RECOMMENDATION SYSTEMS, CONSUMER PRIVACY, AND THE ECOSYSTEMS OF ONLINE PLATFORMS

Abstract

My dissertation investigates the challenges and trade-offs that online platforms face in implementing popularity-based and personalized recommendation systems. In the first paper, I focus on the impact of the bestseller recommendation—a widely used popularity-based system—on consumers, sellers, and the online platform when products are not vertically differentiated and consumers are heterogeneous. We model the entire ecosystem, analyzing not only the effects of the bestseller recommendation system but also how the size of the market, number of consumers, and number of sellers influence outcomes. I show that the bestseller recommendation intensifies competition among sellers, resulting in a lower equilibrium price and decreased seller participation. When sellers’ pricing competition induced by the bestseller recommendation is sufficiently intense, adopting the bestseller recommendation system is not profitable for the platform. Various extensions are analyzed to demonstrate the robustness of the key mechanism and intuition. These results highlight the importance of accounting for the strategic response of the sellers and consumers before an online platform implements the bestseller recommendation system. The second paper studies the impact of the bestseller recommendation system in a setting with homogeneous consumers, represented by a single representative agent, in order to isolate how product valuation differences affect outcomes. In a two-period model, competing sellers set prices while consumers form consideration sets and search for products before purchasing. Without a recommendation system, consumers select sellers randomly in both periods. With the bestseller recommendation, the bestselling product from the first period is always included in the consumer’s consideration set in the second period. The model explores both the informational effect (higher perceived value for the bestseller) and the search effect (guaranteed spot for the bestseller in the next period). I show that the bestseller recommendation system can increase price competition among sellers, leading to lower equilibrium prices compared to a no-recommendation scenario. This intensified competition results from both information and search effects: while the bestseller gains an advantage in the second period, non-bestsellers lower prices aggressively to remain competitive. As a result, the bestseller recommendation system generally hurts sellers and the platform, but it can become profitable when consumers have a viable outside option due to its demand-enhancing effect. These findings underscore the complexity of implementing a bestseller recommendation system and the importance of considering product differentiation and consumer search behavior. The third paper investigates the intricate dynamics between consumer privacy, personalized pricing, and platform profitability in digital markets. By modeling the interaction between consumer characteristics—such as willingness to pay and taste—privacy choices, and platform strategies, I provide a comprehensive analysis of how these elements influence one another. The findings reveal that increased privacy protection decreases platform profit and may lead to lower consumer utility, depending on the correlation between con- sumer characteristics. Notably, personalized pricing, while beneficial to platforms, often harms consumers and can diminish social welfare. The study also explores the impact of privacy on product variety, showing that high privacy levels may lead sellers to offer less diverse product offerings. Additionally, scenarios are examined in which consumers are unaware of the correlation between their taste and willingness to pay, leading to oversharing of information and increased platform profit, with mixed effects on social welfare. This study highlights the complex trade-offs in designing privacy policies and pricing strategies, offering crucial insights for platforms, regulators, and policymakers.