eye-level effect. Moving products from stretch-level (stoop-level) to eye-level increased sales by 5.6% (10%) under a 1-swap reorganization, but did not change sales under 2-swap. We attribute this to partial visual search, where some customers scan only the stretch-level and eye- level shelves. Different planogram reorganizations create distinct choice sets for these customers, resulting in different levels of preference for products placed at eye-level. The difference in eye-level effects between 1-swap and 2-swap also impacts overall sales, with the effective reorganization type varying by shelving unit. When moving a product to eye-level, retailers should customize planogram reorganization based on product preference patterns within each shelving unit. Our counterfactual analysis reveals that compared to assuming no moderation effect of planogram reorganization, uniformly applying the dominant 1-swap strategy across all shelves can boost sales by 3.85%, while tailoring the reorganization for each shelving unit can increase sales by 4.32%.

Problem definition: Are customers loyal to a ride-hailing platform or they see this service as a commodity and multihome (i.e., check several platforms before booking a ride)? Using a large panel dataset on ride-hailing transactions, we investigate to what extent customers multihome. Our dataset offers a unique opportunity to study this question as we observe the repeated choices of riders for both Uber and Lyft. Our dataset comprises more than 1.4 million rides completed by 162 thousand riders in NYC in 2018. Methodology/results: We develop a comprehensive structural model that incorporates both operational (price and waiting time) and behavioral factors (e.g., platform stickiness) to explain riders' choices. Our model also accounts for the dynamic interactions between customers and platforms by assuming that riders update their beliefs on price and waiting time in a Bayesian fashion. Finally, the riders' propensity to multihome is modeled by incorporating the consideration set formation of customers into our framework. We find that riders' choices are not fully explained by operational factors, hence indicating that customers view the platforms as differentiated service providers. While 83.4% of riders took rides with a single platform, our model shows that even the remaining 16.6% , who used both Uber and Lyft at least once, considered both platforms only 43.4% of the time. Managerial implications: It is crucial for ride-hailing platforms to capture this single (or multi)-homing behavior while designing promotions. Specifically, personalized promotions may be ineffective if the platform is not part of the customer’s consideration set. Our results show that targeting customers earlier in their lifecycle can enhance the platform’s market share by 77.56% more than their current promotional strategy. We also find that targeting customers with low search friction results in a 24.78% increase in market share relative to targeting customers with high search friction.

The COVID-19 pandemic has severely disrupted the retail landscape and has accelerated the adoption of innovative technologies. A striking example relates to the proliferation of online grocery orders and the technology deployed to facilitate such logistics. In fact, for many retailers, this disruption was a wake-up call after which they started recognizing the power of data analytics and artificial intelligence (AI). In this article, we discuss the opportunities that AI can offer to retailers in the new normal retail landscape. Some of the techniques described have been applied at scale to adapt previously deployed AI models, whereas in other instances, fresh solutions needed to be developed to help retailers cope with recent disruptions, such as unexpected panic buying, retraining predictive models, and leveraging online-offline synergies.

Airfares evolve dynamically, giving rise to a so-called price path . This price path is controlled via two levers: (i) a fare ladder, which defines a set of airfares before the selling season, and (ii) revenue management algorithms, which control how fares evolve along the ladder during the season. We hypothesize that the current policies to control both levers—which do not account for quality differences between competing airlines—give rise to an inefficient price path and, accordingly, a loss of potential revenue. We substantiate this hypothesis via a field experiment. By partnering with an airline, we introduced quality considerations in the design of fare ladders, across 5,000 itineraries, to show that current ladder-design policies indeed lead to a suboptimal price path. We also show that this inefficiency can be mitigated by incorporating quality differences between competing airlines. This creates a smoother (and more profitable) price path. This paper was accepted by Vishal Gaur, operations management.

Buying display ad impressions via real-time auctions comes with significant allocation and price uncertainties. We design and analyze a contract that mitigates this uncertainty risk by providing guaranteed allocation and prices while maintaining the efficiency of buying in an auction. We study how risk aversion affects the desire for guarantees and how to price a guaranteed allocation. We propose to augment the traditional auction with a programmatic purchase option (which we call a Market-Maker contract ) that removes allocation and price uncertainties. Instead of participating in the auction, advertisers can secure impressions in advance at a fixed premium price offered by the Market-Maker. It is then the responsibility of the Market-Maker to procure these impressions by bidding in the auction. We model buyers as risk-averse agents and analyze the equilibrium outcome when buyers face two purchase options (auction and Market-Maker contract). We derive analytical expressions for the Market-Maker price that reveal insightful relationships with uncertainties in the auction price and buyers’ risk levels. We also show the existence of a Market-Maker price that simultaneously improves the seller’s revenue and the sum of buyers’ utilities. As a building block to our analysis, we establish the truthfulness of the multiunit auction when buyers have nonquasilinear utilities because of risk aversion. Recently, the Google’s Display & Video 360 platform started offering a product akin to Market-Maker called “Guaranteed Packages,” which was inspired by this paper. This paper was accepted by Gabriel Weintraub, revenue management and market analytics. Funding: The authors thank Google Research for its generous support.

We provide new evidence that short-term rental (STR) platforms like Airbnb incentivize residential real estate investment. We exploit two complementary identification strategies. First, we use variation in the timing of STR regulations to estimate the effect of regulation on both Airbnb listings and residential permits. We find that over the first 12 months following the start of the regulation, STR regulations reduce Airbnb listings by 8.9% and residential permits by 10.8%. Second, we show that residential permits decline discontinuously across jurisdictional boundaries in which one side of the boundary has a STR regulation and the other side does not. The effect is especially striking for accessory dwelling units, which decline by 16.5% across regulatory boundaries. Our results imply that STRs incentivize residential investment, and especially so for housing units that are well suited for short-term renting.