This paper provides causal evidence that local social media news significantly promotes small businesses’ adoption of sustainable food innovations, with stronger effects in liberal-leaning regions.
In recent years, an increasing number of firms have joined coalition loyalty programs (CLPs), where they collaborate with other firms in their loyalty program initiatives. However, CLPs remain less studied compared with proprietary loyalty programs (PLPs) offered by individual firms. This paper studies CLPs and compares them with PLPs using an analytical framework where n firms sell a nondurable product to infinitesimal heterogeneous customers over an infinite horizon. We analyze the design of CLPs and show that CLPs can significantly expand the range of market conditions under which offering reward programs is desirable. That is, firms have incentives to join CLPs, even when PLPs are ineffective, as CLPs enable more effective price discrimination. Our research also reveals the critical role of market composition in the effectiveness of CLPs-the optimal design of CLPs depends critically on the relative prominence of customer heterogeneity in shopping intensity and product valuation. Specifically, smaller CLPs are preferred in markets where customers' heterogeneity in shopping intensity is more prominent than that in valuation. This also implies that larger CLPs are not always preferred, offering one explanation for the struggles of some CLPs in their expansion. We check the robustness of our results with extensions.
This paper examines how publishers should set reserve prices in display advertising auctions when advertisers face practical constraints such as campaign reach.
European Union-mandated changes to Google results increased map-related searches but produced little traffic gain for competitors, underscoring Google Maps’ dominance.
This study examines how the dispute between TikTok and Universal Music Group (UMG) impacted music demand on streaming platforms.
Designing centralized matching systems on two-sided platforms entails a tradeoff between matching efficiency and agent autonomy. We investigate this tradeoff in the ride-hailing market by comparing two prevalent centralized dispatch systems: a driveraccept system, where drivers decide whether to accept e-hail requests, and an auto-accept system, where drivers are automatically assigned to these requests. We develop a dynamic structural model of a two-sided market where strategic drivers optimize acceptance and relocation decisions over a shift, and riders choose between transportation modes. We apply this model to Singapore's taxi market, where a leading operator used a driver-accept dispatch system for e-hail trips. We develop an iterative algorithm to solve for the market equilibrium as a fixed point of a nested loop. Our counterfactual analyses reveal that in the short-term equilibrium with fixed demand, switching to an auto-accept system increases driver earnings through higher vehicle utilization, despite the loss of private information caused by removing driver choice. In the long-term equilibrium, drivers' earnings increase further due to market expansion, although benefits are unevenly distributed between e-hail and street-hail services. Consumer surplus increases in both scenarios. Our findings highlight that centralized assignment can enhance overall welfare by mitigating search frictions, even when agents value autonomy.
We investigate organic large language model traffic (oLLM) versus traditional digital channels in e-commerce. Analyzing 12 months of first-party data from 973 websites with $20 billion combined revenue, we examine more than 50,000 transactions from ChatGPT referrals alongside 164 million transactions from traditional channels. Using regression models that account for data sparsity, we assess financial metrics (conversion rate, average order value, revenue per session) and engagement metrics (bounce rate, session duration, page views). Results are consistent across extensive robustness checks. One year after launch, oLLM exhibits conversion rates and revenue per session above paid social but below all other traditional channels. Product complexity moderates the effects: oLLM's financial outcomes and traffic shares are stronger in complex product categories. Engagement metrics show favorable bounce rates but lower session duration and page views. Temporal analysis shows increasing conversion rates but declining average order values, yielding only moderate revenue-per-session gains over time. Cross-website analyses support growing consumer LLM proficiency as the underlying mechanism. The descriptive study positions oLLM as a new and developing channel. With low volumes and modest revenue per session, oLLM currently serves niche informational needs of proficient consumers and does not yet function as a broad conversion channel.
Digital platforms increasingly engage in quality controls by regulating product listings and transactions. Although their immediate effect on the involved sellers is evident, their carryover effect on adopters and their spillover effect on nonadopters remain unclear. We address this gap using a natural experiment on a secondhand platform. The platform introduced a submarket with quality controls in the mobile phone category. The submarket offers quality checks, detailed quality reports, standardized listing information, recommended prices, post-purchase guarantees, and free delivery. We find that the qualitycontrolled submarket not only immediately increases adopters' sales but also has a positive carryover effect. After adopting, sellers learn from the platform's quality standards and improve their subsequent listing behaviors and sales performance, even outside the qualitycontrolled submarket. Moreover, we uncover a U-shaped effect for nonadopters; when the adoption rate is low, nonadopters are disadvantaged because of intensified competition from adopters; however, as adoption increases, they benefit from enhanced buyer perception of platform quality. The competition effect and the perception improvement effect dominate at different stages of adoption. These findings have important implications for platform strategies and policies concerning anticompetitive effects of platform quality controls.
We investigate the informational content of online reviews regarding hygiene standards. Using data from Yelp and New York City restaurant inspections, we show that online reviews are informative about dimensions of hygiene that consumers directly experience but provide limited information on other dimensions. We present causal evidence that restaurant demand responds to hygiene signals in online reviews and that restaurants with higher visibility on review platforms violate less along dimensions for which reviews are more informative. These findings highlight the evolving role of consumer reviews for regulators and firms.
We examine the implications of competitive algorithmic targeting when outcomes of targeting algorithms are the individual consumer-level predicted probabilities of conversion. In these situations, firms implicitly face the well-known precision-recall tradeoff while choosing their targeting strategies. They can choose to target a smaller set of consumers with a high probability of conversion (precision) but miss out on many consumers who might still be interested in their product. Conversely, firms can target a larger set of consumers (recall), but this results in a greater probability that their targeting is wasted on uninterested consumers. We analyze this precision-recall tradeoff under competition between firms that strategically choose their algorithmic targeting policies. We show that competing firms favor a targeting policy that has higher precision but lower recall compared with a monopoly. Firms target fewer consumers when their algorithms are more correlated. They also have the incentive to strategically decrease the precision of their targeting policies in order to reduce competition. If firms endogenously choose their algorithmic correlation, then there is an equilibrium incentive to decrease the correlation.
With the massive growth of social media and other informational platforms that businesses and individuals may use to share their opinions, a society's perspectives of certain contentious issues may be influenced significantly. At the same time, it has been noted that various parties have used paid trolls (i.e., fake comments made by bot accounts) to sway public opinions concerning some hotly debated issues or legal proceedings. This article aims to investigate how public opinions and paid trolls may affect the outcome of legal disputes between opposing parties. Our research indicates that litigating parties involved in lawsuits with social media engagement may purchase paid trolls to influence public opinions, and they do so more significantly when the truth does not match the prior expectation of the public. We also discover that social media users' investigation may be a pitfall for the involved parties. Additionally, the court may be either constrained or aided to identify the truth by having a tendency to render decisions that align with prevailing public opinions, depending on the stake of the lawsuit. Finally, treating public opinions on social media as additional information that aids court decision may backfire.
Generative artificial intelligence (AI) for image synthesis has the potential to transform the digital advertising industry. However, a wide range of uncertainties persists regarding its integration into traditional advertising processes, including finding effective implementations, training methodologies, and achievable performance gains. Specifically, two core challenges limit its practical adoption: a search problem of finding highperforming visuals in a vast creative space, and an alignment problem of ensuring brand and campaign compatibility. This paper proposes a novel end-to-end framework that combines a generative AI with two predictive Bayesian neural networks to identify highperformance and brand-acceptable visuals. We develop a cost-effective Bayesian active learning approach solving simultaneously the dual objectives of performance and alignment. We test the framework in a live advertising campaign for an outdoor activities company. Our system generated a portfolio of visuals achieving a higher mean click-through rate and more consistency (lower variance) than creatives from both a professional human designer and a competing AI model optimizing purely for aesthetics. This research provides a validated methodology that bridges the gap between the theoretical potential of generative AI and its practical application, offering a cost-effective solution to the critical search and alignment problems in creative design.
Class pricing describes a widespread practice of assigning a few price points to a large set of differentiated products. Although previous literature proposes firms' costly price setting activities as a friction-based explanation, I offer a consumer-driven rationale rooted in reference-dependent and loss-averse consumer behaviors. I develop a model incorporating a monopolistic firm selling multiple products to a continuum of consumers with heterogeneous tastes and employ the expectations-based prospect theory proposed by Koszegi and Rabin to depict customers' reference points as endogenously determined via aligning their optimal choices to their rational expectations about consumption outcomes. I find that although product cost variations motivate unequal prices to stimulate demand for lower-cost products, loss aversion constrains this practice; the resulting demand shift asymmetrically diminishes consumers' willingness to pay (WTP) for lower-cost products more significantly than it can elevate WTP for higher-cost alternatives. This asymmetry reduces the firm's total profit from multiple products and drives the adoption of class pricing. My research contributes to a better understanding of class pricing for both academic and managerial practice, and it provides insights on when more prices are not necessarily advantageous in an era of information and emerging artificial intelligence technologies.
Although there are well-established model selection methods (e.g., Bayesian Information Criterion (BIC)), they commonly condition on a priori selected data and parameter granularities. That is, researchers think they are doing model selection, but what they are really doing is model selection conditional on their chosen granularities. We propose a new method, Bayesian dual clustering (BDC), that infers both data and parameter granularities by sampling over their posterior distribution. BDC represents data and parameters as two separate collections of nodes (e.g., stock keeping unit (SKU)) with each node being the unit of analysis. The method then clusters the nodes in each collection and infers the corresponding data and parameter granularities while providing a high degree of interpretability regarding why certain granularities are selected. Notably, BDC can handle large collections, accommodate parameter restrictions (e.g., data need to be at least as granular as parameters) using a split-merge sampler, and relate to other extant methods (e.g., latentclass analysis). We apply BDC to a frequently purchased grocery category. The results show that BDC inferred granularities differ from those from extant approaches, which, in turn, leads to different demand elasticities and optimal actions. We conclude by highlighting the generalizability of BDC to a broad array of marketing problems.
Given the increasing importance of user engagement on digital platforms, this paper proposes a Bayesian deep-learning model called the Multi-Dynamic Neural Poisson System (MDNPS), which can capture the multifaceted nature of user sparse but highdimensional activities on such platforms. MDNPS yields semantically interpretable factors underlying high-dimensional items, quantifies user preferences at different granularity levels, and adeptly captures their temporal and cross-activity dependencies. This model is scalable to large empirical data and can be inferred efficiently with a stochastic variational Bayes algorithm. We apply MDNPS to the largest knowledge-sharing platform in China, focusing on the dynamics in user content consumption and contribution behaviors. We show that MDNPS significantly outperforms benchmark factor models in factor quality and out-of-sample data fitting. Our platform-level and individual-level estimates show rich and interpretable insights about consumer preference dynamics and their relationships with user popularity. These insights offer valuable managerial implications for platforms to understand user segments, personalize content offerings, and enhance user engagement.
We descriptively document a robust U-shaped relationship between income and counterfeit consumption using large-scale field data on U.S. consumers. Relative to the middle-income cohort ($50k-$75k), both low-income (< $15K) and high-income (> $150K) consumers are more active in counterfeit markets: they purchase more, buy more repeatedly, and disproportionately choose higher-priced and niche listings as well as brands' classic product series. The tails differ in composition: relative to the middle-income cohort, low-income demand loads more on lower-tier brands, whereas high-income demand tilts toward ultraluxury and toward higher-priced listings within a series, consistent with greater willingness to pay for counterfeit quality. A cross-brand copurchase network corroborates these patterns, with pronounced clustering by product category and brand position. We discuss implications for theory and practice.
This study introduces a novel instrumental variable (IV) for estimating the causal effects of linear television (TV) advertising using large-scale panel data that link household second-by-second show viewership and ad exposure with daily purchase behavior. We exploit an institutional feature of linear TV: Although advertisers choose which shows to target, networks quasi-randomly determine within-show ad airing times. This creates exogenous variation in focal brand ad exposure among partial show viewers, which we nonparametrically extract to construct a household-show-level IV. We establish the IV's validity in the presence of endogeneity arising from advertisers' show targeting decisions and households' TV viewing behavior. Our IV offers a generalizable and flexible solution for household-level linear TV ad effect measurement using modern single-source data. Applying this method to data from a major food delivery platform, we estimate an ad response model in which both baseline purchase propensity and ad responsiveness vary with purchase history. Na & iuml;ve estimates overstate ad elasticities by 55% compared with IV-corrected estimates. We also find that ad responsiveness is nonmonotonic with respect to purchase frequency and recency. These findings underscore the importance of addressing endogeneity in observational household TV ad exposure data and highlight the potential of behaviorally targeted TV advertising.
In this paper, we propose a Deep-DiD method that incorporates two deep neural networks in a difference-in-difference (DiD) framework to estimate heterogeneous treatment effects (HTEs). The dual-network architecture contains one neural network modeling HTEs as a nonparametric function of pretreatment features and another neural network capturing individual and time fixed effects. Through a series of simulations, we show that our method can uncover the true HTEs with high accuracy under various settings and demonstrates more robust estimation performance compared with existing methods like linear models and random forests. We apply this method to an empirical setting where a large videosharing platform introduced a "Creator Signing Program" aimed at signing creators and motivating them to generate more high-quality video content. Leveraging a matched data set of signed and unsigned creators, we employ our Deep-DiD method to estimate the HTEs of the signing program. Our method can help the platform optimize creator selection by identifying creators with the highest-estimated treatment effects. Through out-of-sample tests, we show that creators selected by the Deep-DiD method experience substantially larger actual performance jumps than those selected by the platform. Creator selection based on the Deep-DiD method also consistently outperforms that based on linear models.
Motivated by numerous real-world examples, where a centralized channel, comprising a centralized manufacturer, competes against a decentralized channel consisting of a decentralized manufacturer and a retailer (e.g., Tesla versus General Motors), we investigate the strategic value of capacity commitment in such asymmetric channels competition. Utilizing a game-theoretic model, we demonstrate that capacity commitment by the centralized manufacturer serves as a form of strategic communication to the decentralized manufacturer, leading to an increase in its wholesale price. Consequently, price competition between the centralized manufacturer and the retailer is mitigated, providing benefits to both the centralized channel and the decentralized channel. The decentralized manufacturer profits from a higher wholesale price, albeit a lower sales quantity by the retailer, potentially harming the retailer. The overall impact on the retailer, considering the mitigated competition and enhanced wholesale price, is contingent upon the degree of differentiation between the products. Ironically, although the decentralized manufacturer gains from capacity commitment by the competing centralized manufacturer, it never benefits by committing to capacity itself when capacity commitment is costless once the centralized manufacturer commits to capacity. We further generalize our model to show that the channel structure of the rival is a fundamental driver in unlocking the strategic value of capacity commitment for a firm. Additionally, we explore various extensions to validate the robustness of our findings.
